search for: dagiselemitt

...lvm-2.9/utils/TableGen/DAGISelMatch > erOpt.cpp:267 > #1063 0x000000000049172a in llvm::OptimizeMatcher (TheMatcher=0x689a5f0, > CGP=...) > at > C:/MinGW/msys/1.0/home/xchen/llvm2.9/llvm-2.9/utils/TableGen/DAGISelMatch > erOpt.cpp:512 > #1064 0x0000000000480e3a in llvm::DAGISelEmitter::run (this=0x22f760, > OS=...) > at > C:/MinGW/msys/1.0/home/xchen/llvm2.9/llvm-2.9/utils/TableGen/DAGISelEmitt > er.cpp:155 > #1065 0x0000000000526188 in main (argc=11, argv=0x322520) > at > C:/MinGW/msys/1.0/home/xchen/llvm2.9/llvm-2.9/utils/TableGen/TableGen.cpp &g...

...orming TableGenStep" > phase). > > Any help/pointers would be appreciated. It breaks for me because the usage of INT64_C, which was introduced on r57663 and 57668. Some googling around indicates that, on Windows, one should #define INT64_C(val) val##i64 It is present on TableGen/DAGISelEmitter.cpp and TableGen/Record.cpp. -- Oscar

...(vector_shuffle > (v8f32 undef), > (v8f32 SRCREGCLASS:$src1), > VEXTRACTF128_shuffle_mask:$src2), > x86_subreg_128bit)))], Well, it's conflicting with the hard-coded case statement from DAGISelEmitter.cpp. What's the best way to resolve this? Introduce another DAG operator that means the same thing as ISD::EXTRACT_SUBREG but that can be used as the top-level operator in a pattern? -Dave

Allow me to rephrase my question: How much agony and gnashing of teeth will I cause if I commit this patch to tblgen (fully tested and changes to DAGISelEmitter.cpp also committed)? -scooter On Feb 8, 2008, at 6:43 PM, Scott Michel wrote: > Question: How hard should tblgen try to fit constants into a > particular > type? > > My case is an xor with i8 immediate where I'm using 0xff in as the > immediate value. 0xff should fit i...

...eGen/DAGISelMatcherGen.cpp:559:0 #21 0x00000000005b8f4a llvm::ConvertPatternToMatcher(llvm::PatternToMatch const&, unsigned int, llvm::CodeGenDAGPatterns const&) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelMatcherGen.cpp:1026:0 #22 0x000000000059db91 (anonymous namespace)::DAGISelEmitter::run(llvm::raw_ostream&) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:169:0 #23 0x000000000059dd1b llvm::EmitDAGISel(llvm::RecordKeeper&, llvm::raw_ostream&) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:187:0 #24 0x00000000007357d1...

Folks, It appears the Windows build has regressed over the past week. The build fails quite early (during the "Performing TableGenStep" phase). Any help/pointers would be appreciated. Thanks, snaroff (a clang developer)

I was building llvm2.9 using MinGW64 on windows, msys was 32 bit so I specified --host option for a cross compiling. Following are my configure options: ../llvm2.9/configure --prefix=/home/AutoESL/llvm-obj --host=x86_64-w64-mingw32 --disable-multilib The error: make[1]: Entering directory `/home/AutoESL/llvm-obj/lib/VMCore' make[1]: ***

...n::InferAllTypes (this=0x7c0fb6c0) at CodeGenDAGPatterns.cpp:1182 #7 0x1c0d66ac in llvm::CodeGenDAGPatterns::ParsePatternFragments (this=0xcfbd1330) at CodeGenDAGPatterns.cpp:1368 #8 0x1c0d47b9 in CodeGenDAGPatterns (this=0xcfbd1330, R=@0x3c04465c) at CodeGenDAGPatterns.cpp:1225 #9 0x1c174c65 in DAGISelEmitter (this=0xcfbd1328, R=@0x3c04465c) at DAGISelEmitter.h:30 #10 0x1c173e40 in main (argc=11, argv=0xcfbd15a4) at TableGen.cpp:178 (I believe Andrew added a line to insert a breakpoint for gdb in the above backtrace) gmake[3]: Entering directory `/usr/ports/devel/llvm/w-llvm-2.3/llvm-2.3/lib/Targe...

...elMatcherGen.cpp:559:0 #21 0x00000000005b8f4a llvm::ConvertPatternToMatcher(llvm::PatternToMatch const&, unsigned int, llvm::CodeGenDAGPatterns const&) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelMatcherGen.cpp:1026:0 #22 0x000000000059db91 (anonymous namespace)::DAGISelEmitter::run(llvm::raw_ostream&) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:169:0 #23 0x000000000059dd1b llvm::EmitDAGISel(llvm::RecordKeeper&, llvm::raw_ostream&) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:187:0 #24 0x000000...

...vm/trunk_nolibtool/obj/utils/TableGen/Debug/CodeGenDAGPatterns.o /Users/jyasskin/src/llvm/trunk_nolibtool/obj/utils/TableGen/Debug/CodeGenInstruction.o /Users/jyasskin/src/llvm/trunk_nolibtool/obj/utils/TableGen/Debug/CodeGenTarget.o /Users/jyasskin/src/llvm/trunk_nolibtool/obj/utils/TableGen/Debug/DAGISelEmitter.o /Users/jyasskin/src/llvm/trunk_nolibtool/obj/utils/TableGen/Debug/FastISelEmitter.o /Users/jyasskin/src/llvm/trunk_nolibtool/obj/utils/TableGen/Debug/InstrEnumEmitter.o /Users/jyasskin/src/llvm/trunk_nolibtool/obj/utils/TableGen/Debug/InstrInfoEmitter.o /Users/jyasskin/src/llvm/trunk_nolibtool/...

On the trunk version of LLVM I am getting the following build error on Snow Leopard (v10.6.2): llvm[3]: Building X86.td DAG instruction selector implementation with tblgen Assertion failed: (LHS->ID != RHS->ID), function operator(), file /Users/peckw/Projects/llvm/llvm-pristine/utils/TableGen/DAGISelEmitter.cpp, line 183. ... <cut - see full_stack_trace.txt> 0. Program arguments: /Users/peckw/Projects/llvm/build/pristine/Debug+Checks/bin/tblgen -I /Users/peckw/Projects/llvm/llvm-pristine/lib/Target/X86 -I /Users/peckw/Projects/llvm/llvm-pristine/include -I /Users/peckw/Projects/llvm/llvm-pri...

Question: How hard should tblgen try to fit constants into a particular type? My case is an xor with i8 immediate where I'm using 0xff in as the immediate value. 0xff should fit into i8 if treated as unsigned, but CodeGenDAGPatterns.cpp assumes that any and all integers less than 32-bits are signed. Should tblgen try to see if the sign-extended version of the constant could fit into the

...elMatcherGen.cpp:559:0 #21 0x00000000005b8f4a llvm::ConvertPatternToMatcher(llvm::PatternToMatch const&amp;, unsigned int, llvm::CodeGenDAGPatterns const&amp;) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelMatcherGen.cpp:1026:0 #22 0x000000000059db91 (anonymous namespace)::DAGISelEmitter::run(llvm::raw_ostream&amp;) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:169:0 #23 0x000000000059dd1b llvm::EmitDAGISel(llvm::RecordKeeper&amp;, llvm::raw_ostream&amp;) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:187:0 #24 0x000000...

...o CMakeFiles/obj.llvm-tblgen.dir/CodeGenInstruction.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenMapTable.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenRegisters.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenSchedule.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenTarget.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelEmitter.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherEmitter.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherGen.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherOpt.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcher.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DFAPacketizerEmitter.cpp.o CMakeFiles/o...

...o CMakeFiles/obj.llvm-tblgen.dir/CodeGenInstruction.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenMapTable.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenRegisters.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenSchedule.cpp.o CMakeFiles/obj.llvm-tblgen.dir/CodeGenTarget.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelEmitter.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherEmitter.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherGen.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherOpt.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcher.cpp.o CMakeFiles/obj.llvm-tblgen.dir/DFAPacketizerEmitter.cpp.o CMakeFiles/o...

...elMatcherGen.cpp:559:0 #21 0x00000000005b8f4a llvm::ConvertPatternToMatcher(llvm::PatternToMatch const&amp;, unsigned int, llvm::CodeGenDAGPatterns const&amp;) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelMatcherGen.cpp:1026:0 #22 0x000000000059db91 (anonymous namespace)::DAGISelEmitter::run(llvm::raw_ostream&amp;) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:169:0 #23 0x000000000059dd1b llvm::EmitDAGISel(llvm::RecordKeeper&amp;, llvm::raw_ostream&amp;) /home/nancy/work/rpp_clang/llvm/utils/TableGen/DAGISelEmitter.cpp:187:0 #24 0x000000...

For those of you who haven't noticed, I'm planning to commit a major change to the Makefile rules tomorrow evening (Tuesday) if there are no complaints about it between now and then. This needs testing on Darwin. I've heard back from Linux on many platforms and even FreeBSD, which is fantastic, but I'm told that lots of LLVMers are running Darwin and I haven't heard a peep

...;-regalloc=local/fast" (roughly 2x slower for some of my benchmarks) is, unfortunately, not an acceptable option. I noticed that a semi-recent commit added the "STATISTIC(NumDAGIselRetries,"Number of times dag isel has to try another path");" as well as a sorting change in DAGISelEmitter.cpp. This suggests to me that there is work being done. Are there other similar efforts, but perhaps aiming for bigger gains, that I should be aware of? From the other end, are there efforts to make the fast-path produce better quality code? Some small things I've noticed: - LegalizeVector...

...uction.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenMapTable.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenRegisters.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenSchedule.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenTarget.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelEmitter.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherEmitter.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherGen.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherOpt.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcher.cpp.o >> CMakeFiles/obj.llvm-tblge...

...uction.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenMapTable.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenRegisters.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenSchedule.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/CodeGenTarget.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelEmitter.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherEmitter.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherGen.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcherOpt.cpp.o >> CMakeFiles/obj.llvm-tblgen.dir/DAGISelMatcher.cpp.o >> CMakeFiles/obj.llvm-tblge...