- %X = trunc i32 257 to i8 ; yields i8:1 - %Y = trunc i32 123 to i1 ; yields i1:true - %Z = trunc i32 122 to i1 ; yields i1:false + %X = trunc i32 257 to i8 ; yields i8:1 + %Y = trunc i32 123 to i1 ; yields i1:true + %Z = trunc i32 122 to i1 ; yields i1:false + %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> ; yields <i8 8, i8 7>

Build Chart

+ + {% for row in builds %} + + + + {% endfor %} +

+ {% for item in row %} + +

+ {% endfor %} +

References: <20110224224612.21E8B2A6C12C@llvm.org> <9DA53217-C6FC-42E4-8D03-218F85893512@apple.com>

Message-ID: On Feb 25, 2011, at 7:26 AM, Benjamin Kramer wrote: > > On 25.02.2011, at 08:14, Evan Cheng wrote: > >> Speculation is nice, but I would feel a lot better if there are hard numbers to back this up. Are we sure these changes are improving performance? > > My unscientific test case (gcc compiled with clang -O3 parsing itself) showed a 0.25% speedup with the original patch (all-constant GEPs only, median of 5 runs, i386, penryn). I just retested with the change Frits suggested and it slowed down ~0.6% compared to the original run so I reverted it. Thanks. What tests are you using to evaluate this? Evan > > I didn't test if it would be beneficial to apply the optimization to GEPs with a scale of one but I doubt it'll have a measurable impact. From benny.kra at googlemail.com Fri Feb 25 11:48:40 2011 From: benny.kra at googlemail.com (Benjamin Kramer) Date: Fri, 25 Feb 2011 18:48:40 +0100 Subject: [llvm-commits] [llvm] r126445 - in /llvm/trunk: lib/Transforms/Utils/SimplifyCFG.cpp test/Transforms/SimplifyCFG/select-gep.ll In-Reply-To: References: <20110224224612.21E8B2A6C12C@llvm.org> <9DA53217-C6FC-42E4-8D03-218F85893512@apple.com>

Message-ID: On 25.02.2011, at 18:25, Evan Cheng wrote: > > On Feb 25, 2011, at 7:26 AM, Benjamin Kramer wrote: > >> >> On 25.02.2011, at 08:14, Evan Cheng wrote: >> >>> Speculation is nice, but I would feel a lot better if there are hard numbers to back this up. Are we sure these changes are improving performance? >> >> My unscientific test case (gcc compiled with clang -O3 parsing itself) showed a 0.25% speedup with the original patch (all-constant GEPs only, median of 5 runs, i386, penryn). I just retested with the change Frits suggested and it slowed down ~0.6% compared to the original run so I reverted it. > > Thanks. What tests are you using to evaluate this? I routinely use the nice C files from http://people.csail.mit.edu/smcc/projects/single-file-programs/ which are easy to handle and large enough to get meaningful results for both compile time and run time. They're extracted from programs that are also part of SPEC CINT, so I'm confident the improvements will also show up there :) From justin.holewinski at gmail.com Fri Feb 25 07:38:40 2011 From: justin.holewinski at gmail.com (Justin Holewinski) Date: Fri, 25 Feb 2011 08:38:40 -0500 Subject: [llvm-commits] [PATCH] PTX Backend: Add basic f32 support Message-ID: This patch adds preliminary 32-bit floating-point support to the PTX Backend. Floating-point add, sub, and mul instructions can now be emitted, with the appropriate infrastructure to support the .f32 type in PTX. I have also added updated test cases for all of the added functionality. As this is my first patch submitted to the list, please let me know if I am missing anything important needed for this patch to be accepted. :) -- Thanks, Justin Holewinski -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/c88bdad8/attachment.html -------------- next part -------------- A non-text attachment was scrubbed... Name: ptx-f32-prelim.patch Type: text/x-patch Size: 18624 bytes Desc: not available Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/c88bdad8/attachment.bin From daniel at zuster.org Fri Feb 25 10:45:22 2011 From: daniel at zuster.org (Daniel Dunbar) Date: Fri, 25 Feb 2011 16:45:22 -0000 Subject: [llvm-commits] [zorg] r126491 - /zorg/trunk/llvmlab/llvmlab/ci/status.py Message-ID: <20110225164522.0D2452A6C12C@llvm.org> Author: ddunbar Date: Fri Feb 25 10:45:21 2011 New Revision: 126491 URL: http://llvm.org/viewvc/llvm-project?rev=126491&view=rev Log: llvmlab.ci.status: Track build slave information. Modified: zorg/trunk/llvmlab/llvmlab/ci/status.py Modified: zorg/trunk/llvmlab/llvmlab/ci/status.py URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/llvmlab/llvmlab/ci/status.py?rev=126491&r1=126490&r2=126491&view=diff ============================================================================== --- zorg/trunk/llvmlab/llvmlab/ci/status.py (original) +++ zorg/trunk/llvmlab/llvmlab/ci/status.py Fri Feb 25 10:45:21 2011 @@ -14,29 +14,36 @@ @staticmethod def fromdata(data): version = data['version'] - if version != 0: + if version not in (0, 1): raise ValueError, "Unknown version" + if version == 0: + slave = None + else: + slave = data['slave'] return BuildStatus(data['name'], data['number'], data['source_stamp'], - data['result'], data['start_time'], data['end_time']) + data['result'], data['start_time'], data['end_time'], + slave) def todata(self): - return { 'version' : 0, + return { 'version' : 1, 'name' : self.name, 'number' : self.number, 'source_stamp' : self.source_stamp, 'result' : self.result, 'start_time' : self.start_time, - 'end_time' : self.end_time } + 'end_time' : self.end_time, + 'slave' : self.slave } def __init__(self, name, number, source_stamp, - result, start_time, end_time): + result, start_time, end_time, slave): self.name = name self.number = number self.source_stamp = source_stamp self.result = result self.start_time = start_time self.end_time = end_time + self.slave = slave class StatusMonitor(threading.Thread): def __init__(self, app, status): @@ -97,7 +104,8 @@ add_build = False if build is None: add_build = True - build = BuildStatus(name, id, None, None, None, None) + build = BuildStatus(name, id, None, None, None, None, + None) # Get the build information. try: @@ -111,6 +119,7 @@ build.source_stamp = res['sourceStamp']['revision'] build.start_time = res['times'][0] build.end_time = res['times'][1] + build.slave = res['slave'] if add_build: # Add to the builds list, maintaining order. From resistor at mac.com Fri Feb 25 12:35:42 2011 From: resistor at mac.com (Owen Anderson) Date: Fri, 25 Feb 2011 10:35:42 -0800 Subject: [llvm-commits] [llvm] r126426 - in /llvm/trunk/tools/lto: LTOCodeGenerator.cpp LTOCodeGenerator.h lto.cpp In-Reply-To: <20110224210406.87C712A6C12D@llvm.org> References: <20110224210406.87C712A6C12D@llvm.org> Message-ID: <8D0F51AA-584B-43ED-BAE2-5E19B858BA3F@mac.com> Should this apply to all targets, or only the same ones for which -integrated-as is the default? --Owen On Feb 24, 2011, at 1:04 PM, Rafael Espindola wrote: > Author: rafael > Date: Thu Feb 24 15:04:06 2011 > New Revision: 126426 > > URL: http://llvm.org/viewvc/llvm-project?rev=126426&view=rev > Log: > Switch LTO to use MC. This takes the linking of libxul.so from about 7m to > 6m30. > > Modified: > llvm/trunk/tools/lto/LTOCodeGenerator.cpp > llvm/trunk/tools/lto/LTOCodeGenerator.h > llvm/trunk/tools/lto/lto.cpp > > Modified: llvm/trunk/tools/lto/LTOCodeGenerator.cpp > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/lto/LTOCodeGenerator.cpp?rev=126426&r1=126425&r2=126426&view=diff > ============================================================================== > --- llvm/trunk/tools/lto/LTOCodeGenerator.cpp (original) > +++ llvm/trunk/tools/lto/LTOCodeGenerator.cpp Thu Feb 24 15:04:06 2011 > @@ -71,10 +71,11 @@ > _linker("LinkTimeOptimizer", "ld-temp.o", _context), _target(NULL), > _emitDwarfDebugInfo(false), _scopeRestrictionsDone(false), > _codeModel(LTO_CODEGEN_PIC_MODEL_DYNAMIC), > - _nativeObjectFile(NULL), _assemblerPath(NULL) > + _nativeObjectFile(NULL) > { > InitializeAllTargets(); > InitializeAllAsmPrinters(); > + InitializeAllAsmParsers(); > } > > LTOCodeGenerator::~LTOCodeGenerator() > @@ -126,21 +127,6 @@ > _mCpu = mCpu; > } > > -void LTOCodeGenerator::setAssemblerPath(const char* path) > -{ > - if ( _assemblerPath ) > - delete _assemblerPath; > - _assemblerPath = new sys::Path(path); > -} > - > -void LTOCodeGenerator::setAssemblerArgs(const char** args, int nargs) > -{ > - for (int i = 0; i < nargs; ++i) { > - const char *arg = args[i]; > - _assemblerArgs.push_back(arg); > - } > -} > - > void LTOCodeGenerator::addMustPreserveSymbol(const char* sym) > { > _mustPreserveSymbols[sym] = 1; > @@ -183,55 +169,42 @@ > > const void* LTOCodeGenerator::compile(size_t* length, std::string& errMsg) > { > - // make unique temp .s file to put generated assembly code > - sys::Path uniqueAsmPath("lto-llvm.s"); > - if ( uniqueAsmPath.createTemporaryFileOnDisk(false, &errMsg) ) > - return NULL; > - sys::RemoveFileOnSignal(uniqueAsmPath); > - > - // generate assembly code > - bool genResult = false; > - { > - tool_output_file asmFile(uniqueAsmPath.c_str(), errMsg); > - if (!errMsg.empty()) > - return NULL; > - genResult = this->generateAssemblyCode(asmFile.os(), errMsg); > - asmFile.os().close(); > - if (asmFile.os().has_error()) { > - asmFile.os().clear_error(); > - return NULL; > - } > - asmFile.keep(); > - } > - if ( genResult ) { > - uniqueAsmPath.eraseFromDisk(); > - return NULL; > - } > - > // make unique temp .o file to put generated object file > sys::PathWithStatus uniqueObjPath("lto-llvm.o"); > if ( uniqueObjPath.createTemporaryFileOnDisk(false, &errMsg) ) { > - uniqueAsmPath.eraseFromDisk(); > + uniqueObjPath.eraseFromDisk(); > return NULL; > } > sys::RemoveFileOnSignal(uniqueObjPath); > > - // assemble the assembly code > - const std::string& uniqueObjStr = uniqueObjPath.str(); > - bool asmResult = this->assemble(uniqueAsmPath.str(), uniqueObjStr, errMsg); > - if ( !asmResult ) { > - // remove old buffer if compile() called twice > - delete _nativeObjectFile; > - > - // read .o file into memory buffer > - OwningPtr BuffPtr; > - if (error_code ec = MemoryBuffer::getFile(uniqueObjStr.c_str(),BuffPtr)) > - errMsg = ec.message(); > - _nativeObjectFile = BuffPtr.take(); > + // generate object file > + bool genResult = false; > + tool_output_file objFile(uniqueObjPath.c_str(), errMsg); > + if (!errMsg.empty()) > + return NULL; > + genResult = this->generateObjectFile(objFile.os(), errMsg); > + objFile.os().close(); > + if (objFile.os().has_error()) { > + objFile.os().clear_error(); > + return NULL; > + } > + objFile.keep(); > + if ( genResult ) { > + uniqueObjPath.eraseFromDisk(); > + return NULL; > } > > + const std::string& uniqueObjStr = uniqueObjPath.str(); > + // remove old buffer if compile() called twice > + delete _nativeObjectFile; > + > + // read .o file into memory buffer > + OwningPtr BuffPtr; > + if (error_code ec = MemoryBuffer::getFile(uniqueObjStr.c_str(),BuffPtr)) > + errMsg = ec.message(); > + _nativeObjectFile = BuffPtr.take(); > + > // remove temp files > - uniqueAsmPath.eraseFromDisk(); > uniqueObjPath.eraseFromDisk(); > > // return buffer, unless error > @@ -241,67 +214,6 @@ > return _nativeObjectFile->getBufferStart(); > } > > - > -bool LTOCodeGenerator::assemble(const std::string& asmPath, > - const std::string& objPath, std::string& errMsg) > -{ > - sys::Path tool; > - bool needsCompilerOptions = true; > - if ( _assemblerPath ) { > - tool = *_assemblerPath; > - needsCompilerOptions = false; > - } else { > - // find compiler driver > - tool = sys::Program::FindProgramByName("gcc"); > - if ( tool.isEmpty() ) { > - errMsg = "can't locate gcc"; > - return true; > - } > - } > - > - // build argument list > - std::vector args; > - llvm::Triple targetTriple(_linker.getModule()->getTargetTriple()); > - const char *arch = targetTriple.getArchNameForAssembler(); > - > - args.push_back(tool.c_str()); > - > - if (targetTriple.getOS() == Triple::Darwin) { > - // darwin specific command line options > - if (arch != NULL) { > - args.push_back("-arch"); > - args.push_back(arch); > - } > - // add -static to assembler command line when code model requires > - if ( (_assemblerPath != NULL) && > - (_codeModel == LTO_CODEGEN_PIC_MODEL_STATIC) ) > - args.push_back("-static"); > - } > - if ( needsCompilerOptions ) { > - args.push_back("-c"); > - args.push_back("-x"); > - args.push_back("assembler"); > - } else { > - for (std::vector::iterator I = _assemblerArgs.begin(), > - E = _assemblerArgs.end(); I != E; ++I) { > - args.push_back(I->c_str()); > - } > - } > - args.push_back("-o"); > - args.push_back(objPath.c_str()); > - args.push_back(asmPath.c_str()); > - args.push_back(0); > - > - // invoke assembler > - if ( sys::Program::ExecuteAndWait(tool, &args[0], 0, 0, 0, 0, &errMsg) ) { > - errMsg = "error in assembly"; > - return true; > - } > - return false; // success > -} > - > - > - > bool LTOCodeGenerator::determineTarget(std::string& errMsg) > { > if ( _target == NULL ) { > @@ -385,8 +297,8 @@ > } > > /// Optimize merged modules using various IPO passes > -bool LTOCodeGenerator::generateAssemblyCode(raw_ostream& out, > - std::string& errMsg) > +bool LTOCodeGenerator::generateObjectFile(raw_ostream& out, > + std::string& errMsg) > { > if ( this->determineTarget(errMsg) ) > return true; > @@ -423,7 +335,7 @@ > formatted_raw_ostream Out(out); > > if (_target->addPassesToEmitFile(*codeGenPasses, Out, > - TargetMachine::CGFT_AssemblyFile, > + TargetMachine::CGFT_ObjectFile, > CodeGenOpt::Aggressive)) { > errMsg = "target file type not supported"; > return true; > > Modified: llvm/trunk/tools/lto/LTOCodeGenerator.h > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/lto/LTOCodeGenerator.h?rev=126426&r1=126425&r2=126426&view=diff > ============================================================================== > --- llvm/trunk/tools/lto/LTOCodeGenerator.h (original) > +++ llvm/trunk/tools/lto/LTOCodeGenerator.h Thu Feb 24 15:04:06 2011 > @@ -37,18 +37,14 @@ > bool setDebugInfo(lto_debug_model, std::string& errMsg); > bool setCodePICModel(lto_codegen_model, std::string& errMsg); > void setCpu(const char *cpu); > - void setAssemblerPath(const char* path); > - void setAssemblerArgs(const char** args, int nargs); > void addMustPreserveSymbol(const char* sym); > bool writeMergedModules(const char* path, > std::string& errMsg); > const void* compile(size_t* length, std::string& errMsg); > void setCodeGenDebugOptions(const char *opts); > private: > - bool generateAssemblyCode(llvm::raw_ostream& out, > - std::string& errMsg); > - bool assemble(const std::string& asmPath, > - const std::string& objPath, std::string& errMsg); > + bool generateObjectFile(llvm::raw_ostream& out, > + std::string& errMsg); > void applyScopeRestrictions(); > bool determineTarget(std::string& errMsg); > > @@ -63,9 +59,7 @@ > StringSet _mustPreserveSymbols; > llvm::MemoryBuffer* _nativeObjectFile; > std::vector _codegenOptions; > - llvm::sys::Path* _assemblerPath; > std::string _mCpu; > - std::vector _assemblerArgs; > }; > > #endif // LTO_CODE_GENERATOR_H > > Modified: llvm/trunk/tools/lto/lto.cpp > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/lto/lto.cpp?rev=126426&r1=126425&r2=126426&view=diff > ============================================================================== > --- llvm/trunk/tools/lto/lto.cpp (original) > +++ llvm/trunk/tools/lto/lto.cpp Thu Feb 24 15:04:06 2011 > @@ -231,7 +231,7 @@ > // > void lto_codegen_set_assembler_path(lto_code_gen_t cg, const char* path) > { > - cg->setAssemblerPath(path); > + // In here only for backwards compatibility. We use MC now. > } > > > @@ -241,7 +241,7 @@ > void lto_codegen_set_assembler_args(lto_code_gen_t cg, const char** args, > int nargs) > { > - cg->setAssemblerArgs(args, nargs); > + // In here only for backwards compatibility. We use MC now. > } > > // > > > _______________________________________________ > llvm-commits mailing list > llvm-commits at cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits From gkistanova at gmail.com Fri Feb 25 13:01:40 2011 From: gkistanova at gmail.com (Galina Kistanova) Date: Fri, 25 Feb 2011 11:01:40 -0800 Subject: [llvm-commits] Possible bug - starting from the revion 124059 llvm-gcc cross to ARM build failing with the ld.exe crash In-Reply-To: <4D672101.7030207@gmail.com> References: <4D66165D.909@free.fr> <4D672101.7030207@gmail.com> Message-ID: TotT builds OK now. Here is builder with this configuration: http://google1.osuosl.org:8011/builders/llvm-gcc-mingw32-cross-arm-linux-gnueabi-hard-float Thanks Galina 2011/2/24 Rafael ?vila de Esp?ndola : > On 2011-02-24 20:34, Galina Kistanova wrote: >>> This commit looks like it can be reverted independently. >>> Galina, does reverting it fix the problem? >> >> Yes, after reverting tot builds OK. > > Can you build without reverting by syncing past 126421? > >> Thanks >> >> Galina > > Cheers, > Rafael > _______________________________________________ > llvm-commits mailing list > llvm-commits at cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits > From rafael.espindola at gmail.com Fri Feb 25 13:09:12 2011 From: rafael.espindola at gmail.com (Rafael Avila de Espindola) Date: Fri, 25 Feb 2011 14:09:12 -0500 Subject: [llvm-commits] [llvm] r126426 - in /llvm/trunk/tools/lto: LTOCodeGenerator.cpp LTOCodeGenerator.h lto.cpp In-Reply-To: <8D0F51AA-584B-43ED-BAE2-5E19B858BA3F@mac.com> References: <20110224210406.87C712A6C12D@llvm.org> <8D0F51AA-584B-43ED-BAE2-5E19B858BA3F@mac.com> Message-ID: <4D67FE58.9010008@gmail.com> On 11-02-25 01:35 PM, Owen Anderson wrote: > Should this apply to all targets, or only the same ones for which -integrated-as is the default? Do you have a target for which you need LTO but MC is not working yet? This patch was intended as a stepping stone for requiring MC for finding definitions and uses in global asm, so if there is still a target with no MC that needs LTO, it is probably better to just revert this for now :-( > --Owen Cheers, Rafael From sabre at nondot.org Fri Feb 25 13:06:35 2011 From: sabre at nondot.org (Chris Lattner) Date: Fri, 25 Feb 2011 19:06:35 -0000 Subject: [llvm-commits] [llvm] r126500 - in /llvm/trunk/test/MC: ARM/bracket-exprs.s ELF/bracket-exprs.s Message-ID: <20110225190635.C3B312A6C12C@llvm.org> Author: lattner Date: Fri Feb 25 13:06:35 2011 New Revision: 126500 URL: http://llvm.org/viewvc/llvm-project?rev=126500&view=rev Log: split this test into arch specific pieces, so the ARM test isn't run when the arm backend isn't built. This fixes PR9327 Added: llvm/trunk/test/MC/ARM/bracket-exprs.s Modified: llvm/trunk/test/MC/ELF/bracket-exprs.s Added: llvm/trunk/test/MC/ARM/bracket-exprs.s URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/ARM/bracket-exprs.s?rev=126500&view=auto ============================================================================== --- llvm/trunk/test/MC/ARM/bracket-exprs.s (added) +++ llvm/trunk/test/MC/ARM/bracket-exprs.s Fri Feb 25 13:06:35 2011 @@ -0,0 +1,15 @@ +// RUN: llvm-mc -triple arm-unknown-linux %s | FileCheck %s + +// CHECK: .byte 1 +.if [~0 >> 1] == -1 +.byte 1 +.else +.byte 2 +.endif + +// CHECK: .byte 3 +.if 4 * [4 + (3 + [2 * 2] + 1)] == 48 +.byte 3 +.else +.byte 4 +.endif Modified: llvm/trunk/test/MC/ELF/bracket-exprs.s URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/ELF/bracket-exprs.s?rev=126500&r1=126499&r2=126500&view=diff ============================================================================== --- llvm/trunk/test/MC/ELF/bracket-exprs.s (original) +++ llvm/trunk/test/MC/ELF/bracket-exprs.s Fri Feb 25 13:06:35 2011 @@ -1,5 +1,4 @@ // RUN: llvm-mc -triple i386-unknown-unknown %s | FileCheck %s -// RUN: llvm-mc -triple arm-unknown-linux %s | FileCheck %s // CHECK: .byte 1 .if [~0 >> 1] == -1 From resistor at mac.com Fri Feb 25 13:18:52 2011 From: resistor at mac.com (Owen Anderson) Date: Fri, 25 Feb 2011 11:18:52 -0800 Subject: [llvm-commits] [llvm] r126426 - in /llvm/trunk/tools/lto: LTOCodeGenerator.cpp LTOCodeGenerator.h lto.cpp In-Reply-To: <4D67FE58.9010008@gmail.com> References: <20110224210406.87C712A6C12D@llvm.org> <8D0F51AA-584B-43ED-BAE2-5E19B858BA3F@mac.com> <4D67FE58.9010008@gmail.com> Message-ID: <54BBC0F6-E4D8-439A-B233-EF4FEE08B915@mac.com> On Feb 25, 2011, at 11:09 AM, Rafael Avila de Espindola wrote: > On 11-02-25 01:35 PM, Owen Anderson wrote: >> Should this apply to all targets, or only the same ones for which -integrated-as is the default? > > Do you have a target for which you need LTO but MC is not working yet? This patch was intended as a stepping stone for requiring MC for finding definitions and uses in global asm, so if there is still a target with no MC that needs LTO, it is probably better to just revert this for now :-( There are shipping LLVM compilers with LTO support that target ARM, but ARM MC is not 100% functional yet. That said, this does make sense. My point was just that it should be enabled by default on a per-target basis, just like clang's -integrated-as is. --Owen -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/4abae506/attachment.html From rafael.espindola at gmail.com Fri Feb 25 13:24:52 2011 From: rafael.espindola at gmail.com (Rafael Avila de Espindola) Date: Fri, 25 Feb 2011 14:24:52 -0500 Subject: [llvm-commits] [llvm] r126426 - in /llvm/trunk/tools/lto: LTOCodeGenerator.cpp LTOCodeGenerator.h lto.cpp In-Reply-To: <54BBC0F6-E4D8-439A-B233-EF4FEE08B915@mac.com> References: <20110224210406.87C712A6C12D@llvm.org> <8D0F51AA-584B-43ED-BAE2-5E19B858BA3F@mac.com> <4D67FE58.9010008@gmail.com> <54BBC0F6-E4D8-439A-B233-EF4FEE08B915@mac.com> Message-ID: <4D680204.2080102@gmail.com> > There are shipping LLVM compilers with LTO support that target ARM, but > ARM MC is not 100% functional yet. > > That said, this /does/ make sense. My point was just that it should be > enabled by default on a per-target basis, just like clang's > -integrated-as is. Well, the problem is that I should soon replace the .globl grepping with a custom streamer, so it doesn't make a lot of sense to require a working asm parser in libLTO and not use it while producing .o files. If this is an off tree compiler, it is probably reasonable to ask them to just not pull this patch, no? > --Owen Cheers, Rafael From jan_sjodin at yahoo.com Fri Feb 25 13:37:26 2011 From: jan_sjodin at yahoo.com (Jan Sjodin) Date: Fri, 25 Feb 2011 11:37:26 -0800 (PST) Subject: [llvm-commits] Change DwarfUsesAbsoluteLabelForStmtList to false for X86ELFMCAsmInfo. Message-ID: <133510.39411.qm@web55604.mail.re4.yahoo.com> Dwarfdump gives an error and gdb fails when generating code in memory (but happens to accept it in a .o file) in Linux when DwarfUsesAbsoluteLabelForStmtList = true, this patch sets it to false to fix this issue. I don't know if this is true for other targets, but the change is limited to X86ELFMCAsmInfo. - Jan -------------- next part -------------- A non-text attachment was scrubbed... Name: 0039_dwarflabel.patch Type: application/octet-stream Size: 449 bytes Desc: not available Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/02087c56/attachment.obj From rafael.espindola at gmail.com Fri Feb 25 13:46:10 2011 From: rafael.espindola at gmail.com (Rafael Avila de Espindola) Date: Fri, 25 Feb 2011 14:46:10 -0500 Subject: [llvm-commits] Code refactoring of MCELFStreamer.cpp and ELFObjectWriter.cpp In-Reply-To: <960821.36942.qm@web55604.mail.re4.yahoo.com> References: <960821.36942.qm@web55604.mail.re4.yahoo.com> Message-ID: <4D680702.2040101@gmail.com> On 11-02-25 12:02 PM, Jan Sjodin wrote: > Removed duplicated code by putting shared functions in MCELF.h/cpp and made the > rest of the C-style static functions static methods in the corresponding > classes. Moving the duplicated functions to a new file is OK, but why convert the function only used in the ELFObjectWriter to static methods? > - Jan Cheers, Rafael From jan_sjodin at yahoo.com Fri Feb 25 14:05:02 2011 From: jan_sjodin at yahoo.com (Jan Sjodin) Date: Fri, 25 Feb 2011 12:05:02 -0800 (PST) Subject: [llvm-commits] Code refactoring of MCELFStreamer.cpp and ELFObjectWriter.cpp In-Reply-To: <4D680702.2040101@gmail.com> References: <960821.36942.qm@web55604.mail.re4.yahoo.com> <4D680702.2040101@gmail.com> Message-ID: <470208.84552.qm@web55602.mail.re4.yahoo.com> Just to keep the code more object-oriented. If a sub-class (I intend to inherit from these classes) needs the functions, they will not be available if they are hidden as static functions. - Jan ----- Original Message ---- > From: Rafael Avila de Espindola > To: llvm-commits at cs.uiuc.edu > Sent: Fri, February 25, 2011 2:46:10 PM > Subject: Re: [llvm-commits] Code refactoring of MCELFStreamer.cpp and >ELFObjectWriter.cpp > > On 11-02-25 12:02 PM, Jan Sjodin wrote: > > Removed duplicated code by putting shared functions in MCELF.h/cpp and made >the > > rest of the C-style static functions static methods in the corresponding > > classes. > > Moving the duplicated functions to a new file is OK, but why convert the > function only used in the ELFObjectWriter to static methods? > > > - Jan > > Cheers, > Rafael > _______________________________________________ > llvm-commits mailing list > llvm-commits at cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits > From dimitry at andric.com Fri Feb 25 14:21:21 2011 From: dimitry at andric.com (Dimitry Andric) Date: Fri, 25 Feb 2011 21:21:21 +0100 Subject: [llvm-commits] [patch] Fix getDriver().getInstalledDir() comparisons Message-ID: <4D680F41.1070008@andric.com> Since getDriver().getInstalledDir() returns a const char *, don't try to compare it with getDriver().Dir.c_str(), since that is a pointer comparison, not a "are these strings equal" comparison. Instead, just compare with getDriver().Dir directly, so both sides will get promoted to std::string, and the regular std::string comparison operator applies. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: clang-driver-dir-1.diff Url: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/8c7710aa/attachment.pl From dimitry at andric.com Fri Feb 25 14:21:30 2011 From: dimitry at andric.com (Dimitry Andric) Date: Fri, 25 Feb 2011 21:21:30 +0100 Subject: [llvm-commits] [patch] Set default CPU on FreeBSD Message-ID: <4D680F4A.7040402@andric.com> The default CPU on FreeBSD for 32-bit should be i486, not pentium4. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: clang-freebsd-defcpu-1.diff Url: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/fbf69efe/attachment.pl From dimitry at andric.com Fri Feb 25 14:21:35 2011 From: dimitry at andric.com (Dimitry Andric) Date: Fri, 25 Feb 2011 21:21:35 +0100 Subject: [llvm-commits] [patch] Don't add /usr/local/include to default include path on FreeBSD Message-ID: <4D680F4F.4090403@andric.com> On FreeBSD, make sure /usr/local/include is *not* in the default include path. This avoids accidentally including the wrong headers. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: clang-freebsd-nolocal-1.diff Url: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/c8015ada/attachment.pl From dimitry at andric.com Fri Feb 25 14:21:40 2011 From: dimitry at andric.com (Dimitry Andric) Date: Fri, 25 Feb 2011 21:21:40 +0100 Subject: [llvm-commits] [patch] Cleanup FilePaths and ProgramPaths for FreeBSD Message-ID: <4D680F54.3040009@andric.com> On FreeBSD, we don't want 'getDriver().Dir + "/../lib"' added to the ToolChain's FilePaths. If clang is installed as a port in /usr/local, it is *not* supposed to use /usr/local/lib by default, for example. Additionally, there are no clang-related executables in either /usr/libexec, or getDriver().Dir + "/../libexec", anymore, so remove that from the ToolChain's ProgramPaths. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: clang-freebsd-tcpaths-1.diff Url: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/ba23c9f6/attachment.pl From dimitry at andric.com Fri Feb 25 14:21:48 2011 From: dimitry at andric.com (Dimitry Andric) Date: Fri, 25 Feb 2011 21:21:48 +0100 Subject: [llvm-commits] [patch] Fix library path on linker command line for FreeBSD Message-ID: <4D680F5C.2070006@andric.com> For linking on FreeBSD, don't add a hardcoded "-L/usr/lib", but retrieve the library paths from the ToolChain object instead. Copy the relevant code from linuxtools::Link::ConstructJob(), and replace the std::string stuff with llvm::StringRef, while we're here. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: clang-freebsd-libpath-1.diff Url: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/65642b91/attachment.pl From gkistanova at gmail.com Fri Feb 25 14:22:42 2011 From: gkistanova at gmail.com (Galina Kistanova) Date: Fri, 25 Feb 2011 20:22:42 -0000 Subject: [llvm-commits] [llvm-gcc-4.2] r126506 - /llvm-gcc-4.2/trunk/extras/build-self-4-mingw32 Message-ID: <20110225202242.327612A6C12C@llvm.org> Author: gkistanova Date: Fri Feb 25 14:22:42 2011 New Revision: 126506 URL: http://llvm.org/viewvc/llvm-project?rev=126506&view=rev Log: Some build script code cleaning for two-stage scripted builder. Modified: llvm-gcc-4.2/trunk/extras/build-self-4-mingw32 Modified: llvm-gcc-4.2/trunk/extras/build-self-4-mingw32 URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/extras/build-self-4-mingw32?rev=126506&r1=126505&r2=126506&view=diff ============================================================================== --- llvm-gcc-4.2/trunk/extras/build-self-4-mingw32 (original) +++ llvm-gcc-4.2/trunk/extras/build-self-4-mingw32 Fri Feb 25 14:22:42 2011 @@ -40,8 +40,8 @@ LLVM_GCC_src=llvm-gcc.src # The LLVM-GCC source code root directory name. LLVM_obj_1=llvm-1.obj # The LLVM build root directory name. LLVM_GCC_obj_1=llvm-gcc-1.obj # The LLVM-GCC build root directory name. -LLVM_obj_2=llvm-2.obj # The LLVM build root directory name for Stage2. -LLVM_GCC_obj_2=llvm-gcc-2.obj # The LLVM-GCC build root directory name for Stage2. +LLVM_obj_2=llvm-2.obj # The LLVM build directory name for Stage2. +LLVM_GCC_obj_2=llvm-gcc-2.obj # The LLVM-GCC build directory name for Stage2. INSTALL=install # Where the result will be installed. # CFLAGS and CXXFLAGS must not be set during the building of cross-tools. @@ -51,6 +51,12 @@ BUILD_ROOT=$PWD # Where build happens. PRIVATE_INSTALL=${BUILD_ROOT}/${INSTALL} # Where the result will be installed. +export PATH=/opt/local/bin:/opt/local/sbin:/bin:/sbin:/usr/bin:/usr/sbin +export PATH=$PATH:/mingw_build_tools/install_with_gcc/bin + +# Make sure our just built tools will be selected first. +export PATH=${PRIVATE_INSTALL}/bin:$PATH + #------------------------------------------------------------------------------ # Define build steps, parse and validate input parameters #------------------------------------------------------------------------------ @@ -66,7 +72,6 @@ do_install_llvmgcc_1=no # Install LLVM-GCC. do_configure_llvm_2=no # Configure LLVM for Stage2. do_make_llvm_2=no # Make LLVM for Stage2. -do_test_llvm_2=no # Test LLVM for Stage2. do_configure_llvmgcc_2=no # Configure LLVM-GCC for Stage2. do_make_llvmgcc_2=no # Make LLVM-GCC for Stage2. do_install_llvmgcc_2=no # Install LLVM-GCC for Stage2. @@ -91,7 +96,6 @@ install_llvmgcc_1 | \ configure_llvm_2 | \ make_llvm_2 | \ - test_llvm_2 | \ configure_llvmgcc_2 | \ make_llvmgcc_2 | \ install_llvmgcc_2 | \ @@ -118,7 +122,6 @@ do_install_llvmgcc_1=yes do_configure_llvm_2=yes do_make_llvm_2=yes - do_test_llvm_2=yes do_configure_llvmgcc_2=yes do_make_llvmgcc_2=yes do_install_llvmgcc_2=yes @@ -143,7 +146,8 @@ # We need a local copy of binutils, system libraries and headers, # since we will be installing there. - cp -RL /cross-tools/ ${PRIVATE_INSTALL} + + cp -RL /mingw_build_tools/install_with_gcc/ ${PRIVATE_INSTALL} fi @@ -249,10 +253,6 @@ #------------------------------------------------------------------------------ if [ "$do_configure_llvm_2" == "yes" ] ; then - # Make sure our just built tools will be selected first. - # Note: Always set this, even when do_all has been requested. - export PATH=${PRIVATE_INSTALL}/bin:${PATH} - # Remove previous build files if any. rm -rf ${BUILD_ROOT}/${LLVM_obj_2} mkdir -p ${BUILD_ROOT}/${LLVM_obj_2} @@ -273,11 +273,6 @@ #------------------------------------------------------------------------------ if [ "$do_make_llvm_2" == "yes" ] ; then - if [ "$do_all" == "no" ] ; then - # Make sure our just built tools will be selected first. - export PATH=${PRIVATE_INSTALL}/bin:${PATH} - fi - cd ${BUILD_ROOT}/${LLVM_obj_2} # NOTE: Do not use make ENABLE_OPTIMIZED=1. Some tests fail because of that. nice -n 20 make VERBOSE=1 \ @@ -286,31 +281,10 @@ fi #------------------------------------------------------------------------------ -# Step: Stage2. Test LLVM. -#------------------------------------------------------------------------------ -if [ "$do_test_llvm_2" == "yes" ] ; then - - if [ "$do_all" == "no" ] ; then - # Make sure our just built tools will be selected first. - export PATH=${PRIVATE_INSTALL}/bin:${PATH} - fi - - cd ${BUILD_ROOT}/${LLVM_obj_2} - make check-lit VERBOSE=1 \ - $@ # Extra args if any - -fi - -#------------------------------------------------------------------------------ # Step: Stage2. Configure LLVM-GCC. #------------------------------------------------------------------------------ if [ "$do_configure_llvmgcc_2" == "yes" ] ; then - if [ "$do_all" == "no" ] ; then - # Make sure our just built tools will be selected first. - export PATH=${PRIVATE_INSTALL}/bin:${PATH} - fi - # Remove previous build files if any. rm -rf ${BUILD_ROOT}/${LLVM_GCC_obj_2} mkdir -p ${BUILD_ROOT}/${LLVM_GCC_obj_2} @@ -334,11 +308,6 @@ #------------------------------------------------------------------------------ if [ "$do_make_llvmgcc_2" == "yes" ] ; then - if [ "$do_all" == "no" ] ; then - # Make sure our just built tools will be selected first. - export PATH=${PRIVATE_INSTALL}/bin:${PATH} - fi - cd ${BUILD_ROOT}/${LLVM_GCC_obj_2} nice -n 20 make \ $@ # Extra args if any @@ -350,11 +319,6 @@ #------------------------------------------------------------------------------ if [ "$do_install_llvmgcc_2" == "yes" ] ; then - if [ "$do_all" == "no" ] ; then - # Make sure our just built tools will be selected first. - export PATH=${PRIVATE_INSTALL}/bin:${PATH} - fi - cd ${BUILD_ROOT}/${LLVM_GCC_obj_2} nice -n 20 make install \ $@ # Extra args if any From stuart at apple.com Fri Feb 25 14:42:39 2011 From: stuart at apple.com (Stuart Hastings) Date: Fri, 25 Feb 2011 20:42:39 -0000 Subject: [llvm-commits] [llvm] r126507 - /llvm/trunk/utils/buildit/build_llvm Message-ID: <20110225204239.DD1722A6C12C@llvm.org> Author: stuart Date: Fri Feb 25 14:42:39 2011 New Revision: 126507 URL: http://llvm.org/viewvc/llvm-project?rev=126507&view=rev Log: Omit lto.h from the llvmCore result; henceforth, this will be supplied by clang. Radar 9042056. Modified: llvm/trunk/utils/buildit/build_llvm Modified: llvm/trunk/utils/buildit/build_llvm URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/buildit/build_llvm?rev=126507&r1=126506&r2=126507&view=diff ============================================================================== --- llvm/trunk/utils/buildit/build_llvm (original) +++ llvm/trunk/utils/buildit/build_llvm Fri Feb 25 14:42:39 2011 @@ -341,6 +341,9 @@ fi rm -f lib/libLTO.a lib/libLTO.la +// Omit lto.h from the result. Clang will supply. +find $DEST_DIR$DEST_ROOT -name lto.h -delete + ################################################################################ # Remove debugging information from DEST_DIR. From greened at obbligato.org Fri Feb 25 14:51:27 2011 From: greened at obbligato.org (David Greene) Date: Fri, 25 Feb 2011 20:51:27 -0000 Subject: [llvm-commits] [llvm] r126510 - /llvm/trunk/utils/llvmbuild Message-ID: <20110225205127.A05AC2A6C12C@llvm.org> Author: greened Date: Fri Feb 25 14:51:27 2011 New Revision: 126510 URL: http://llvm.org/viewvc/llvm-project?rev=126510&view=rev Log: Add some options for building LLVM in different environments: --force-configure to force running configure before building. --extra-llvm-config-flags --extra-llvm-gcc-config-flags --extra-gcc-config-flags Pass additional argument to the various configure invocations. This also eliminates a default build flavor because explicitly specifying builds could result in build flavors being run repeatedly. Finally, turn off fortran builds for the moment because install appears to be broken. Modified: llvm/trunk/utils/llvmbuild Modified: llvm/trunk/utils/llvmbuild URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/llvmbuild?rev=126510&r1=126509&r2=126510&view=diff ============================================================================== --- llvm/trunk/utils/llvmbuild (original) +++ llvm/trunk/utils/llvmbuild Fri Feb 25 14:51:27 2011 @@ -163,7 +163,7 @@ " [default: %default]")) parser.add_option("--src", action="append", help=("Top-level source directory [default: %default]")) - parser.add_option("--build", action="append", default=["debug"], + parser.add_option("--build", action="append", help=("Build types to run [default: %default]")) parser.add_option("--branch", action="append", help=("Source branch to build [default: %default]")) @@ -181,6 +181,14 @@ help=("Root install directory [default: %default]")) parser.add_option("--builddir", help=("Root build directory [default: %default]")) + parser.add_option("--extra-llvm-config-flags", default="", + help=("Extra flags to pass to llvm configure [default: %default]")) + parser.add_option("--extra-llvm-gcc-config-flags", default="", + help=("Extra flags to pass to llvm-gcc configure [default: %default]")) + parser.add_option("--extra-gcc-config-flags", default="", + help=("Extra flags to pass to gcc configure [default: %default]")) + parser.add_option("--force-configure", default=False, action="store_true", + help=("Force reconfigure of all components")) return def check_options(parser, options, valid_builds): @@ -284,18 +292,20 @@ class ExecutableNotFound(Exception): pass class FileNotExecutable(Exception): pass - def __init__(self, work_queue, jobs, cc, cxx, build_abbrev, source_abbrev, - branch_abbrev, build_prefix, install_prefix): + def __init__(self, work_queue, jobs, + build_abbrev, source_abbrev, branch_abbrev, + options): super().__init__() self.work_queue = work_queue self.jobs = jobs - self.cc = cc - self.cxx = cxx + self.cc = options.cc + self.cxx = options.cxx self.build_abbrev = build_abbrev self.source_abbrev = source_abbrev self.branch_abbrev = branch_abbrev - self.build_prefix = build_prefix - self.install_prefix = install_prefix + self.build_prefix = options.builddir + self.install_prefix = options.prefix + self.options = options self.component_abbrev = dict( llvm="llvm", llvm_gcc="lgcc", @@ -399,13 +409,16 @@ configure_flags = dict( llvm=dict(debug=["--prefix=" + self.install_prefix, + "--with-extra-options=-Werror", "--with-cxx-include-root=" + cxxroot, "--with-cxx-include-arch=" + cxxarch], release=["--prefix=" + self.install_prefix, + "--with-extra-options=-Werror", "--enable-optimized", "--with-cxx-include-root=" + cxxroot, "--with-cxx-include-arch=" + cxxarch], paranoid=["--prefix=" + self.install_prefix, + "--with-extra-options=-Werror", "--enable-expensive-checks", "--with-cxx-include-root=" + cxxroot, "--with-cxx-include-arch=" + cxxarch]), @@ -413,26 +426,35 @@ "--enable-checking", "--program-prefix=llvm-", "--enable-llvm=" + self.build_prefix + "/llvm/" + build_suffix, - "--enable-languages=c,c++,fortran"], +# Fortran install seems to be broken. +# "--enable-languages=c,c++,fortran"], + "--enable-languages=c,c++"], release=["--prefix=" + self.install_prefix, "--program-prefix=llvm-", "--enable-llvm=" + self.build_prefix + "/llvm/" + build_suffix, - "--enable-languages=c,c++,fortran"], +# Fortran install seems to be broken. +# "--enable-languages=c,c++,fortran"], + "--enable-languages=c,c++"], paranoid=["--prefix=" + self.install_prefix, "--enable-checking", "--program-prefix=llvm-", "--enable-llvm=" + self.build_prefix + "/llvm/" + build_suffix, - "--enable-languages=c,c++,fortran"]), +# Fortran install seems to be broken. +# "--enable-languages=c,c++,fortran"]), + "--enable-languages=c,c++"]), llvm2=dict(debug=["--prefix=" + self.install_prefix, + "--with-extra-options=-Werror", "--with-llvmgccdir=" + self.install_prefix + "/bin", "--with-cxx-include-root=" + cxxroot, "--with-cxx-include-arch=" + cxxarch], release=["--prefix=" + self.install_prefix, + "--with-extra-options=-Werror", "--enable-optimized", "--with-llvmgccdir=" + self.install_prefix + "/bin", "--with-cxx-include-root=" + cxxroot, "--with-cxx-include-arch=" + cxxarch], paranoid=["--prefix=" + self.install_prefix, + "--with-extra-options=-Werror", "--enable-expensive-checks", "--with-llvmgccdir=" + self.install_prefix + "/bin", "--with-cxx-include-root=" + cxxroot, @@ -599,28 +621,37 @@ if (branch is not None): srcdir += "/" + branch + comp_key = comp.replace("-", "_") + + config_args = configure_flags[comp_key][build][:] + config_args.extend(getattr(self.options, + "extra_" + comp_key + + "_config_flags").split()) + self.logger.info("Configuring " + component + " in " + builddir) self.configure(component, srcdir, builddir, - configure_flags[comp.replace("-", "_")][build], - configure_env[comp.replace("-", "_")][build]) + config_args, + configure_env[comp_key][build]) self.logger.info("Building " + component + " in " + builddir) self.make(component, srcdir, builddir, - make_flags[comp.replace("-", "_")][build], - make_env[comp.replace("-", "_")][build]) + make_flags[comp_key][build], + make_env[comp_key][build]) self.logger.info("Installing " + component + " in " + installdir) self.make(component, srcdir, builddir, - make_install_flags[comp.replace("-", "_")][build], - make_install_env[comp.replace("-", "_")][build]) + make_install_flags[comp_key][build], + make_install_env[comp_key][build]) self.logger.info("Testing " + component + " in " + builddir) self.make(component, srcdir, builddir, - make_check_flags[comp.replace("-", "_")][build], - make_check_env[comp.replace("-", "_")][build]) + make_check_flags[comp_key][build], + make_check_env[comp_key][build]) def configure(self, component, srcdir, builddir, flags, env): + self.logger.debug("Configure " + str(flags)) + configure_files = dict( llvm=[(srcdir + "/configure", builddir + "/Makefile")], llvm_gcc=[(srcdir + "/configure", builddir + "/Makefile"), @@ -630,8 +661,11 @@ (srcdir + "/gcc/configure", builddir + "/gcc/Makefile")], dragonegg=[()]) + doconfig = False for conf, mf in configure_files[component.replace("-", "_")]: + if not os.path.exists(conf): + return if os.path.exists(conf) and os.path.exists(mf): confstat = os.stat(conf) makestat = os.stat(mf) @@ -642,7 +676,7 @@ doconfig = True break - if not doconfig: + if not doconfig and not self.options.force_configure: return program = srcdir + "/configure" @@ -689,9 +723,9 @@ for t in range(options.threads): jobs = options.jobs // options.threads - builder = Builder(work_queue, jobs, options.cc.strip(), options.cxx.strip(), + builder = Builder(work_queue, jobs, build_abbrev, source_abbrev, branch_abbrev, - options.builddir.strip(), options.prefix.strip()) + options) builder.daemon = True builder.start() From baldrick at free.fr Fri Feb 25 14:55:08 2011 From: baldrick at free.fr (Duncan Sands) Date: Fri, 25 Feb 2011 20:55:08 -0000 Subject: [llvm-commits] [dragonegg] r126511 - /dragonegg/trunk/llvm-convert.cpp Message-ID: <20110225205508.C1DA52A6C12C@llvm.org> Author: baldrick Date: Fri Feb 25 14:55:08 2011 New Revision: 126511 URL: http://llvm.org/viewvc/llvm-project?rev=126511&view=rev Log: Reject multiple alternative asm constraints which are not consistent in the number of alternatives. Use the same error message as GCC rather than asserting. Modified: dragonegg/trunk/llvm-convert.cpp Modified: dragonegg/trunk/llvm-convert.cpp URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-convert.cpp?rev=126511&r1=126510&r2=126511&view=diff ============================================================================== --- dragonegg/trunk/llvm-convert.cpp (original) +++ dragonegg/trunk/llvm-convert.cpp Fri Feb 25 14:55:08 2011 @@ -6776,8 +6776,11 @@ *p; ++p) if (*p == ',') ++NumInputChoices; - assert((!NumChoices || NumChoices == NumInputChoices) && - "invalid constraints!"); + if (NumChoices && (NumInputChoices != NumChoices)) { + error_at(gimple_location(stmt), "operand constraints for % differ " + "in number of alternatives"); + return; + } if (NumChoices == 0) NumChoices = NumInputChoices; } @@ -6788,8 +6791,11 @@ *p; ++p) if (*p == ',') ++NumOutputChoices; - assert((!NumChoices || NumChoices == NumOutputChoices) && - "invalid constraints!"); + if (NumChoices && (NumOutputChoices != NumChoices)) { + error_at(gimple_location(stmt), "operand constraints for % differ " + "in number of alternatives"); + return; + } if (NumChoices == 0) NumChoices = NumOutputChoices; } From baldrick at free.fr Fri Feb 25 15:16:45 2011 From: baldrick at free.fr (Duncan Sands) Date: Fri, 25 Feb 2011 21:16:45 -0000 Subject: [llvm-commits] [dragonegg] r126512 - /dragonegg/trunk/llvm-convert.cpp Message-ID: <20110225211645.876EC2A6C12C@llvm.org> Author: baldrick Date: Fri Feb 25 15:16:45 2011 New Revision: 126512 URL: http://llvm.org/viewvc/llvm-project?rev=126512&view=rev Log: Add some infrastructure which allows us to alter the result type of an inline asm and have it be corrected to the expected type before being used. Modified: dragonegg/trunk/llvm-convert.cpp Modified: dragonegg/trunk/llvm-convert.cpp URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-convert.cpp?rev=126512&r1=126511&r2=126512&view=diff ============================================================================== --- dragonegg/trunk/llvm-convert.cpp (original) +++ dragonegg/trunk/llvm-convert.cpp Fri Feb 25 15:16:45 2011 @@ -6814,14 +6814,18 @@ std::vector CallOps; std::vector CallArgTypes; - // StoreCallResultAddr - The pointer to store the result of the call through. - SmallVector StoreCallResultAddrs; - // CallResultTypes - The inline asm call may return one or more results. The // types of the results are recorded here along with a flag indicating whether // the corresponding GCC type is signed. SmallVector, 4> CallResultTypes; + // CallResultDests - Each result returned by the inline asm call is stored in + // a memory location. These are listed here along with a flag indicating if + // the GCC type corresponding to the memory location is signed. The type of + // the memory location is allowed to differ from the type of the call result, + // in which case the result is converted before being stored. + SmallVector, 4> CallResultDests; + SmallVector, 4> OutputLocations; // ConstraintStr - The string of constraints in LLVM format. @@ -6877,25 +6881,25 @@ } LValue Dest; - const Type *DestValTy; + const Type *DestValTy = ConvertType(TREE_TYPE(Operand)); if (TREE_CODE(Operand) == SSA_NAME) { // The ASM is defining an ssa name. Store the output to a temporary, then // load it out again later as the ssa name. - DestValTy = ConvertType(TREE_TYPE(Operand)); MemRef TmpLoc = CreateTempLoc(DestValTy); SSADefinitions.push_back(std::make_pair(Operand, TmpLoc)); Dest = LValue(TmpLoc); } else { Dest = EmitLV(Operand); - DestValTy = cast(Dest.Ptr->getType())->getElementType(); + assert(cast(Dest.Ptr->getType())->getElementType() == + DestValTy && "LValue has wrong type!"); } assert(!Dest.isBitfield() && "Cannot assign into a bitfield!"); if (!AllowsMem && DestValTy->isSingleValueType()) {// Reg dest -> asm return - StoreCallResultAddrs.push_back(Dest.Ptr); ConstraintStr += ",="; ConstraintStr += SimplifiedConstraint; bool IsSigned = !TYPE_UNSIGNED(TREE_TYPE(Operand)); + CallResultDests.push_back(std::make_pair(Dest.Ptr, IsSigned)); CallResultTypes.push_back(std::make_pair(DestValTy, IsSigned)); OutputLocations.push_back(std::make_pair(true, CallResultTypes.size()-1)); } else { @@ -7160,13 +7164,17 @@ CV->setDoesNotThrow(); // If the call produces a value, store it into the destination. - if (StoreCallResultAddrs.size() == 1) - Builder.CreateStore(CV, StoreCallResultAddrs[0]); - else if (unsigned NumResults = StoreCallResultAddrs.size()) { - for (unsigned i = 0; i != NumResults; ++i) { - Value *ValI = Builder.CreateExtractValue(CV, i, "asmresult"); - Builder.CreateStore(ValI, StoreCallResultAddrs[i]); - } + for (unsigned i = 0, NumResults = CallResultTypes.size(); i != NumResults; + ++i) { + Value *Val = NumResults == 1 ? + CV : Builder.CreateExtractValue(CV, i, "asmresult"); + bool ValIsSigned = CallResultTypes[i].second; + + Value *Dest = CallResultDests[i].first; + const Type *DestTy = cast(Dest->getType())->getElementType(); + bool DestIsSigned = CallResultDests[i].second; + Val = CastToAnyType(Val, ValIsSigned, DestTy, DestIsSigned); + Builder.CreateStore(Val, Dest); } // If the call defined any ssa names, associate them with their value. From resistor at mac.com Fri Feb 25 15:41:48 2011 From: resistor at mac.com (Owen Anderson) Date: Fri, 25 Feb 2011 21:41:48 -0000 Subject: [llvm-commits] [llvm] r126518 - in /llvm/trunk: include/llvm/Target/ lib/CodeGen/SelectionDAG/ lib/Target/Alpha/ lib/Target/Blackfin/ lib/Target/CellSPU/ lib/Target/MSP430/ lib/Target/PowerPC/ lib/Target/SystemZ/ lib/Target/X86/ lib/Target/XCore/ Message-ID: <20110225214149.7F04A2A6C12C@llvm.org> Author: resistor Date: Fri Feb 25 15:41:48 2011 New Revision: 126518 URL: http://llvm.org/viewvc/llvm-project?rev=126518&view=rev Log: Allow targets to specify a the type of the RHS of a shift parameterized on the type of the LHS. Modified: llvm/trunk/include/llvm/Target/TargetLowering.h llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp llvm/trunk/lib/Target/Alpha/AlphaISelLowering.cpp llvm/trunk/lib/Target/Alpha/AlphaISelLowering.h llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.cpp llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.h llvm/trunk/lib/Target/CellSPU/SPUISelLowering.cpp llvm/trunk/lib/Target/CellSPU/SPUISelLowering.h llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.cpp llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.h llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp llvm/trunk/lib/Target/PowerPC/PPCISelLowering.h llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.cpp llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.h llvm/trunk/lib/Target/X86/X86ISelLowering.cpp llvm/trunk/lib/Target/X86/X86ISelLowering.h llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Modified: llvm/trunk/include/llvm/Target/TargetLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/include/llvm/Target/TargetLowering.h (original) +++ llvm/trunk/include/llvm/Target/TargetLowering.h Fri Feb 25 15:41:48 2011 @@ -111,7 +111,7 @@ bool isBigEndian() const { return !IsLittleEndian; } bool isLittleEndian() const { return IsLittleEndian; } MVT getPointerTy() const { return PointerTy; } - MVT getShiftAmountTy() const { return ShiftAmountTy; } + virtual MVT getShiftAmountTy(EVT LHSTy) const; /// isSelectExpensive - Return true if the select operation is expensive for /// this target. @@ -210,7 +210,7 @@ /// ValueTypeActions - For each value type, keep a LegalizeAction enum /// that indicates how instruction selection should deal with the type. uint8_t ValueTypeActions[MVT::LAST_VALUETYPE]; - + LegalizeAction getExtendedTypeAction(EVT VT) const { // Handle non-vector integers. if (!VT.isVector()) { @@ -260,17 +260,17 @@ ValueTypeActionImpl() { std::fill(ValueTypeActions, array_endof(ValueTypeActions), 0); } - + LegalizeAction getTypeAction(EVT VT) const { if (!VT.isExtended()) return getTypeAction(VT.getSimpleVT()); return getExtendedTypeAction(VT); } - + LegalizeAction getTypeAction(MVT VT) const { return (LegalizeAction)ValueTypeActions[VT.SimpleTy]; } - + void setTypeAction(EVT VT, LegalizeAction Action) { unsigned I = VT.getSimpleVT().SimpleTy; ValueTypeActions[I] = Action; @@ -291,7 +291,7 @@ LegalizeAction getTypeAction(MVT VT) const { return ValueTypeActions.getTypeAction(VT); } - + /// getTypeToTransformTo - For types supported by the target, this is an /// identity function. For types that must be promoted to larger types, this /// returns the larger type to promote to. For integer types that are larger @@ -324,7 +324,7 @@ EVT NVT = VT.getRoundIntegerType(Context); if (NVT == VT) // Size is a power of two - expand to half the size. return EVT::getIntegerVT(Context, VT.getSizeInBits() / 2); - + // Promote to a power of two size, avoiding multi-step promotion. return getTypeAction(NVT) == Promote ? getTypeToTransformTo(Context, NVT) : NVT; @@ -997,10 +997,6 @@ // protected: - /// setShiftAmountType - Describe the type that should be used for shift - /// amounts. This type defaults to the pointer type. - void setShiftAmountType(MVT VT) { ShiftAmountTy = VT; } - /// setBooleanContents - Specify how the target extends the result of a /// boolean value from i1 to a wider type. See getBooleanContents. void setBooleanContents(BooleanContent Ty) { BooleanContents = Ty; } @@ -1047,12 +1043,12 @@ /// SelectIsExpensive - Tells the code generator not to expand operations /// into sequences that use the select operations if possible. - void setSelectIsExpensive(bool isExpensive = true) { - SelectIsExpensive = isExpensive; + void setSelectIsExpensive(bool isExpensive = true) { + SelectIsExpensive = isExpensive; } - /// JumpIsExpensive - Tells the code generator not to expand sequence of - /// operations into a seperate sequences that increases the amount of + /// JumpIsExpensive - Tells the code generator not to expand sequence of + /// operations into a seperate sequences that increases the amount of /// flow control. void setJumpIsExpensive(bool isExpensive = true) { JumpIsExpensive = isExpensive; @@ -1369,7 +1365,7 @@ CW_Good = 1, // Good weight. CW_Better = 2, // Better weight. CW_Best = 3, // Best weight. - + // Well-known weights. CW_SpecificReg = CW_Okay, // Specific register operands. CW_Register = CW_Good, // Register operands. @@ -1422,21 +1418,21 @@ CallOperandVal(0), ConstraintVT(MVT::Other) { } }; - + typedef std::vector AsmOperandInfoVector; - + /// ParseConstraints - Split up the constraint string from the inline /// assembly value into the specific constraints and their prefixes, /// and also tie in the associated operand values. /// If this returns an empty vector, and if the constraint string itself /// isn't empty, there was an error parsing. virtual AsmOperandInfoVector ParseConstraints(ImmutableCallSite CS) const; - + /// Examine constraint type and operand type and determine a weight value. /// The operand object must already have been set up with the operand type. virtual ConstraintWeight getMultipleConstraintMatchWeight( AsmOperandInfo &info, int maIndex) const; - + /// Examine constraint string and operand type and determine a weight value. /// The operand object must already have been set up with the operand type. virtual ConstraintWeight getSingleConstraintMatchWeight( @@ -1446,7 +1442,7 @@ /// type to use for the specific AsmOperandInfo, setting /// OpInfo.ConstraintCode and OpInfo.ConstraintType. If the actual operand /// being passed in is available, it can be passed in as Op, otherwise an - /// empty SDValue can be passed. + /// empty SDValue can be passed. virtual void ComputeConstraintToUse(AsmOperandInfo &OpInfo, SDValue Op, SelectionDAG *DAG = 0) const; @@ -1660,10 +1656,6 @@ /// llvm.longjmp. Defaults to false. bool UseUnderscoreLongJmp; - /// ShiftAmountTy - The type to use for shift amounts, usually i8 or whatever - /// PointerTy is. - MVT ShiftAmountTy; - /// BooleanContents - Information about the contents of the high-bits in /// boolean values held in a type wider than i1. See getBooleanContents. BooleanContent BooleanContents; Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Fri Feb 25 15:41:48 2011 @@ -279,8 +279,8 @@ /// getShiftAmountTy - Returns a type large enough to hold any valid /// shift amount - before type legalization these can be huge. - EVT getShiftAmountTy() { - return LegalTypes ? TLI.getShiftAmountTy() : TLI.getPointerTy(); + EVT getShiftAmountTy(EVT LHSTy) { + return LegalTypes ? TLI.getShiftAmountTy(LHSTy) : TLI.getPointerTy(); } /// isTypeLegal - This method returns true if we are running before type @@ -670,7 +670,7 @@ if (LoadSDNode *LD = dyn_cast(Op)) { EVT MemVT = LD->getMemoryVT(); ISD::LoadExtType ExtType = ISD::isNON_EXTLoad(LD) - ? (TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT) ? ISD::ZEXTLOAD + ? (TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT) ? ISD::ZEXTLOAD : ISD::EXTLOAD) : LD->getExtensionType(); Replace = true; @@ -894,7 +894,7 @@ LoadSDNode *LD = cast(N); EVT MemVT = LD->getMemoryVT(); ISD::LoadExtType ExtType = ISD::isNON_EXTLoad(LD) - ? (TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT) ? ISD::ZEXTLOAD + ? (TLI.isLoadExtLegal(ISD::ZEXTLOAD, MemVT) ? ISD::ZEXTLOAD : ISD::EXTLOAD) : LD->getExtensionType(); SDValue NewLD = DAG.getExtLoad(ExtType, dl, PVT, @@ -1521,7 +1521,7 @@ // Since it may not be valid to emit a fold to zero for vector initializers // check if we can before folding. static SDValue tryFoldToZero(DebugLoc DL, const TargetLowering &TLI, EVT VT, - SelectionDAG &DAG, bool LegalOperations) { + SelectionDAG &DAG, bool LegalOperations) { if (!VT.isVector()) { return DAG.getConstant(0, VT); } else if (!LegalOperations || TLI.isOperationLegal(ISD::BUILD_VECTOR, VT)) { @@ -1647,7 +1647,7 @@ if (N1C && N1C->getAPIntValue().isPowerOf2()) return DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0, DAG.getConstant(N1C->getAPIntValue().logBase2(), - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); // fold (mul x, -(1 << c)) -> -(x << c) or (-x) << c if (N1C && (-N1C->getAPIntValue()).isPowerOf2()) { unsigned Log2Val = (-N1C->getAPIntValue()).logBase2(); @@ -1656,7 +1656,8 @@ return DAG.getNode(ISD::SUB, N->getDebugLoc(), VT, DAG.getConstant(0, VT), DAG.getNode(ISD::SHL, N->getDebugLoc(), VT, N0, - DAG.getConstant(Log2Val, getShiftAmountTy()))); + DAG.getConstant(Log2Val, + getShiftAmountTy(N0.getValueType())))); } // (mul (shl X, c1), c2) -> (mul X, c2 << c1) if (N1C && N0.getOpcode() == ISD::SHL && @@ -1753,18 +1754,18 @@ // Splat the sign bit into the register SDValue SGN = DAG.getNode(ISD::SRA, N->getDebugLoc(), VT, N0, DAG.getConstant(VT.getSizeInBits()-1, - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); AddToWorkList(SGN.getNode()); // Add (N0 < 0) ? abs2 - 1 : 0; SDValue SRL = DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, SGN, DAG.getConstant(VT.getSizeInBits() - lg2, - getShiftAmountTy())); + getShiftAmountTy(SGN.getValueType()))); SDValue ADD = DAG.getNode(ISD::ADD, N->getDebugLoc(), VT, N0, SRL); AddToWorkList(SRL.getNode()); AddToWorkList(ADD.getNode()); // Divide by pow2 SDValue SRA = DAG.getNode(ISD::SRA, N->getDebugLoc(), VT, ADD, - DAG.getConstant(lg2, getShiftAmountTy())); + DAG.getConstant(lg2, getShiftAmountTy(ADD.getValueType()))); // If we're dividing by a positive value, we're done. Otherwise, we must // negate the result. @@ -1814,7 +1815,7 @@ if (N1C && N1C->getAPIntValue().isPowerOf2()) return DAG.getNode(ISD::SRL, N->getDebugLoc(), VT, N0, DAG.getConstant(N1C->getAPIntValue().logBase2(), - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); // fold (udiv x, (shl c, y)) -> x >>u (log2(c)+y) iff c is power of 2 if (N1.getOpcode() == ISD::SHL) { if (ConstantSDNode *SHC = dyn_cast(N1.getOperand(0))) { @@ -1955,7 +1956,7 @@ if (N1C && N1C->getAPIntValue() == 1) return DAG.getNode(ISD::SRA, N->getDebugLoc(), N0.getValueType(), N0, DAG.getConstant(N0.getValueType().getSizeInBits() - 1, - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); // fold (mulhs x, undef) -> 0 if (N0.getOpcode() == ISD::UNDEF || N1.getOpcode() == ISD::UNDEF) return DAG.getConstant(0, VT); @@ -1971,11 +1972,11 @@ N1 = DAG.getNode(ISD::SIGN_EXTEND, DL, NewVT, N1); N1 = DAG.getNode(ISD::MUL, DL, NewVT, N0, N1); N1 = DAG.getNode(ISD::SRL, DL, NewVT, N1, - DAG.getConstant(SimpleSize, getShiftAmountTy())); + DAG.getConstant(SimpleSize, getShiftAmountTy(N1.getValueType()))); return DAG.getNode(ISD::TRUNCATE, DL, VT, N1); } } - + return SDValue(); } @@ -2007,11 +2008,11 @@ N1 = DAG.getNode(ISD::ZERO_EXTEND, DL, NewVT, N1); N1 = DAG.getNode(ISD::MUL, DL, NewVT, N0, N1); N1 = DAG.getNode(ISD::SRL, DL, NewVT, N1, - DAG.getConstant(SimpleSize, getShiftAmountTy())); + DAG.getConstant(SimpleSize, getShiftAmountTy(N1.getValueType()))); return DAG.getNode(ISD::TRUNCATE, DL, VT, N1); } } - + return SDValue(); } @@ -2090,14 +2091,14 @@ Lo = DAG.getNode(ISD::MUL, DL, NewVT, Lo, Hi); // Compute the high part as N1. Hi = DAG.getNode(ISD::SRL, DL, NewVT, Lo, - DAG.getConstant(SimpleSize, getShiftAmountTy())); + DAG.getConstant(SimpleSize, getShiftAmountTy(Lo.getValueType()))); Hi = DAG.getNode(ISD::TRUNCATE, DL, VT, Hi); // Compute the low part as N0. Lo = DAG.getNode(ISD::TRUNCATE, DL, VT, Lo); return CombineTo(N, Lo, Hi); } } - + return SDValue(); } @@ -2107,7 +2108,7 @@ EVT VT = N->getValueType(0); DebugLoc DL = N->getDebugLoc(); - + // If the type twice as wide is legal, transform the mulhu to a wider multiply // plus a shift. if (VT.isSimple() && !VT.isVector()) { @@ -2120,14 +2121,14 @@ Lo = DAG.getNode(ISD::MUL, DL, NewVT, Lo, Hi); // Compute the high part as N1. Hi = DAG.getNode(ISD::SRL, DL, NewVT, Lo, - DAG.getConstant(SimpleSize, getShiftAmountTy())); + DAG.getConstant(SimpleSize, getShiftAmountTy(Lo.getValueType()))); Hi = DAG.getNode(ISD::TRUNCATE, DL, VT, Hi); // Compute the low part as N0. Lo = DAG.getNode(ISD::TRUNCATE, DL, VT, Lo); return CombineTo(N, Lo, Hi); } } - + return SDValue(); } @@ -3004,7 +3005,7 @@ N0.getOpcode() == ISD::SIGN_EXTEND) && N0.getOperand(0).getOpcode() == ISD::SHL && isa(N0.getOperand(0)->getOperand(1))) { - uint64_t c1 = + uint64_t c1 = cast(N0.getOperand(0)->getOperand(1))->getZExtValue(); uint64_t c2 = N1C->getZExtValue(); EVT InnerShiftVT = N0.getOperand(0).getValueType(); @@ -3133,7 +3134,8 @@ TLI.isOperationLegalOrCustom(ISD::TRUNCATE, VT) && TLI.isTruncateFree(VT, TruncVT)) { - SDValue Amt = DAG.getConstant(ShiftAmt, getShiftAmountTy()); + SDValue Amt = DAG.getConstant(ShiftAmt, + getShiftAmountTy(N0.getOperand(0).getValueType())); SDValue Shift = DAG.getNode(ISD::SRL, N0.getDebugLoc(), VT, N0.getOperand(0), Amt); SDValue Trunc = DAG.getNode(ISD::TRUNCATE, N0.getDebugLoc(), TruncVT, @@ -3180,7 +3182,7 @@ LargeShiftAmt->getZExtValue()) { SDValue Amt = DAG.getConstant(LargeShiftAmt->getZExtValue() + N1C->getZExtValue(), - getShiftAmountTy()); + getShiftAmountTy(N0.getOperand(0).getOperand(0).getValueType())); SDValue SRA = DAG.getNode(ISD::SRA, N->getDebugLoc(), LargeVT, N0.getOperand(0).getOperand(0), Amt); return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, SRA); @@ -3245,7 +3247,7 @@ if (N1C && N0.getOpcode() == ISD::TRUNCATE && N0.getOperand(0).getOpcode() == ISD::SRL && isa(N0.getOperand(0)->getOperand(1))) { - uint64_t c1 = + uint64_t c1 = cast(N0.getOperand(0)->getOperand(1))->getZExtValue(); uint64_t c2 = N1C->getZExtValue(); EVT InnerShiftVT = N0.getOperand(0).getValueType(); @@ -3256,7 +3258,7 @@ if (c1 + c2 >= InnerShiftSize) return DAG.getConstant(0, VT); return DAG.getNode(ISD::TRUNCATE, N0->getDebugLoc(), VT, - DAG.getNode(ISD::SRL, N0->getDebugLoc(), InnerShiftVT, + DAG.getNode(ISD::SRL, N0->getDebugLoc(), InnerShiftVT, N0.getOperand(0)->getOperand(0), DAG.getConstant(c1 + c2, ShiftCountVT))); } @@ -3320,7 +3322,7 @@ if (ShAmt) { Op = DAG.getNode(ISD::SRL, N0.getDebugLoc(), VT, Op, - DAG.getConstant(ShAmt, getShiftAmountTy())); + DAG.getConstant(ShAmt, getShiftAmountTy(Op.getValueType()))); AddToWorkList(Op.getNode()); } @@ -4025,11 +4027,11 @@ } DebugLoc DL = N->getDebugLoc(); - - // Ensure that the shift amount is wide enough for the shifted value. + + // Ensure that the shift amount is wide enough for the shifted value. if (VT.getSizeInBits() >= 256) ShAmt = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i32, ShAmt); - + return DAG.getNode(N0.getOpcode(), DL, VT, DAG.getNode(ISD::ZERO_EXTEND, DL, VT, N0.getOperand(0)), ShAmt); @@ -4278,12 +4280,12 @@ return SDValue(); unsigned EVTBits = ExtVT.getSizeInBits(); - + // Do not generate loads of non-round integer types since these can // be expensive (and would be wrong if the type is not byte sized). if (!ExtVT.isRound()) return SDValue(); - + unsigned ShAmt = 0; if (N0.getOpcode() == ISD::SRL && N0.hasOneUse()) { if (ConstantSDNode *N01 = dyn_cast(N0.getOperand(1))) { @@ -4298,7 +4300,7 @@ // At this point, we must have a load or else we can't do the transform. if (!isa(N0)) return SDValue(); - + // If the shift amount is larger than the input type then we're not // accessing any of the loaded bytes. If the load was a zextload/extload // then the result of the shift+trunc is zero/undef (handled elsewhere). @@ -4319,18 +4321,18 @@ N0 = N0.getOperand(0); } } - + // If we haven't found a load, we can't narrow it. Don't transform one with // multiple uses, this would require adding a new load. if (!isa(N0) || !N0.hasOneUse() || // Don't change the width of a volatile load. cast(N0)->isVolatile()) return SDValue(); - + // Verify that we are actually reducing a load width here. if (cast(N0)->getMemoryVT().getSizeInBits() < EVTBits) return SDValue(); - + LoadSDNode *LN0 = cast(N0); EVT PtrType = N0.getOperand(1).getValueType(); @@ -4368,7 +4370,7 @@ // Shift the result left, if we've swallowed a left shift. SDValue Result = Load; if (ShLeftAmt != 0) { - EVT ShImmTy = getShiftAmountTy(); + EVT ShImmTy = getShiftAmountTy(Result.getValueType()); if (!isUIntN(ShImmTy.getSizeInBits(), ShLeftAmt)) ShImmTy = VT; Result = DAG.getNode(ISD::SHL, N0.getDebugLoc(), VT, @@ -5984,7 +5986,8 @@ // shifted by ByteShift and truncated down to NumBytes. if (ByteShift) IVal = DAG.getNode(ISD::SRL, IVal->getDebugLoc(), IVal.getValueType(), IVal, - DAG.getConstant(ByteShift*8, DC->getShiftAmountTy())); + DAG.getConstant(ByteShift*8, + DC->getShiftAmountTy(IVal.getValueType()))); // Figure out the offset for the store and the alignment of the access. unsigned StOffset; @@ -6399,7 +6402,7 @@ EVT VT = InVec.getValueType(); - // If we can't generate a legal BUILD_VECTOR, exit + // If we can't generate a legal BUILD_VECTOR, exit if (LegalOperations && !TLI.isOperationLegal(ISD::BUILD_VECTOR, VT)) return SDValue(); @@ -7107,7 +7110,8 @@ if (N2C && ((N2C->getAPIntValue() & (N2C->getAPIntValue()-1)) == 0)) { unsigned ShCtV = N2C->getAPIntValue().logBase2(); ShCtV = XType.getSizeInBits()-ShCtV-1; - SDValue ShCt = DAG.getConstant(ShCtV, getShiftAmountTy()); + SDValue ShCt = DAG.getConstant(ShCtV, + getShiftAmountTy(N0.getValueType())); SDValue Shift = DAG.getNode(ISD::SRL, N0.getDebugLoc(), XType, N0, ShCt); AddToWorkList(Shift.getNode()); @@ -7123,7 +7127,7 @@ SDValue Shift = DAG.getNode(ISD::SRA, N0.getDebugLoc(), XType, N0, DAG.getConstant(XType.getSizeInBits()-1, - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); AddToWorkList(Shift.getNode()); if (XType.bitsGT(AType)) { @@ -7151,13 +7155,15 @@ // Shift the tested bit over the sign bit. APInt AndMask = ConstAndRHS->getAPIntValue(); SDValue ShlAmt = - DAG.getConstant(AndMask.countLeadingZeros(), getShiftAmountTy()); + DAG.getConstant(AndMask.countLeadingZeros(), + getShiftAmountTy(AndLHS.getValueType())); SDValue Shl = DAG.getNode(ISD::SHL, N0.getDebugLoc(), VT, AndLHS, ShlAmt); // Now arithmetic right shift it all the way over, so the result is either // all-ones, or zero. SDValue ShrAmt = - DAG.getConstant(AndMask.getBitWidth()-1, getShiftAmountTy()); + DAG.getConstant(AndMask.getBitWidth()-1, + getShiftAmountTy(Shl.getValueType())); SDValue Shr = DAG.getNode(ISD::SRA, N0.getDebugLoc(), VT, Shl, ShrAmt); return DAG.getNode(ISD::AND, DL, VT, Shr, N3); @@ -7201,7 +7207,7 @@ // shl setcc result by log2 n2c return DAG.getNode(ISD::SHL, DL, N2.getValueType(), Temp, DAG.getConstant(N2C->getAPIntValue().logBase2(), - getShiftAmountTy())); + getShiftAmountTy(Temp.getValueType()))); } // Check to see if this is the equivalent of setcc @@ -7224,7 +7230,7 @@ SDValue Ctlz = DAG.getNode(ISD::CTLZ, N0.getDebugLoc(), XType, N0); return DAG.getNode(ISD::SRL, DL, XType, Ctlz, DAG.getConstant(Log2_32(XType.getSizeInBits()), - getShiftAmountTy())); + getShiftAmountTy(Ctlz.getValueType()))); } // fold (setgt X, 0) -> (srl (and (-X, ~X), size(X)-1)) if (N1C && N1C->isNullValue() && CC == ISD::SETGT) { @@ -7234,13 +7240,13 @@ return DAG.getNode(ISD::SRL, DL, XType, DAG.getNode(ISD::AND, DL, XType, NegN0, NotN0), DAG.getConstant(XType.getSizeInBits()-1, - getShiftAmountTy())); + getShiftAmountTy(XType))); } // fold (setgt X, -1) -> (xor (srl (X, size(X)-1), 1)) if (N1C && N1C->isAllOnesValue() && CC == ISD::SETGT) { SDValue Sign = DAG.getNode(ISD::SRL, N0.getDebugLoc(), XType, N0, DAG.getConstant(XType.getSizeInBits()-1, - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); return DAG.getNode(ISD::XOR, DL, XType, Sign, DAG.getConstant(1, XType)); } } @@ -7267,7 +7273,7 @@ SDValue Shift = DAG.getNode(ISD::SRA, N0.getDebugLoc(), XType, N0, DAG.getConstant(XType.getSizeInBits()-1, - getShiftAmountTy())); + getShiftAmountTy(N0.getValueType()))); SDValue Add = DAG.getNode(ISD::ADD, N0.getDebugLoc(), XType, N0, Shift); AddToWorkList(Shift.getNode()); Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp Fri Feb 25 15:41:48 2011 @@ -87,7 +87,7 @@ // If someone requests legalization of the new node, return itself. if (From != To) LegalizedNodes.insert(std::make_pair(To, To)); - + // Transfer SDDbgValues. DAG.TransferDbgValues(From, To); } @@ -498,7 +498,8 @@ int IncrementSize = NumBits / 8; // Divide the stored value in two parts. - SDValue ShiftAmount = DAG.getConstant(NumBits, TLI.getShiftAmountTy()); + SDValue ShiftAmount = DAG.getConstant(NumBits, + TLI.getShiftAmountTy(Val.getValueType())); SDValue Lo = Val; SDValue Hi = DAG.getNode(ISD::SRL, dl, VT, Val, ShiftAmount); @@ -645,7 +646,8 @@ } // aggregate the two parts - SDValue ShiftAmount = DAG.getConstant(NumBits, TLI.getShiftAmountTy()); + SDValue ShiftAmount = DAG.getConstant(NumBits, + TLI.getShiftAmountTy(Hi.getValueType())); SDValue Result = DAG.getNode(ISD::SHL, dl, VT, Hi, ShiftAmount); Result = DAG.getNode(ISD::OR, dl, VT, Result, Lo); @@ -1264,7 +1266,8 @@ // Move the top bits to the right place. Hi = DAG.getNode(ISD::SHL, dl, Hi.getValueType(), Hi, - DAG.getConstant(RoundWidth, TLI.getShiftAmountTy())); + DAG.getConstant(RoundWidth, + TLI.getShiftAmountTy(Hi.getValueType()))); // Join the hi and lo parts. Result = DAG.getNode(ISD::OR, dl, Node->getValueType(0), Lo, Hi); @@ -1293,7 +1296,8 @@ // Move the top bits to the right place. Hi = DAG.getNode(ISD::SHL, dl, Hi.getValueType(), Hi, - DAG.getConstant(ExtraWidth, TLI.getShiftAmountTy())); + DAG.getConstant(ExtraWidth, + TLI.getShiftAmountTy(Hi.getValueType()))); // Join the hi and lo parts. Result = DAG.getNode(ISD::OR, dl, Node->getValueType(0), Lo, Hi); @@ -1482,7 +1486,8 @@ Tmp2 = DAG.getNode(ISD::ADD, dl, Tmp2.getValueType(), Tmp2, DAG.getIntPtrConstant(IncrementSize)); Hi = DAG.getNode(ISD::SRL, dl, Tmp3.getValueType(), Tmp3, - DAG.getConstant(RoundWidth, TLI.getShiftAmountTy())); + DAG.getConstant(RoundWidth, + TLI.getShiftAmountTy(Tmp3.getValueType()))); Hi = DAG.getTruncStore(Tmp1, dl, Hi, Tmp2, ST->getPointerInfo().getWithOffset(IncrementSize), ExtraVT, isVolatile, isNonTemporal, @@ -1492,7 +1497,8 @@ // TRUNCSTORE:i24 X -> TRUNCSTORE:i16 (srl X, 8), TRUNCSTORE at +2:i8 X // Store the top RoundWidth bits. Hi = DAG.getNode(ISD::SRL, dl, Tmp3.getValueType(), Tmp3, - DAG.getConstant(ExtraWidth, TLI.getShiftAmountTy())); + DAG.getConstant(ExtraWidth, + TLI.getShiftAmountTy(Tmp3.getValueType()))); Hi = DAG.getTruncStore(Tmp1, dl, Hi, Tmp2, ST->getPointerInfo(), RoundVT, isVolatile, isNonTemporal, Alignment); @@ -1727,7 +1733,8 @@ assert(BitShift < LoadTy.getSizeInBits() && "Pointer advanced wrong?"); if (BitShift) SignBit = DAG.getNode(ISD::SHL, dl, LoadTy, SignBit, - DAG.getConstant(BitShift,TLI.getShiftAmountTy())); + DAG.getConstant(BitShift, + TLI.getShiftAmountTy(SignBit.getValueType()))); } } // Now get the sign bit proper, by seeing whether the value is negative. @@ -2207,7 +2214,8 @@ if (!isSigned) { SDValue Fast = DAG.getNode(ISD::SINT_TO_FP, dl, MVT::f32, Op0); - SDValue ShiftConst = DAG.getConstant(1, TLI.getShiftAmountTy()); + SDValue ShiftConst = + DAG.getConstant(1, TLI.getShiftAmountTy(Op0.getValueType())); SDValue Shr = DAG.getNode(ISD::SRL, dl, MVT::i64, Op0, ShiftConst); SDValue AndConst = DAG.getConstant(1, MVT::i64); SDValue And = DAG.getNode(ISD::AND, dl, MVT::i64, Op0, AndConst); @@ -2226,7 +2234,6 @@ } // Otherwise, implement the fully general conversion. - EVT SHVT = TLI.getShiftAmountTy(); SDValue And = DAG.getNode(ISD::AND, dl, MVT::i64, Op0, DAG.getConstant(UINT64_C(0xfffffffffffff800), MVT::i64)); @@ -2241,6 +2248,7 @@ Op0, DAG.getConstant(UINT64_C(0x0020000000000000), MVT::i64), ISD::SETUGE); SDValue Sel2 = DAG.getNode(ISD::SELECT, dl, MVT::i64, Ge, Sel, Op0); + EVT SHVT = TLI.getShiftAmountTy(Sel2.getValueType()); SDValue Sh = DAG.getNode(ISD::SRL, dl, MVT::i64, Sel2, DAG.getConstant(32, SHVT)); @@ -2387,7 +2395,7 @@ /// SDValue SelectionDAGLegalize::ExpandBSWAP(SDValue Op, DebugLoc dl) { EVT VT = Op.getValueType(); - EVT SHVT = TLI.getShiftAmountTy(); + EVT SHVT = TLI.getShiftAmountTy(VT); SDValue Tmp1, Tmp2, Tmp3, Tmp4, Tmp5, Tmp6, Tmp7, Tmp8; switch (VT.getSimpleVT().SimpleTy) { default: assert(0 && "Unhandled Expand type in BSWAP!"); @@ -2450,7 +2458,7 @@ default: assert(0 && "Cannot expand this yet!"); case ISD::CTPOP: { EVT VT = Op.getValueType(); - EVT ShVT = TLI.getShiftAmountTy(); + EVT ShVT = TLI.getShiftAmountTy(VT); unsigned Len = VT.getSizeInBits(); assert(VT.isInteger() && Len <= 128 && Len % 8 == 0 && @@ -2487,7 +2495,7 @@ Op = DAG.getNode(ISD::SRL, dl, VT, DAG.getNode(ISD::MUL, dl, VT, Op, Mask01), DAG.getConstant(Len - 8, ShVT)); - + return Op; } case ISD::CTLZ: { @@ -2501,7 +2509,7 @@ // // but see also: http://www.hackersdelight.org/HDcode/nlz.cc EVT VT = Op.getValueType(); - EVT ShVT = TLI.getShiftAmountTy(); + EVT ShVT = TLI.getShiftAmountTy(VT); unsigned len = VT.getSizeInBits(); for (unsigned i = 0; (1U << i) <= (len / 2); ++i) { SDValue Tmp3 = DAG.getConstant(1ULL << i, ShVT); @@ -2737,7 +2745,7 @@ // SAR. However, it is doubtful that any exist. EVT ExtraVT = cast(Node->getOperand(1))->getVT(); EVT VT = Node->getValueType(0); - EVT ShiftAmountTy = TLI.getShiftAmountTy(); + EVT ShiftAmountTy = TLI.getShiftAmountTy(VT); if (VT.isVector()) ShiftAmountTy = VT; unsigned BitsDiff = VT.getScalarType().getSizeInBits() - @@ -2901,7 +2909,7 @@ // 1 -> Hi Tmp1 = DAG.getNode(ISD::SRL, dl, OpTy, Node->getOperand(0), DAG.getConstant(OpTy.getSizeInBits()/2, - TLI.getShiftAmountTy())); + TLI.getShiftAmountTy(Node->getOperand(0).getValueType()))); Tmp1 = DAG.getNode(ISD::TRUNCATE, dl, Node->getValueType(0), Tmp1); } else { // 0 -> Lo @@ -3260,7 +3268,7 @@ assert(LC != RTLIB::UNKNOWN_LIBCALL && "Cannot expand this operation!"); LHS = DAG.getNode(Ops[isSigned][2], dl, WideVT, LHS); RHS = DAG.getNode(Ops[isSigned][2], dl, WideVT, RHS); - + SDValue Ret = ExpandLibCall(LC, Node, isSigned); BottomHalf = DAG.getNode(ISD::TRUNCATE, dl, VT, Ret); TopHalf = DAG.getNode(ISD::SRL, dl, Ret.getValueType(), Ret, @@ -3268,7 +3276,8 @@ TopHalf = DAG.getNode(ISD::TRUNCATE, dl, VT, TopHalf); } if (isSigned) { - Tmp1 = DAG.getConstant(VT.getSizeInBits() - 1, TLI.getShiftAmountTy()); + Tmp1 = DAG.getConstant(VT.getSizeInBits() - 1, + TLI.getShiftAmountTy(BottomHalf.getValueType())); Tmp1 = DAG.getNode(ISD::SRA, dl, VT, BottomHalf, Tmp1); TopHalf = DAG.getSetCC(dl, TLI.getSetCCResultType(VT), TopHalf, Tmp1, ISD::SETNE); @@ -3286,7 +3295,7 @@ Tmp2 = DAG.getNode(ISD::ANY_EXTEND, dl, PairTy, Node->getOperand(1)); Tmp2 = DAG.getNode(ISD::SHL, dl, PairTy, Tmp2, DAG.getConstant(PairTy.getSizeInBits()/2, - TLI.getShiftAmountTy())); + TLI.getShiftAmountTy(PairTy))); Results.push_back(DAG.getNode(ISD::OR, dl, PairTy, Tmp1, Tmp2)); break; } @@ -3464,7 +3473,7 @@ Tmp1 = DAG.getNode(ISD::ZERO_EXTEND, dl, NVT, Node->getOperand(0)); Tmp1 = DAG.getNode(ISD::BSWAP, dl, NVT, Tmp1); Tmp1 = DAG.getNode(ISD::SRL, dl, NVT, Tmp1, - DAG.getConstant(DiffBits, TLI.getShiftAmountTy())); + DAG.getConstant(DiffBits, TLI.getShiftAmountTy(NVT))); Results.push_back(Tmp1); break; } Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp Fri Feb 25 15:41:48 2011 @@ -177,25 +177,27 @@ // First get the sign bit of second operand. SDValue SignBit = DAG.getNode(ISD::SHL, dl, RVT, DAG.getConstant(1, RVT), DAG.getConstant(RSize - 1, - TLI.getShiftAmountTy())); + TLI.getShiftAmountTy(RVT))); SignBit = DAG.getNode(ISD::AND, dl, RVT, RHS, SignBit); // Shift right or sign-extend it if the two operands have different types. int SizeDiff = RVT.getSizeInBits() - LVT.getSizeInBits(); if (SizeDiff > 0) { SignBit = DAG.getNode(ISD::SRL, dl, RVT, SignBit, - DAG.getConstant(SizeDiff, TLI.getShiftAmountTy())); + DAG.getConstant(SizeDiff, + TLI.getShiftAmountTy(SignBit.getValueType()))); SignBit = DAG.getNode(ISD::TRUNCATE, dl, LVT, SignBit); } else if (SizeDiff < 0) { SignBit = DAG.getNode(ISD::ANY_EXTEND, dl, LVT, SignBit); SignBit = DAG.getNode(ISD::SHL, dl, LVT, SignBit, - DAG.getConstant(-SizeDiff, TLI.getShiftAmountTy())); + DAG.getConstant(-SizeDiff, + TLI.getShiftAmountTy(SignBit.getValueType()))); } // Clear the sign bit of the first operand. SDValue Mask = DAG.getNode(ISD::SHL, dl, LVT, DAG.getConstant(1, LVT), DAG.getConstant(LSize - 1, - TLI.getShiftAmountTy())); + TLI.getShiftAmountTy(LVT))); Mask = DAG.getNode(ISD::SUB, dl, LVT, Mask, DAG.getConstant(1, LVT)); LHS = DAG.getNode(ISD::AND, dl, LVT, LHS, Mask); Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp Fri Feb 25 15:41:48 2011 @@ -1420,7 +1420,7 @@ /// the target's desired shift amount type. SDValue SelectionDAG::getShiftAmountOperand(SDValue Op) { EVT OpTy = Op.getValueType(); - MVT ShTy = TLI.getShiftAmountTy(); + MVT ShTy = TLI.getShiftAmountTy(OpTy); if (OpTy == ShTy || OpTy.isVector()) return Op; ISD::NodeType Opcode = OpTy.bitsGT(ShTy) ? ISD::TRUNCATE : ISD::ZERO_EXTEND; @@ -2048,7 +2048,7 @@ return; } break; - + default: // Allow the target to implement this method for its nodes. if (Op.getOpcode() >= ISD::BUILTIN_OP_END) { @@ -2292,12 +2292,12 @@ if ((Op.getOpcode() != ISD::ADD && Op.getOpcode() != ISD::OR) || !isa(Op.getOperand(1))) return false; - - if (Op.getOpcode() == ISD::OR && + + if (Op.getOpcode() == ISD::OR && !MaskedValueIsZero(Op.getOperand(0), cast(Op.getOperand(1))->getAPIntValue())) return false; - + return true; } @@ -2748,7 +2748,7 @@ // i8, which is easy to fall into in generic code that uses // TLI.getShiftAmount(). assert(N2.getValueType().getSizeInBits() >= - Log2_32_Ceil(N1.getValueType().getSizeInBits()) && + Log2_32_Ceil(N1.getValueType().getSizeInBits()) && "Invalid use of small shift amount with oversized value!"); // Always fold shifts of i1 values so the code generator doesn't need to Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Fri Feb 25 15:41:48 2011 @@ -909,7 +909,7 @@ Val.getResNo(), Offset, dl, DbgSDNodeOrder); DAG.AddDbgValue(SDV, Val.getNode(), false); } - } else + } else DEBUG(dbgs() << "Dropping debug info for " << DI); DanglingDebugInfoMap[V] = DanglingDebugInfo(); } @@ -1418,7 +1418,7 @@ // jle foo // if (const BinaryOperator *BOp = dyn_cast(CondVal)) { - if (!TLI.isJumpExpensive() && + if (!TLI.isJumpExpensive() && BOp->hasOneUse() && (BOp->getOpcode() == Instruction::And || BOp->getOpcode() == Instruction::Or)) { @@ -2409,19 +2409,19 @@ void SelectionDAGBuilder::visitShift(const User &I, unsigned Opcode) { SDValue Op1 = getValue(I.getOperand(0)); SDValue Op2 = getValue(I.getOperand(1)); - - MVT ShiftTy = TLI.getShiftAmountTy(); - + + MVT ShiftTy = TLI.getShiftAmountTy(Op2.getValueType()); + // Coerce the shift amount to the right type if we can. if (!I.getType()->isVectorTy() && Op2.getValueType() != ShiftTy) { unsigned ShiftSize = ShiftTy.getSizeInBits(); unsigned Op2Size = Op2.getValueType().getSizeInBits(); DebugLoc DL = getCurDebugLoc(); - + // If the operand is smaller than the shift count type, promote it. if (ShiftSize > Op2Size) Op2 = DAG.getNode(ISD::ZERO_EXTEND, DL, ShiftTy, Op2); - + // If the operand is larger than the shift count type but the shift // count type has enough bits to represent any shift value, truncate // it now. This is a common case and it exposes the truncate to Modified: llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp Fri Feb 25 15:41:48 2011 @@ -563,7 +563,7 @@ setOperationAction(ISD::TRAP, MVT::Other, Expand); IsLittleEndian = TD->isLittleEndian(); - ShiftAmountTy = PointerTy = MVT::getIntegerVT(8*TD->getPointerSize()); + PointerTy = MVT::getIntegerVT(8*TD->getPointerSize()); memset(RegClassForVT, 0,MVT::LAST_VALUETYPE*sizeof(TargetRegisterClass*)); memset(TargetDAGCombineArray, 0, array_lengthof(TargetDAGCombineArray)); maxStoresPerMemset = maxStoresPerMemcpy = maxStoresPerMemmove = 8; @@ -596,6 +596,10 @@ delete &TLOF; } +MVT TargetLowering::getShiftAmountTy(EVT LHSTy) const { + return MVT::getIntegerVT(8*TD->getPointerSize()); +} + /// canOpTrap - Returns true if the operation can trap for the value type. /// VT must be a legal type. bool TargetLowering::canOpTrap(unsigned Op, EVT VT) const { @@ -1401,7 +1405,7 @@ BitWidth - InnerVT.getSizeInBits()) & DemandedMask) == 0 && isTypeDesirableForOp(ISD::SHL, InnerVT)) { - EVT ShTy = getShiftAmountTy(); + EVT ShTy = getShiftAmountTy(InnerVT); if (!APInt(BitWidth, ShAmt).isIntN(ShTy.getSizeInBits())) ShTy = InnerVT; SDValue NarrowShl = @@ -2188,7 +2192,7 @@ if (ConstantSDNode *AndRHS = dyn_cast(N0.getOperand(1))) { EVT ShiftTy = DCI.isBeforeLegalize() ? - getPointerTy() : getShiftAmountTy(); + getPointerTy() : getShiftAmountTy(N0.getValueType()); if (Cond == ISD::SETNE && C1 == 0) {// (X & 8) != 0 --> (X & 8) >> 3 // Perform the xform if the AND RHS is a single bit. if (AndRHS->getAPIntValue().isPowerOf2()) { @@ -2359,7 +2363,7 @@ // (Z-X) == X --> Z == X<<1 SDValue SH = DAG.getNode(ISD::SHL, dl, N1.getValueType(), N1, - DAG.getConstant(1, getShiftAmountTy())); + DAG.getConstant(1, getShiftAmountTy(N1.getValueType()))); if (!DCI.isCalledByLegalizer()) DCI.AddToWorklist(SH.getNode()); return DAG.getSetCC(dl, VT, N0.getOperand(0), SH, Cond); @@ -2381,7 +2385,7 @@ assert(N1.getOpcode() == ISD::SUB && "Unexpected operation!"); // X == (Z-X) --> X<<1 == Z SDValue SH = DAG.getNode(ISD::SHL, dl, N1.getValueType(), N0, - DAG.getConstant(1, getShiftAmountTy())); + DAG.getConstant(1, getShiftAmountTy(N0.getValueType()))); if (!DCI.isCalledByLegalizer()) DCI.AddToWorklist(SH.getNode()); return DAG.getSetCC(dl, VT, SH, N1.getOperand(0), Cond); @@ -2493,7 +2497,7 @@ } } } - + return false; } @@ -3141,14 +3145,14 @@ // Shift right algebraic if shift value is nonzero if (magics.s > 0) { Q = DAG.getNode(ISD::SRA, dl, VT, Q, - DAG.getConstant(magics.s, getShiftAmountTy())); + DAG.getConstant(magics.s, getShiftAmountTy(Q.getValueType()))); if (Created) Created->push_back(Q.getNode()); } // Extract the sign bit and add it to the quotient SDValue T = DAG.getNode(ISD::SRL, dl, VT, Q, DAG.getConstant(VT.getSizeInBits()-1, - getShiftAmountTy())); + getShiftAmountTy(Q.getValueType()))); if (Created) Created->push_back(T.getNode()); return DAG.getNode(ISD::ADD, dl, VT, Q, T); @@ -3192,19 +3196,19 @@ assert(magics.s < N1C->getAPIntValue().getBitWidth() && "We shouldn't generate an undefined shift!"); return DAG.getNode(ISD::SRL, dl, VT, Q, - DAG.getConstant(magics.s, getShiftAmountTy())); + DAG.getConstant(magics.s, getShiftAmountTy(Q.getValueType()))); } else { SDValue NPQ = DAG.getNode(ISD::SUB, dl, VT, N->getOperand(0), Q); if (Created) Created->push_back(NPQ.getNode()); NPQ = DAG.getNode(ISD::SRL, dl, VT, NPQ, - DAG.getConstant(1, getShiftAmountTy())); + DAG.getConstant(1, getShiftAmountTy(NPQ.getValueType()))); if (Created) Created->push_back(NPQ.getNode()); NPQ = DAG.getNode(ISD::ADD, dl, VT, NPQ, Q); if (Created) Created->push_back(NPQ.getNode()); return DAG.getNode(ISD::SRL, dl, VT, NPQ, - DAG.getConstant(magics.s-1, getShiftAmountTy())); + DAG.getConstant(magics.s-1, getShiftAmountTy(NPQ.getValueType()))); } } Modified: llvm/trunk/lib/Target/Alpha/AlphaISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Alpha/AlphaISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/Alpha/AlphaISelLowering.cpp (original) +++ llvm/trunk/lib/Target/Alpha/AlphaISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -48,7 +48,6 @@ : TargetLowering(TM, new TargetLoweringObjectFileELF()) { // Set up the TargetLowering object. //I am having problems with shr n i8 1 - setShiftAmountType(MVT::i64); setBooleanContents(ZeroOrOneBooleanContent); addRegisterClass(MVT::i64, Alpha::GPRCRegisterClass); Modified: llvm/trunk/lib/Target/Alpha/AlphaISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Alpha/AlphaISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/Alpha/AlphaISelLowering.h (original) +++ llvm/trunk/lib/Target/Alpha/AlphaISelLowering.h Fri Feb 25 15:41:48 2011 @@ -31,25 +31,25 @@ /// GPRelHi/GPRelLo - These represent the high and low 16-bit /// parts of a global address respectively. - GPRelHi, GPRelLo, + GPRelHi, GPRelLo, /// RetLit - Literal Relocation of a Global RelLit, /// GlobalRetAddr - used to restore the return address GlobalRetAddr, - + /// CALL - Normal call. CALL, /// DIVCALL - used for special library calls for div and rem DivCall, - + /// return flag operand RET_FLAG, /// CHAIN = COND_BRANCH CHAIN, OPC, (G|F)PRC, DESTBB [, INFLAG] - This - /// corresponds to the COND_BRANCH pseudo instruction. + /// corresponds to the COND_BRANCH pseudo instruction. /// *PRC is the input register to compare to zero, /// OPC is the branch opcode to use (e.g. Alpha::BEQ), /// DESTBB is the destination block to branch to, and INFLAG is @@ -62,7 +62,9 @@ class AlphaTargetLowering : public TargetLowering { public: explicit AlphaTargetLowering(TargetMachine &TM); - + + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i64; } + /// getSetCCResultType - Get the SETCC result ValueType virtual MVT::SimpleValueType getSetCCResultType(EVT VT) const; @@ -92,7 +94,7 @@ ConstraintWeight getSingleConstraintMatchWeight( AsmOperandInfo &info, const char *constraint) const; - std::vector + std::vector getRegClassForInlineAsmConstraint(const std::string &Constraint, EVT VT) const; Modified: llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.cpp (original) +++ llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -41,7 +41,6 @@ BlackfinTargetLowering::BlackfinTargetLowering(TargetMachine &TM) : TargetLowering(TM, new TargetLoweringObjectFileELF()) { - setShiftAmountType(MVT::i16); setBooleanContents(ZeroOrOneBooleanContent); setStackPointerRegisterToSaveRestore(BF::SP); setIntDivIsCheap(false); Modified: llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.h (original) +++ llvm/trunk/lib/Target/Blackfin/BlackfinISelLowering.h Fri Feb 25 15:41:48 2011 @@ -32,6 +32,7 @@ class BlackfinTargetLowering : public TargetLowering { public: BlackfinTargetLowering(TargetMachine &TM); + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i16; } virtual MVT::SimpleValueType getSetCCResultType(EVT VT) const; virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const; virtual void ReplaceNodeResults(SDNode *N, Modified: llvm/trunk/lib/Target/CellSPU/SPUISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPUISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/CellSPU/SPUISelLowering.cpp (original) +++ llvm/trunk/lib/Target/CellSPU/SPUISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -435,7 +435,6 @@ setOperationAction(ISD::FDIV, MVT::v4f32, Legal); - setShiftAmountType(MVT::i32); setBooleanContents(ZeroOrNegativeOneBooleanContent); setStackPointerRegisterToSaveRestore(SPU::R1); @@ -2190,7 +2189,7 @@ { SDValue N0 = Op.getOperand(0); // Everything has at least one operand DebugLoc dl = Op.getDebugLoc(); - EVT ShiftVT = TLI.getShiftAmountTy(); + EVT ShiftVT = TLI.getShiftAmountTy(N0.getValueType()); assert(Op.getValueType() == MVT::i8); switch (Opc) { @@ -3112,7 +3111,7 @@ switch (*constraint) { default: weight = TargetLowering::getSingleConstraintMatchWeight(info, constraint); - break; + break; //FIXME: Seems like the supported constraint letters were just copied // from PPC, as the following doesn't correspond to the GCC docs. // I'm leaving it so until someone adds the corresponding lowering support. Modified: llvm/trunk/lib/Target/CellSPU/SPUISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPUISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/CellSPU/SPUISelLowering.h (original) +++ llvm/trunk/lib/Target/CellSPU/SPUISelLowering.h Fri Feb 25 15:41:48 2011 @@ -109,6 +109,8 @@ /// getSetCCResultType - Return the ValueType for ISD::SETCC virtual MVT::SimpleValueType getSetCCResultType(EVT VT) const; + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i32; } + //! Custom lowering hooks virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const; @@ -179,9 +181,9 @@ virtual bool isLegalICmpImmediate(int64_t Imm) const; - virtual bool isLegalAddressingMode(const AddrMode &AM, + virtual bool isLegalAddressingMode(const AddrMode &AM, const Type *Ty) const; - + /// After allocating this many registers, the allocator should feel /// register pressure. The value is a somewhat random guess, based on the /// number of non callee saved registers in the C calling convention. Modified: llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.cpp (original) +++ llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -77,10 +77,6 @@ // Division is expensive setIntDivIsCheap(false); - // Even if we have only 1 bit shift here, we can perform - // shifts of the whole bitwidth 1 bit per step. - setShiftAmountType(MVT::i8); - setStackPointerRegisterToSaveRestore(MSP430::SPW); setBooleanContents(ZeroOrOneBooleanContent); setSchedulingPreference(Sched::Latency); @@ -330,7 +326,7 @@ // Arguments passed in registers EVT RegVT = VA.getLocVT(); switch (RegVT.getSimpleVT().SimpleTy) { - default: + default: { #ifndef NDEBUG errs() << "LowerFormalArguments Unhandled argument type: " Modified: llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.h (original) +++ llvm/trunk/lib/Target/MSP430/MSP430ISelLowering.h Fri Feb 25 15:41:48 2011 @@ -73,6 +73,8 @@ public: explicit MSP430TargetLowering(MSP430TargetMachine &TM); + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i8; } + /// LowerOperation - Provide custom lowering hooks for some operations. virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const; Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp (original) +++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -362,7 +362,6 @@ setOperationAction(ISD::BUILD_VECTOR, MVT::v4f32, Custom); } - setShiftAmountType(MVT::i32); setBooleanContents(ZeroOrOneBooleanContent); if (TM.getSubtarget().isPPC64()) { Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.h (original) +++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.h Fri Feb 25 15:41:48 2011 @@ -29,36 +29,36 @@ /// FSEL - Traditional three-operand fsel node. /// FSEL, - + /// FCFID - The FCFID instruction, taking an f64 operand and producing /// and f64 value containing the FP representation of the integer that /// was temporarily in the f64 operand. FCFID, - - /// FCTI[D,W]Z - The FCTIDZ and FCTIWZ instructions, taking an f32 or f64 + + /// FCTI[D,W]Z - The FCTIDZ and FCTIWZ instructions, taking an f32 or f64 /// operand, producing an f64 value containing the integer representation /// of that FP value. FCTIDZ, FCTIWZ, - + /// STFIWX - The STFIWX instruction. The first operand is an input token /// chain, then an f64 value to store, then an address to store it to. STFIWX, - + // VMADDFP, VNMSUBFP - The VMADDFP and VNMSUBFP instructions, taking // three v4f32 operands and producing a v4f32 result. VMADDFP, VNMSUBFP, - + /// VPERM - The PPC VPERM Instruction. /// VPERM, - + /// Hi/Lo - These represent the high and low 16-bit parts of a global /// address respectively. These nodes have two operands, the first of /// which must be a TargetGlobalAddress, and the second of which must be a /// Constant. Selected naively, these turn into 'lis G+C' and 'li G+C', /// though these are usually folded into other nodes. Hi, Lo, - + TOC_ENTRY, /// The following three target-specific nodes are used for calls through @@ -80,37 +80,37 @@ /// This instruction is lowered in PPCRegisterInfo::eliminateFrameIndex to /// compute an allocation on the stack. DYNALLOC, - + /// GlobalBaseReg - On Darwin, this node represents the result of the mflr /// at function entry, used for PIC code. GlobalBaseReg, - + /// These nodes represent the 32-bit PPC shifts that operate on 6-bit /// shift amounts. These nodes are generated by the multi-precision shift /// code. SRL, SRA, SHL, - + /// EXTSW_32 - This is the EXTSW instruction for use with "32-bit" /// registers. EXTSW_32, /// CALL - A direct function call. CALL_Darwin, CALL_SVR4, - + /// NOP - Special NOP which follows 64-bit SVR4 calls. NOP, /// CHAIN,FLAG = MTCTR(VAL, CHAIN[, INFLAG]) - Directly corresponds to a /// MTCTR instruction. MTCTR, - + /// CHAIN,FLAG = BCTRL(CHAIN, INFLAG) - Directly corresponds to a /// BCTRL instruction. BCTRL_Darwin, BCTRL_SVR4, - + /// Return with a flag operand, matched by 'blr' RET_FLAG, - + /// R32 = MFCR(CRREG, INFLAG) - Represents the MFCRpseud/MFOCRF /// instructions. This copies the bits corresponding to the specified /// CRREG into the resultant GPR. Bits corresponding to other CR regs @@ -122,20 +122,20 @@ /// encoding for the OPC field to identify the compare. For example, 838 /// is VCMPGTSH. VCMP, - + /// RESVEC, OUTFLAG = VCMPo(LHS, RHS, OPC) - Represents one of the - /// altivec VCMP*o instructions. For lack of better number, we use the + /// altivec VCMP*o instructions. For lack of better number, we use the /// opcode number encoding for the OPC field to identify the compare. For /// example, 838 is VCMPGTSH. VCMPo, - + /// CHAIN = COND_BRANCH CHAIN, CRRC, OPC, DESTBB [, INFLAG] - This /// corresponds to the COND_BRANCH pseudo instruction. CRRC is the /// condition register to branch on, OPC is the branch opcode to use (e.g. /// PPC::BLE), DESTBB is the destination block to branch to, and INFLAG is /// an optional input flag argument. COND_BRANCH, - + // The following 5 instructions are used only as part of the // long double-to-int conversion sequence. @@ -150,7 +150,7 @@ MTFSB1, /// F8RC, OUTFLAG = FADDRTZ F8RC, F8RC, INFLAG - This is an FADD done with - /// rounding towards zero. It has flags added so it won't move past the + /// rounding towards zero. It has flags added so it won't move past the /// FPSCR-setting instructions. FADDRTZ, @@ -174,14 +174,14 @@ /// STD_32 - This is the STD instruction for use with "32-bit" registers. STD_32 = ISD::FIRST_TARGET_MEMORY_OPCODE, - - /// CHAIN = STBRX CHAIN, GPRC, Ptr, Type - This is a + + /// CHAIN = STBRX CHAIN, GPRC, Ptr, Type - This is a /// byte-swapping store instruction. It byte-swaps the low "Type" bits of /// the GPRC input, then stores it through Ptr. Type can be either i16 or /// i32. - STBRX, - - /// GPRC, CHAIN = LBRX CHAIN, Ptr, Type - This is a + STBRX, + + /// GPRC, CHAIN = LBRX CHAIN, Ptr, Type - This is a /// byte-swapping load instruction. It loads "Type" bits, byte swaps it, /// then puts it in the bottom bits of the GPRC. TYPE can be either i16 /// or i32. @@ -194,7 +194,7 @@ /// isVPKUHUMShuffleMask - Return true if this is the shuffle mask for a /// VPKUHUM instruction. bool isVPKUHUMShuffleMask(ShuffleVectorSDNode *N, bool isUnary); - + /// isVPKUWUMShuffleMask - Return true if this is the shuffle mask for a /// VPKUWUM instruction. bool isVPKUWUMShuffleMask(ShuffleVectorSDNode *N, bool isUnary); @@ -208,16 +208,16 @@ /// a VRGH* instruction with the specified unit size (1,2 or 4 bytes). bool isVMRGHShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize, bool isUnary); - + /// isVSLDOIShuffleMask - If this is a vsldoi shuffle mask, return the shift /// amount, otherwise return -1. int isVSLDOIShuffleMask(SDNode *N, bool isUnary); - + /// isSplatShuffleMask - Return true if the specified VECTOR_SHUFFLE operand /// specifies a splat of a single element that is suitable for input to /// VSPLTB/VSPLTH/VSPLTW. bool isSplatShuffleMask(ShuffleVectorSDNode *N, unsigned EltSize); - + /// isAllNegativeZeroVector - Returns true if all elements of build_vector /// are -0.0. bool isAllNegativeZeroVector(SDNode *N); @@ -225,24 +225,26 @@ /// getVSPLTImmediate - Return the appropriate VSPLT* immediate to splat the /// specified isSplatShuffleMask VECTOR_SHUFFLE mask. unsigned getVSPLTImmediate(SDNode *N, unsigned EltSize); - + /// get_VSPLTI_elt - If this is a build_vector of constants which can be /// formed by using a vspltis[bhw] instruction of the specified element /// size, return the constant being splatted. The ByteSize field indicates /// the number of bytes of each element [124] -> [bhw]. SDValue get_VSPLTI_elt(SDNode *N, unsigned ByteSize, SelectionDAG &DAG); } - + class PPCTargetLowering : public TargetLowering { const PPCSubtarget &PPCSubTarget; public: explicit PPCTargetLowering(PPCTargetMachine &TM); - + /// getTargetNodeName() - This method returns the name of a target specific /// DAG node. virtual const char *getTargetNodeName(unsigned Opcode) const; + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i32; } + /// getSetCCResultType - Return the ISD::SETCC ValueType virtual MVT::SimpleValueType getSetCCResultType(EVT VT) const; @@ -253,19 +255,19 @@ SDValue &Offset, ISD::MemIndexedMode &AM, SelectionDAG &DAG) const; - + /// SelectAddressRegReg - Given the specified addressed, check to see if it /// can be represented as an indexed [r+r] operation. Returns false if it /// can be more efficiently represented with [r+imm]. bool SelectAddressRegReg(SDValue N, SDValue &Base, SDValue &Index, SelectionDAG &DAG) const; - + /// SelectAddressRegImm - Returns true if the address N can be represented /// by a base register plus a signed 16-bit displacement [r+imm], and if it /// is not better represented as reg+reg. bool SelectAddressRegImm(SDValue N, SDValue &Disp, SDValue &Base, SelectionDAG &DAG) const; - + /// SelectAddressRegRegOnly - Given the specified addressed, force it to be /// represented as an indexed [r+r] operation. bool SelectAddressRegRegOnly(SDValue N, SDValue &Base, SDValue &Index, @@ -277,7 +279,7 @@ bool SelectAddressRegImmShift(SDValue N, SDValue &Disp, SDValue &Base, SelectionDAG &DAG) const; - + /// LowerOperation - Provide custom lowering hooks for some operations. /// virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const; @@ -289,10 +291,10 @@ SelectionDAG &DAG) const; virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const; - + virtual void computeMaskedBitsForTargetNode(const SDValue Op, const APInt &Mask, - APInt &KnownZero, + APInt &KnownZero, APInt &KnownOne, const SelectionDAG &DAG, unsigned Depth = 0) const; @@ -300,13 +302,13 @@ virtual MachineBasicBlock * EmitInstrWithCustomInserter(MachineInstr *MI, MachineBasicBlock *MBB) const; - MachineBasicBlock *EmitAtomicBinary(MachineInstr *MI, + MachineBasicBlock *EmitAtomicBinary(MachineInstr *MI, MachineBasicBlock *MBB, bool is64Bit, unsigned BinOpcode) const; - MachineBasicBlock *EmitPartwordAtomicBinary(MachineInstr *MI, - MachineBasicBlock *MBB, + MachineBasicBlock *EmitPartwordAtomicBinary(MachineInstr *MI, + MachineBasicBlock *MBB, bool is8bit, unsigned Opcode) const; - + ConstraintType getConstraintType(const std::string &Constraint) const; /// Examine constraint string and operand type and determine a weight value. @@ -314,7 +316,7 @@ ConstraintWeight getSingleConstraintMatchWeight( AsmOperandInfo &info, const char *constraint) const; - std::pair + std::pair getRegForInlineAsmConstraint(const std::string &Constraint, EVT VT) const; @@ -329,11 +331,11 @@ char ConstraintLetter, std::vector &Ops, SelectionDAG &DAG) const; - + /// isLegalAddressingMode - Return true if the addressing mode represented /// by AM is legal for this target, for a load/store of the specified type. virtual bool isLegalAddressingMode(const AddrMode &AM, const Type *Ty)const; - + /// isLegalAddressImmediate - Return true if the integer value can be used /// as the offset of the target addressing mode for load / store of the /// given type. @@ -344,7 +346,7 @@ virtual bool isLegalAddressImmediate(GlobalValue *GV) const; virtual bool isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const; - + /// getOptimalMemOpType - Returns the target specific optimal type for load /// and store operations as a result of memset, memcpy, and memmove /// lowering. If DstAlign is zero that means it's safe to destination Modified: llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.cpp (original) +++ llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -59,9 +59,6 @@ // Compute derived properties from the register classes computeRegisterProperties(); - // Set shifts properties - setShiftAmountType(MVT::i64); - // Provide all sorts of operation actions setLoadExtAction(ISD::SEXTLOAD, MVT::i1, Promote); setLoadExtAction(ISD::ZEXTLOAD, MVT::i1, Promote); Modified: llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.h (original) +++ llvm/trunk/lib/Target/SystemZ/SystemZISelLowering.h Fri Feb 25 15:41:48 2011 @@ -57,6 +57,8 @@ public: explicit SystemZTargetLowering(SystemZTargetMachine &TM); + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i64; } + /// LowerOperation - Provide custom lowering hooks for some operations. virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const; Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original) +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -220,7 +220,6 @@ static MVT IntVTs[] = { MVT::i8, MVT::i16, MVT::i32, MVT::i64 }; // X86 is weird, it always uses i8 for shift amounts and setcc results. - setShiftAmountType(MVT::i8); setBooleanContents(ZeroOrOneBooleanContent); setSchedulingPreference(Sched::RegPressure); setStackPointerRegisterToSaveRestore(X86StackPtr); @@ -4181,7 +4180,8 @@ SrcOp = DAG.getNode(ISD::BITCAST, dl, ShVT, SrcOp); return DAG.getNode(ISD::BITCAST, dl, VT, DAG.getNode(Opc, dl, ShVT, SrcOp, - DAG.getConstant(NumBits, TLI.getShiftAmountTy()))); + DAG.getConstant(NumBits, + TLI.getShiftAmountTy(SrcOp.getValueType())))); } SDValue @@ -4330,15 +4330,15 @@ // For AVX-length vectors, build the individual 128-bit pieces and // use shuffles to put them in place. - if (VT.getSizeInBits() > 256 && - Subtarget->hasAVX() && + if (VT.getSizeInBits() > 256 && + Subtarget->hasAVX() && !ISD::isBuildVectorAllZeros(Op.getNode())) { SmallVector V; V.resize(NumElems); for (unsigned i = 0; i < NumElems; ++i) { V[i] = Op.getOperand(i); } - + EVT HVT = EVT::getVectorVT(*DAG.getContext(), ExtVT, NumElems/2); // Build the lower subvector. @@ -5046,7 +5046,8 @@ DAG.getIntPtrConstant(Elt1 / 2)); if ((Elt1 & 1) == 0) InsElt = DAG.getNode(ISD::SHL, dl, MVT::i16, InsElt, - DAG.getConstant(8, TLI.getShiftAmountTy())); + DAG.getConstant(8, + TLI.getShiftAmountTy(InsElt.getValueType()))); else if (Elt0 >= 0) InsElt = DAG.getNode(ISD::AND, dl, MVT::i16, InsElt, DAG.getConstant(0xFF00, MVT::i16)); @@ -5060,7 +5061,8 @@ Elt0Src, DAG.getIntPtrConstant(Elt0 / 2)); if ((Elt0 & 1) != 0) InsElt0 = DAG.getNode(ISD::SRL, dl, MVT::i16, InsElt0, - DAG.getConstant(8, TLI.getShiftAmountTy())); + DAG.getConstant(8, + TLI.getShiftAmountTy(InsElt0.getValueType()))); else if (Elt1 >= 0) InsElt0 = DAG.getNode(ISD::AND, dl, MVT::i16, InsElt0, DAG.getConstant(0x00FF, MVT::i16)); @@ -5477,7 +5479,7 @@ // Both of them can't be memory operations though. if (MayFoldVectorLoad(V1) && MayFoldVectorLoad(V2)) CanFoldLoad = false; - + if (CanFoldLoad) { if (HasSSE2 && NumElems == 2) return getTargetShuffleNode(X86ISD::MOVLPD, dl, VT, V1, V2, DAG); @@ -6090,7 +6092,7 @@ SDValue ScaledN2 = N2; if (Upper) ScaledN2 = DAG.getNode(ISD::SUB, dl, N2.getValueType(), N2, - DAG.getConstant(NumElems / + DAG.getConstant(NumElems / (VT.getSizeInBits() / 128), N2.getValueType())); Op = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, SubN0.getValueType(), SubN0, Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/X86/X86ISelLowering.h (original) +++ llvm/trunk/lib/Target/X86/X86ISelLowering.h Fri Feb 25 15:41:48 2011 @@ -159,16 +159,16 @@ /// PSHUFB - Shuffle 16 8-bit values within a vector. PSHUFB, - + /// PANDN - and with not'd value. PANDN, - + /// PSIGNB/W/D - Copy integer sign. - PSIGNB, PSIGNW, PSIGND, - + PSIGNB, PSIGNW, PSIGND, + /// PBLENDVB - Variable blend PBLENDVB, - + /// FMAX, FMIN - Floating point max and min. /// FMAX, FMIN, @@ -212,7 +212,7 @@ // ADD, SUB, SMUL, etc. - Arithmetic operations with FLAGS results. ADD, SUB, ADC, SBB, SMUL, INC, DEC, OR, XOR, AND, - + UMUL, // LOW, HI, FLAGS = umul LHS, RHS // MUL_IMM - X86 specific multiply by immediate. @@ -467,6 +467,8 @@ virtual unsigned getJumpTableEncoding() const; + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i8; } + virtual const MCExpr * LowerCustomJumpTableEntry(const MachineJumpTableInfo *MJTI, const MachineBasicBlock *MBB, unsigned uid, Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original) +++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Fri Feb 25 15:41:48 2011 @@ -42,9 +42,9 @@ using namespace llvm; const char *XCoreTargetLowering:: -getTargetNodeName(unsigned Opcode) const +getTargetNodeName(unsigned Opcode) const { - switch (Opcode) + switch (Opcode) { case XCoreISD::BL : return "XCoreISD::BL"; case XCoreISD::PCRelativeWrapper : return "XCoreISD::PCRelativeWrapper"; @@ -77,7 +77,6 @@ // Division is expensive setIntDivIsCheap(false); - setShiftAmountType(MVT::i32); setStackPointerRegisterToSaveRestore(XCore::SP); setSchedulingPreference(Sched::RegPressure); @@ -95,7 +94,7 @@ // Stop the combiner recombining select and set_cc setOperationAction(ISD::SELECT_CC, MVT::Other, Expand); - + // 64bit setOperationAction(ISD::ADD, MVT::i64, Custom); setOperationAction(ISD::SUB, MVT::i64, Custom); @@ -106,14 +105,14 @@ setOperationAction(ISD::SHL_PARTS, MVT::i32, Expand); setOperationAction(ISD::SRA_PARTS, MVT::i32, Expand); setOperationAction(ISD::SRL_PARTS, MVT::i32, Expand); - + // Bit Manipulation setOperationAction(ISD::CTPOP, MVT::i32, Expand); setOperationAction(ISD::ROTL , MVT::i32, Expand); setOperationAction(ISD::ROTR , MVT::i32, Expand); - + setOperationAction(ISD::TRAP, MVT::Other, Legal); - + // Jump tables. setOperationAction(ISD::BR_JT, MVT::Other, Custom); @@ -122,7 +121,7 @@ // Thread Local Storage setOperationAction(ISD::GlobalTLSAddress, MVT::i32, Custom); - + // Conversion of i64 -> double produces constantpool nodes setOperationAction(ISD::ConstantPool, MVT::i32, Custom); @@ -143,7 +142,7 @@ setOperationAction(ISD::VACOPY, MVT::Other, Expand); setOperationAction(ISD::VAARG, MVT::Other, Custom); setOperationAction(ISD::VASTART, MVT::Other, Custom); - + // Dynamic stack setOperationAction(ISD::STACKSAVE, MVT::Other, Expand); setOperationAction(ISD::STACKRESTORE, MVT::Other, Expand); @@ -163,7 +162,7 @@ SDValue XCoreTargetLowering:: LowerOperation(SDValue Op, SelectionDAG &DAG) const { - switch (Op.getOpcode()) + switch (Op.getOpcode()) { case ISD::GlobalAddress: return LowerGlobalAddress(Op, DAG); case ISD::GlobalTLSAddress: return LowerGlobalTLSAddress(Op, DAG); @@ -414,7 +413,7 @@ SDValue Chain = LD->getChain(); SDValue BasePtr = LD->getBasePtr(); DebugLoc DL = Op.getDebugLoc(); - + SDValue Base; int64_t Offset; if (!LD->isVolatile() && @@ -437,10 +436,10 @@ SDValue HighOffset = DAG.getConstant((Offset & ~0x3) + 4, MVT::i32); SDValue LowShift = DAG.getConstant((Offset & 0x3) * 8, MVT::i32); SDValue HighShift = DAG.getConstant(32 - (Offset & 0x3) * 8, MVT::i32); - + SDValue LowAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, LowOffset); SDValue HighAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, HighOffset); - + SDValue Low = DAG.getLoad(getPointerTy(), DL, Chain, LowAddr, MachinePointerInfo(), false, false, 0); SDValue High = DAG.getLoad(getPointerTy(), DL, Chain, @@ -453,7 +452,7 @@ SDValue Ops[] = { Result, Chain }; return DAG.getMergeValues(Ops, 2, DL); } - + if (LD->getAlignment() == 2) { SDValue Low = DAG.getExtLoad(ISD::ZEXTLOAD, DL, MVT::i32, Chain, BasePtr, LD->getPointerInfo(), MVT::i16, @@ -473,16 +472,16 @@ SDValue Ops[] = { Result, Chain }; return DAG.getMergeValues(Ops, 2, DL); } - + // Lower to a call to __misaligned_load(BasePtr). const Type *IntPtrTy = getTargetData()->getIntPtrType(*DAG.getContext()); TargetLowering::ArgListTy Args; TargetLowering::ArgListEntry Entry; - + Entry.Ty = IntPtrTy; Entry.Node = BasePtr; Args.push_back(Entry); - + std::pair CallResult = LowerCallTo(Chain, IntPtrTy, false, false, false, false, 0, CallingConv::C, false, @@ -515,7 +514,7 @@ SDValue BasePtr = ST->getBasePtr(); SDValue Value = ST->getValue(); DebugLoc dl = Op.getDebugLoc(); - + if (ST->getAlignment() == 2) { SDValue Low = Value; SDValue High = DAG.getNode(ISD::SRL, dl, MVT::i32, Value, @@ -532,19 +531,19 @@ ST->isNonTemporal(), 2); return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, StoreLow, StoreHigh); } - + // Lower to a call to __misaligned_store(BasePtr, Value). const Type *IntPtrTy = getTargetData()->getIntPtrType(*DAG.getContext()); TargetLowering::ArgListTy Args; TargetLowering::ArgListEntry Entry; - + Entry.Ty = IntPtrTy; Entry.Node = BasePtr; Args.push_back(Entry); - + Entry.Node = Value; Args.push_back(Entry); - + std::pair CallResult = LowerCallTo(Chain, Type::getVoidTy(*DAG.getContext()), false, false, false, false, 0, CallingConv::C, false, @@ -722,7 +721,7 @@ } DebugLoc dl = N->getDebugLoc(); - + // Extract components SDValue LHSL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32, N->getOperand(0), DAG.getConstant(0, MVT::i32)); @@ -732,7 +731,7 @@ N->getOperand(1), DAG.getConstant(0, MVT::i32)); SDValue RHSH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32, N->getOperand(1), DAG.getConstant(1, MVT::i32)); - + // Expand unsigned Opcode = (N->getOpcode() == ISD::ADD) ? XCoreISD::LADD : XCoreISD::LSUB; @@ -740,7 +739,7 @@ SDValue Carry = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32), LHSL, RHSL, Zero); SDValue Lo(Carry.getNode(), 1); - + SDValue Ignored = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32), LHSH, RHSH, Carry); SDValue Hi(Ignored.getNode(), 1); @@ -761,8 +760,8 @@ Node->getOperand(1), MachinePointerInfo(V), false, false, 0); // Increment the pointer, VAList, to the next vararg - SDValue Tmp3 = DAG.getNode(ISD::ADD, dl, getPointerTy(), VAList, - DAG.getConstant(VT.getSizeInBits(), + SDValue Tmp3 = DAG.getNode(ISD::ADD, dl, getPointerTy(), VAList, + DAG.getConstant(VT.getSizeInBits(), getPointerTy())); // Store the incremented VAList to the legalized pointer Tmp3 = DAG.getStore(VAList.getValue(1), dl, Tmp3, Node->getOperand(1), @@ -781,20 +780,20 @@ MachineFunction &MF = DAG.getMachineFunction(); XCoreFunctionInfo *XFI = MF.getInfo(); SDValue Addr = DAG.getFrameIndex(XFI->getVarArgsFrameIndex(), MVT::i32); - return DAG.getStore(Op.getOperand(0), dl, Addr, Op.getOperand(1), + return DAG.getStore(Op.getOperand(0), dl, Addr, Op.getOperand(1), MachinePointerInfo(), false, false, 0); } SDValue XCoreTargetLowering::LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const { DebugLoc dl = Op.getDebugLoc(); - // Depths > 0 not supported yet! + // Depths > 0 not supported yet! if (cast(Op.getOperand(0))->getZExtValue() > 0) return SDValue(); - + MachineFunction &MF = DAG.getMachineFunction(); const TargetRegisterInfo *RegInfo = getTargetMachine().getRegisterInfo(); - return DAG.getCopyFromReg(DAG.getEntryNode(), dl, + return DAG.getCopyFromReg(DAG.getEntryNode(), dl, RegInfo->getFrameRegister(MF), MVT::i32); } @@ -919,7 +918,7 @@ // Get a count of how many bytes are to be pushed on the stack. unsigned NumBytes = CCInfo.getNextStackOffset(); - Chain = DAG.getCALLSEQ_START(Chain,DAG.getConstant(NumBytes, + Chain = DAG.getCALLSEQ_START(Chain,DAG.getConstant(NumBytes, getPointerTy(), true)); SmallVector, 4> RegsToPass; @@ -944,8 +943,8 @@ Arg = DAG.getNode(ISD::ANY_EXTEND, dl, VA.getLocVT(), Arg); break; } - - // Arguments that can be passed on register must be kept at + + // Arguments that can be passed on register must be kept at // RegsToPass vector if (VA.isRegLoc()) { RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg)); @@ -954,7 +953,7 @@ int Offset = VA.getLocMemOffset(); - MemOpChains.push_back(DAG.getNode(XCoreISD::STWSP, dl, MVT::Other, + MemOpChains.push_back(DAG.getNode(XCoreISD::STWSP, dl, MVT::Other, Chain, Arg, DAG.getConstant(Offset/4, MVT::i32))); } @@ -963,16 +962,16 @@ // Transform all store nodes into one single node because // all store nodes are independent of each other. if (!MemOpChains.empty()) - Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, + Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, &MemOpChains[0], MemOpChains.size()); - // Build a sequence of copy-to-reg nodes chained together with token + // Build a sequence of copy-to-reg nodes chained together with token // chain and flag operands which copy the outgoing args into registers. // The InFlag in necessary since all emited instructions must be // stuck together. SDValue InFlag; for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) { - Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first, + Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first, RegsToPass[i].second, InFlag); InFlag = Chain.getValue(1); } @@ -986,7 +985,7 @@ Callee = DAG.getTargetExternalSymbol(E->getSymbol(), MVT::i32); // XCoreBranchLink = #chain, #target_address, #opt_in_flags... - // = Chain, Callee, Reg#1, Reg#2, ... + // = Chain, Callee, Reg#1, Reg#2, ... // // Returns a chain & a flag for retval copy to use. SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue); @@ -994,7 +993,7 @@ Ops.push_back(Chain); Ops.push_back(Callee); - // Add argument registers to the end of the list so that they are + // Add argument registers to the end of the list so that they are // known live into the call. for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) Ops.push_back(DAG.getRegister(RegsToPass[i].first, @@ -1098,11 +1097,11 @@ unsigned StackSlotSize = XCoreFrameLowering::stackSlotSize(); unsigned LRSaveSize = StackSlotSize; - + for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) { CCValAssign &VA = ArgLocs[i]; - + if (VA.isRegLoc()) { // Arguments passed in registers EVT RegVT = VA.getLocVT(); @@ -1139,12 +1138,12 @@ // Create the SelectionDAG nodes corresponding to a load //from this parameter SDValue FIN = DAG.getFrameIndex(FI, MVT::i32); - InVals.push_back(DAG.getLoad(VA.getLocVT(), dl, Chain, FIN, + InVals.push_back(DAG.getLoad(VA.getLocVT(), dl, Chain, FIN, MachinePointerInfo::getFixedStack(FI), false, false, 0)); } } - + if (isVarArg) { /* Argument registers */ static const unsigned ArgRegs[] = { @@ -1186,7 +1185,7 @@ true)); } } - + return Chain; } @@ -1222,7 +1221,7 @@ // Analize return values. CCInfo.AnalyzeReturn(Outs, RetCC_XCore); - // If this is the first return lowered for this function, add + // If this is the first return lowered for this function, add // the regs to the liveout set for the function. if (DAG.getMachineFunction().getRegInfo().liveout_empty()) { for (unsigned i = 0; i != RVLocs.size(); ++i) @@ -1237,7 +1236,7 @@ CCValAssign &VA = RVLocs[i]; assert(VA.isRegLoc() && "Can only return in registers!"); - Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), + Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), OutVals[i], Flag); // guarantee that all emitted copies are @@ -1265,7 +1264,7 @@ DebugLoc dl = MI->getDebugLoc(); assert((MI->getOpcode() == XCore::SELECT_CC) && "Unexpected instr type to insert"); - + // To "insert" a SELECT_CC instruction, we actually have to insert the diamond // control-flow pattern. The incoming instruction knows the destination vreg // to set, the condition code register to branch on, the true/false values to @@ -1273,7 +1272,7 @@ const BasicBlock *LLVM_BB = BB->getBasicBlock(); MachineFunction::iterator It = BB; ++It; - + // thisMBB: // ... // TrueVal = ... @@ -1296,7 +1295,7 @@ // Next, add the true and fallthrough blocks as its successors. BB->addSuccessor(copy0MBB); BB->addSuccessor(sinkMBB); - + BuildMI(BB, dl, TII.get(XCore::BRFT_lru6)) .addReg(MI->getOperand(1).getReg()).addMBB(sinkMBB); @@ -1304,10 +1303,10 @@ // %FalseValue = ... // # fallthrough to sinkMBB BB = copy0MBB; - + // Update machine-CFG edges BB->addSuccessor(sinkMBB); - + // sinkMBB: // %Result = phi [ %FalseValue, copy0MBB ], [ %TrueValue, thisMBB ] // ... @@ -1316,7 +1315,7 @@ TII.get(XCore::PHI), MI->getOperand(0).getReg()) .addReg(MI->getOperand(3).getReg()).addMBB(copy0MBB) .addReg(MI->getOperand(2).getReg()).addMBB(thisMBB); - + MI->eraseFromParent(); // The pseudo instruction is gone now. return BB; } @@ -1354,7 +1353,7 @@ // fold (ladd x, 0, y) -> 0, add x, y iff carry is unused and y has only the // low bit set - if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { + if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { APInt KnownZero, KnownOne; APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(), VT.getSizeInBits() - 1); @@ -1377,7 +1376,7 @@ EVT VT = N0.getValueType(); // fold (lsub 0, 0, x) -> x, -x iff x has only the low bit set - if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) { + if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) { APInt KnownZero, KnownOne; APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(), VT.getSizeInBits() - 1); @@ -1393,7 +1392,7 @@ // fold (lsub x, 0, y) -> 0, sub x, y iff borrow is unused and y has only the // low bit set - if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { + if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { APInt KnownZero, KnownOne; APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(), VT.getSizeInBits() - 1); @@ -1557,7 +1556,7 @@ /// isLegalAddressingMode - Return true if the addressing mode represented /// by AM is legal for this target, for a load/store of the specified type. bool -XCoreTargetLowering::isLegalAddressingMode(const AddrMode &AM, +XCoreTargetLowering::isLegalAddressingMode(const AddrMode &AM, const Type *Ty) const { if (Ty->getTypeID() == Type::VoidTyID) return AM.Scale == 0 && isImmUs(AM.BaseOffs) && isImmUs4(AM.BaseOffs); @@ -1568,7 +1567,7 @@ return Size >= 4 && !AM.HasBaseReg && AM.Scale == 0 && AM.BaseOffs%4 == 0; } - + switch (Size) { case 1: // reg + imm @@ -1593,7 +1592,7 @@ // reg + reg<<2 return AM.Scale == 4 && AM.BaseOffs == 0; } - + return false; } @@ -1603,7 +1602,7 @@ std::vector XCoreTargetLowering:: getRegClassForInlineAsmConstraint(const std::string &Constraint, - EVT VT) const + EVT VT) const { if (Constraint.size() != 1) return std::vector(); @@ -1611,9 +1610,9 @@ switch (Constraint[0]) { default : break; case 'r': - return make_vector(XCore::R0, XCore::R1, XCore::R2, - XCore::R3, XCore::R4, XCore::R5, - XCore::R6, XCore::R7, XCore::R8, + return make_vector(XCore::R0, XCore::R1, XCore::R2, + XCore::R3, XCore::R4, XCore::R5, + XCore::R6, XCore::R7, XCore::R8, XCore::R9, XCore::R10, XCore::R11, 0); break; } Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=126518&r1=126517&r2=126518&view=diff ============================================================================== --- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original) +++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Fri Feb 25 15:41:48 2011 @@ -20,11 +20,11 @@ #include "XCore.h" namespace llvm { - + // Forward delcarations class XCoreSubtarget; class XCoreTargetMachine; - + namespace XCoreISD { enum NodeType { // Start the numbering where the builtin ops and target ops leave off. @@ -38,16 +38,16 @@ // dp relative address DPRelativeWrapper, - + // cp relative address CPRelativeWrapper, - + // Store word to stack STWSP, // Corresponds to retsp instruction RETSP, - + // Corresponds to LADD instruction LADD, @@ -74,13 +74,14 @@ //===--------------------------------------------------------------------===// // TargetLowering Implementation //===--------------------------------------------------------------------===// - class XCoreTargetLowering : public TargetLowering + class XCoreTargetLowering : public TargetLowering { public: explicit XCoreTargetLowering(XCoreTargetMachine &TM); virtual unsigned getJumpTableEncoding() const; + virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i32; } /// LowerOperation - Provide custom lowering hooks for some operations. virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG) const; @@ -91,10 +92,10 @@ virtual void ReplaceNodeResults(SDNode *N, SmallVectorImpl&Results, SelectionDAG &DAG) const; - /// getTargetNodeName - This method returns the name of a target specific + /// getTargetNodeName - This method returns the name of a target specific // DAG node. virtual const char *getTargetNodeName(unsigned Opcode) const; - + virtual MachineBasicBlock * EmitInstrWithCustomInserter(MachineInstr *MI, MachineBasicBlock *MBB) const; @@ -108,7 +109,7 @@ private: const XCoreTargetMachine &TM; const XCoreSubtarget &Subtarget; - + // Lower Operand helpers SDValue LowerCCCArguments(SDValue Chain, CallingConv::ID CallConv, @@ -148,12 +149,12 @@ SDValue LowerSMUL_LOHI(SDValue Op, SelectionDAG &DAG) const; SDValue LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const; SDValue LowerTRAMPOLINE(SDValue Op, SelectionDAG &DAG) const; - + // Inline asm support std::vector getRegClassForInlineAsmConstraint(const std::string &Constraint, EVT VT) const; - + // Expand specifics SDValue TryExpandADDWithMul(SDNode *Op, SelectionDAG &DAG) const; SDValue ExpandADDSUB(SDNode *Op, SelectionDAG &DAG) const; From clattner at apple.com Fri Feb 25 15:49:23 2011 From: clattner at apple.com (Chris Lattner) Date: Fri, 25 Feb 2011 13:49:23 -0800 Subject: [llvm-commits] [llvm] r126518 - in /llvm/trunk: include/llvm/Target/ lib/CodeGen/SelectionDAG/ lib/Target/Alpha/ lib/Target/Blackfin/ lib/Target/CellSPU/ lib/Target/MSP430/ lib/Target/PowerPC/ lib/Target/SystemZ/ lib/Target/X86/ lib/Target/XCore/ In-Reply-To: <20110225214149.7F04A2A6C12C@llvm.org> References: <20110225214149.7F04A2A6C12C@llvm.org> Message-ID: On Feb 25, 2011, at 1:41 PM, Owen Anderson wrote: > Author: resistor > Date: Fri Feb 25 15:41:48 2011 > New Revision: 126518 > > URL: http://llvm.org/viewvc/llvm-project?rev=126518&view=rev > Log: > Allow targets to specify a the type of the RHS of a shift parameterized on the type of the LHS. Ok, but please don't conflate massive "remove trailing whitespace" changes into patches like this. -Chris From dmalyshev at accesssoftek.com Fri Feb 25 16:14:38 2011 From: dmalyshev at accesssoftek.com (Danil Malyshev) Date: Fri, 25 Feb 2011 14:14:38 -0800 Subject: [llvm-commits] Console input should be opened in binary mode on Windows (Python setting) Message-ID: <6AE1604EE3EC5F4296C096518C6B77EE17E1229176@mail.accesssoftek.com> Hello All, Some tests on Windows fail because these tests use python scripts for analyze binary output from console, but python open it in the text mode. Attached patch fix it. The patch fixes the following tests: LLVM :: CodeGen/ARM/2010-12-13-reloc-pic.ll LLVM :: MC/ARM/elf-reloc-01.ll LLVM :: MC/COFF/symbol-alias.s LLVM :: MC/ELF/align.s LLVM :: MC/ELF/cfi.s LLVM :: MC/ELF/relocation.s LLVM :: MC/ELF/tls-i386.s LLVM :: MC/ELF/weakref.s Tested in MinGW and Ubuntu. Thank you, Danil -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/1d367d09/attachment.html -------------- next part -------------- A non-text attachment was scrubbed... Name: python-binary-console-01.patch Type: application/octet-stream Size: 503 bytes Desc: python-binary-console-01.patch Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/1d367d09/attachment.obj From dmalyshev at accesssoftek.com Fri Feb 25 16:17:02 2011 From: dmalyshev at accesssoftek.com (Danil Malyshev) Date: Fri, 25 Feb 2011 14:17:02 -0800 Subject: [llvm-commits] ExecutionEngine regress tests should be skipped for cross LLVM build when host CPU different then target CPU Message-ID: <6AE1604EE3EC5F4296C096518C6B77EE17E1229178@mail.accesssoftek.com> Hello All, These tests use LLI interpreter, but this interpreter cannot normally work if LLVM target CPU different current system CPU. So, these tests should be skipped. Please, review the attached patch with this feature. Unit tests ExecutionEngine should be skipped too, I will do it in the next patch. Thank you, Danil -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/fdf3d384/attachment.html -------------- next part -------------- A non-text attachment was scrubbed... Name: skip-tests-01.patch Type: application/octet-stream Size: 2436 bytes Desc: skip-tests-01.patch Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110225/fdf3d384/attachment.obj From stuart at apple.com Fri Feb 25 16:47:58 2011 From: stuart at apple.com (Stuart Hastings) Date: Fri, 25 Feb 2011 22:47:58 -0000 Subject: [llvm-commits] [llvm] r126525 - /llvm/trunk/utils/buildit/build_llvm Message-ID: <20110225224758.7A5D32A6C12C@llvm.org> Author: stuart Date: Fri Feb 25 16:47:58 2011 New Revision: 126525 URL: http://llvm.org/viewvc/llvm-project?rev=126525&view=rev Log: Fix bad comment marker. Modified: llvm/trunk/utils/buildit/build_llvm Modified: llvm/trunk/utils/buildit/build_llvm URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/buildit/build_llvm?rev=126525&r1=126524&r2=126525&view=diff ============================================================================== --- llvm/trunk/utils/buildit/build_llvm (original) +++ llvm/trunk/utils/buildit/build_llvm Fri Feb 25 16:47:58 2011 @@ -341,7 +341,7 @@ fi rm -f lib/libLTO.a lib/libLTO.la -// Omit lto.h from the result. Clang will supply. +# Omit lto.h from the result. Clang will supply. find $DEST_DIR$DEST_ROOT -name lto.h -delete ################################################################################ From grosbach at apple.com Fri Feb 25 16:53:20 2011 From: grosbach at apple.com (Jim Grosbach) Date: Fri, 25 Feb 2011 22:53:20 -0000 Subject: [llvm-commits] [llvm] r126526 - /llvm/trunk/lib/CodeGen/LowerSubregs.cpp Message-ID: <20110225225320.F3DC02A6C12C@llvm.org> Author: grosbach Date: Fri Feb 25 16:53:20 2011 New Revision: 126526 URL: http://llvm.org/viewvc/llvm-project?rev=126526&view=rev Log: Trailing whitespace. Modified: llvm/trunk/lib/CodeGen/LowerSubregs.cpp Modified: llvm/trunk/lib/CodeGen/LowerSubregs.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LowerSubregs.cpp?rev=126526&r1=126525&r2=126526&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/LowerSubregs.cpp (original) +++ llvm/trunk/lib/CodeGen/LowerSubregs.cpp Fri Feb 25 16:53:20 2011 @@ -37,7 +37,7 @@ public: static char ID; // Pass identification, replacement for typeid LowerSubregsInstructionPass() : MachineFunctionPass(ID) {} - + const char *getPassName() const { return "Subregister lowering instruction pass"; } @@ -64,8 +64,8 @@ char LowerSubregsInstructionPass::ID = 0; } -FunctionPass *llvm::createLowerSubregsPass() { - return new LowerSubregsInstructionPass(); +FunctionPass *llvm::createLowerSubregsPass() { + return new LowerSubregsInstructionPass(); } /// TransferDeadFlag - MI is a pseudo-instruction with DstReg dead, @@ -192,9 +192,9 @@ /// copies. /// bool LowerSubregsInstructionPass::runOnMachineFunction(MachineFunction &MF) { - DEBUG(dbgs() << "Machine Function\n" + DEBUG(dbgs() << "Machine Function\n" << "********** LOWERING SUBREG INSTRS **********\n" - << "********** Function: " + << "********** Function: " << MF.getFunction()->getName() << '\n'); TRI = MF.getTarget().getRegisterInfo(); TII = MF.getTarget().getInstrInfo(); From jasonwkim at google.com Fri Feb 25 17:40:42 2011 From: jasonwkim at google.com (Jason Kim) Date: Fri, 25 Feb 2011 15:40:42 -0800 Subject: [llvm-commits] Code refactoring of MCELFStreamer.cpp and ELFObjectWriter.cpp In-Reply-To: <470208.84552.qm@web55602.mail.re4.yahoo.com> References: <960821.36942.qm@web55604.mail.re4.yahoo.com> <4D680702.2040101@gmail.com> <470208.84552.qm@web55602.mail.re4.yahoo.com> Message-ID: On Fri, Feb 25, 2011 at 12:05 PM, Jan Sjodin wrote: > Just to keep the code more object-oriented. If a sub-class (I intend to inherit > from these classes) needs the functions, they will not be available if they are > hidden as static functions. Hi Jan, Just out of curiosity, what do you plan to do with MCELF inheritance wise? Thanks -jason > > - Jan > > > ----- Original Message ---- >> From: Rafael Avila de Espindola >> To: llvm-commits at cs.uiuc.edu >> Sent: Fri, February 25, 2011 2:46:10 PM >> Subject: Re: [llvm-commits] Code refactoring of MCELFStreamer.cpp and >>ELFObjectWriter.cpp >> >> On 11-02-25 12:02 PM, Jan Sjodin wrote: >> > Removed duplicated code by ?putting shared functions in MCELF.h/cpp and made >>the >> > rest of the C-style ?static functions static methods in the corresponding >> > ?classes. >> >> Moving the duplicated functions to a new file is OK, but why ?convert the >> function only used in the ELFObjectWriter to static ?methods? >> >> > - ?Jan >> >> Cheers, >> Rafael >> _______________________________________________ >> llvm-commits ?mailing list >> llvm-commits at cs.uiuc.edu >> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits >> > _______________________________________________ > llvm-commits mailing list > llvm-commits at cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits > From aggarwa4 at illinois.edu Fri Feb 25 17:37:46 2011 From: aggarwa4 at illinois.edu (Arushi Aggarwal) Date: Fri, 25 Feb 2011 23:37:46 -0000 Subject: [llvm-commits] [poolalloc] r126532 - /poolalloc/trunk/lib/DSA/AddressTakenAnalysis.cpp Message-ID: <20110225233746.720A32A6C12C@llvm.org> Author: aggarwa4 Date: Fri Feb 25 17:37:46 2011 New Revision: 126532 URL: http://llvm.org/viewvc/llvm-project?rev=126532&view=rev Log: call (bitcast F to ...) () should not cause F to be considered address taken, as that value is never really stored, and used elsewhere. Modified: poolalloc/trunk/lib/DSA/AddressTakenAnalysis.cpp Modified: poolalloc/trunk/lib/DSA/AddressTakenAnalysis.cpp URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/AddressTakenAnalysis.cpp?rev=126532&r1=126531&r2=126532&view=diff ============================================================================== --- poolalloc/trunk/lib/DSA/AddressTakenAnalysis.cpp (original) +++ poolalloc/trunk/lib/DSA/AddressTakenAnalysis.cpp Fri Feb 25 17:37:46 2011 @@ -48,8 +48,14 @@ if(isAddressTaken(U)) return true; } else { - // FIXME handle bitcasts to see if the resultant value - // is ever used in an address taken fashion + if (Constant *C = dyn_cast(U)) { + if (ConstantExpr *CE = dyn_cast(C)) { + if (CE->getOpcode() == Instruction::BitCast) { + return isAddressTaken(CE); + } + } + } + return true; } From dimitry at andric.com Fri Feb 25 17:41:33 2011 From: dimitry at andric.com (Dimitry Andric) Date: Sat, 26 Feb 2011 00:41:33 +0100 Subject: [llvm-commits] [patch] Fix getDriver().getInstalledDir() comparisons In-Reply-To: <4D680F41.1070008@andric.com> References: <4D680F41.1070008@andric.com> Message-ID: <4D683E2D.9050800@andric.com> Apologies for posting these patches here, please disregard them, they're really about clang. :) Copied to cfe-commits instead... From bob.wilson at apple.com Fri Feb 25 17:42:03 2011 From: bob.wilson at apple.com (Bob Wilson) Date: Fri, 25 Feb 2011 23:42:03 -0000 Subject: [llvm-commits] [llvm] r126533 - /llvm/trunk/utils/buildit/GNUmakefile Message-ID: <20110225234203.F04B72A6C12C@llvm.org> Author: bwilson Date: Fri Feb 25 17:42:03 2011 New Revision: 126533 URL: http://llvm.org/viewvc/llvm-project?rev=126533&view=rev Log: Add a new "Embedded" makefile target for Apple-style builds. This one just installs the default build into a different destination directory. Modified: llvm/trunk/utils/buildit/GNUmakefile Modified: llvm/trunk/utils/buildit/GNUmakefile URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/buildit/GNUmakefile?rev=126533&r1=126532&r2=126533&view=diff ============================================================================== --- llvm/trunk/utils/buildit/GNUmakefile (original) +++ llvm/trunk/utils/buildit/GNUmakefile Fri Feb 25 17:42:03 2011 @@ -80,6 +80,10 @@ export MACOSX_DEPLOYMENT_TARGET=`sw_vers -productVersion`; \ $(MAKE) IOS_SIM_BUILD=yes PREFIX=$(SDKROOT)/usr/local install +Embedded: + ARM_PLATFORM=`xcodebuild -version -sdk iphoneos PlatformPath` && \ + $(MAKE) DSTROOT=$(DSTROOT)$$ARM_PLATFORM install + # installhdrs does nothing, because the headers aren't useful until # the compiler is installed. installhdrs: @@ -128,4 +132,4 @@ $(OBJROOT) $(SYMROOT) $(DSTROOT): mkdir -p $@ -.PHONY: install installsrc clean EmbeddedHosted EmbeddedSim +.PHONY: install installsrc clean EmbeddedHosted EmbeddedSim Embedded From rafael.espindola at gmail.com Fri Feb 25 18:15:27 2011 From: rafael.espindola at gmail.com (=?ISO-8859-1?Q?Rafael_=C1vila_de_Esp=EDndola?=) Date: Fri, 25 Feb 2011 19:15:27 -0500 Subject: [llvm-commits] Code refactoring of MCELFStreamer.cpp and ELFObjectWriter.cpp In-Reply-To: <470208.84552.qm@web55602.mail.re4.yahoo.com> References: <960821.36942.qm@web55604.mail.re4.yahoo.com> <4D680702.2040101@gmail.com> <470208.84552.qm@web55602.mail.re4.yahoo.com> Message-ID: <4D68461F.2090807@gmail.com> On 2011-02-25 15:05, Jan Sjodin wrote: > Just to keep the code more object-oriented. If a sub-class (I intend to inherit > from these classes) needs the functions, they will not be available if they are > hidden as static functions. So you intent to move the ELFObjectWriter class declaration to a header? I have the same question as Jason, what do you want to do with it? :-) > - Jan Cheers, Rafael From bob.wilson at apple.com Fri Feb 25 18:22:18 2011 From: bob.wilson at apple.com (Bob Wilson) Date: Sat, 26 Feb 2011 00:22:18 -0000 Subject: [llvm-commits] [llvm] r126534 - /llvm/trunk/utils/buildit/build_llvm Message-ID: <20110226002218.175762A6C12C@llvm.org> Author: bwilson Date: Fri Feb 25 18:22:17 2011 New Revision: 126534 URL: http://llvm.org/viewvc/llvm-project?rev=126534&view=rev Log: Removed unnecessary dylibs from Apple builds, with or without "lib" prefix. Radar 9056686 Modified: llvm/trunk/utils/buildit/build_llvm Modified: llvm/trunk/utils/buildit/build_llvm URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/buildit/build_llvm?rev=126534&r1=126533&r2=126534&view=diff ============================================================================== --- llvm/trunk/utils/buildit/build_llvm (original) +++ llvm/trunk/utils/buildit/build_llvm Fri Feb 25 18:22:17 2011 @@ -267,8 +267,9 @@ # The Hello dylib is an example of how to build a pass. # The BugpointPasses module is only used to test bugpoint. # These unversioned dylibs cause verification failures, so do not install them. -rm $DEST_DIR$DEST_ROOT/lib/libLLVMHello.dylib -rm $DEST_DIR$DEST_ROOT/lib/libBugpointPasses.dylib +# (The wildcards are used to match a "lib" prefix if it is present.) +rm $DEST_DIR$DEST_ROOT/lib/*LLVMHello.dylib +rm $DEST_DIR$DEST_ROOT/lib/*BugpointPasses.dylib # Compress manpages MDIR=$DEST_DIR$DEST_ROOT/share/man/man1 From jan_sjodin at yahoo.com Fri Feb 25 19:54:30 2011 From: jan_sjodin at yahoo.com (Jan Sjodin) Date: Fri, 25 Feb 2011 17:54:30 -0800 (PST) Subject: [llvm-commits] Code refactoring of MCELFStreamer.cpp and ELFObjectWriter.cpp In-Reply-To: <4D68461F.2090807@gmail.com> References: <960821.36942.qm@web55604.mail.re4.yahoo.com> <4D680702.2040101@gmail.com> <470208.84552.qm@web55602.mail.re4.yahoo.com> <4D68461F.2090807@gmail.com> Message-ID: <267631.76534.qm@web55603.mail.re4.yahoo.com> The JIT implementation I am working on (MCJIT which was renamed to JunkJIT) became a lot nicer when I decided to inherit MCELFStreamer and ELFObjectWriter. This was needed to make the debug information available to gdb since it expects ELF format. I basically reuse everything, the only thing that differs is that I use a mem_raw_outstream to emit code to memory and make sure that all sections are created before or during the post layout binding so that when RecordRelocation is called all symbol addresses are known. In the ELFObjectWriter, the code in WriteObject that creates sections is moved to the post layout binding, and the code that emits the object stays in place. This is perhaps not very clean, since it is not formalized in the interface, but it works okay and I only need to shuffle around the code a little. You could imagine that if section creation was restricted by the interface to only happenen before/during post layout, then a generic JunkJIT could be written for any object format. Each object writer would be required to compute offsets for the sections. I am currently not aiming to do this since the MCJIT proposal by Daniel separates the Streamer/ObjectWriter from figuring out relocations and external symbols. However, the code refactoring patches I have/will submit are fairly benign and enable a JIT (without lazy compilation) using the MC framework. - Jan ----- Original Message ---- > From: Rafael ?vila de Esp?ndola > To: Jan Sjodin > Cc: llvm-commits at cs.uiuc.edu > Sent: Fri, February 25, 2011 7:15:27 PM > Subject: Re: [llvm-commits] Code refactoring of MCELFStreamer.cpp and >ELFObjectWriter.cpp > > On 2011-02-25 15:05, Jan Sjodin wrote: > > Just to keep the code more object-oriented. If a sub-class (I intend to >inherit > > > from these classes) needs the functions, they will not be available if they >are > > > hidden as static functions. > > So you intent to move the ELFObjectWriter class declaration to a header? > I have the same question as Jason, what do you want to do with it? :-) > > > - Jan > > Cheers, > Rafael > From isanbard at gmail.com Fri Feb 25 21:09:12 2011 From: isanbard at gmail.com (Bill Wendling) Date: Sat, 26 Feb 2011 03:09:12 -0000 Subject: [llvm-commits] [llvm] r126538 - in /llvm/trunk/utils/TableGen: AsmWriterEmitter.cpp AsmWriterEmitter.h Message-ID: <20110226030912.858682A6C12C@llvm.org> Author: void Date: Fri Feb 25 21:09:12 2011 New Revision: 126538 URL: http://llvm.org/viewvc/llvm-project?rev=126538&view=rev Log: A new TableGen feature! (Not turned on just yet.) InstAlias<{alias}, {aliasee}>; The InstAlias instruction should be able to go from the MCInst to the {alias}. All of the information is there to match the MCInst with the {aliasee}. From there, it's a simple matter to emit the {alias}, with the correct operands from the {aliasee}. The code this patch generates can be used by the InstPrinter to automatically print out the alias without having to write special C++ code to handle the situation. This is a WIP, and therefore are several limitations. For instance, it cannot handle AsmOperands at the moment. It also doesn't know what to do when two {alias}es match the same {aliasee}. (Currently, it just ignores those two cases and allows the printInstruction method to handle them.) Modified: llvm/trunk/utils/TableGen/AsmWriterEmitter.cpp llvm/trunk/utils/TableGen/AsmWriterEmitter.h Modified: llvm/trunk/utils/TableGen/AsmWriterEmitter.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/AsmWriterEmitter.cpp?rev=126538&r1=126537&r2=126538&view=diff ============================================================================== --- llvm/trunk/utils/TableGen/AsmWriterEmitter.cpp (original) +++ llvm/trunk/utils/TableGen/AsmWriterEmitter.cpp Fri Feb 25 21:09:12 2011 @@ -542,7 +542,255 @@ << "}\n\n#endif\n"; } +void AsmWriterEmitter::EmitPrintAliasInstruction(raw_ostream &O) { + CodeGenTarget Target(Records); + Record *AsmWriter = Target.getAsmWriter(); + O << "\n#ifdef PRINT_ALIAS_INSTR\n"; + O << "#undef PRINT_ALIAS_INSTR\n\n"; + + // Enumerate the register classes. + const std::vector &RegisterClasses = + Target.getRegisterClasses(); + + O << "namespace { // Register classes\n"; + O << " enum RegClass {\n"; + + // Emit the register enum value for each RegisterClass. + for (unsigned I = 0, E = RegisterClasses.size(); I != E; ++I) { + if (I != 0) O << ",\n"; + O << " RC_" << RegisterClasses[I].TheDef->getName(); + } + + O << "\n };\n"; + O << "} // end anonymous namespace\n\n"; + + // Emit a function that returns 'true' if a regsiter is part of a particular + // register class. I.e., RAX is part of GR64 on X86. + O << "static bool regIsInRegisterClass" + << "(unsigned RegClass, unsigned Reg) {\n"; + + // Emit the switch that checks if a register belongs to a particular register + // class. + O << " switch (RegClass) {\n"; + O << " default: break;\n"; + + for (unsigned I = 0, E = RegisterClasses.size(); I != E; ++I) { + const CodeGenRegisterClass &RC = RegisterClasses[I]; + + // Give the register class a legal C name if it's anonymous. + std::string Name = RC.TheDef->getName(); + O << " case RC_" << Name << ":\n"; + + // Emit the register list now. + unsigned IE = RC.Elements.size(); + if (IE == 1) { + O << " if (Reg == " << getQualifiedName(RC.Elements[0]) << ")\n"; + O << " return true;\n"; + } else { + O << " switch (Reg) {\n"; + O << " default: break;\n"; + + for (unsigned II = 0; II != IE; ++II) { + Record *Reg = RC.Elements[II]; + O << " case " << getQualifiedName(Reg) << ":\n"; + } + + O << " return true;\n"; + O << " }\n"; + } + + O << " break;\n"; + } + + O << " }\n\n"; + O << " return false;\n"; + O << "}\n\n"; + + // Emit the method that prints the alias instruction. + std::string ClassName = AsmWriter->getValueAsString("AsmWriterClassName"); + + bool isMC = AsmWriter->getValueAsBit("isMCAsmWriter"); + const char *MachineInstrClassName = isMC ? "MCInst" : "MachineInstr"; + + O << "bool " << Target.getName() << ClassName + << "::printAliasInstr(const " << MachineInstrClassName + << " *MI, raw_ostream &OS) {\n"; + + std::vector AllInstAliases = + Records.getAllDerivedDefinitions("InstAlias"); + + // Create a map from the qualified name to a list of potential matches. + std::map > AliasMap; + for (std::vector::iterator + I = AllInstAliases.begin(), E = AllInstAliases.end(); I != E; ++I) { + CodeGenInstAlias *Alias = new CodeGenInstAlias(*I, Target); + const Record *R = *I; + const DagInit *DI = R->getValueAsDag("ResultInst"); + const DefInit *Op = dynamic_cast(DI->getOperator()); + AliasMap[getQualifiedName(Op->getDef())].push_back(Alias); + } + + if (AliasMap.empty() || !isMC) { + // FIXME: Support MachineInstr InstAliases? + O << " return true;\n"; + O << "}\n\n"; + O << "#endif // PRINT_ALIAS_INSTR\n"; + return; + } + + O << " StringRef AsmString;\n"; + O << " std::map OpMap;\n"; + O << " switch (MI->getOpcode()) {\n"; + O << " default: return true;\n"; + + for (std::map >::iterator + I = AliasMap.begin(), E = AliasMap.end(); I != E; ++I) { + std::vector &Aliases = I->second; + + std::map CondCount; + std::map BodyMap; + + std::string AsmString = ""; + + for (std::vector::iterator + II = Aliases.begin(), IE = Aliases.end(); II != IE; ++II) { + const CodeGenInstAlias *CGA = *II; + AsmString = CGA->AsmString; + unsigned Indent = 8; + unsigned LastOpNo = CGA->ResultInstOperandIndex.size(); + + std::string Cond; + raw_string_ostream CondO(Cond); + + CondO << "if (MI->getNumOperands() == " << LastOpNo; + + std::map OpMap; + bool CantHandle = false; + + for (unsigned i = 0, e = LastOpNo; i != e; ++i) { + const CodeGenInstAlias::ResultOperand &RO = CGA->ResultOperands[i]; + + switch (RO.Kind) { + default: assert(0 && "unexpected InstAlias operand kind"); + case CodeGenInstAlias::ResultOperand::K_Record: { + const Record *Rec = RO.getRecord(); + StringRef ROName = RO.getName(); + + if (Rec->isSubClassOf("RegisterClass")) { + CondO << " &&\n"; + CondO.indent(Indent) << "MI->getOperand(" <ResultOperands[i].getRecord()->getName() + << ", MI->getOperand(" <getOperand(" <getOperand(" + << OpMap[ROName] << ").getReg()"; + } + } else { + assert(Rec->isSubClassOf("Operand") && "Unexpected operand!"); + // FIXME: We need to handle these situations. + CantHandle = true; + break; + } + + break; + } + case CodeGenInstAlias::ResultOperand::K_Imm: + CondO << " &&\n"; + CondO.indent(Indent) << "MI->getOperand(" <ResultOperands[i].getImm(); + break; + case CodeGenInstAlias::ResultOperand::K_Reg: + CondO << " &&\n"; + CondO.indent(Indent) << "MI->getOperand(" <ResultOperands[i].getRegister()->getName(); + break; + } + + if (CantHandle) break; + } + + if (CantHandle) continue; + + CondO << ")"; + + std::string Body; + raw_string_ostream BodyO(Body); + + BodyO << " // " << CGA->Result->getAsString() << "\n"; + BodyO << " AsmString = \"" << AsmString << "\";\n"; + + for (std::map::iterator + III = OpMap.begin(), IIE = OpMap.end(); III != IIE; ++III) + BodyO << " OpMap[\"" << III->first << "\"] = " + << III->second << ";\n"; + + ++CondCount[CondO.str()]; + BodyMap[CondO.str()] = BodyO.str(); + } + + std::string Code; + raw_string_ostream CodeO(Code); + + bool EmitElse = false; + for (std::map::iterator + II = CondCount.begin(), IE = CondCount.end(); II != IE; ++II) { + if (II->second != 1) continue; + CodeO << " "; + if (EmitElse) CodeO << "} else "; + CodeO << II->first << " {\n"; + CodeO << BodyMap[II->first]; + EmitElse = true; + } + + if (CodeO.str().empty()) continue; + + O << " case " << I->first << ":\n"; + O << CodeO.str(); + O << " }\n"; + O << " break;\n"; + } + + O << " }\n\n"; + + // Code that prints the alias, replacing the operands with the ones from the + // MCInst. + O << " if (AsmString.empty()) return true;\n"; + O << " std::pair ASM = AsmString.split(' ');\n"; + O << " OS << '\\t' << ASM.first;\n"; + + O << " if (!ASM.second.empty()) {\n"; + O << " OS << '\\t';\n"; + O << " for (StringRef::iterator\n"; + O << " I = ASM.second.begin(), E = ASM.second.end(); I != E; ) {\n"; + O << " if (*I == '$') {\n"; + O << " StringRef::iterator Start = ++I;\n"; + O << " while (I != E &&\n"; + O << " ((*I >= 'a' && *I <= 'z') ||\n"; + O << " (*I >= 'A' && *I <= 'Z') ||\n"; + O << " (*I >= '0' && *I <= '9') ||\n"; + O << " *I == '_'))\n"; + O << " ++I;\n"; + O << " StringRef Name(Start, I - Start);\n"; + O << " printOperand(MI, OpMap[Name], OS);\n"; + O << " } else {\n"; + O << " OS << *I++;\n"; + O << " }\n"; + O << " }\n"; + O << " }\n\n"; + + O << " return false;\n"; + O << "}\n\n"; + + O << "#endif // PRINT_ALIAS_INSTR\n"; +} void AsmWriterEmitter::run(raw_ostream &O) { EmitSourceFileHeader("Assembly Writer Source Fragment", O); @@ -550,5 +798,6 @@ EmitPrintInstruction(O); EmitGetRegisterName(O); EmitGetInstructionName(O); + EmitPrintAliasInstruction(O); } Modified: llvm/trunk/utils/TableGen/AsmWriterEmitter.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/AsmWriterEmitter.h?rev=126538&r1=126537&r2=126538&view=diff ============================================================================== --- llvm/trunk/utils/TableGen/AsmWriterEmitter.h (original) +++ llvm/trunk/utils/TableGen/AsmWriterEmitter.h Fri Feb 25 21:09:12 2011 @@ -38,6 +38,7 @@ void EmitPrintInstruction(raw_ostream &o); void EmitGetRegisterName(raw_ostream &o); void EmitGetInstructionName(raw_ostream &o); + void EmitPrintAliasInstruction(raw_ostream &O); AsmWriterInst *getAsmWriterInstByID(unsigned ID) const { assert(ID < NumberedInstructions.size()); From zwarich at apple.com Sat Feb 26 02:03:28 2011 From: zwarich at apple.com (Cameron Zwarich) Date: Sat, 26 Feb 2011 00:03:28 -0800 Subject: [llvm-commits] [llvm] r126425 - in /llvm/trunk: include/llvm/MC/ lib/CodeGen/AsmPrinter/ lib/MC/ lib/Target/PTX/ lib/Target/X86/ In-Reply-To: <20110224210400.8E73B2A6C12C@llvm.org> References: <20110224210400.8E73B2A6C12C@llvm.org> Message-ID: <35842674-99C6-4386-8E41-A1B17D7F30DE@apple.com> Unfortunately, this change caused failures on the selfhosting Linux bots, so I rolled it out. Hopefully you can diagnose the problem from the logs and submit a patch that fixes the problem. Cameron From evan.cheng at apple.com Sat Feb 26 02:23:42 2011 From: evan.cheng at apple.com (Evan Cheng) Date: Sat, 26 Feb 2011 00:23:42 -0800 Subject: [llvm-commits] [llvm] r126445 - in /llvm/trunk: lib/Transforms/Utils/SimplifyCFG.cpp test/Transforms/SimplifyCFG/select-gep.ll In-Reply-To: References: <20110224224612.21E8B2A6C12C@llvm.org> <9DA53217-C6FC-42E4-8D03-218F85893512@apple.com>

Message-ID: <9D7E7806-AE6F-45BA-AB25-AFCB675AAFA3@apple.com> On Feb 25, 2011, at 9:48 AM, Benjamin Kramer wrote: > > On 25.02.2011, at 18:25, Evan Cheng wrote: > >> >> On Feb 25, 2011, at 7:26 AM, Benjamin Kramer wrote: >> >>> >>> On 25.02.2011, at 08:14, Evan Cheng wrote: >>> >>>> Speculation is nice, but I would feel a lot better if there are hard numbers to back this up. Are we sure these changes are improving performance? >>> >>> My unscientific test case (gcc compiled with clang -O3 parsing itself) showed a 0.25% speedup with the original patch (all-constant GEPs only, median of 5 runs, i386, penryn). I just retested with the change Frits suggested and it slowed down ~0.6% compared to the original run so I reverted it. >> >> Thanks. What tests are you using to evaluate this? > > I routinely use the nice C files from http://people.csail.mit.edu/smcc/projects/single-file-programs/ > which are easy to handle and large enough to get meaningful results for both compile time and run > time. They're extracted from programs that are also part of SPEC CINT, so I'm confident the > improvements will also show up there :) These are nice, but not quite enough. In this future, please at least run the llvm testsuite for anything that might affect performance. Thanks, Evan From anton at korobeynikov.info Sat Feb 26 06:36:00 2011 From: anton at korobeynikov.info (Anton Korobeynikov) Date: Sat, 26 Feb 2011 15:36:00 +0300 Subject: [llvm-commits] Change DwarfUsesAbsoluteLabelForStmtList to false for X86ELFMCAsmInfo. In-Reply-To: <133510.39411.qm@web55604.mail.re4.yahoo.com> References: <133510.39411.qm@web55604.mail.re4.yahoo.com> Message-ID: Hello Jan, > Dwarfdump gives an error and gdb fails when generating code in memory (but > happens to accept it in a .o file) in Linux when > DwarfUsesAbsoluteLabelForStmtList = true, this patch sets it to false to fix > this issue. I don't know if this is true for other targets, but the change is > limited to X86ELFMCAsmInfo. How the patch was tested? How we might be sure this won't break non JIT case? -- With best regards, Anton Korobeynikov Faculty of Mathematics and Mechanics, Saint Petersburg State University From baldrick at free.fr Sat Feb 26 07:39:45 2011 From: baldrick at free.fr (Duncan Sands) Date: Sat, 26 Feb 2011 13:39:45 -0000 Subject: [llvm-commits] [dragonegg] r126545 - /dragonegg/trunk/llvm-convert.cpp Message-ID: <20110226133945.145D12A6C12C@llvm.org> Author: baldrick Date: Sat Feb 26 07:39:44 2011 New Revision: 126545 URL: http://llvm.org/viewvc/llvm-project?rev=126545&view=rev Log: If an error occurred then be cool and don't require every SSA name to have a proper definition. This prevents the compiler from crashing if we bailed out on an error condition without taking care to define some SSA name. Modified: dragonegg/trunk/llvm-convert.cpp Modified: dragonegg/trunk/llvm-convert.cpp URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-convert.cpp?rev=126545&r1=126544&r2=126545&view=diff ============================================================================== --- dragonegg/trunk/llvm-convert.cpp (original) +++ dragonegg/trunk/llvm-convert.cpp Sat Feb 26 07:39:44 2011 @@ -1008,15 +1008,33 @@ TheDebugInfo->EmitFunctionEnd(true); } -#ifndef NDEBUG - if (!errorcount && !sorrycount) +#ifdef NDEBUG + // When processing broken code it can be awkward to ensure that every SSA name + // that was used has a definition. So in this case we play it cool and create + // an artificial definition for such SSA names. The choice of definition does + // not matter because the compiler is going to exit with an error anyway. + if (errorcount || sorrycount) +#else + // When checks are enabled, complain if an SSA name was used but not defined. +#endif for (DenseMap >::const_iterator I = SSANames.begin(), - E = SSANames.end(); I != E; ++I) - if (isSSAPlaceholder(I->second)) { + E = SSANames.end(); I != E; ++I) { + Value *NameDef = I->second; + // If this is not a placeholder then the SSA name was defined. + if (!isSSAPlaceholder(NameDef)) + continue; + + // If an error occurred then replace the placeholder with undef. Thanks + // to this we can just bail out on errors, without having to worry about + // whether we defined every SSA name. + if (errorcount || sorrycount) { + NameDef->replaceAllUsesWith(UndefValue::get(NameDef->getType())); + delete NameDef; + } else { debug_tree(I->first); llvm_unreachable("SSA name never defined!"); } -#endif + } return Fn; } From jan_sjodin at yahoo.com Sat Feb 26 08:40:31 2011 From: jan_sjodin at yahoo.com (Jan Sjodin) Date: Sat, 26 Feb 2011 06:40:31 -0800 (PST) Subject: [llvm-commits] Change DwarfUsesAbsoluteLabelForStmtList to false for X86ELFMCAsmInfo. In-Reply-To: References: <133510.39411.qm@web55604.mail.re4.yahoo.com> Message-ID: <512414.14506.qm@web55606.mail.re4.yahoo.com> ----- Original Message ---- > From: Anton Korobeynikov > To: Jan Sjodin > Cc: llvm-commits at cs.uiuc.edu > Sent: Sat, February 26, 2011 7:36:00 AM > Subject: Re: [llvm-commits] Change DwarfUsesAbsoluteLabelForStmtList to false >for X86ELFMCAsmInfo. > > Hello Jan, > > > Dwarfdump gives an error and gdb fails when generating code in memory (but > > happens to accept it in a .o file) in Linux when > > DwarfUsesAbsoluteLabelForStmtList = true, this patch sets it to false to fix > > this issue. I don't know if this is true for other targets, but the change >is > > limited to X86ELFMCAsmInfo. > How the patch was tested? How we might be sure this won't break non JIT case? > > -- > With best regards, Anton Korobeynikov > Faculty of Mathematics and Mechanics, Saint Petersburg State University > The question you are really asking is if there is a bug in dwarfdump or a bug in llvm, because dwarfdump always complains. I tried this on Linux 64-bit with clang and llvm trunk: file empty.c: --------------------------------------------------------------- int main() { return 42; } --------------------------------------------------------------- Compile and run dwarfdump: --------------------------------------------------------------- > clang -c -g -emit-llvm -o empty.bc empty.c > llc -filetype=obj empty.bc > dwarfdump empty.o .debug_info COMPILE_UNIT

: <0>< 11> DW_TAG_compile_unit DW_AT_producer clang version 2.9 (trunk 126545) DW_AT_language DW_LANG_C99 DW_AT_name empty.c DW_AT_entry_pc 0x0 DW_AT_stmt_list 0x0 DW_AT_comp_dir /home/jsjodin/Work/LLVM/Sandboxes/TOT/test LOCAL_SYMBOLS: <1>< 114> DW_TAG_base_type DW_AT_encoding DW_ATE_signed DW_AT_name int DW_AT_byte_size 4 <1>< 121> DW_TAG_subprogram DW_AT_name main DW_AT_decl_file 1 DW_AT_decl_line 2 DW_AT_type <114> DW_AT_external yes(1) DW_AT_low_pc 0x0 DW_AT_high_pc 0xe DW_AT_frame_base DW_OP_reg7 AT of 16359 (0x3fe7) is unknown to dwarfdump. Continuing. yes(1) .debug_line: line number info for a single cu dwarfdump ERROR: dwarf_srclines: DW_DLE_ATTR_FORM_BAD (114) --------------------------------------------------------------- dwarfdump cleary doesn't like what llc has produced. With the patch applied we try again: --------------------------------------------------------------- .debug_info COMPILE_UNIT

: <0>< 11> DW_TAG_compile_unit DW_AT_producer clang version 2.9 (trunk 126545) DW_AT_language DW_LANG_C99 DW_AT_name empty.c DW_AT_entry_pc 0x0 DW_AT_stmt_list 0 DW_AT_comp_dir /home/jsjodin/Work/LLVM/Sandboxes/TOT/test LOCAL_SYMBOLS: <1>< 110> DW_TAG_base_type DW_AT_encoding DW_ATE_signed DW_AT_name int DW_AT_byte_size 4 <1>< 117> DW_TAG_subprogram DW_AT_name main DW_AT_decl_file 1 /home/jsjodin/Work/LLVM/Sandboxes/TOT/test/empty.c DW_AT_decl_line 2 DW_AT_type <110> DW_AT_external yes(1) DW_AT_low_pc 0x0 DW_AT_high_pc 0xe DW_AT_frame_base DW_OP_reg7 AT of 16359 (0x3fe7) is unknown to dwarfdump. Continuing. yes(1) .debug_line: line number info for a single cu Source lines (from CU-DIE at .debug_info offset 11): [row,column] // means index section-offset begin-addr end-addr length-of-block-entry .debug_abbrev < 1>< 0> DW_TAG_compile_unit DW_children_yes < 3> DW_AT_producer DW_FORM_string < 5> DW_AT_language DW_FORM_data2 < 7> DW_AT_name DW_FORM_string < 9> DW_AT_entry_pc DW_FORM_addr < 11> DW_AT_stmt_list DW_FORM_data4 < 13> DW_AT_comp_dir DW_FORM_string < 2>< 17> DW_TAG_base_type DW_children_no < 20> DW_AT_encoding DW_FORM_data1 < 22> DW_AT_name DW_FORM_string < 24> DW_AT_byte_size DW_FORM_data1 < 3>< 28> DW_TAG_subprogram DW_children_no < 31> DW_AT_name DW_FORM_string < 33> DW_AT_decl_file DW_FORM_data1 < 35> DW_AT_decl_line DW_FORM_data1 < 37> DW_AT_type DW_FORM_ref4 < 39> DW_AT_external DW_FORM_flag < 41> DW_AT_low_pc DW_FORM_addr < 43> DW_AT_high_pc DW_FORM_addr < 45> DW_AT_frame_base DW_FORM_block1 < 47> AT of 16359 (0x3fe7) is unknown to dwarfdump. Continuing. DW_FORM_flag < 4>< 52> null .debug_abbrev entry .debug_string .debug_aranges .debug_frame fde: < 0><0x0:0xe>

0x00000000: .debug_static_func .debug_static_vars .debug_pubtypes .debug_weaknames --------------------------------------------------------------- From rafael.espindola at gmail.com Sat Feb 26 09:10:16 2011 From: rafael.espindola at gmail.com (=?ISO-8859-1?Q?Rafael_=C1vila_de_Esp=EDndola?=) Date: Sat, 26 Feb 2011 10:10:16 -0500 Subject: [llvm-commits] Code refactoring of MCELFStreamer.cpp and ELFObjectWriter.cpp In-Reply-To: <267631.76534.qm@web55603.mail.re4.yahoo.com> References: <960821.36942.qm@web55604.mail.re4.yahoo.com> <4D680702.2040101@gmail.com> <470208.84552.qm@web55602.mail.re4.yahoo.com> <4D68461F.2090807@gmail.com> <267631.76534.qm@web55603.mail.re4.yahoo.com> Message-ID: <4D6917D8.6030506@gmail.com> On 11-02-25 8:54 PM, Jan Sjodin wrote: > The JIT implementation I am working on (MCJIT which was renamed to JunkJIT) > became a lot nicer when I decided to inherit MCELFStreamer and ELFObjectWriter. > This was needed to make the debug information available to gdb since it expects > ELF format. I basically reuse everything, the only thing that differs is that I > use a mem_raw_outstream to emit code to memory and make sure that all sections > are created before or during the post layout binding so that when > RecordRelocation is called all symbol addresses are known. In the > ELFObjectWriter, the code in WriteObject that creates sections is moved to the > post layout binding, and the code that emits the object stays in place. This is > perhaps not very clean, since it is not formalized in the interface, but it > works okay and I only need to shuffle around the code a little. You could > imagine that if section creation was restricted by the interface to only > happenen before/during post layout, then a generic JunkJIT could be written for > any object format. Each object writer would be required to compute offsets for > the sections. I am currently not aiming to do this since the MCJIT proposal by > Daniel separates the Streamer/ObjectWriter from figuring out relocations and > external symbols. However, the code refactoring patches I have/will submit are > fairly benign and enable a JIT (without lazy compilation) using the MC > framework. Cool. The rafactoring patch is OK. > - Jan > Cheers, Rafael From rafael.espindola at gmail.com Sat Feb 26 09:49:38 2011 From: rafael.espindola at gmail.com (=?ISO-8859-1?Q?Rafael_=C1vila_de_Esp=EDndola?=) Date: Sat, 26 Feb 2011 10:49:38 -0500 Subject: [llvm-commits] [patch][lto] Use MC to scan for symbol definitions and uses Message-ID: <4D692112.5030707@gmail.com> The attached patch changes LTO to use MC for scanning symbol definitions and uses. It is not as complete as it could be, but it already handles some interesting cases: * A single .globl produces an undefined reference * Uses with no definitions produce undefined references * Definitions are correctly tracked as being local or global * Macro expansions are handled correctly :-) The patch also changes LLVM's handling of module level asm to make sure it always ends in a new line. This makes it easy to just pass that string to MC and match what we would see if printing to a file. Cheers, Rafael -------------- next part -------------- A non-text attachment was scrubbed... Name: mc.patch Type: text/x-patch Size: 15447 bytes Desc: not available Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110226/b2157e50/attachment-0001.bin From rafael.espindola at gmail.com Sat Feb 26 10:44:13 2011 From: rafael.espindola at gmail.com (Rafael Espindola) Date: Sat, 26 Feb 2011 16:44:13 -0000 Subject: [llvm-commits] [llvm] r126546 - /llvm/trunk/tools/gold/gold-plugin.cpp Message-ID: <20110226164413.D593F2A6C12C@llvm.org> Author: rafael Date: Sat Feb 26 10:44:13 2011 New Revision: 126546 URL: http://llvm.org/viewvc/llvm-project?rev=126546&view=rev Log: LTO uses MC now. Modified: llvm/trunk/tools/gold/gold-plugin.cpp Modified: llvm/trunk/tools/gold/gold-plugin.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/gold/gold-plugin.cpp?rev=126546&r1=126545&r2=126546&view=diff ============================================================================== --- llvm/trunk/tools/gold/gold-plugin.cpp (original) +++ llvm/trunk/tools/gold/gold-plugin.cpp Sat Feb 26 10:44:13 2011 @@ -73,8 +73,6 @@ static generate_bc generate_bc_file = BC_NO; static std::string bc_path; static std::string obj_path; - static std::string as_path; - static std::vector as_args; static std::vector pass_through; static std::string extra_library_path; static std::string triple; @@ -96,16 +94,6 @@ generate_api_file = true; } else if (opt.startswith("mcpu=")) { mcpu = opt.substr(strlen("mcpu=")); - } else if (opt.startswith("as=")) { - if (!as_path.empty()) { - (*message)(LDPL_WARNING, "Path to as specified twice. " - "Discarding %s", opt_); - } else { - as_path = opt.substr(strlen("as=")); - } - } else if (opt.startswith("as-arg=")) { - llvm::StringRef item = opt.substr(strlen("as-arg=")); - as_args.push_back(item.str()); } else if (opt.startswith("extra-library-path=")) { extra_library_path = opt.substr(strlen("extra_library_path=")); } else if (opt.startswith("pass-through=")) { @@ -432,18 +420,6 @@ lto_codegen_set_pic_model(code_gen, output_type); lto_codegen_set_debug_model(code_gen, LTO_DEBUG_MODEL_DWARF); - if (!options::as_path.empty()) { - sys::Path p = sys::Program::FindProgramByName(options::as_path); - lto_codegen_set_assembler_path(code_gen, p.c_str()); - } - if (!options::as_args.empty()) { - std::vector as_args_p; - for (std::vector::iterator I = options::as_args.begin(), - E = options::as_args.end(); I != E; ++I) { - as_args_p.push_back(I->c_str()); - } - lto_codegen_set_assembler_args(code_gen, &as_args_p[0], as_args_p.size()); - } if (!options::mcpu.empty()) lto_codegen_set_cpu(code_gen, options::mcpu.c_str()); From fvbommel at gmail.com Sat Feb 26 11:10:33 2011 From: fvbommel at gmail.com (Frits van Bommel) Date: Sat, 26 Feb 2011 18:10:33 +0100 Subject: [llvm-commits] PATCH: instcombine switch on select of constants to br In-Reply-To: References: <893CFB3C-832C-4959-AE72-4186FCD61E99@gmail.com> <4D2A730F.2090207@mxc.ca> <130DB463-EF73-40A2-8166-5D3A014D9D12@gmail.com>

Message-ID: Since there has been no progress here for over six weeks, I updated this patch myself. Does anyone mind if I commit this version? -------------- next part -------------- A non-text attachment was scrubbed... Name: simplify-switch-on-select.patch Type: text/x-patch Size: 5729 bytes Desc: not available Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110226/a36072e0/attachment.bin From baldrick at free.fr Sat Feb 26 11:08:15 2011 From: baldrick at free.fr (Duncan Sands) Date: Sat, 26 Feb 2011 17:08:15 -0000 Subject: [llvm-commits] [dragonegg] r126547 - /dragonegg/trunk/llvm-convert.cpp Message-ID: <20110226170815.3E0DA2A6C12C@llvm.org> Author: baldrick Date: Sat Feb 26 11:08:15 2011 New Revision: 126547 URL: http://llvm.org/viewvc/llvm-project?rev=126547&view=rev Log: Fix PR3373: if an asm output is smaller than a tied input, promote the output to the size of the input. This is safe to do as long as the output is not explicitly mentioned in the asm string, so check this. (If it is explicitly mentioned then it might still be safe, but that requires a detailed analysis of the asm string). The logic already handled the case of an input that is smaller than a tied output, but did not check that the input is not explicitly mentioned in the asm string, so add this check. Modified: dragonegg/trunk/llvm-convert.cpp Modified: dragonegg/trunk/llvm-convert.cpp URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-convert.cpp?rev=126547&r1=126546&r2=126547&view=diff ============================================================================== --- dragonegg/trunk/llvm-convert.cpp (original) +++ dragonegg/trunk/llvm-convert.cpp Sat Feb 26 11:08:15 2011 @@ -3206,6 +3206,39 @@ } } +/// isOperandMentioned - Return true if the given operand is explicitly +/// mentioned in the asm string. For example if passed operand 1 then +/// this routine checks that the asm string does not contain "%1". +static bool isOperandMentioned(gimple stmt, unsigned OpNum) { + // If this is a non-extended ASM then the contents of the asm string are not + // to be interpreted. + if (gimple_asm_input_p(stmt)) + return false; + // Search for a non-escaped '%' character followed by OpNum. + for (const char *AsmStr = gimple_asm_string(stmt); *AsmStr; ++AsmStr) { + if (*AsmStr != '%') + // Not a '%', move on to next character. + continue; + char Next = AsmStr[1]; + // If this is "%%" then the '%' is escaped - skip both '%' characters. + if (Next == '%') { + ++AsmStr; + continue; + } + // Whitespace is not allowed between the '%' and the number, so check that + // the next character is a digit. + if (!ISDIGIT(Next)) + continue; + char *EndPtr; + // If this is an explicit reference to OpNum then we are done. + if (OpNum == strtoul(AsmStr+1, &EndPtr, 10)) + return true; + // Otherwise, skip over the number and keep scanning. + AsmStr = EndPtr - 1; + } + return false; +} + /// CanonicalizeConstraint - If we can canonicalize the constraint into /// something simpler, do so now. This turns register classes with a single /// register into the register itself, expands builtin constraints to multiple @@ -6934,6 +6967,7 @@ tree Input = gimple_asm_input_op(stmt, i); tree Val = TREE_VALUE(Input); tree type = TREE_TYPE(Val); + bool IsSigned = !TYPE_UNSIGNED(type); const char *Constraint = Constraints[NumOutputs+i]; @@ -7022,25 +7056,38 @@ } unsigned OTyBits = TD.getTypeSizeInBits(OTy); unsigned OpTyBits = TD.getTypeSizeInBits(OpTy); - if (OTyBits == 0 || OpTyBits == 0 || OTyBits < OpTyBits) { - // It's tempting to implement the OTyBits < OpTyBits case by - // truncating Op down to OTy, however that breaks in the case of an - // inline asm constraint that corresponds to a single register, - // because the user can write code that assumes the whole register - // is defined, despite the output operand being only a subset of the - // register. For example: - // - // asm ("sarl $10, %%eax" : "=a"(c) : "0"(1000000)); - // - // The expected behavior is for %eax to be fully defined with the - // value 1000000 immediately before the asm. - error_at(gimple_location(stmt), - "unsupported inline asm: input constraint with a matching " - "output constraint of incompatible type!"); + if (OTyBits == 0 || OpTyBits == 0) { + error_at(gimple_location(stmt), "unsupported inline asm: input " + "constraint with a matching output constraint of " + "incompatible type!"); return; + } else if (OTyBits < OpTyBits) { + // The output is smaller than the input. If the output is not a + // register then bail out. Likewise, if the output is explicitly + // mentioned in the asm string then we cannot safely promote it, + // so bail out in this case too. + if (!OutputLocations[Match].first || + isOperandMentioned(stmt, Match)) { + error_at(gimple_location(stmt), "unsupported inline asm: input " + "constraint with a matching output constraint of " + "incompatible type!"); + return; + } + // Use the input type for the output, and arrange for the result to + // be truncated to the original output type after the asm call. + CallResultTypes[OutputIndex] = std::make_pair(OpTy, IsSigned); } else if (OTyBits > OpTyBits) { - Op = CastToAnyType(Op, !TYPE_UNSIGNED(type), - OTy, CallResultTypes[OutputIndex].second); + // The input is smaller than the output. If the input is explicitly + // mentioned in the asm string then we cannot safely promote it, so + // bail out. + if (isOperandMentioned(stmt, NumOutputs + i)) { + error_at(gimple_location(stmt), "unsupported inline asm: input " + "constraint with a matching output constraint of " + "incompatible type!"); + return; + } + Op = CastToAnyType(Op, IsSigned, OTy, + CallResultTypes[OutputIndex].second); if (BYTES_BIG_ENDIAN) { Constant *ShAmt = ConstantInt::get(Op->getType(), OTyBits-OpTyBits); From aggarwa4 at illinois.edu Sat Feb 26 11:50:26 2011 From: aggarwa4 at illinois.edu (Arushi Aggarwal) Date: Sat, 26 Feb 2011 17:50:26 -0000 Subject: [llvm-commits] [poolalloc] r126548 - /poolalloc/trunk/lib/AssistDS/MergeGEP.cpp Message-ID: <20110226175026.0BABA2A6C12C@llvm.org> Author: aggarwa4 Date: Sat Feb 26 11:50:25 2011 New Revision: 126548 URL: http://llvm.org/viewvc/llvm-project?rev=126548&view=rev Log: Fixed formatting. Modified: poolalloc/trunk/lib/AssistDS/MergeGEP.cpp Modified: poolalloc/trunk/lib/AssistDS/MergeGEP.cpp URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/AssistDS/MergeGEP.cpp?rev=126548&r1=126547&r2=126548&view=diff ============================================================================== --- poolalloc/trunk/lib/AssistDS/MergeGEP.cpp (original) +++ poolalloc/trunk/lib/AssistDS/MergeGEP.cpp Sat Feb 26 11:50:25 2011 @@ -44,67 +44,67 @@ GetElementPtrInst *GEP = cast(I); if(!isa(GEP->getType()->getElementType())) continue; - std::vector worklist; - for (Value::use_iterator UI = GEP->use_begin(), - UE = GEP->use_end(); UI != UE; ++UI){ - if(!isa(UI)) - break; - GetElementPtrInst *GEPUse = cast(UI); - worklist.push_back(GEPUse); - } - while(!worklist.empty()){ - GetElementPtrInst *GEPUse = worklist.back(); - worklist.pop_back(); - SmallVector Indices; - Indices.append(GEP->op_begin()+1, GEP->op_end()); - Indices.append(GEPUse->idx_begin()+1, GEPUse->idx_end()); - GetElementPtrInst *GEPNew = GetElementPtrInst::Create(GEP->getOperand(0), - Indices.begin(), - Indices.end(), - GEPUse->getName()+ "moda", - GEPUse); - GEPUse->replaceAllUsesWith(GEPNew); - GEPUse->eraseFromParent(); - found = true; - changed = true; - } + std::vector worklist; + for (Value::use_iterator UI = GEP->use_begin(), + UE = GEP->use_end(); UI != UE; ++UI){ + if(!isa(UI)) + break; + GetElementPtrInst *GEPUse = cast(UI); + worklist.push_back(GEPUse); + } + while(!worklist.empty()){ + GetElementPtrInst *GEPUse = worklist.back(); + worklist.pop_back(); + SmallVector Indices; + Indices.append(GEP->op_begin()+1, GEP->op_end()); + Indices.append(GEPUse->idx_begin()+1, GEPUse->idx_end()); + GetElementPtrInst *GEPNew = GetElementPtrInst::Create(GEP->getOperand(0), + Indices.begin(), + Indices.end(), + GEPUse->getName()+ "moda", + GEPUse); + GEPUse->replaceAllUsesWith(GEPNew); + GEPUse->eraseFromParent(); + found = true; + changed = true; + } } } - std::vector worklist; - for (Function::iterator B = F->begin(), FE = F->end(); B != FE; ++B) { - for (BasicBlock::iterator I = B->begin(), BE = B->end(); I != BE; I++) { - if(!(isa(I))) - continue; - GetElementPtrInst *GEP1 = cast(I); - if(!isa(GEP1->getType()->getElementType())) + std::vector worklist; + for (Function::iterator B = F->begin(), FE = F->end(); B != FE; ++B) { + for (BasicBlock::iterator I = B->begin(), BE = B->end(); I != BE; I++) { + if(!(isa(I))) + continue; + GetElementPtrInst *GEP1 = cast(I); + if(!isa(GEP1->getType()->getElementType())) continue; - if (Constant *C = dyn_cast(GEP1->getOperand(0))) { - if (ConstantExpr *CE = dyn_cast(C)) { - if (CE->getOpcode() == Instruction::GetElementPtr) { + if (Constant *C = dyn_cast(GEP1->getOperand(0))) { + if (ConstantExpr *CE = dyn_cast(C)) { + if (CE->getOpcode() == Instruction::GetElementPtr) { worklist.push_back(GEP1); + } } } } } - } - while(!worklist.empty()) { - GetElementPtrInst *GEP1 = worklist.back(); - worklist.pop_back(); - Constant *C = cast(GEP1->getOperand(0)); - ConstantExpr *CE = cast(C); - SmallVector Indices; - Indices.append(CE->op_begin()+1, CE->op_end()); - Indices.append(GEP1->idx_begin()+1, GEP1->idx_end()); - GetElementPtrInst *GEPNew = GetElementPtrInst::Create(CE->getOperand(0), - Indices.begin(), - Indices.end(), - GEP1->getName()+ "modb", - GEP1); - GEP1->replaceAllUsesWith(GEPNew); - GEP1->eraseFromParent(); - changed = true; - found = true; - } + while(!worklist.empty()) { + GetElementPtrInst *GEP1 = worklist.back(); + worklist.pop_back(); + Constant *C = cast(GEP1->getOperand(0)); + ConstantExpr *CE = cast(C); + SmallVector Indices; + Indices.append(CE->op_begin()+1, CE->op_end()); + Indices.append(GEP1->idx_begin()+1, GEP1->idx_end()); + GetElementPtrInst *GEPNew = GetElementPtrInst::Create(CE->getOperand(0), + Indices.begin(), + Indices.end(), + GEP1->getName()+ "modb", + GEP1); + GEP1->replaceAllUsesWith(GEPNew); + GEP1->eraseFromParent(); + changed = true; + found = true; + } } }while(found); return changed; From aggarwa4 at illinois.edu Sat Feb 26 11:51:03 2011 From: aggarwa4 at illinois.edu (Arushi Aggarwal) Date: Sat, 26 Feb 2011 17:51:03 -0000 Subject: [llvm-commits] [poolalloc] r126549 - /poolalloc/trunk/lib/DSA/StdLibPass.cpp Message-ID: <20110226175103.3FD752A6C12C@llvm.org> Author: aggarwa4 Date: Sat Feb 26 11:51:03 2011 New Revision: 126549 URL: http://llvm.org/viewvc/llvm-project?rev=126549&view=rev Log: Fixed formatting. Modified: poolalloc/trunk/lib/DSA/StdLibPass.cpp Modified: poolalloc/trunk/lib/DSA/StdLibPass.cpp URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/StdLibPass.cpp?rev=126549&r1=126548&r2=126549&view=diff ============================================================================== --- poolalloc/trunk/lib/DSA/StdLibPass.cpp (original) +++ poolalloc/trunk/lib/DSA/StdLibPass.cpp Sat Feb 26 11:51:03 2011 @@ -77,7 +77,6 @@ #define NRET_YNARGS {0,1,0,0,0,0,0,0,0,0} #define YRET_YNARGS {1,1,0,0,0,0,0,0,0,0} - const struct { const char* name; libAction action; @@ -92,79 +91,79 @@ {"setrlimit", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, {"getcwd", {NRET_NYARGS, YRET_YNARGS, NRET_NARGS, YRET_YNARGS, false}}, - {"remove", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"rename", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"unlink", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"fileno", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"create", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"write", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"read", {NRET_YARGS, YRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, - {"truncate", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"open", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"remove", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"rename", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"unlink", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"fileno", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"create", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"write", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"read", {NRET_YARGS, YRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, + {"truncate", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"open", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"chdir", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"mkdir", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"rmdir", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"chdir", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"mkdir", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"rmdir", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"chmod", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"fchmod", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"chmod", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"fchmod", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"kill", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - - {"execl", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"execlp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"execle", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"execv", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"execvp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"kill", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"time", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"times", {NRET_YARGS, YRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, - {"ctime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"asctime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"utime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"localtime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"gmtime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"ftime", {NRET_YARGS, NRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, + {"execl", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"execlp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"execle", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"execv", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"execvp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + + {"time", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"times", {NRET_YARGS, YRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, + {"ctime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"asctime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"utime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"localtime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"gmtime", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"ftime", {NRET_YARGS, NRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, // printf not strictly true, %n could cause a write - {"printf", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"fprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, - {"fprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, - {"sprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, - {"snprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, - {"vsnprintf", {NRET_YARGS, YRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, - {"sscanf", {NRET_YARGS, YRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, - {"scanf", {NRET_YARGS, YRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, - {"fscanf", {NRET_YARGS, YRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, + {"printf", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"fprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, + {"fprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, + {"sprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, + {"snprintf", {NRET_YARGS, NRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, + {"vsnprintf", {NRET_YARGS, YRET_YNARGS, NRET_NARGS, NRET_NARGS, false}}, + {"sscanf", {NRET_YARGS, YRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, + {"scanf", {NRET_YARGS, YRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, + {"fscanf", {NRET_YARGS, YRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, + + {"calloc", {NRET_NARGS, YRET_NARGS, YRET_NARGS, NRET_NARGS, false}}, + {"malloc", {NRET_NARGS, YRET_NARGS, YRET_NARGS, NRET_NARGS, false}}, + {"valloc", {NRET_NARGS, YRET_NARGS, YRET_NARGS, NRET_NARGS, false}}, + {"realloc", {NRET_NARGS, YRET_NARGS, YRET_YNARGS, YRET_YNARGS,false}}, + {"free", {NRET_NARGS, NRET_NARGS, NRET_YNARGS, NRET_NARGS, false}}, - {"calloc", {NRET_NARGS, YRET_NARGS, YRET_NARGS, NRET_NARGS, false}}, - {"malloc", {NRET_NARGS, YRET_NARGS, YRET_NARGS, NRET_NARGS, false}}, - {"valloc", {NRET_NARGS, YRET_NARGS, YRET_NARGS, NRET_NARGS, false}}, - {"realloc", {NRET_NARGS, YRET_NARGS, YRET_YNARGS, YRET_YNARGS,false}}, - {"free", {NRET_NARGS, NRET_NARGS, NRET_YNARGS, NRET_NARGS, false}}, - - {"strdup", {NRET_YARGS, YRET_NARGS, YRET_NARGS, YRET_YARGS, false}}, - {"__strdup", {NRET_YARGS, YRET_NARGS, YRET_NARGS, YRET_YARGS, false}}, - {"wcsdup", {NRET_YARGS, YRET_NARGS, YRET_NARGS, YRET_YARGS, false}}, - - {"strlen", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"wcslen", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - - {"atoi", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"atof", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"atol", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"atoll", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"atoq", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - - {"memcmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"strcmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"wcscmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"strncmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"wcsncmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"strcasecmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"wcscasecmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"strncasecmp",{NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"wcsncasecmp",{NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"strdup", {NRET_YARGS, YRET_NARGS, YRET_NARGS, YRET_YARGS, false}}, + {"__strdup", {NRET_YARGS, YRET_NARGS, YRET_NARGS, YRET_YARGS, false}}, + {"wcsdup", {NRET_YARGS, YRET_NARGS, YRET_NARGS, YRET_YARGS, false}}, + + {"strlen", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"wcslen", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + + {"atoi", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"atof", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"atol", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"atoll", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"atoq", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + + {"memcmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"strcmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"wcscmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"strncmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"wcsncmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"strcasecmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"wcscasecmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"strncasecmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"wcsncasecmp", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, {"strcat", {NRET_YARGS, YRET_YARGS, NRET_NARGS, YRET_YARGS, true}}, {"strncat", {NRET_YARGS, YRET_YARGS, NRET_NARGS, YRET_YARGS, true}}, @@ -219,11 +218,11 @@ {"fseek", {NRET_YARGS, NRET_YARGS, NRET_NARGS, NRET_NARGS, true}}, {"rewind", {NRET_YARGS, NRET_YARGS, NRET_NARGS, NRET_NARGS, true}}, {"ferror", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"fwrite", {NRET_YARGS, NRET_NYARGS, NRET_NARGS, NRET_NARGS, false}}, + {"fwrite", {NRET_YARGS, NRET_NYARGS,NRET_NARGS, NRET_NARGS, false}}, {"fread", {NRET_NYARGS, NRET_YARGS, NRET_NARGS, NRET_NARGS, false}}, {"fdopen", {NRET_YARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, - {"__errno_location", {NRET_NARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, + {"__errno_location", {NRET_NARGS, YRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, {"puts", {NRET_YARGS, NRET_NARGS, NRET_NARGS, NRET_NARGS, false}}, {"gets", {NRET_NARGS, YRET_YARGS, NRET_NARGS, YRET_YNARGS, false}}, @@ -377,7 +376,6 @@ if (!I->isDeclaration()) getOrCreateGraph(&*I); - //FIXME: Should this happen in a "StdLib" Pass??? // // Erase direct calls to functions that don't return a pointer and are marked // with the readnone annotation. From aggarwa4 at illinois.edu Sat Feb 26 12:00:25 2011 From: aggarwa4 at illinois.edu (Arushi Aggarwal) Date: Sat, 26 Feb 2011 18:00:25 -0000 Subject: [llvm-commits] [poolalloc] r126550 - /poolalloc/trunk/lib/AssistDS/MergeArrayIndexGEP.cpp Message-ID: <20110226180025.E548B2A6C12C@llvm.org> Author: aggarwa4 Date: Sat Feb 26 12:00:25 2011 New Revision: 126550 URL: http://llvm.org/viewvc/llvm-project?rev=126550&view=rev Log: Merge GEPs. Code copied from InstCombine. Added: poolalloc/trunk/lib/AssistDS/MergeArrayIndexGEP.cpp Added: poolalloc/trunk/lib/AssistDS/MergeArrayIndexGEP.cpp URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/AssistDS/MergeArrayIndexGEP.cpp?rev=126550&view=auto ============================================================================== --- poolalloc/trunk/lib/AssistDS/MergeArrayIndexGEP.cpp (added) +++ poolalloc/trunk/lib/AssistDS/MergeArrayIndexGEP.cpp Sat Feb 26 12:00:25 2011 @@ -0,0 +1,125 @@ +//===-- MergeGEP.cpp - Merge GEPs for indexing in arrays ------------ ----===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// +//===----------------------------------------------------------------------===// +#define DEBUG_TYPE "mergearraygep" + +#include "llvm/Instructions.h" +#include "llvm/Module.h" +#include "llvm/Pass.h" +#include "llvm/Instructions.h" +#include "llvm/Constants.h" +#include "llvm/Support/GetElementPtrTypeIterator.h" +#include "llvm/Transforms/Utils/Cloning.h" +#include "llvm/ADT/Statistic.h" +#include "llvm/Support/FormattedStream.h" +#include "llvm/Support/Debug.h" +#include + +using namespace llvm; + + +namespace { + class MergeArrayGEP : public ModulePass { + public: + static char ID; + MergeArrayGEP() : ModulePass(&ID) {} + bool runOnModule(Module& M) { + for (Module::iterator F = M.begin(); F != M.end(); ++F){ + for (Function::iterator B = F->begin(), FE = F->end(); B != FE; ++B) { + for (BasicBlock::iterator I = B->begin(), BE = B->end(); I != BE;) { + GetElementPtrInst *GEP = dyn_cast(I); + I++; + if(GEP == NULL) + continue; + simplifyGEP(GEP); + } + } + } + return true; + } + static void simplifyGEP(GetElementPtrInst *GEP) { + Value *PtrOp = GEP->getOperand(0); + if (GEPOperator *Src = dyn_cast(PtrOp)) { + // Note that if our source is a gep chain itself that we wait for that + // chain to be resolved before we perform this transformation. This + // avoids us creating a TON of code in some cases. + // + if (GetElementPtrInst *SrcGEP = + dyn_cast(Src->getOperand(0))) + if (SrcGEP->getNumOperands() == 2) + return; // Wait until our source is folded to completion. + + SmallVector Indices; + + // Find out whether the last index in the source GEP is a sequential idx. + bool EndsWithSequential = false; + for (gep_type_iterator I = gep_type_begin(*Src), E = gep_type_end(*Src); + I != E; ++I) + EndsWithSequential = !(*I)->isStructTy(); + + // Can we combine the two pointer arithmetics offsets? + if (EndsWithSequential) { + // Replace: gep (gep %P, long B), long A, ... + // With: T = long A+B; gep %P, T, ... + // + Value *Sum; + Value *SO1 = Src->getOperand(Src->getNumOperands()-1); + Value *GO1 = GEP->getOperand(1); + if (SO1 == Constant::getNullValue(SO1->getType())) { + Sum = GO1; + } else if (GO1 == Constant::getNullValue(GO1->getType())) { + Sum = SO1; + } else { + // If they aren't the same type, then the input hasn't been processed + // by the loop above yet (which canonicalizes sequential index types to + // intptr_t). Just avoid transforming this until the input has been + // normalized. + if (SO1->getType() != GO1->getType()) + return; + Sum = llvm::BinaryOperator::Create(BinaryOperator::Add,SO1, GO1, PtrOp->getName()+".sum",GEP); + } + + // Update the GEP in place if possible. + if (Src->getNumOperands() == 2) { + GEP->setOperand(0, Src->getOperand(0)); + GEP->setOperand(1, Sum); + return; + } + Indices.append(Src->op_begin()+1, Src->op_end()-1); + Indices.push_back(Sum); + Indices.append(GEP->op_begin()+2, GEP->op_end()); + } else if (isa(GEP->idx_begin()) && + cast(GEP->idx_begin())->isNullValue() && + Src->getNumOperands() != 1) { + // Otherwise we can do the fold if the first index of the GEP is a zero + Indices.append(Src->op_begin()+1, Src->op_end()); + Indices.append(GEP->idx_begin()+1, GEP->idx_end()); + } + + if (!Indices.empty()){ + GetElementPtrInst *GEPNew = (GEP->isInBounds() && Src->isInBounds()) ? + GetElementPtrInst::CreateInBounds(Src->getOperand(0), Indices.begin(), + Indices.end(), GEP->getName(), GEP) : + GetElementPtrInst::Create(Src->getOperand(0), Indices.begin(), + Indices.end(), GEP->getName(), GEP); + GEP->replaceAllUsesWith(GEPNew); + GEP->eraseFromParent(); + } + } + + + } + }; +} + +char MergeArrayGEP::ID = 0; +static RegisterPass +X("mergearrgep", "Merge GEPs for arrays indexing"); From aggarwa4 at illinois.edu Sat Feb 26 12:01:00 2011 From: aggarwa4 at illinois.edu (Arushi Aggarwal) Date: Sat, 26 Feb 2011 18:01:00 -0000 Subject: [llvm-commits] [poolalloc] r126551 - /poolalloc/trunk/test/TEST.types.Makefile Message-ID: <20110226180100.77E402A6C12C@llvm.org> Author: aggarwa4 Date: Sat Feb 26 12:01:00 2011 New Revision: 126551 URL: http://llvm.org/viewvc/llvm-project?rev=126551&view=rev Log: Add the new mergearrgep flag. Modified: poolalloc/trunk/test/TEST.types.Makefile Modified: poolalloc/trunk/test/TEST.types.Makefile URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/test/TEST.types.Makefile?rev=126551&r1=126550&r2=126551&view=diff ============================================================================== --- poolalloc/trunk/test/TEST.types.Makefile (original) +++ poolalloc/trunk/test/TEST.types.Makefile Sat Feb 26 12:01:00 2011 @@ -45,9 +45,13 @@ Output/%.temp1.bc: Output/%.llvm1.bc -$(RUNTOOLSAFELY) $(LLVMLD) -disable-opt $(SAFE_OPTS) -link-as-library $< $(PA_PRE_RT) -o $@ +$(PROGRAMS_TO_TEST:%=Output/%.opt1.bc): \ +Output/%.opt1.bc: Output/%.llvm1.bc $(LOPT) $(ASSIST_SO) + -$(RUNOPT) -load $(ASSIST_SO) -disable-opt -info-output-file=$(CURDIR)/$@.info -instnamer -internalize -varargsfunc -indclone -funcspec -ipsccp -deadargelim -simplifygep -die -mergegep -mergearrgep -die -globaldce -simplifycfg -deadargelim -arg-simplify -varargsfunc -indclone -funcspec -deadargelim -globaldce -die -simplifycfg -gep-args -deadargelim -die -mergegep -die -globaldce -stats -time-passes $< -f -o $@ + $(PROGRAMS_TO_TEST:%=Output/%.opt.bc): \ Output/%.opt.bc: Output/%.llvm1.bc $(LOPT) $(ASSIST_SO) - -$(RUNOPT) -load $(ASSIST_SO) -disable-opt -info-output-file=$(CURDIR)/$@.info -instnamer -internalize -varargsfunc -indclone -funcspec -ipsccp -deadargelim -simplifygep -die -mergegep -die -globaldce -simplifycfg -deadargelim -arg-simplify -varargsfunc -indclone -funcspec -deadargelim -globaldce -die -simplifycfg -gep-args -deadargelim -die -mergegep -die -globaldce -stats -time-passes $< -f -o $@ + -$(RUNOPT) -load $(ASSIST_SO) -disable-opt -info-output-file=$(CURDIR)/$@.info -instnamer -internalize -varargsfunc -indclone -funcspec -ipsccp -deadargelim -simplifygep -die -mergegep -die -mergearrgep -die -globaldce -simplifycfg -deadargelim -arg-simplify -die -varargsfunc -die -simplifycfg -globaldce -indclone -funcspec -deadargelim -globaldce -die -simplifycfg -gep-args -deadargelim -die -mergegep -die -mergearrgep -die -globaldce -stats -time-passes $< -f -o $@ $(PROGRAMS_TO_TEST:%=Output/%.temp2.bc): \ Output/%.temp2.bc: Output/%.temp1.bc $(LOPT) $(ASSIST_SO) @@ -203,6 +207,9 @@ @/bin/echo -n "CALLS1: " >> $@ - at grep 'Number of calls that could not be resolved' $@.time.1 >> $@ @echo >> $@ + @-if test -f Output/$*.opt.diff-nat; then \ + printf "RUN: 1" >> $@;\ + fi $(PROGRAMS_TO_TEST:%=test.$(TEST).%): \ From clattner at apple.com Sat Feb 26 15:29:21 2011 From: clattner at apple.com (Chris Lattner) Date: Sat, 26 Feb 2011 13:29:21 -0800 Subject: [llvm-commits] [patch] Switch LTO to use MC In-Reply-To: <4D67D60A.9040401@gmail.com> References: <4D65DCFB.90205@gmail.com> <99D6D76F-3DAD-461D-93DE-F64B2557F0E1@apple.com> <4D67237B.6060306@gmail.com> <3B27AD7C-DE66-4564-BA79-3C958D6D3E72@apple.com> <4D67D60A.9040401@gmail.com> Message-ID: On Feb 25, 2011, at 8:17 AM, Rafael Avila de Espindola wrote: >>> Doing llvm-link of all the IL files is about 3m if I remember >>> correctly. >> >> Wow. Out of curiosity, is much of that time spent in type resolution >> stuff? I expect that to disappear when the (post 2.9) type system >> rewrite happens. I think we can make the "loading + linking" process >> much faster :) > > Not sure. I did noticed that it looks like the merge is done two modules at a time with no state saved. It might be possible to save some time in there. > > Btw, I have seen the 3.0 planned type changes mentioned some times, but I don't know what they are in detail. Is there a doc somewhere? I wrote them up and sent a link to llvmdev. >> Aha, very nice. That should be possible by just reusing a >> memorybuffer or something. > > Yes, I just need a way to represent MemoryBuffers that are not NULL terminated :-) I agree that having null termination be optional would be useful. Clang really wants this (to make its lexer faster) but I don't think anyone else cares. I'd be fine with eliminating this requirement in the MemBuffer API and giving MemoryBuffer::getFileOrSTDIN a flags argument, where one flag would be "RequiresNullTerminatedBuffer" or something like that. -Chris From benny.kra at googlemail.com Sat Feb 26 16:48:07 2011 From: benny.kra at googlemail.com (Benjamin Kramer) Date: Sat, 26 Feb 2011 22:48:07 -0000 Subject: [llvm-commits] [llvm] r126557 - in /llvm/trunk: lib/CodeGen/SelectionDAG/DAGCombiner.cpp lib/Target/README.txt test/CodeGen/X86/adde-carry.ll Message-ID: <20110226224807.808922A6C12C@llvm.org> Author: d0k Date: Sat Feb 26 16:48:07 2011 New Revision: 126557 URL: http://llvm.org/viewvc/llvm-project?rev=126557&view=rev Log: Add some DAGCombines for (adde 0, 0, glue), which are useful to optimize legalized code for large integer arithmetic. 1. Inform users of ADDEs with two 0 operands that it never sets carry 2. Fold other ADDs or ADDCs into the ADDE if possible It would be neat if we could do the same thing for SETCC+ADD eventually, but we can't do that in target independent code. Added: llvm/trunk/test/CodeGen/X86/adde-carry.ll Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp llvm/trunk/lib/Target/README.txt Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=126557&r1=126556&r2=126557&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Sat Feb 26 16:48:07 2011 @@ -1290,6 +1290,16 @@ return SDValue(); } +/// isCarryMaterialization - Returns true if V is an ADDE node that is known to +/// return 0 or 1 depending on the carry flag. +static bool isCarryMaterialization(SDValue V) { + if (V.getOpcode() != ISD::ADDE) + return false; + + ConstantSDNode *C = dyn_cast(V.getOperand(0)); + return C && C->isNullValue() && V.getOperand(0) == V.getOperand(1); +} + SDValue DAGCombiner::visitADD(SDNode *N) { SDValue N0 = N->getOperand(0); SDValue N1 = N->getOperand(1); @@ -1453,6 +1463,18 @@ return DAG.getNode(ISD::SUB, DL, VT, N1, ZExt); } + // add (adde 0, 0, glue), X -> adde X, 0, glue + if (N0->hasOneUse() && isCarryMaterialization(N0)) + return DAG.getNode(ISD::ADDE, N->getDebugLoc(), + DAG.getVTList(VT, MVT::Glue), N1, N0.getOperand(0), + N0.getOperand(2)); + + // add X, (adde 0, 0, glue) -> adde X, 0, glue + if (N1->hasOneUse() && isCarryMaterialization(N1)) + return DAG.getNode(ISD::ADDE, N->getDebugLoc(), + DAG.getVTList(VT, MVT::Glue), N0, N1.getOperand(0), + N1.getOperand(2)); + return SDValue(); } @@ -1496,6 +1518,16 @@ N->getDebugLoc(), MVT::Glue)); } + // addc (adde 0, 0, glue), X -> adde X, 0, glue + if (N0->hasOneUse() && isCarryMaterialization(N0)) + return DAG.getNode(ISD::ADDE, N->getDebugLoc(), N->getVTList(), N1, + DAG.getConstant(0, VT), N0.getOperand(2)); + + // addc X, (adde 0, 0, glue) -> adde X, 0, glue + if (N1->hasOneUse() && isCarryMaterialization(N1)) + return DAG.getNode(ISD::ADDE, N->getDebugLoc(), N->getVTList(), N0, + DAG.getConstant(0, VT), N1.getOperand(2)); + return SDValue(); } @@ -1506,6 +1538,12 @@ ConstantSDNode *N0C = dyn_cast(N0); ConstantSDNode *N1C = dyn_cast(N1); + // If both operands are null we know that carry out will always be false. + if (N0C && N0C->isNullValue() && N0 == N1) + DAG.ReplaceAllUsesOfValueWith(SDValue(N, 1), DAG.getNode(ISD::CARRY_FALSE, + N->getDebugLoc(), + MVT::Glue)); + // canonicalize constant to RHS if (N0C && !N1C) return DAG.getNode(ISD::ADDE, N->getDebugLoc(), N->getVTList(), Modified: llvm/trunk/lib/Target/README.txt URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/README.txt?rev=126557&r1=126556&r2=126557&view=diff ============================================================================== --- llvm/trunk/lib/Target/README.txt (original) +++ llvm/trunk/lib/Target/README.txt Sat Feb 26 16:48:07 2011 @@ -1780,43 +1780,6 @@ //===---------------------------------------------------------------------===// -Take the following testcase on x86-64 (similar testcases exist for all targets -with addc/adde): - -define void @a(i64* nocapture %s, i64* nocapture %t, i64 %a, i64 %b, -i64 %c) nounwind { -entry: - %0 = zext i64 %a to i128 ; [#uses=1] - %1 = zext i64 %b to i128 ; [#uses=1] - %2 = add i128 %1, %0 ; [#uses=2] - %3 = zext i64 %c to i128 ; [#uses=1] - %4 = shl i128 %3, 64 ; [#uses=1] - %5 = add i128 %4, %2 ; [#uses=1] - %6 = lshr i128 %5, 64 ; [#uses=1] - %7 = trunc i128 %6 to i64 ; [#uses=1] - store i64 %7, i64* %s, align 8 - %8 = trunc i128 %2 to i64 ; [#uses=1] - store i64 %8, i64* %t, align 8 - ret void -} - -Generated code: - addq %rcx, %rdx - sbbq %rax, %rax - subq %rax, %r8 - movq %r8, (%rdi) - movq %rdx, (%rsi) - ret - -Expected code: - addq %rcx, %rdx - adcq $0, %r8 - movq %r8, (%rdi) - movq %rdx, (%rsi) - ret - -//===---------------------------------------------------------------------===// - Switch lowering generates less than ideal code for the following switch: define void @a(i32 %x) nounwind { entry: Added: llvm/trunk/test/CodeGen/X86/adde-carry.ll URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/adde-carry.ll?rev=126557&view=auto ============================================================================== --- llvm/trunk/test/CodeGen/X86/adde-carry.ll (added) +++ llvm/trunk/test/CodeGen/X86/adde-carry.ll Sat Feb 26 16:48:07 2011 @@ -0,0 +1,26 @@ +; RUN: llc -march=x86-64 < %s | FileCheck %s -check-prefix=CHECK-64 +; RUN: llc -march=x86 < %s | FileCheck %s -check-prefix=CHECK-32 + +define void @a(i64* nocapture %s, i64* nocapture %t, i64 %a, i64 %b, i64 %c) nounwind { +entry: + %0 = zext i64 %a to i128 + %1 = zext i64 %b to i128 + %2 = add i128 %1, %0 + %3 = zext i64 %c to i128 + %4 = shl i128 %3, 64 + %5 = add i128 %4, %2 + %6 = lshr i128 %5, 64 + %7 = trunc i128 %6 to i64 + store i64 %7, i64* %s, align 8 + %8 = trunc i128 %2 to i64 + store i64 %8, i64* %t, align 8 + ret void + +; CHECK-32: addl +; CHECK-32: adcl +; CHECK-32: adcl $0 +; CHECK-32: adcl $0 + +; CHECK-64: addq +; CHECK-64: adcq $0 +} From daniel at zuster.org Sat Feb 26 17:07:34 2011 From: daniel at zuster.org (Daniel Dunbar) Date: Sat, 26 Feb 2011 15:07:34 -0800 Subject: [llvm-commits] [patch][lto] Use MC to scan for symbol definitions and uses In-Reply-To: <4D692112.5030707@gmail.com> References: <4D692112.5030707@gmail.com> Message-ID: Very nice! Patch looks quite reasonable to me, my only concern would be whether it should wait for after 2.9 branch. It's probably safe, but it might be better to take it afterwards in case it needs to bake. Are you aware of situations the current code handles that this one doesn't? - Daniel 2011/2/26 Rafael ?vila de Esp?ndola : > The attached patch changes LTO to use MC for scanning symbol definitions > and uses. It is not as complete as it could be, but it already handles > some interesting cases: > > * A single .globl produces an undefined reference > * Uses with no definitions produce undefined references > * Definitions are correctly tracked as being local or global > * Macro expansions are handled correctly :-) > > The patch also changes LLVM's handling of module level asm to make sure > it always ends in a new line. This makes it easy to just pass that > string to MC and match what we would see if printing to a file. > > Cheers, > Rafael > > _______________________________________________ > llvm-commits mailing list > llvm-commits at cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits > > From daniel at zuster.org Sat Feb 26 17:17:12 2011 From: daniel at zuster.org (Daniel Dunbar) Date: Sat, 26 Feb 2011 23:17:12 -0000 Subject: [llvm-commits] [llvm] r126558 - in /llvm/trunk: include/llvm/ADT/Statistic.h lib/Support/Statistic.cpp Message-ID: <20110226231712.A3B352A6C12C@llvm.org> Author: ddunbar Date: Sat Feb 26 17:17:12 2011 New Revision: 126558 URL: http://llvm.org/viewvc/llvm-project?rev=126558&view=rev Log: Support: Add llvm::AreStatisticsEnabled(). Modified: llvm/trunk/include/llvm/ADT/Statistic.h llvm/trunk/lib/Support/Statistic.cpp Modified: llvm/trunk/include/llvm/ADT/Statistic.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/Statistic.h?rev=126558&r1=126557&r2=126558&view=diff ============================================================================== --- llvm/trunk/include/llvm/ADT/Statistic.h (original) +++ llvm/trunk/include/llvm/ADT/Statistic.h Sat Feb 26 17:17:12 2011 @@ -121,6 +121,9 @@ /// \brief Enable the collection and printing of statistics. void EnableStatistics(); +/// \brief Check if statistics are enabled. +bool AreStatisticsEnabled(); + /// \brief Print statistics to the file returned by CreateInfoOutputFile(). void PrintStatistics(); Modified: llvm/trunk/lib/Support/Statistic.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/Statistic.cpp?rev=126558&r1=126557&r2=126558&view=diff ============================================================================== --- llvm/trunk/lib/Support/Statistic.cpp (original) +++ llvm/trunk/lib/Support/Statistic.cpp Sat Feb 26 17:17:12 2011 @@ -101,6 +101,10 @@ Enabled.setValue(true); } +bool llvm::AreStatisticsEnabled() { + return Enabled; +} + void llvm::PrintStatistics(raw_ostream &OS) { StatisticInfo &Stats = *StatInfo; From clattner at apple.com Sat Feb 26 20:36:04 2011 From: clattner at apple.com (Chris Lattner) Date: Sat, 26 Feb 2011 18:36:04 -0800 Subject: [llvm-commits] [Review request][Win64] Use pushq/popq GPRs in function prologue/epilogue In-Reply-To: References:

Message-ID: Hi Takumi, Evan said (verbally) that this looks ok to him, go ahead and commit it, thanks! -Chris On Feb 21, 2011, at 6:23 PM, NAKAMURA Takumi wrote: > Ping. > > > On Fri, Feb 18, 2011 at 2:02 PM, NAKAMURA Takumi wrote: >> Ping. >> >> It would also resolve potential issue if other x86-64 targets had a >> new calling conversions to preserve XMMs. >> >> ...Takumi >> >> >> On Mon, Feb 7, 2011 at 8:49 PM, NAKAMURA Takumi wrote: >>> Anton, >>> >>> I attached an updated patch, refined and comments added, thank you. >>> It is at; https://github.com/chapuni/LLVM/commit/8da7419f8170e5a11063eb8e992ecb1ad40f9f6a >>> >>> >>> On the github, Anton wrote; >>>> View Commit: https://github.com/chapuni/LLVM/commit/d719f1b1ba5f01823e00904c1599adf356089a1d >>> >>>> This will pessimize non-win64 targets. Basically, the problem is that you cannot use pushq/popq for high xmm regs, which are callee-saved on win64-only (but not on linux / darwin, for example). >>> >>> I thought the patch does not affect to non-win64 targets, but it might >>> be dubious to check whether regs are GPR or not. >>> I rewrote checking expressions. >>> >>> I am sorry if I missed your point. >>> >>>> Another problem is function prologue emission. Have you verified that you always have proper stack frame? Even in case when high xmm regs (xmm5, etc.) are spilled? >>> >>> As far as spiller emits in order [pushq GPRs...] and [mov xmm to fp], >>> and restorer emits in order [mov xmm from fp] and [popq GPRs], >>> emitPrologue() and emitEpilogue() will emit adjusting %rsp onto proper place. >>> And I can expect spiller can place XMMs i128 aligned. >>> >>> ; for example >>> define void @foo() nounwind { >>> entry: >>> tail call void (...)* @bar() nounwind >>> tail call void asm sideeffect "nop", >>> "~{si},~{di},~{xmm13},~{xmm11},~{xmm15},~{dirflag},~{fpsr},~{flags}"() >>> nounwind >>> ret void >>> } >>> declare void @bar(...) >>> >>> #### -mtriple=x86_64-mingw32 >>> foo: >>> pushq %rsi >>> pushq %rdi >>> subq $88, %rsp >>> movaps %xmm15, 32(%rsp) # 16-byte Spill >>> movaps %xmm13, 48(%rsp) # 16-byte Spill >>> movaps %xmm11, 64(%rsp) # 16-byte Spill >>> callq bar >>> #APP >>> nop >>> #NO_APP >>> movaps 64(%rsp), %xmm11 # 16-byte Reload >>> movaps 48(%rsp), %xmm13 # 16-byte Reload >>> movaps 32(%rsp), %xmm15 # 16-byte Reload >>> addq $88, %rsp >>> popq %rdi >>> popq %rsi >>> ret >>> >>> And also, I have checked to build clang by 3 stage. >>> (x64-clang can build and test clang and llvm) >>> >>> >>> ...Takumi >>> >> > > _______________________________________________ > llvm-commits mailing list > llvm-commits at cs.uiuc.edu > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits From daniel at zuster.org Sat Feb 26 21:22:36 2011 From: daniel at zuster.org (Daniel Dunbar) Date: Sun, 27 Feb 2011 03:22:36 -0000 Subject: [llvm-commits] [zorg] r126561 - /zorg/trunk/zorg/buildbot/builders/ClangBuilder.py Message-ID: <20110227032236.235DE2A6C12C@llvm.org> Author: ddunbar Date: Sat Feb 26 21:22:35 2011 New Revision: 126561 URL: http://llvm.org/viewvc/llvm-project?rev=126561&view=rev Log: buildbot: Add ClangBuilder.addClangGDBTests() Modified: zorg/trunk/zorg/buildbot/builders/ClangBuilder.py Modified: zorg/trunk/zorg/buildbot/builders/ClangBuilder.py URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/builders/ClangBuilder.py?rev=126561&r1=126560&r2=126561&view=diff ============================================================================== --- zorg/trunk/zorg/buildbot/builders/ClangBuilder.py (original) +++ zorg/trunk/zorg/buildbot/builders/ClangBuilder.py Sat Feb 26 21:22:35 2011 @@ -346,8 +346,8 @@ return f -def addClangTests(f, ignores={}, install_prefix="%(builddir)s/llvm.install", - languages = ('gcc', 'g++', 'objc', 'obj-c++')): +def addClangGCCTests(f, ignores={}, install_prefix="%(builddir)s/llvm.install", + languages = ('gcc', 'g++', 'objc', 'obj-c++')): make_vars = [WithProperties( 'CC_UNDER_TEST=%s/bin/clang' % install_prefix), WithProperties( @@ -365,6 +365,26 @@ logfiles={ 'dg.sum' : 'obj/%s/%s.sum' % (lang, lang) }, ignore=gcc_dg_ignores.get(lang, []))) +def addClangGDBTests(f, ignores={}, install_prefix="%(builddir)s/llvm.install"): + make_vars = [WithProperties( + 'CC_UNDER_TEST=%s/bin/clang' % install_prefix), + WithProperties( + 'CXX_UNDER_TEST=%s/bin/clang++' % install_prefix)] + f.addStep(SVN(name='svn-clang-tests', mode='update', + baseURL='http://llvm.org/svn/llvm-project/clang-tests/', + defaultBranch='trunk', workdir='clang-tests')) + gdb_dg_ignores = ignores.get('gdb-1472-testsuite', {}) + f.addStep(DejaGNUCommand.DejaGNUCommand( + name='test-gdb-1472-testsuite', + command=["make", "-k", "check"] + make_vars, + description="gdb-1472-testsuite", + workdir='clang-tests/gdb-1472-testsuite', + logfiles={ 'dg.sum' : 'obj/gdb.sum' }, + ignore=gdb_dg_ignores)) + +# FIXME: Deprecated. +addClangTests = addClangGCCTests + def getClangTestsIgnoresFromPath(path, key): def readList(path): if not os.path.exists(path): @@ -387,4 +407,12 @@ readList(os.path.join(lang_path, 'XPASS.txt'))) ignores['gcc-4_2-testsuite' ] = gcc_dg_ignores + ignores_path = os.path.join(path, 'gdb-1472-testsuite', 'expected_results', + key) + gdb_dg_ignores = ( + readList(os.path.join(ignores_path, 'FAIL.txt')) + + readList(os.path.join(ignores_path, 'UNRESOLVED.txt')) + + readList(os.path.join(ignores_path, 'XPASS.txt'))) + ignores['gdb-1472-testsuite' ] = gdb_dg_ignores + return ignores From grosser at fim.uni-passau.de Sat Feb 26 22:11:03 2011 From: grosser at fim.uni-passau.de (Tobias Grosser) Date: Sun, 27 Feb 2011 04:11:03 -0000 Subject: [llvm-commits] [llvm] r126562 - in /llvm/trunk: include/llvm/Support/DOTGraphTraits.h include/llvm/Support/GraphWriter.h lib/CodeGen/ScheduleDAGPrinter.cpp lib/CodeGen/SelectionDAG/SelectionDAGPrinter.cpp Message-ID: <20110227041103.611FB2A6C12C@llvm.org> Author: grosser Date: Sat Feb 26 22:11:03 2011 New Revision: 126562 URL: http://llvm.org/viewvc/llvm-project?rev=126562&view=rev Log: Pass the graph to the DOTGraphTraits.getEdgeAttributes(). This follows the interface of getNodeAttributes. Modified: llvm/trunk/include/llvm/Support/DOTGraphTraits.h llvm/trunk/include/llvm/Support/GraphWriter.h llvm/trunk/lib/CodeGen/ScheduleDAGPrinter.cpp llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGPrinter.cpp Modified: llvm/trunk/include/llvm/Support/DOTGraphTraits.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/DOTGraphTraits.h?rev=126562&r1=126561&r2=126562&view=diff ============================================================================== --- llvm/trunk/include/llvm/Support/DOTGraphTraits.h (original) +++ llvm/trunk/include/llvm/Support/DOTGraphTraits.h Sat Feb 26 22:11:03 2011 @@ -89,8 +89,9 @@ /// If you want to override the dot attributes printed for a particular edge, /// override this method. - template - static std::string getEdgeAttributes(const void *Node, EdgeIter EI) { + template + static std::string getEdgeAttributes(const void *Node, EdgeIter EI, + const GraphType& Graph) { return ""; } Modified: llvm/trunk/include/llvm/Support/GraphWriter.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/GraphWriter.h?rev=126562&r1=126561&r2=126562&view=diff ============================================================================== --- llvm/trunk/include/llvm/Support/GraphWriter.h (original) +++ llvm/trunk/include/llvm/Support/GraphWriter.h Sat Feb 26 22:11:03 2011 @@ -240,7 +240,7 @@ emitEdge(static_cast(Node), edgeidx, static_cast(TargetNode), DestPort, - DTraits.getEdgeAttributes(Node, EI)); + DTraits.getEdgeAttributes(Node, EI, G)); } } Modified: llvm/trunk/lib/CodeGen/ScheduleDAGPrinter.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ScheduleDAGPrinter.cpp?rev=126562&r1=126561&r2=126562&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/ScheduleDAGPrinter.cpp (original) +++ llvm/trunk/lib/CodeGen/ScheduleDAGPrinter.cpp Sat Feb 26 22:11:03 2011 @@ -51,7 +51,8 @@ /// If you want to override the dot attributes printed for a particular /// edge, override this method. static std::string getEdgeAttributes(const SUnit *Node, - SUnitIterator EI) { + SUnitIterator EI, + const ScheduleDAG *Graph) { if (EI.isArtificialDep()) return "color=cyan,style=dashed"; if (EI.isCtrlDep()) Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGPrinter.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGPrinter.cpp?rev=126562&r1=126561&r2=126562&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGPrinter.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGPrinter.cpp Sat Feb 26 22:11:03 2011 @@ -90,7 +90,8 @@ /// If you want to override the dot attributes printed for a particular /// edge, override this method. template - static std::string getEdgeAttributes(const void *Node, EdgeIter EI) { + static std::string getEdgeAttributes(const void *Node, EdgeIter EI, + const SelectionDAG *Graph) { SDValue Op = EI.getNode()->getOperand(EI.getOperand()); EVT VT = Op.getValueType(); if (VT == MVT::Glue) From grosser at fim.uni-passau.de Sat Feb 26 22:11:05 2011 From: grosser at fim.uni-passau.de (Tobias Grosser) Date: Sun, 27 Feb 2011 04:11:05 -0000 Subject: [llvm-commits] [llvm] r126563 - in /llvm/trunk: cmake/config-ix.cmake include/llvm/Config/config.h.cmake include/llvm/Config/llvm-config.h.cmake Message-ID: <20110227041105.A65582A6C12D@llvm.org> Author: grosser Date: Sat Feb 26 22:11:05 2011 New Revision: 126563 URL: http://llvm.org/viewvc/llvm-project?rev=126563&view=rev Log: cmake: Add xdot.py support as it already exists in autoconf. Modified: llvm/trunk/cmake/config-ix.cmake llvm/trunk/include/llvm/Config/config.h.cmake llvm/trunk/include/llvm/Config/llvm-config.h.cmake Modified: llvm/trunk/cmake/config-ix.cmake URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/cmake/config-ix.cmake?rev=126563&r1=126562&r2=126563&view=diff ============================================================================== --- llvm/trunk/cmake/config-ix.cmake (original) +++ llvm/trunk/cmake/config-ix.cmake Sat Feb 26 22:11:05 2011 @@ -224,6 +224,7 @@ # available programs checks function(llvm_find_program name) string(TOUPPER ${name} NAME) + string(REGEX REPLACE "\\." "_" NAME ${NAME}) find_program(LLVM_PATH_${NAME} ${name}) mark_as_advanced(LLVM_PATH_${NAME}) if(LLVM_PATH_${NAME}) @@ -241,6 +242,7 @@ llvm_find_program(fdp) llvm_find_program(dot) llvm_find_program(dotty) +llvm_find_program(xdot.py) if( LLVM_ENABLE_FFI ) find_path(FFI_INCLUDE_PATH ffi.h PATHS ${FFI_INCLUDE_DIR}) Modified: llvm/trunk/include/llvm/Config/config.h.cmake URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Config/config.h.cmake?rev=126563&r1=126562&r2=126563&view=diff ============================================================================== --- llvm/trunk/include/llvm/Config/config.h.cmake (original) +++ llvm/trunk/include/llvm/Config/config.h.cmake Sat Feb 26 22:11:05 2011 @@ -453,7 +453,7 @@ #cmakedefine HAVE_WRITEV ${HAVE_WRITEV} /* Define if the xdot.py program is available */ -#undef HAVE_XDOT_PY +#cmakedefine HAVE_XDOT_PY ${HAVE_XDOT_PY} /* Have host's _alloca */ #cmakedefine HAVE__ALLOCA ${HAVE__ALLOCA} @@ -585,7 +585,7 @@ #cmakedefine LLVM_PATH_TWOPI "${LLVM_PATH_TWOPI}" /* Define to path to xdot.py program if found or 'echo xdot.py' otherwise */ -#undef LLVM_PATH_XDOT_PY +#cmakedefine LLVM_PATH_XDOT_PY "${LLVM_PATH_XDOT_PY}" /* Installation prefix directory */ #cmakedefine LLVM_PREFIX "${LLVM_PREFIX}" Modified: llvm/trunk/include/llvm/Config/llvm-config.h.cmake URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Config/llvm-config.h.cmake?rev=126563&r1=126562&r2=126563&view=diff ============================================================================== --- llvm/trunk/include/llvm/Config/llvm-config.h.cmake (original) +++ llvm/trunk/include/llvm/Config/llvm-config.h.cmake Sat Feb 26 22:11:05 2011 @@ -91,6 +91,9 @@ /* Define to path to twopi program if found or 'echo twopi' otherwise */ #cmakedefine LLVM_PATH_TWOPI "${LLVM_PATH_TWOPI}" +/* Define to path to xdot.py program if found or 'echo xdot.py' otherwise */ +#cmakedefine LLVM_PATH_XDOT_PY "${LLVM_PATH_XDOT.PY}" + /* Installation prefix directory */ #cmakedefine LLVM_PREFIX "${LLVM_PREFIX}" From grosser at fim.uni-passau.de Sat Feb 26 22:11:07 2011 From: grosser at fim.uni-passau.de (Tobias Grosser) Date: Sun, 27 Feb 2011 04:11:07 -0000 Subject: [llvm-commits] [llvm] r126564 - /llvm/trunk/lib/Analysis/RegionPrinter.cpp Message-ID: <20110227041107.92B962A6C12E@llvm.org> Author: grosser Date: Sat Feb 26 22:11:07 2011 New Revision: 126564 URL: http://llvm.org/viewvc/llvm-project?rev=126564&view=rev Log: RegionPrinter: Ignore back edges when layouting the graph Modified: llvm/trunk/lib/Analysis/RegionPrinter.cpp Modified: llvm/trunk/lib/Analysis/RegionPrinter.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionPrinter.cpp?rev=126564&r1=126563&r2=126564&view=diff ============================================================================== --- llvm/trunk/lib/Analysis/RegionPrinter.cpp (original) +++ llvm/trunk/lib/Analysis/RegionPrinter.cpp Sat Feb 26 22:11:07 2011 @@ -70,6 +70,32 @@ G->getTopLevelRegion()); } + std::string getEdgeAttributes(RegionNode *srcNode, + GraphTraits::ChildIteratorType CI, RegionInfo *RI) { + + RegionNode *destNode = *CI; + + if (srcNode->isSubRegion() || destNode->isSubRegion()) + return ""; + + // In case of a backedge, do not use it to define the layout of the nodes. + BasicBlock *srcBB = srcNode->getNodeAs(); + BasicBlock *destBB = destNode->getNodeAs(); + + Region *R = RI->getRegionFor(destBB); + + while (R && R->getParent()) + if (R->getParent()->getEntry() == destBB) + R = R->getParent(); + else + break; + + if (R->getEntry() == destBB && R->contains(srcBB)) + return "constraint=false"; + + return ""; + } + // Print the cluster of the subregions. This groups the single basic blocks // and adds a different background color for each group. static void printRegionCluster(const Region *R, GraphWriter &GW, From nadav.rotem at intel.com Sun Feb 27 01:40:43 2011 From: nadav.rotem at intel.com (Nadav Rotem) Date: Sun, 27 Feb 2011 07:40:43 -0000 Subject: [llvm-commits] [llvm] r126565 - /llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Message-ID: <20110227074043.DF08A2A6C12C@llvm.org> Author: nadav Date: Sun Feb 27 01:40:43 2011 New Revision: 126565 URL: http://llvm.org/viewvc/llvm-project?rev=126565&view=rev Log: Fix typos in the comments. Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=126565&r1=126564&r2=126565&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Sun Feb 27 01:40:43 2011 @@ -3726,7 +3726,8 @@ // fold (sext (load x)) -> (sext (truncate (sextload x))) // None of the supported targets knows how to perform load and sign extend - // in one instruction. We only perform this transformation on scalars. + // on vectors in one instruction. We only perform this transformation on + // scalars. if (ISD::isNON_EXTLoad(N0.getNode()) && !VT.isVector() && ((!LegalOperations && !cast(N0)->isVolatile()) || TLI.isLoadExtLegal(ISD::SEXTLOAD, N0.getValueType()))) { @@ -3930,7 +3931,8 @@ // fold (zext (load x)) -> (zext (truncate (zextload x))) // None of the supported targets knows how to perform load and vector_zext - // in one instruction. We only perform this transformation on scalar zext. + // on vectors in one instruction. We only perform this transformation on + // scalars. if (ISD::isNON_EXTLoad(N0.getNode()) && !VT.isVector() && ((!LegalOperations && !cast(N0)->isVolatile()) || TLI.isLoadExtLegal(ISD::ZEXTLOAD, N0.getValueType()))) { @@ -4139,7 +4141,8 @@ // fold (aext (load x)) -> (aext (truncate (extload x))) // None of the supported targets knows how to perform load and any_ext - // in one instruction. We only perform this transformation on scalars. + // on vectors in one instruction. We only perform this transformation on + // scalars. if (ISD::isNON_EXTLoad(N0.getNode()) && !VT.isVector() && ((!LegalOperations && !cast(N0)->isVolatile()) || TLI.isLoadExtLegal(ISD::EXTLOAD, N0.getValueType()))) { @@ -4552,7 +4555,7 @@ // See if we can simplify the input to this truncate through knowledge that // only the low bits are being used. // For example "trunc (or (shl x, 8), y)" // -> trunc y - // Currenly we only perform this optimization on scalars because vectors + // Currently we only perform this optimization on scalars because vectors // may have different active low bits. if (!VT.isVector()) { SDValue Shorter = From zwarich at apple.com Sun Feb 27 02:05:57 2011 From: zwarich at apple.com (Cameron Zwarich) Date: Sun, 27 Feb 2011 08:05:57 -0000 Subject: [llvm-commits] [llvm] r126566 - in /llvm/trunk/test/CodeGen/X86: phi-bit-propagation.ll phi-constants.ll Message-ID: <20110227080557.CD64F2A6C12C@llvm.org> Author: zwarich Date: Sun Feb 27 02:05:57 2011 New Revision: 126566 URL: http://llvm.org/viewvc/llvm-project?rev=126566&view=rev Log: Give a test file a more sensible name so that it can hold more test cases. Added: llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll - copied, changed from r126565, llvm/trunk/test/CodeGen/X86/phi-constants.ll Removed: llvm/trunk/test/CodeGen/X86/phi-constants.ll Copied: llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll (from r126565, llvm/trunk/test/CodeGen/X86/phi-constants.ll) URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll?p2=llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll&p1=llvm/trunk/test/CodeGen/X86/phi-constants.ll&r1=126565&r2=126566&rev=126566&view=diff ============================================================================== (empty) Removed: llvm/trunk/test/CodeGen/X86/phi-constants.ll URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/phi-constants.ll?rev=126565&view=auto ============================================================================== --- llvm/trunk/test/CodeGen/X86/phi-constants.ll (original) +++ llvm/trunk/test/CodeGen/X86/phi-constants.ll (removed) @@ -1,35 +0,0 @@ -; RUN: llc < %s -march=x86-64 | FileCheck %s - -%"class.std::bitset" = type { [8 x i8] } - -define zeroext i1 @_Z3fooPjmS_mRSt6bitsetILm32EE(i32* nocapture %a, i64 %asize, i32* nocapture %b, i64 %bsize, %"class.std::bitset"* %bits) nounwind readonly ssp noredzone { -entry: - %tmp.i.i.i.i = bitcast %"class.std::bitset"* %bits to i64* - br label %for.cond - -for.cond: ; preds = %for.inc, %entry - %0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ] - %conv = zext i32 %0 to i64 - %cmp = icmp eq i64 %conv, %bsize - br i1 %cmp, label %return, label %for.body - -for.body: ; preds = %for.cond - %arrayidx = getelementptr inbounds i32* %b, i64 %conv - %tmp5 = load i32* %arrayidx, align 4 - %conv6 = zext i32 %tmp5 to i64 - %rem.i.i.i.i = and i64 %conv6, 63 - %tmp3.i = load i64* %tmp.i.i.i.i, align 8 - %shl.i.i = shl i64 1, %rem.i.i.i.i - %and.i = and i64 %shl.i.i, %tmp3.i - %cmp.i = icmp eq i64 %and.i, 0 - br i1 %cmp.i, label %for.inc, label %return - -for.inc: ; preds = %for.body - %inc = add i32 %0, 1 - br label %for.cond - -return: ; preds = %for.body, %for.cond -; CHECK-NOT: and - %retval.0 = phi i1 [ true, %for.body ], [ false, %for.cond ] - ret i1 %retval.0 -} From zwarich at apple.com Sun Feb 27 02:06:01 2011 From: zwarich at apple.com (Cameron Zwarich) Date: Sun, 27 Feb 2011 08:06:01 -0000 Subject: [llvm-commits] [llvm] r126567 - in /llvm/trunk: include/llvm/CodeGen/FunctionLoweringInfo.h test/CodeGen/X86/phi-bit-propagation.ll Message-ID: <20110227080601.A40302A6C12D@llvm.org> Author: zwarich Date: Sun Feb 27 02:06:01 2011 New Revision: 126567 URL: http://llvm.org/viewvc/llvm-project?rev=126567&view=rev Log: Fix PR9324 / by handling the case where a PHI has no uses. Modified: llvm/trunk/include/llvm/CodeGen/FunctionLoweringInfo.h llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll Modified: llvm/trunk/include/llvm/CodeGen/FunctionLoweringInfo.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/FunctionLoweringInfo.h?rev=126567&r1=126566&r2=126567&view=diff ============================================================================== --- llvm/trunk/include/llvm/CodeGen/FunctionLoweringInfo.h (original) +++ llvm/trunk/include/llvm/CodeGen/FunctionLoweringInfo.h Sun Feb 27 02:06:01 2011 @@ -187,7 +187,12 @@ /// InvalidatePHILiveOutRegInfo - Invalidates a PHI's LiveOutInfo, to be /// called when a block is visited before all of its predecessors. void InvalidatePHILiveOutRegInfo(const PHINode *PN) { - unsigned Reg = ValueMap[PN]; + // PHIs with no uses have no ValueMap entry. + DenseMap::const_iterator It = ValueMap.find(PN); + if (It == ValueMap.end()) + return; + + unsigned Reg = It->second; LiveOutRegInfo.grow(Reg); LiveOutRegInfo[Reg].IsValid = false; } Modified: llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll?rev=126567&r1=126566&r2=126567&view=diff ============================================================================== --- llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll (original) +++ llvm/trunk/test/CodeGen/X86/phi-bit-propagation.ll Sun Feb 27 02:06:01 2011 @@ -33,3 +33,23 @@ %retval.0 = phi i1 [ true, %for.body ], [ false, %for.cond ] ret i1 %retval.0 } + +; This test case caused an assertion failure; see PR9324. +define void @func_37() noreturn nounwind ssp { +entry: + br i1 undef, label %lbl_919, label %entry.for.inc_crit_edge + +entry.for.inc_crit_edge: ; preds = %entry + br label %for.inc + +lbl_919: ; preds = %for.cond7.preheader, %entry + br label %for.cond7.preheader + +for.cond7.preheader: ; preds = %for.inc, %lbl_919 + %storemerge.ph = phi i8 [ 0, %lbl_919 ], [ %add, %for.inc ] + br i1 undef, label %for.inc, label %lbl_919 + +for.inc: ; preds = %for.cond7.preheader, %entry.for.inc_crit_edge + %add = add i8 undef, 1 + br label %for.cond7.preheader +} From geek4civic at gmail.com Sun Feb 27 02:47:19 2011 From: geek4civic at gmail.com (NAKAMURA Takumi) Date: Sun, 27 Feb 2011 08:47:19 -0000 Subject: [llvm-commits] [llvm] r126568 - in /llvm/trunk: lib/Target/X86/X86FrameLowering.cpp test/CodeGen/X86/tailcallstack64.ll Message-ID: <20110227084719.99DDD2A6C12D@llvm.org> Author: chapuni Date: Sun Feb 27 02:47:19 2011 New Revision: 126568 URL: http://llvm.org/viewvc/llvm-project?rev=126568&view=rev Log: Target/X86: Always emit "push/pop GPRs" in prologue/epilogue and emit "spill/reload frames" for XMMs. It improves Win64's prologue/epilogue but it would not affect ia32 and amd64 (lack of nonvolatile XMMs). Modified: llvm/trunk/lib/Target/X86/X86FrameLowering.cpp llvm/trunk/test/CodeGen/X86/tailcallstack64.ll Modified: llvm/trunk/lib/Target/X86/X86FrameLowering.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FrameLowering.cpp?rev=126568&r1=126567&r2=126568&view=diff ============================================================================== --- llvm/trunk/lib/Target/X86/X86FrameLowering.cpp (original) +++ llvm/trunk/lib/Target/X86/X86FrameLowering.cpp Sun Feb 27 02:47:19 2011 @@ -892,7 +892,6 @@ MachineFunction &MF = *MBB.getParent(); - bool isWin64 = STI.isTargetWin64(); unsigned SlotSize = STI.is64Bit() ? 8 : 4; unsigned FPReg = TRI->getFrameRegister(MF); unsigned CalleeFrameSize = 0; @@ -900,25 +899,39 @@ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo(); X86MachineFunctionInfo *X86FI = MF.getInfo(); + // Push GPRs. It increases frame size. unsigned Opc = STI.is64Bit() ? X86::PUSH64r : X86::PUSH32r; for (unsigned i = CSI.size(); i != 0; --i) { unsigned Reg = CSI[i-1].getReg(); + if (!X86::GR64RegClass.contains(Reg) && + !X86::GR32RegClass.contains(Reg)) + continue; // Add the callee-saved register as live-in. It's killed at the spill. MBB.addLiveIn(Reg); if (Reg == FPReg) // X86RegisterInfo::emitPrologue will handle spilling of frame register. continue; - if (!X86::VR128RegClass.contains(Reg) && !isWin64) { - CalleeFrameSize += SlotSize; - BuildMI(MBB, MI, DL, TII.get(Opc)).addReg(Reg, RegState::Kill); - } else { - const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); - TII.storeRegToStackSlot(MBB, MI, Reg, true, CSI[i-1].getFrameIdx(), - RC, TRI); - } + CalleeFrameSize += SlotSize; + BuildMI(MBB, MI, DL, TII.get(Opc)).addReg(Reg, RegState::Kill); } X86FI->setCalleeSavedFrameSize(CalleeFrameSize); + + // Make XMM regs spilled. X86 does not have ability of push/pop XMM. + // It can be done by spilling XMMs to stack frame. + // Note that only Win64 ABI might spill XMMs. + for (unsigned i = CSI.size(); i != 0; --i) { + unsigned Reg = CSI[i-1].getReg(); + if (X86::GR64RegClass.contains(Reg) || + X86::GR32RegClass.contains(Reg)) + continue; + // Add the callee-saved register as live-in. It's killed at the spill. + MBB.addLiveIn(Reg); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.storeRegToStackSlot(MBB, MI, Reg, true, CSI[i-1].getFrameIdx(), + RC, TRI); + } + return true; } @@ -933,21 +946,30 @@ MachineFunction &MF = *MBB.getParent(); const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo(); + + // Reload XMMs from stack frame. + for (unsigned i = 0, e = CSI.size(); i != e; ++i) { + unsigned Reg = CSI[i].getReg(); + if (X86::GR64RegClass.contains(Reg) || + X86::GR32RegClass.contains(Reg)) + continue; + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.loadRegFromStackSlot(MBB, MI, Reg, CSI[i].getFrameIdx(), + RC, TRI); + } + + // POP GPRs. unsigned FPReg = TRI->getFrameRegister(MF); - bool isWin64 = STI.isTargetWin64(); unsigned Opc = STI.is64Bit() ? X86::POP64r : X86::POP32r; for (unsigned i = 0, e = CSI.size(); i != e; ++i) { unsigned Reg = CSI[i].getReg(); + if (!X86::GR64RegClass.contains(Reg) && + !X86::GR32RegClass.contains(Reg)) + continue; if (Reg == FPReg) // X86RegisterInfo::emitEpilogue will handle restoring of frame register. continue; - if (!X86::VR128RegClass.contains(Reg) && !isWin64) { - BuildMI(MBB, MI, DL, TII.get(Opc), Reg); - } else { - const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); - TII.loadRegFromStackSlot(MBB, MI, Reg, CSI[i].getFrameIdx(), - RC, TRI); - } + BuildMI(MBB, MI, DL, TII.get(Opc), Reg); } return true; } Modified: llvm/trunk/test/CodeGen/X86/tailcallstack64.ll URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcallstack64.ll?rev=126568&r1=126567&r2=126568&view=diff ============================================================================== --- llvm/trunk/test/CodeGen/X86/tailcallstack64.ll (original) +++ llvm/trunk/test/CodeGen/X86/tailcallstack64.ll Sun Feb 27 02:47:19 2011 @@ -2,7 +2,7 @@ ; RUN: llc < %s -tailcallopt -mtriple=x86_64-win32 -post-RA-scheduler=true | FileCheck %s ; FIXME: Redundant unused stack allocation could be eliminated. -; CHECK: subq ${{24|88}}, %rsp +; CHECK: subq ${{24|72}}, %rsp ; Check that lowered arguments on the stack do not overwrite each other. ; Add %in1 %p1 to a different temporary register (%eax). From geek4civic at gmail.com Sun Feb 27 02:51:59 2011 From: geek4civic at gmail.com (NAKAMURA Takumi) Date: Sun, 27 Feb 2011 17:51:59 +0900 Subject: [llvm-commits] [Review request][Win64] Use pushq/popq GPRs in function prologue/epilogue In-Reply-To: References:

Message-ID: Applied in r126568, thank you! ...Takumi From baldrick at free.fr Sun Feb 27 07:24:02 2011 From: baldrick at free.fr (Duncan Sands) Date: Sun, 27 Feb 2011 13:24:02 -0000 Subject: [llvm-commits] [dragonegg] r126569 - /dragonegg/trunk/llvm-convert.cpp Message-ID: <20110227132403.0BDF92A6C12C@llvm.org> Author: baldrick Date: Sun Feb 27 07:24:02 2011 New Revision: 126569 URL: http://llvm.org/viewvc/llvm-project?rev=126569&view=rev Log: Calculate the CallArgTypes vector from the call operands, eliminating the possibility of forgetting to update it. Modified: dragonegg/trunk/llvm-convert.cpp Modified: dragonegg/trunk/llvm-convert.cpp URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-convert.cpp?rev=126569&r1=126568&r2=126569&view=diff ============================================================================== --- dragonegg/trunk/llvm-convert.cpp (original) +++ dragonegg/trunk/llvm-convert.cpp Sun Feb 27 07:24:02 2011 @@ -6863,7 +6863,6 @@ bool HasSideEffects = gimple_asm_volatile_p(stmt) || (NumOutputs == 0); std::vector CallOps; - std::vector CallArgTypes; // CallResultTypes - The inline asm call may return one or more results. The // types of the results are recorded here along with a flag indicating whether @@ -6957,8 +6956,7 @@ ConstraintStr += ",=*"; ConstraintStr += SimplifiedConstraint; CallOps.push_back(Dest.Ptr); - CallArgTypes.push_back(Dest.Ptr->getType()); - OutputLocations.push_back(std::make_pair(false, CallArgTypes.size()-1)); + OutputLocations.push_back(std::make_pair(false, CallOps.size()-1)); } } @@ -7041,7 +7039,7 @@ if (OutputLocations[Match].first) OTy = CallResultTypes[OutputIndex].first; else { - OTy = CallArgTypes[OutputIndex]; + OTy = CallOps[OutputIndex]->getType(); assert(OTy->isPointerTy() && "Expected pointer type!"); OTy = cast(OTy)->getElementType(); } @@ -7098,14 +7096,12 @@ } CallOps.push_back(Op); - CallArgTypes.push_back(Op->getType()); } else { // Memory operand. mark_addressable(TREE_VALUE(Input)); isIndirect = true; LValue Src = EmitLV(Val); assert(!Src.isBitfield() && "Cannot read from a bitfield!"); CallOps.push_back(Src.Ptr); - CallArgTypes.push_back(Src.Ptr->getType()); } ConstraintStr += ','; @@ -7209,6 +7205,12 @@ break; } + // Compute the types of the arguments to the asm call. + std::vector CallArgTypes(CallOps.size()); + for (unsigned i = 0, e = CallOps.size(); i != e; ++i) + CallArgTypes[i] = CallOps[i]->getType(); + + // Get the type of the called asm "function". const FunctionType *FTy = FunctionType::get(CallResultType, CallArgTypes, false); From baldrick at free.fr Sun Feb 27 07:29:05 2011 From: baldrick at free.fr (Duncan Sands) Date: Sun, 27 Feb 2011 13:29:05 -0000 Subject: [llvm-commits] [dragonegg] r126570 - /dragonegg/trunk/llvm-convert.cpp Message-ID: <20110227132905.58F242A6C12C@llvm.org> Author: baldrick Date: Sun Feb 27 07:29:05 2011 New Revision: 126570 URL: http://llvm.org/viewvc/llvm-project?rev=126570&view=rev Log: Add descriptions of some variables using when processing inline asm. Modified: dragonegg/trunk/llvm-convert.cpp Modified: dragonegg/trunk/llvm-convert.cpp URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-convert.cpp?rev=126570&r1=126569&r2=126570&view=diff ============================================================================== --- dragonegg/trunk/llvm-convert.cpp (original) +++ dragonegg/trunk/llvm-convert.cpp Sun Feb 27 07:29:05 2011 @@ -6862,8 +6862,6 @@ // side effects. bool HasSideEffects = gimple_asm_volatile_p(stmt) || (NumOutputs == 0); - std::vector CallOps; - // CallResultTypes - The inline asm call may return one or more results. The // types of the results are recorded here along with a flag indicating whether // the corresponding GCC type is signed. @@ -6876,10 +6874,15 @@ // in which case the result is converted before being stored. SmallVector, 4> CallResultDests; - SmallVector, 4> OutputLocations; + // CallOps - The operands pass to the inline asm call. + std::vector CallOps; - // ConstraintStr - The string of constraints in LLVM format. - std::string ConstraintStr; + // OutputLocations - For each output holds an index into CallOps (if the flag + // is false) or into CallResultTypes (if the flag is true). Outputs returned + // in memory are passed to the asm as an operand and thus appear in CallOps. + // Those returned in registers are obtained as one of the results of the asm + // call and thus correspond to an entry in CallResultTypes. + SmallVector, 4> OutputLocations; // SSADefinitions - If the asm defines an SSA name then the SSA name and a // memory location are recorded here. The asm result defining the SSA name @@ -6887,6 +6890,9 @@ // to define the SSA name. SmallVector, 4> SSADefinitions; + // ConstraintStr - The string of constraints in LLVM format. + std::string ConstraintStr; + // Process outputs. for (unsigned i = 0; i != NumOutputs; ++i) { tree Output = gimple_asm_output_op(stmt, i); From ofv at wanadoo.es Sun Feb 27 07:32:51 2011 From: ofv at wanadoo.es (Oscar Fuentes) Date: Sun, 27 Feb 2011 13:32:51 -0000 Subject: [llvm-commits] [llvm] r126571 - /llvm/trunk/CMakeLists.txt Message-ID: <20110227133251.210962A6C12C@llvm.org> Author: ofv Date: Sun Feb 27 07:32:50 2011 New Revision: 126571 URL: http://llvm.org/viewvc/llvm-project?rev=126571&view=rev Log: Install include/llvm/Support/LICENSE.TXT. PR9321. Modified: llvm/trunk/CMakeLists.txt Modified: llvm/trunk/CMakeLists.txt URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/CMakeLists.txt?rev=126571&r1=126570&r2=126571&view=diff ============================================================================== --- llvm/trunk/CMakeLists.txt (original) +++ llvm/trunk/CMakeLists.txt Sun Feb 27 07:32:50 2011 @@ -260,6 +260,7 @@ PATTERN "*.h" PATTERN "*.td" PATTERN "*.inc" + PATTERN "LICENSE.TXT" PATTERN ".svn" EXCLUDE ) From baldrick at free.fr Sun Feb 27 07:54:02 2011 From: baldrick at free.fr (Duncan Sands) Date: Sun, 27 Feb 2011 13:54:02 -0000 Subject: [llvm-commits] [llvm] r126573 - /llvm/trunk/docs/tutorial/LangImpl5.html Message-ID: <20110227135402.39E7B2A6C12C@llvm.org> Author: baldrick Date: Sun Feb 27 07:54:01 2011 New Revision: 126573 URL: http://llvm.org/viewvc/llvm-project?rev=126573&view=rev Log: Fix typo pointed out in pr9339. Modified: llvm/trunk/docs/tutorial/LangImpl5.html Modified: llvm/trunk/docs/tutorial/LangImpl5.html URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/tutorial/LangImpl5.html?rev=126573&r1=126572&r2=126573&view=diff ============================================================================== --- llvm/trunk/docs/tutorial/LangImpl5.html (original) +++ llvm/trunk/docs/tutorial/LangImpl5.html Sun Feb 27 07:54:01 2011 @@ -72,7 +72,7 @@

Extending Kaleidoscope to support if/then/else is quite straightforward. It -basically requires adding lexer support for this "new" concept to the lexer, +basically requires adding support for this "new" concept to the lexer, parser, AST, and LLVM code emitter. This example is nice, because it shows how easy it is to "grow" a language over time, incrementally extending it as new ideas are discovered.

From baldrick at free.fr Sun Feb 27 08:41:27 2011 From: baldrick at free.fr (Duncan Sands) Date: Sun, 27 Feb 2011 14:41:27 -0000 Subject: [llvm-commits] [llvm] r126574 - in /llvm/trunk: lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp test/CodeGen/Generic/2011-02-27-Fpextend.ll Message-ID: <20110227144127.B356F2A6C12C@llvm.org> Author: baldrick Date: Sun Feb 27 08:41:27 2011 New Revision: 126574 URL: http://llvm.org/viewvc/llvm-project?rev=126574&view=rev Log: Legalize support for fpextend of vector. PR9309. Added: llvm/trunk/test/CodeGen/Generic/2011-02-27-Fpextend.ll Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp?rev=126574&r1=126573&r2=126574&view=diff ============================================================================== --- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp (original) +++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp Sun Feb 27 08:41:27 2011 @@ -440,6 +440,7 @@ case ISD::FCEIL: case ISD::FRINT: case ISD::FNEARBYINT: + case ISD::FP_EXTEND: case ISD::FP_TO_SINT: case ISD::FP_TO_UINT: case ISD::SINT_TO_FP: @@ -2005,6 +2006,7 @@ case ISD::STORE: Res = WidenVecOp_STORE(N); break; case ISD::FP_ROUND: + case ISD::FP_EXTEND: case ISD::FP_TO_SINT: case ISD::FP_TO_UINT: case ISD::SINT_TO_FP: Added: llvm/trunk/test/CodeGen/Generic/2011-02-27-Fpextend.ll URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Generic/2011-02-27-Fpextend.ll?rev=126574&view=auto ============================================================================== --- llvm/trunk/test/CodeGen/Generic/2011-02-27-Fpextend.ll (added) +++ llvm/trunk/test/CodeGen/Generic/2011-02-27-Fpextend.ll Sun Feb 27 08:41:27 2011 @@ -0,0 +1,7 @@ +; RUN: llc < %s +; PR9309 + +define <4 x double> @f_fu(<4 x float>) nounwind { + %float2double.i = fpext <4 x float> %0 to <4 x double> + ret <4 x double> %float2double.i +} From rafael.espindola at gmail.com Sun Feb 27 10:51:18 2011 From: rafael.espindola at gmail.com (=?ISO-8859-1?Q?Rafael_=C1vila_de_Esp=EDndola?=) Date: Sun, 27 Feb 2011 11:51:18 -0500 Subject: [llvm-commits] [patch][lto] Use MC to scan for symbol definitions and uses In-Reply-To: References: <4D692112.5030707@gmail.com> Message-ID: <4D6A8106.9020809@gmail.com> On 11-02-26 6:07 PM, Daniel Dunbar wrote: > Very nice! > > Patch looks quite reasonable to me, my only concern would be whether > it should wait for after 2.9 branch. It's probably safe, but it might > be better to take it afterwards in case it needs to bake. > > Are you aware of situations the current code handles that this one doesn't? No, but it is possible that some users have dependency on the old behavior. For example, we now require that assembly being passed by libLTO be valid. With the git mirrors in place it should not be a big problem to help the patch for 10 days :-) > - Daniel > Cheers, Rafael From rafael.espindola at gmail.com Sun Feb 27 10:55:23 2011 From: rafael.espindola at gmail.com (=?ISO-8859-1?Q?Rafael_=C1vila_de_Esp=EDndola?=) Date: Sun, 27 Feb 2011 11:55:23 -0500 Subject: [llvm-commits] [patch] Switch LTO to use MC In-Reply-To: References: <4D65DCFB.90205@gmail.com> <99D6D76F-3DAD-461D-93DE-F64B2557F0E1@apple.com> <4D67237B.6060306@gmail.com> <3B27AD7C-DE66-4564-BA79-3C958D6D3E72@apple.com> <4D67D60A.9040401@gmail.com> Message-ID: <4D6A81FB.8070207@gmail.com> >> Btw, I have seen the 3.0 planned type changes mentioned some times, >> but I don't know what they are in detail. Is there a doc >> somewhere? > > I wrote them up and sent a link to llvmdev. Thanks! >> Yes, I just need a way to represent MemoryBuffers that are not NULL >> terminated :-) > > I agree that having null termination be optional would be useful. > Clang really wants this (to make its lexer faster) but I don't think > anyone else cares. I'd be fine with eliminating this requirement in > the MemBuffer API and giving MemoryBuffer::getFileOrSTDIN a flags > argument, where one flag would be "RequiresNullTerminatedBuffer" or > something like that. OK, I will give that a try. I would probably still wrap MemoryBuffer in a ZMemoryBuffer that just sets RequiresNullTerminatedBuffer to true so that people don't accidentally pass the wrong type of buffer to the lexer. > -Chris Cheers, Rafael From benny.kra at googlemail.com Sun Feb 27 12:13:53 2011 From: benny.kra at googlemail.com (Benjamin Kramer) Date: Sun, 27 Feb 2011 18:13:53 -0000 Subject: [llvm-commits] [llvm] r126578 - /llvm/trunk/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c Message-ID: <20110227181353.350092A6C12C@llvm.org> Author: d0k Date: Sun Feb 27 12:13:53 2011 New Revision: 126578 URL: http://llvm.org/viewvc/llvm-project?rev=126578&view=rev Log: Silence enum conversion warnings. Modified: llvm/trunk/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c Modified: llvm/trunk/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c?rev=126578&r1=126577&r2=126578&view=diff ============================================================================== --- llvm/trunk/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c (original) +++ llvm/trunk/lib/Target/X86/Disassembler/X86DisassemblerDecoder.c Sun Feb 27 12:13:53 2011 @@ -749,7 +749,7 @@ insn->sibIndex = SIB_INDEX_NONE; break; default: - insn->sibIndex = (EABase)(sibIndexBase + index); + insn->sibIndex = (SIBIndex)(sibIndexBase + index); if (insn->sibIndex == SIB_INDEX_sib || insn->sibIndex == SIB_INDEX_sib64) insn->sibIndex = SIB_INDEX_NONE; @@ -796,7 +796,7 @@ } break; default: - insn->sibBase = (EABase)(sibBaseBase + base); + insn->sibBase = (SIBBase)(sibBaseBase + base); break; } From clattner at apple.com Sun Feb 27 13:47:02 2011 From: clattner at apple.com (Chris Lattner) Date: Sun, 27 Feb 2011 11:47:02 -0800 Subject: [llvm-commits] [patch] Switch LTO to use MC In-Reply-To: <4D6A81FB.8070207@gmail.com> References: <4D65DCFB.90205@gmail.com> <99D6D76F-3DAD-461D-93DE-F64B2557F0E1@apple.com> <4D67237B.6060306@gmail.com> <3B27AD7C-DE66-4564-BA79-3C958D6D3E72@apple.com> <4D67D60A.9040401@gmail.com> <4D6A81FB.8070207@gmail.com> Message-ID: <4CB06DCC-CD72-4848-A270-4C89BE775D97@apple.com> On Feb 27, 2011, at 8:55 AM, Rafael ?vila de Esp?ndola wrote: >>> >>> Yes, I just need a way to represent MemoryBuffers that are not NULL >>> terminated :-) >> >> I agree that having null termination be optional would be useful. >> Clang really wants this (to make its lexer faster) but I don't think >> anyone else cares. I'd be fine with eliminating this requirement in >> the MemBuffer API and giving MemoryBuffer::getFileOrSTDIN a flags >> argument, where one flag would be "RequiresNullTerminatedBuffer" or >> something like that. > > OK, I will give that a try. I would probably still wrap MemoryBuffer in a ZMemoryBuffer that just sets RequiresNullTerminatedBuffer to true so that people don't accidentally pass the wrong type of buffer to the lexer. I think it would be perfectly fine to just have the lexer assert that it is nul terminated. The reason I mentioned adding a flag to the file open APIs is that MemoryBuffer does an optimization for the mmap case, when the file is not a multiple of a page size: it just knows the OS zero fills the rest of the mapped page. -Chris From rafael.espindola at gmail.com Sun Feb 27 14:15:37 2011 From: rafael.espindola at gmail.com (Rafael Espindola) Date: Sun, 27 Feb 2011 20:15:37 -0000 Subject: [llvm-commits] [llvm] r126579 - /llvm/trunk/tools/gold/gold-plugin.cpp Message-ID: <20110227201537.598A82A6C12C@llvm.org> Author: rafael Date: Sun Feb 27 14:15:37 2011 New Revision: 126579 URL: http://llvm.org/viewvc/llvm-project?rev=126579&view=rev Log: bfd was fixed, remove the work around. Modified: llvm/trunk/tools/gold/gold-plugin.cpp Modified: llvm/trunk/tools/gold/gold-plugin.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/gold/gold-plugin.cpp?rev=126579&r1=126578&r2=126579&view=diff ============================================================================== --- llvm/trunk/tools/gold/gold-plugin.cpp (original) +++ llvm/trunk/tools/gold/gold-plugin.cpp Sun Feb 27 14:15:37 2011 @@ -274,19 +274,8 @@ } free(buf); } else { - // FIXME: We should not need to pass -1 as the file size, but there - // is a bug in BFD that causes it to pass 0 to us. Remove this once - // that is fixed. - off_t size = file->filesize ? file->filesize : -1; - - // FIXME: We should not need to reset the position in the file, but there - // is a bug in BFD. Remove this once that is fixed. - off_t old_pos = lseek(file->fd, 0, SEEK_CUR); - lseek(file->fd, 0, SEEK_SET); - M = lto_module_create_from_fd(file->fd, file->name, size); - - lseek(file->fd, old_pos, SEEK_SET); + M = lto_module_create_from_fd(file->fd, file->name, file->filesize); if (!M) return LDPS_OK; } From rafael.espindola at gmail.com Sun Feb 27 14:30:22 2011 From: rafael.espindola at gmail.com (Rafael Espindola) Date: Sun, 27 Feb 2011 20:30:22 -0000 Subject: [llvm-commits] [llvm] r126580 - /llvm/trunk/tools/gold/gold-plugin.cpp Message-ID: <20110227203022.85A042A6C12C@llvm.org> Author: rafael Date: Sun Feb 27 14:30:22 2011 New Revision: 126580 URL: http://llvm.org/viewvc/llvm-project?rev=126580&view=rev Log: Gold now rescans archives as needed, so the pass-through options are not necessary anymore. Modified: llvm/trunk/tools/gold/gold-plugin.cpp Modified: llvm/trunk/tools/gold/gold-plugin.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/gold/gold-plugin.cpp?rev=126580&r1=126579&r2=126580&view=diff ============================================================================== --- llvm/trunk/tools/gold/gold-plugin.cpp (original) +++ llvm/trunk/tools/gold/gold-plugin.cpp Sun Feb 27 14:30:22 2011 @@ -73,7 +73,6 @@ static generate_bc generate_bc_file = BC_NO; static std::string bc_path; static std::string obj_path; - static std::vector pass_through; static std::string extra_library_path; static std::string triple; static std::string mcpu; @@ -96,9 +95,6 @@ mcpu = opt.substr(strlen("mcpu=")); } else if (opt.startswith("extra-library-path=")) { extra_library_path = opt.substr(strlen("extra_library_path=")); - } else if (opt.startswith("pass-through=")) { - llvm::StringRef item = opt.substr(strlen("pass-through=")); - pass_through.push_back(item.str()); } else if (opt.startswith("mtriple=")) { triple = opt.substr(strlen("mtriple=")); } else if (opt.startswith("obj-path=")) { @@ -489,24 +485,6 @@ return LDPS_ERR; } - for (std::vector::iterator i = options::pass_through.begin(), - e = options::pass_through.end(); - i != e; ++i) { - std::string &item = *i; - const char *item_p = item.c_str(); - if (llvm::StringRef(item).startswith("-l")) { - if (add_input_library(item_p + 2) != LDPS_OK) { - (*message)(LDPL_ERROR, "Unable to add library to the link."); - return LDPS_ERR; - } - } else { - if (add_input_file(item_p) != LDPS_OK) { - (*message)(LDPL_ERROR, "Unable to add .o file to the link."); - return LDPS_ERR; - } - } - } - if (options::obj_path.empty()) Cleanup.push_back(sys::Path(objPath)); From sabre at nondot.org Sun Feb 27 16:51:37 2011 From: sabre at nondot.org (Chris Lattner) Date: Sun, 27 Feb 2011 22:51:37 -0000 Subject: [llvm-commits] [llvm] r126590 - /llvm/trunk/include/llvm/ADT/ArrayRef.h Message-ID: <20110227225137.EE8272A6C12C@llvm.org> Author: lattner Date: Sun Feb 27 16:51:37 2011 New Revision: 126590 URL: http://llvm.org/viewvc/llvm-project?rev=126590&view=rev Log: add a data() method. Modified: llvm/trunk/include/llvm/ADT/ArrayRef.h Modified: llvm/trunk/include/llvm/ADT/ArrayRef.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/ArrayRef.h?rev=126590&r1=126589&r2=126590&view=diff ============================================================================== --- llvm/trunk/include/llvm/ADT/ArrayRef.h (original) +++ llvm/trunk/include/llvm/ADT/ArrayRef.h Sun Feb 27 16:51:37 2011 @@ -79,6 +79,8 @@ /// empty - Check if the array is empty. bool empty() const { return Length == 0; } + const T *data() const { return Data; } + /// size - Get the array size. size_t size() const { return Length; } From sabre at nondot.org Sun Feb 27 16:51:57 2011 From: sabre at nondot.org (Chris Lattner) Date: Sun, 27 Feb 2011 22:51:57 -0000 Subject: [llvm-commits] [llvm] r126591 - /llvm/trunk/include/llvm/ADT/ScopedHashTable.h Message-ID: <20110227225158.063CC2A6C12C@llvm.org> Author: lattner Date: Sun Feb 27 16:51:57 2011 New Revision: 126591 URL: http://llvm.org/viewvc/llvm-project?rev=126591&view=rev Log: add the ability to walk the scope tree and insert at not-the-current scope. Modified: llvm/trunk/include/llvm/ADT/ScopedHashTable.h Modified: llvm/trunk/include/llvm/ADT/ScopedHashTable.h URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/ScopedHashTable.h?rev=126591&r1=126590&r2=126591&view=diff ============================================================================== --- llvm/trunk/include/llvm/ADT/ScopedHashTable.h (original) +++ llvm/trunk/include/llvm/ADT/ScopedHashTable.h Sun Feb 27 16:51:57 2011 @@ -96,6 +96,9 @@ ScopedHashTableScope(ScopedHashTable &HT); ~ScopedHashTableScope(); + ScopedHashTableScope *getParentScope() { return PrevScope; } + const ScopedHashTableScope *getParentScope() const { return PrevScope; } + private: friend class ScopedHashTable; ScopedHashTableVal *getLastValInScope() { @@ -141,9 +144,14 @@ template class ScopedHashTable { +public: + /// ScopeTy - This is a helpful typedef that allows clients to get easy access + /// to the name of the scope for this hash table. + typedef ScopedHashTableScope ScopeTy; +private: typedef ScopedHashTableVal ValTy; DenseMap TopLevelMap; - ScopedHashTableScope *CurScope; + ScopeTy *CurScope; AllocatorTy Allocator; @@ -157,9 +165,6 @@ assert(CurScope == 0 && TopLevelMap.empty() && "Scope imbalance!"); } - /// ScopeTy - This is a helpful typedef that allows clients to get easy access - /// to the name of the scope for this hash table. - typedef ScopedHashTableScope ScopeTy; /// Access to the allocator. typedef typename ReferenceAdder::result AllocatorRefTy; @@ -180,13 +185,7 @@ } void insert(const K &Key, const V &Val) { - assert(CurScope && "No scope active!"); - - ScopedHashTableVal *&KeyEntry = TopLevelMap[Key]; - - KeyEntry = ValTy::Create(CurScope->getLastValInScope(), KeyEntry, Key, Val, - Allocator); - CurScope->setLastValInScope(KeyEntry); + insertIntoScope(CurScope, Key, Val); } typedef ScopedHashTableIterator iterator; @@ -199,6 +198,21 @@ if (I == TopLevelMap.end()) return end(); return iterator(I->second); } + + ScopeTy *getCurScope() { return CurScope; } + const ScopeTy *getCurScope() const { return CurScope; } + + /// insertIntoScope - This inserts the specified key/value at the specified + /// (possibly not the current) scope. While it is ok to insert into a scope + /// that isn't the current one, it isn't ok to insert *underneath* an existing + /// value of the specified key. + void insertIntoScope(ScopeTy *S, const K &Key, const V &Val) { + assert(S && "No scope active!"); + ScopedHashTableVal *&KeyEntry = TopLevelMap[Key]; + KeyEntry = ValTy::Create(S->getLastValInScope(), KeyEntry, Key, Val, + Allocator); + S->setLastValInScope(KeyEntry); + } }; /// ScopedHashTableScope ctor - Install this as the current scope for the hash From isanbard at gmail.com Sun Feb 27 19:10:44 2011 From: isanbard at gmail.com (Bill Wendling) Date: Mon, 28 Feb 2011 01:10:44 -0000 Subject: [llvm-commits] [llvm] r126604 - /llvm/trunk/docs/HowToReleaseLLVM.html Message-ID: <20110228011044.E1D5C2A6C12C@llvm.org> Author: void Date: Sun Feb 27 19:10:44 2011 New Revision: 126604 URL: http://llvm.org/viewvc/llvm-project?rev=126604&view=rev Log: Update the documentation on "How to Release LLVM". It lays out a new way of tagging and branching for the release. I will update this more throughout the 2.9 release process. Modified: llvm/trunk/docs/HowToReleaseLLVM.html Modified: llvm/trunk/docs/HowToReleaseLLVM.html URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/HowToReleaseLLVM.html?rev=126604&r1=126603&r2=126604&view=diff ============================================================================== --- llvm/trunk/docs/HowToReleaseLLVM.html (original) +++ llvm/trunk/docs/HowToReleaseLLVM.html Sun Feb 27 19:10:44 2011 @@ -17,7 +17,8 @@

Written by Tanya Lattner, Reid Spencer, - John Criswell + John Criswell, & + Bill Wendling

@@ -26,486 +27,588 @@

- This document collects information about successfully releasing LLVM - (including subprojects llvm-gcc and Clang) to the public. - It is the release manager's responsibility to ensure that a high quality - build of LLVM is released. -

+ +

This document contains information about successfully releasing LLVM — + including subprojects: e.g., llvm-gcc and clang — to + the public. It is the Release Manager's responsibility to ensure that a high + quality build of LLVM is released.

Release Timeline

LLVM is released on a time based schedule (currently every 6 months). We - do not have dot releases because of the nature of LLVM incremental - development philosophy. The release schedule is roughly as follows: -

Set code freeze and branch creation date for 6 months after last code freeze -date. Announce release schedule to the LLVM community and update the website.
Create release branch and begin release process.
Send out pre-release for first round of testing. Testing will last 7-10 days. -During the first round of testing, regressions should be found and fixed. Patches -are merged from mainline to the release branch.
Generate and send out second pre-release. Bugs found during this time will -not be fixed unless absolutely critical. Bugs introduce by patches merged in -will be fixed and if so, a 3rd round of testing is needed.
The release notes should be updated during the first and second round of -pre-release testing.
Finally, release!

LLVM is released on a time based schedule — roughly every 6 months. We + do not normally have dot releases because of the nature of LLVM's incremental + development philosophy. That said, the only thing preventing dot releases for + critical bug fixes from happening is a lack of resources — testers, + machines, time, etc. And, because of the high quality we desire for LLVM + releases, we cannot allow for a truncated form of release qualification.

+ +

The release process is roughly as follows:

+ +

Set code freeze and branch creation date for 6 months after last code + freeze date. Announce release schedule to the LLVM community and update + the website.
Create release branch and begin release process.
Send out release candidate sources for first round of testing. Testing + lasts 7-10 days. During the first round of testing, any regressions found + should be fixed. Patches are merged from mainline into the release + branch. Also, all features need to be completed during this time. Any + features not completed at the end of the first round of testing will be + removed or disabled for the release.
Generate and send out the second release candidate sources. Only + critial bugs found during this testing phase will be fixed. Any + bugs introduced by merged patches will be fixed. If so a third round of + testing is needed.
The release notes are updated.
Finally, release!

+ +

Release Process

+ +

Release Administrative Tasks

Release Administrative Tasks

Create Release Branch
Update Version Numbers

Building the Release

Building the Release

Build the LLVM Source Distributions
Build LLVM
Build the LLVM-GCC Binary Distribution
Build the Clang Binary Distribution
Target Specific Build Details

Release Qualification Criteria

Release Qualification Criteria

Qualify LLVM
Qualify LLVM-GCC
Qualify Clang
Specific Target Qualification Details

Community Testing
Release Patch Rules
Release final tasks

Community Testing
Release Patch Rules
Release final tasks

Update Documentation
Tag the LLVM Release Branch
Tag the LLVM Final Release
Update the LLVM Demo Page
Update the LLVM Website
Announce the Release

-Release Administrative Tasks

Release Administrative Tasks

-This section describes a few administrative tasks that need to be done for the -release process to begin. Specifically, it involves creating the release branch, - resetting version numbers, and creating the release tarballs for the release - team to begin testing. + +

This section describes a few administrative tasks that need to be done for + the release process to begin. Specifically, it involves:

+ +

Creating the release branch,
Setting version numbers, and
Tagging release candidates for the release team to begin testing

Create Release Branch

Branch the Subversion HEAD using the following procedure:

-
Verify that the current Subversion HEAD is in decent shape by examining - nightly tester or buildbot results.
-
Request all developers to refrain from committing. Offenders get commit - rights taken away (temporarily).
-
Create the release branch for llvm, llvm-gcc4.2, - clang, and the test-suite. The branch name will be - release_XX,where XX is the major and minor release numbers. - Clang will have a different release number than llvm/ - llvm-gcc4 since its first release was years later - (still deciding if this will be true or not). These branches - can be created without checking out anything from subversion. -
+ +
Branch the Subversion trunk using the following procedure:
+ +
1. Remind developers that the release branching is imminent and to refrain + from committing patches that might break the build. E.g., new features, + large patches for works in progress, an overhaul of the type system, an + exciting new TableGen feature, etc.
2. Verify that the current Subversion trunk is in decent shape by + examining nightly tester and buildbot results.
3. Create the release branch for llvm, llvm-gcc-4.2, + clang, and the test-suite from the last known good + revision. The branch's name is release_XY, where X is + the major and Y the minor release numbers. The branches should be + created using the following commands:
  -
  +
```
-svn copy https://llvm.org/svn/llvm-project/llvm/trunk \
-         https://llvm.org/svn/llvm-project/llvm/branches/release_XX
-svn copy https://llvm.org/svn/llvm-project/llvm-gcc-4.2/trunk \
-         https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XX
-svn copy https://llvm.org/svn/llvm-project/test-suite/trunk \
-         https://llvm.org/svn/llvm-project/test-suite/branches/release_XX
-svn copy https://llvm.org/svn/llvm-project/cfe/trunk \
-         https://llvm.org/svn/llvm-project/cfe/branches/release_XX
+$ svn copy https://llvm.org/svn/llvm-project/llvm/trunk \
+           https://llvm.org/svn/llvm-project/llvm/branches/release_XY
+
+$ svn copy https://llvm.org/svn/llvm-project/llvm-gcc-4.2/trunk \
+           https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XY
+
+$ svn copy https://llvm.org/svn/llvm-project/test-suite/trunk \
+           https://llvm.org/svn/llvm-project/test-suite/branches/release_XY
+
+$ svn copy https://llvm.org/svn/llvm-project/cfe/trunk \
+           https://llvm.org/svn/llvm-project/cfe/branches/release_XY
 
```
  -
  +
4. -
  Advise developers they can work on Subversion HEAD again.
5. -
  The Release Manager should switch to the release branch (as all changes - to the release will now be done in the branch). The easiest way to do this - is to grab another working copy using the following commands:
  +
6. Advise developers that they may now check their patches into the + Subversion tree again.
7. The Release Manager should switch to the release branch, because all + changes to the release will now be done in the branch. The easiest way to + do this is to grab a working copy using the following commands:
```
-svn co https://llvm.org/svn/llvm-project/llvm/branches/release_XX
-svn co https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XX
-svn co https://llvm.org/svn/llvm-project/test-suite/branches/release_XX
-svn co https://llvm.org/svn/llvm-project/cfe/branches/release_XX
+$ svn co https://llvm.org/svn/llvm-project/llvm/branches/release_XY llvm-X.Y
+
+$ svn co https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XY llvm-gcc-4.2-X.Y
+
+$ svn co https://llvm.org/svn/llvm-project/test-suite/branches/release_XY test-suite-X.Y
+
+$ svn co https://llvm.org/svn/llvm-project/cfe/branches/release_XY clang-X.Y
 
```
-

Update LLVM Version

- After creating the LLVM release branch, update the release branches' - autoconf/configure.ac version from X.Xsvn to just X.X. Update it on mainline - as well to be the next version (X.X+1svn). Regenerated the configure script - for both. This must be done for both llvm and the - test-suite. -

FIXME: Add a note about clang.

In addition, the version number of all the Bugzilla components must be - updated for the next release. -

+ +

After creating the LLVM release branch, update the release branches' + autoconf and configure.ac versions from 'X.Ysvn' + to 'X.Y'. Update it on mainline as well to be the next version + ('X.Y+1svn'). Regenerate the configure scripts for both + llvm and the test-suite.

+ +

In addition, the version numbers of all the Bugzilla components must be + updated for the next release.

Build the LLVM Source Distributions

Build the LLVM Release Candidates

- Create source distributions for LLVM, LLVM-GCC, - clang, and the llvm test-suite by exporting the source from - Subversion and archiving it. This can be done with the following commands: -

+ +

Create release candidates for llvm, llvm-gcc, + clang, and the LLVM test-suite by tagging the branch with + the respective release candidate number. For instance, to create Release + Candidate 1 you would issue the following commands:

-svn export https://llvm.org/svn/llvm-project/llvm/branches/release_XX llvm-X.X
-svn export https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XX llvm-gcc4.2-X.X.source
-svn export https://llvm.org/svn/llvm-project/test-suite/branches/release_XX llvm-test-X.X
-svn export https://llvm.org/svn/llvm-project/cfe/branches/release_XX clang-X.X
-tar -czvf - llvm-X.X          | gzip > llvm-X.X.tar.gz
-tar -czvf - llvm-test-X.X     | gzip > llvm-test-X.X.tar.gz
-tar -czvf - llvm-gcc4.2-X.X.source | gzip > llvm-gcc-4.2-X.X.source.tar.gz
-tar -czvf - clang-X.X | gzip > clang-X.X.tar.gz
+$ svn mkdir https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_XY
+$ svn copy https://llvm.org/svn/llvm-project/llvm/branches/release_XY \
+           https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_XY/RC1
+
+$ svn mkdir https://llvm.org/svn/llvm-project/llvm-gcc-4.2/tags/RELEASE_XY
+$ svn copy https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XY \
+           https://llvm.org/svn/llvm-project/llvm-gcc-4.2/tags/RELEASE_XY/RC1
+
+$ svn mkdir https://llvm.org/svn/llvm-project/test-suite/tags/RELEASE_XY
+$ svn copy https://llvm.org/svn/llvm-project/test-suite/branches/release_XY \
+           https://llvm.org/svn/llvm-project/test-suite/tags/RELEASE_XY/RC1
+
+$ svn mkdir https://llvm.org/svn/llvm-project/cfe/tags/RELEASE_XY
+$ svn copy https://llvm.org/svn/llvm-project/cfe/branches/release_XY \
+           https://llvm.org/svn/llvm-project/cfe/tags/RELEASE_XY/RC1

+ +

Similarly, Release Candidate 2 would be named RC2 and so + on. This keeps a permanent copy of the release candidate around for people to + export and build as they wish. The final released sources will be tagged in + the RELEASE_XY directory as Final + (c.f. Tag the LLVM Final Release).

+ +

The Release Manager may supply pre-packaged source tarballs for users. This + can be done with the following commands:

+ +

+$ svn export https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_XY/RC1 llvm-X.Yrc1
+$ svn export https://llvm.org/svn/llvm-project/llvm-gcc-4.2/tags/RELEASE_XY/RC1 llvm-gcc4.2-X.Yrc1
+$ svn export https://llvm.org/svn/llvm-project/test-suite/tags/RELEASE_XY/RC1 llvm-test-X.Yrc1
+$ svn export https://llvm.org/svn/llvm-project/cfe/tags/RELEASE_XY/RC1 clang-X.Yrc1
+
+$ tar -czvf - llvm-X.Yrc1        | gzip > llvm-X.Yrc1.src.tar.gz
+$ tar -czvf - llvm-test-X.Yrc1   | gzip > llvm-test-X.Yrc1.src.tar.gz
+$ tar -czvf - llvm-gcc4.2-X.Yrc1 | gzip > llvm-gcc-4.2-X.Yrc1.src.tar.gz
+$ tar -czvf - clang-X.Yrc1       | gzip > clang-X.Yrc1.src.tar.gz
+

-Building the Release

Building the Release

-The build of llvm, llvm-gcc, and clang must be free -of errors and warnings in both debug, release+asserts, and release builds. -If all builds are clean, then the release passes build qualification. -

debug: ENABLE_OPTIMIZED=0
release+asserts: ENABLE_OPTIMIZED=1
release: ENABLE_OPTIMIZED=1 DISABLE_ASSERTIONS=1

The builds of llvm, llvm-gcc, and clang + must be free of errors and warnings in Debug, Release+Asserts, and + Release builds. If all builds are clean, then the release passes Build + Qualification.

+ +

The make options for building the different modes:

+ + + + + + +

Mode	Options
Debug	`ENABLE_OPTIMIZED=0`
Release+Asserts	`ENABLE_OPTIMIZED=1`
Release	`ENABLE_OPTIMIZED=1 DISABLE_ASSERTIONS=1`

Build LLVM

- Build both debug, release+asserts (optimized), and release versions of - LLVM on all supported platforms. Direction to build llvm are - here. -

+ +

Build Debug, Release+Asserts, and Release versions + of llvm on all supported platforms. Directions to build + llvm are + here.

Build the LLVM GCC Binary Distribution

- Creating the LLVM GCC binary distribution (release/optimized) requires - performing the following steps for each supported platform: -

- Build the LLVM GCC front-end by following the directions in the README.LLVM - file. The frontend must be compiled with c, c++, objc (mac only), - objc++ (mac only) and fortran support.
Please boostrap as well.
Be sure to build with LLVM_VERSION_INFO=X.X, where X is the major and - minor release numbers. -
- Copy the installation directory to a directory named for the specific target. - For example on Red Hat Enterprise Linux, the directory would be named - llvm-gcc4.2-2.6-x86-linux-RHEL4. Archive and compress the new directory. -

Creating the llvm-gcc binary distribution (Release/Optimized) + requires performing the following steps for each supported platform:

+ +

Build the llvm-gcc front-end by following the directions in + the README.LLVM file. The front-end must be compiled with C, C++, + Objective-C (Mac only), Objective-C++ (Mac only), and Fortran + support.
Boostrapping must be enabled.
Be sure to build with LLVM_VERSION_INFO=X.Y, where X + is the major and Y is the minor release numbers.
Copy the installation directory to a directory named for the specific + target. For example on Red Hat Enterprise Linux, the directory would be + named llvm-gcc4.2-2.6-x86-linux-RHEL4. Archive and compress the + new directory.

Build Clang -Binary Distribution

Build Clang Binary Distribution

- Creating the Clang binary distribution (debug/release/release) requires - performing the following steps for each supported platform: -

- Build clang according to the directions - here. -
Build both a debug and release version of clang, but the binary - will be a release build.
- Package clang (details to follow). -

Creating the clang binary distribution + (Debug/Release+Asserts/Release) requires performing the following steps for + each supported platform:

+ +

Build clang according to the directions + here.
Build both a debug and release version of clang. The binary will be the + release build.
Package clang (details to follow).

+ -

Target Specific Build -Details

Target Specific Build Details

- The table below specifies which compilers are used for each arch/os combination - when qualifying the build of llvm, llvm-gcc, clang. -

- -

- + +
The table below specifies which compilers are used for each Arch/OS + combination when qualifying the build of llvm, llvm-gcc, + and clang.
+ +
- + - -

Architecture OS compiler

x86-32 Mac OS 10.5 gcc 4.0.1

x86-32 Linux gcc 4.2.X, gcc 4.3.X

x86-32 FreeBSD gcc 4.2.X
x86-32 mingw gcc 3.4.5
x86-32 mingw gcc 3.4.5

x86-64 Mac OS 10.5 gcc 4.0.1

x86-64 Linux gcc 4.2.X, gcc 4.3.X

x86-64 FreeBSD gcc 4.2.X
-

Architecture	OS	compiler
x86-32	Mac OS 10.5	gcc 4.0.1
x86-32	Linux	gcc 4.2.X, gcc 4.3.X
x86-32	FreeBSD	gcc 4.2.X
x86-32	mingw	gcc 3.4.5
x86-32	mingw	gcc 3.4.5
x86-64	Mac OS 10.5	gcc 4.0.1
x86-64	Linux	gcc 4.2.X, gcc 4.3.X
x86-64	FreeBSD	gcc 4.2.X

Building the Release

- A release is qualified when it has no regressions from the previous - release (or baseline). Regressions are related to correctness only and not - performance at this time. Regressions are new failures in the set of tests that - are used to qualify each product and only include things on the list. - Ultimately, there is no end to the number of possible bugs in a release. We - need a very concrete and definitive release criteria that ensures we have - monotonically improving quality on some metric. The metric we use is - described below. This doesn't mean that we don't care about other things, - but this are things that must be satisfied before a release can go out -

A release is qualified when it has no regressions from the previous release + (or baseline). Regressions are related to correctness first and performance + second. (We may tolerate some minor performance regressions if they are + deemed necessary for the general quality of the compiler.)

+ +

Regressions are new failures in the set of tests that are used to qualify + each product and only include things on the list. Every release will have + some bugs in it. It is the reality of developing a complex piece of + software. We need a very concrete and definitive release criteria that + ensures we have monotonically improving quality on some metric. The metric we + use is described below. This doesn't mean that we don't care about other + criteria, but these are the criteria which we found to be most important and + which must be satisfied before a release can go out

+ +

Qualify LLVM

- LLVM is qualified when it has a clean dejagnu test run without a frontend and - it has no regressions when using either llvm-gcc or clang - with the test-suite from the previous release. -

+ +

LLVM is qualified when it has a clean test run without a front-end. And it + has no regressions when using either llvm-gcc or clang with + the test-suite from the previous release.

Qualify LLVM-GCC

- LLVM-GCC is qualified when front-end specific tests in the - llvm dejagnu test suite all pass and there are no regressions in - the test-suite.

We do not use the gcc dejagnu test suite as release criteria.

+ +

LLVM-GCC is qualified when front-end specific tests in the + llvm regression test suite all pass and there are no regressions in + the test-suite.

+ +

We do not use the GCC DejaGNU test suite as release criteria.

Qualify Clang

- Clang is qualified when front-end specific tests in the - llvm dejagnu test suite all pass, clang's own test suite passes - cleanly, and there are no regressions in the test-suite.

+ +

Clang is qualified when front-end specific tests in the + llvm dejagnu test suite all pass, clang's own test suite passes + cleanly, and there are no regressions in the test-suite.

Specific Target Qualification Details

- + +
Architecture OS llvm-gcc baseline clang baseline - tests
+ -
Architecture OS llvm-gcc baseline clang baseline tests

x86-32 Linux last release last release llvm dejagnu, clang tests, test-suite (including spec)

x86-32 FreeBSD none last release llvm dejagnu, clang tests, test-suite

x86-32 mingw last release none QT

x86-64 Mac OS 10.X last release last release llvm dejagnu, clang tests, test-suite (including spec)

x86-64 Linux last release last release llvm dejagnu, clang tests, test-suite (including spec)

x86-64 FreeBSD none last release llvm dejagnu, clang tests, test-suite

+ +

Architecture	OS	llvm-gcc baseline	clang baseline	tests
x86-32	Linux	last release	last release	llvm dejagnu, clang tests, test-suite (including spec)
x86-32	FreeBSD	none	last release	llvm dejagnu, clang tests, test-suite
x86-32	mingw	last release	none	QT
x86-64	Mac OS 10.X	last release	last release	llvm dejagnu, clang tests, test-suite (including spec)
x86-64	Linux	last release	last release	llvm dejagnu, clang tests, test-suite (including spec)
x86-64	FreeBSD	none	last release	llvm dejagnu, clang tests, test-suite

Community Testing

- Once all testing has been completed and appropriate bugs filed, the pre-release - tar balls may be put on the website and the LLVM community is notified. Ask that - all LLVM developers test the release in 2 ways:

Download llvm-X.X, llvm-test-X.X, and the appropriate llvm-gcc4 - and/or clang binary. Build LLVM. - Run "make check" and the full llvm-test suite (make TEST=nightly report).
Download llvm-X.X, llvm-test-X.X, and the llvm-gcc4 and/or clang source. - Compile everything. Run "make check" and the full llvm-test suite (make TEST=nightly - report).

Ask LLVM developers to submit the report and make check results to the list. - Attempt to verify that there are no regressions from the previous release. - The results are not used to qualify a release, but to spot other potential - problems. For unsupported targets, verify that make check at least is - clean.

+ +

Once all testing has been completed and appropriate bugs filed, the release + candidate tarballs are put on the website and the LLVM community is + notified. Ask that all LLVM developers test the release in 2 ways:

+ +

Download llvm-X.Y, llvm-test-X.Y, and the + appropriate llvm-gcc and/or clang binary. Build + LLVM. Run make check and the full LLVM test suite (make + TEST=nightly report).
Download llvm-X.Y, llvm-test-X.Y, and the + llvm-gcc and/or clang source. Compile everything. Run + make check and the full LLVM test suite (make TEST=nightly + report).

+ +

Ask LLVM developers to submit the test suite report and make check + results to the list. Verify that there are no regressions from the previous + release. The results are not used to qualify a release, but to spot other + potential problems. For unsupported targets, verify that make check + is at least clean.

During the first round of testing time, - all regressions must be fixed before the second pre-release is created.

During the first round of testing, all regressions must be fixed before the + second release candidate is tagged.

If this is the second round of testing, this is only to ensure the bug - fixes previously merged in have not created new major problems. This is not - the time to solve additional and unrelated bugs. If no patches are merged in, - the release is determined to be ready and the release manager may move onto - the next step. -

If this is the second round of testing, the testing is only to ensure that + bug fixes previously merged in have not created new major problems. This + is not the time to solve additional and unrelated bugs! If no patches are + merged in, the release is determined to be ready and the release manager may + move onto the next stage.

Release Patch Rules -

Release Patch Rules

- Below are the rules regarding patching the release branch.

Patches applied to the release branch are only applied by the release - manager.

During the first round of testing, patches that fix regressions or that - are small and relatively risk free (verified by the appropriate code owner) - are applied to the branch. Code owners are asked to be very conservative in - approving patches for the branch and we reserve the right to reject any patch - that does not fix a regression as previously defined.

During the remaining rounds of testing, only patches that fix regressions - may be applied.

- -

Below are the rules regarding patching the release branch:

+ +

Patches applied to the release branch may only be applied by the + release manager.
During the first round of testing, patches that fix regressions or that + are small and relatively risk free (verified by the appropriate code + owner) are applied to the branch. Code owners are asked to be very + conservative in approving patches for the branch. We reserve the right to + reject any patch that does not fix a regression as previously + defined.
During the remaining rounds of testing, only patches that fix critical + regressions may be applied.

+ +

Release Final Tasks

- The final stages of the release process involving tagging the release branch, - updating documentation that refers to the release, and updating the demo - page.

FIXME: Add a note if anything needs to be done to the clang website. - Eventually the websites will be merged hopefully.

+ +

The final stages of the release process involves tagging the "final" release + branch, updating documentation that refers to the release, and updating the + demo page.

Update Documentation

- Review the documentation and ensure that it is up to date. The Release Notes - must be updated to reflect bug fixes, new known issues, and changes in the - list of supported platforms. The Getting Started Guide should be updated to - reflect the new release version number tag avaiable from Subversion and - changes in basic system requirements. Merge both changes from mainline into - the release branch. -

+ +

Review the documentation and ensure that it is up to date. The "Release + Notes" must be updated to reflect new features, bug fixes, new known issues, + and changes in the list of supported platforms. The "Getting Started Guide" + should be updated to reflect the new release version number tag avaiable from + Subversion and changes in basic system requirements. Merge both changes from + mainline into the release branch.

Tag the Release Branch

Tag the LLVM Final Release

Tag the release branch using the following procedure:

+ +

Tag the final release sources using the following procedure:

-svn copy https://llvm.org/svn/llvm-project/llvm/branches/release_XX \
-         https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_XX
-svn copy https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XX \
-         https://llvm.org/svn/llvm-project/llvm-gcc-4.2/tags/RELEASE_XX
-svn copy https://llvm.org/svn/llvm-project/test-suite/branches/release_XX \
-         https://llvm.org/svn/llvm-project/test-suite/tags/RELEASE_XX
+$ svn copy https://llvm.org/svn/llvm-project/llvm/branches/release_XY \
+           https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_XY/Final
+
+$ svn copy https://llvm.org/svn/llvm-project/llvm-gcc-4.2/branches/release_XY \
+           https://llvm.org/svn/llvm-project/llvm-gcc-4.2/tags/RELEASE_XY/Final
+
+$ svn copy https://llvm.org/svn/llvm-project/test-suite/branches/release_XY \
+           https://llvm.org/svn/llvm-project/test-suite/tags/RELEASE_XY/Final
+
+$ svn copy https://llvm.org/svn/llvm-project/cfe/branches/release_XY \
+           https://llvm.org/svn/llvm-project/cfe/tags/RELEASE_XY/Final

- +

Update the LLVM Demo Page

- The LLVM demo page must be updated to use the new release. This consists of - using the llvm-gcc binary and building LLVM. Update the website demo page - configuration to use the new release.

+ +

The LLVM demo page must be updated to use the new release. This consists of + using the new llvm-gcc binary and building LLVM.

Update the LLVM Website

- The website must be updated before the release announcement is sent out. Here is - what to do:

Check out the website module from CVS.
Create a new subdirectory X.X in the releases directory.
Commit the llvm, test-suite, llvm-gcc source, - clang source, clang binaries, - and llvm-gcc binaries in this new directory.
Copy and commit the llvm/docs and LICENSE.txt - files into this new directory. The docs should be built with BUILD_FOR_WEBSITE=1.
Commit the index.html to the release/X.X directory to redirect (use from previous - release.
Update the releases/download.html file with the new release.
Update the releases/index.html with the new release and link to - release documentation.
Finally, update the main page (index.html and sidebar) to - point to the new release and release announcement. Make sure this all gets - committed back into Subversion.

+ +

The website must be updated before the release announcement is sent out. Here + is what to do:

+ +

Check out the www module from Subversion.
Create a new subdirectory X.Y in the releases directory.
Commit the llvm, test-suite, llvm-gcc source, + clang source, clang binaries, and llvm-gcc + binaries in this new directory.
Copy and commit the llvm/docs and LICENSE.txt files + into this new directory. The docs should be built with + BUILD_FOR_WEBSITE=1.
Commit the index.html to the release/X.Y directory to + redirect (use from previous release.
Update the releases/download.html file with the new release.
Update the releases/index.html with the new release and link to + release documentation.
Finally, update the main page (index.html and sidebar) to point + to the new release and release announcement. Make sure this all gets + committed back into Subversion.

Announce the Release

Have Chris send out the release announcement when everything is finished.

+ +

Have Chris send out the release announcement when everything is finished.

From nicholas at mxc.ca Sun Feb 27 21:01:47 2011 From: nicholas at mxc.ca (Nick Lewycky) Date: Sun, 27 Feb 2011 19:01:47 -0800 Subject: [llvm-commits] PATCH: instcombine switch on select of constants to br In-Reply-To: References: <893CFB3C-832C-4959-AE72-4186FCD61E99@gmail.com> <4D2A730F.2090207@mxc.ca> <130DB463-EF73-40A2-8166-5D3A014D9D12@gmail.com>

Message-ID: <4D6B101B.8070703@mxc.ca> Frits van Bommel wrote: > Since there has been no progress here for over six weeks, I updated > this patch myself. > Does anyone mind if I commit this version? I kinda wish it didn't work this way. Here's what I have in mind: 1. fix llvm.org/PR774 (make the default case optional) 2. one optimization which finds switches that have two destinations and turns them into branches 3. one optimization which sees select (and other things) feeding into a switch and removes the impossible cases from the switch, possibly including the default case. That said, if I can't talk you into doing that much work, then this patch looks fine: @@ -2357,10 +2382,10 @@ return true; } - if (SelectInst *SI = dyn_cast(IBI->getAddress())) { + if (SelectInst *SI = dyn_cast(IBI->getAddress())) if (SimplifyIndirectBrOnSelect(IBI, SI)) return SimplifyCFG(BB) | true; - } + return Changed; } You didn't change any behaviour here? Just leave that out of the commit. Nick From geek4civic at gmail.com Sun Feb 27 22:51:57 2011 From: geek4civic at gmail.com (NAKAMURA Takumi) Date: Mon, 28 Feb 2011 13:51:57 +0900 Subject: [llvm-commits] [PATCH] Don't install libUnitTestMain.a. It might be useless without gtest headers. Message-ID: --- utils/unittest/UnitTestMain/Makefile | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-) ...Takumi -------------- next part -------------- A non-text attachment was scrubbed... Name: 0001-Don-t-install-libUnitTestMain.a.-It-might-be-use.patch.txt Type: text/x-patch Size: 359 bytes Desc: not available Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110228/ee874332/attachment.bin From geek4civic at gmail.com Sun Feb 27 22:53:07 2011 From: geek4civic at gmail.com (NAKAMURA Takumi) Date: Mon, 28 Feb 2011 13:53:07 +0900 Subject: [llvm-commits] [PATCH] Makefile.rules: [PR9321] "make install" may install include files along explicit pattern. Message-ID: --- Makefile.rules | 23 ++++++++++++++++++++--- 1 files changed, 20 insertions(+), 3 deletions(-) I prefer adding explicit patterns rather than adding exclusive patterns, to suppress installing *.cmake. ...Takumi -------------- next part -------------- A non-text attachment was scrubbed... Name: 0002-Makefile.rules-PR9321-make-install-may-install-i.patch.txt Type: text/x-patch Size: 1592 bytes Desc: not available Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20110228/9b415fb5/attachment.bin From geek4civic at gmail.com Sun Feb 27 23:18:07 2011 From: geek4civic at gmail.com (NAKAMURA Takumi) Date: Mon, 28 Feb 2011 05:18:07 -0000 Subject: [llvm-commits] [llvm] r126632 - /llvm/trunk/utils/unittest/UnitTestMain/Makefile Message-ID: <20110228051807.2DA8D2A6C12C@llvm.org> Author: chapuni Date: Sun Feb 27 23:18:07 2011 New Revision: 126632 URL: http://llvm.org/viewvc/llvm-project?rev=126632&view=rev Log: Don't install libUnitTestMain.a. It might be useless without gtest headers. Modified: llvm/trunk/utils/unittest/UnitTestMain/Makefile Modified: llvm/trunk/utils/unittest/UnitTestMain/Makefile URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/unittest/UnitTestMain/Makefile?rev=126632&r1=126631&r2=126632&view=diff ============================================================================== --- llvm/trunk/utils/unittest/UnitTestMain/Makefile (original) +++ llvm/trunk/utils/unittest/UnitTestMain/Makefile Sun Feb 27 23:18:07 2011 @@ -27,4 +27,6 @@ CPP.Flags += -DGTEST_HAS_PTHREAD=0 endif +NO_INSTALL = 1 + include $(LEVEL)/Makefile.common From geek4civic at gmail.com Sun Feb 27 23:22:38 2011 From: geek4civic at gmail.com (NAKAMURA Takumi) Date: Mon, 28 Feb 2011 14:22:38 +0900 Subject: [llvm-commits] [PATCH] Don't install libUnitTestMain.a. It might be useless without gtest headers. In-Reply-To: References: Message-ID: Committed in r126632, thank you Chandler! ...Takumi

Arguments:

Semantics:

Example:

Arguments:

Semantics:

Build Chart