search for: tmp56

...nwind > %tmp45.i = fmul float %tmp33.i, %tmp33.i > %tmp47.i = fdiv float %tmp45.i, 2.000000e+00 > %tmp48.i = fadd float %tmp27.i, %tmp47.i > %tmp50.i = fmul float %tmp48.i, %tmp21.i > %tmp51.i = fadd float %call41.i, %tmp50.i > %tmp53.i = fdiv float %tmp51.i, %tmp37.i > %tmp56.i = fsub float %tmp53.i, %tmp37.i > %tmp59.i = fsub float -0.000000e+00, %tmp27.i > %tmp61.i = fmul float %tmp21.i, %tmp59.i > %call62.i = tail call float @llvm.exp.f32(float %tmp61.i) nounwind > %tmp63.i = fmul float %tmp9.i, %call62.i > %call.i.i = tail call float @fabs(float...

Hello everybody, I have stumbled upon a test case (the attached module is a slightly reduced version) that shows extremely reduced performance on linux compared to windows when executed using LLVM's JIT. We narrowed the problem down to the actual code being generated, the source IR on both systems is the same. Try compiling the attached module: llc -O3 -filetype=asm -o BAD.s BAD.ll Under

...nts/20070711/63f1e212/attachment.html> -------------- next part -------------- llc -debug -print-machineinstrs -soft-float -disable-fp-elim -march=mips sieve.bc -f -o llvm_sieve.s MERGING MOSTLY EMPTY BLOCKS - BEFORE: bb54: ; preds = %bb49 br label %bb55 bb55: ; preds = %bb54, %cond_next %tmp56 = load i32* %NUM ; <i32> [#uses=1] %tmp57 = sub i32 %tmp56, 1 ; <i32> [#uses=1] store i32 %tmp57, i32* %NUM %tmp58 = load i32* %NUM ; <i32> [#uses=1] %tmp59 = icmp ne i32 %tmp58, -1 ; <i1> [#uses=1] %tmp5960 = zext i1 %tmp59 to i8 ; <i8> [#uses=1] %toBool61...

...p1, i8** %saved_stack.9, align 4 %tmp2 = load i32* %n_addr, align 4 ; <i32> [#uses=1] store i32 %tmp2, i32* %n.0, align 4 %tmp3 = load i32* %n.0, align 4 ; <i32> [#uses=1] %tmp4 = sub i32 %tmp3, 1 ; <i32> [#uses=0] %tmp5 = load i32* %n.0, align 4 ; <i32> [#uses=1] %tmp56 = zext i32 %tmp5 to i64 ; <i64> [#uses=1] %mask = and i64 %tmp56, 68719476735 ; <i64> [#uses=2] %mask7 = and i64 %mask, 68719476735 ; <i64> [#uses=0] %mask8 = and i64 %mask, 68719476735 ; <i64> [#uses=1] %tmp9 = mul i64 %mask8, 8 ; <i64> [#uses=1] %mask10 = a...

...ringNotEqualZero13: br label %l12 StringEqualZero13: call i32 @yyText(i32 %yybegin, i32 %yyend) %tmp54 = ptrtoint i8* @str0 to i32 %tmp55 = icmp eq i32 %tmp0, 0 br i1 %tmp55, label %PredicateEqualZero0, label %PredicateNotEqualZero0 PredicateEqualZero0: br label %l12 PredicateNotEqualZero0: %tmp56 = call i32 @yymatchChar(i8 105) %tmp57 = icmp ne i32 %tmp105, 0 br i1 %tmp57, label %StringNotEqualZero14, label %StringEqualZero14 StringNotEqualZero14: br label %l12 StringEqualZero14: call i32 @yyText(i32 %yybegin, i32 %yyend) %tmp58 = ptrtoint i8* @str1 to i32 %tmp59 = icmp eq i32 %tmp1,...