> Do you have a small test case for this instead of running the entire > test suite? I'm not sure how much time the full test suite takes.
Don't have a small test case, but I do have a big one. I've been verifying that the patch helps in cross-compiles from opensuse where it's a lot faster to build a patched gcc (and the test case) than in cygwin. Large test case (takes about 15 minutes on my year-old laptop): curl -L http://llvm.org/releases/3.7.1/llvm-3.7.1.src.tar.xz | tar -xJf - mkdir -p llvmbuild/bin cd llvmbuild echo 'set(CMAKE_C_COMPILER gcc)' > NATIVE.cmake echo 'set(CMAKE_CXX_COMPILER g++)' >> NATIVE.cmake cp /usr/i686-w64-mingw32/sys-root/mingw/bin/*.dll bin cmake ../llvm-3.7.1.src -DCMAKE_SYSTEM_NAME=Windows \ -DCMAKE_BUILD_TYPE=Release -DLLVM_TARGETS_TO_BUILD=host \ -DCMAKE_C_COMPILER=i686-w64-mingw32-gcc \ -DCMAKE_CXX_COMPILER=i686-w64-mingw32-g++ \ -DCROSS_TOOLCHAIN_FLAGS_NATIVE=-DCMAKE_TOOLCHAIN_FILE=$PWD/NATIVE.cmake make -j`nproc` opt bin/opt.exe -slp-vectorizer \ -S ../llvm-3.7.1.src/test/Transforms/SLPVectorizer/X86/vector.ll That'll give the following sane LLVM IR output on gcc 4.9, or crash on unpatched 5+. ; ModuleID = '../llvm-3.7.1.src/test/Transforms/SLPVectorizer/X86/vector.ll' target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target triple = "x86_64-apple-macosx10.8.0" define void @test(<4 x i32> %in, <4 x i32> %in2) { %k = icmp eq <4 x i32> %in, %in2 ret void } > I'll to include the update in the next 5.x release. Thanks! We're mainly trying to double-check that the patch doesn't introduce any regressions, cause more problems than it solves. -Tony -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple