diff options
Diffstat (limited to 'llvm/test/Transforms')
5 files changed, 146 insertions, 114 deletions
diff --git a/llvm/test/Transforms/CodeGenPrepare/ARM/bitreverse-recognize.ll b/llvm/test/Transforms/CodeGenPrepare/ARM/bitreverse-recognize.ll new file mode 100644 index 00000000000..36440da2162 --- /dev/null +++ b/llvm/test/Transforms/CodeGenPrepare/ARM/bitreverse-recognize.ll @@ -0,0 +1,37 @@ +; RUN: opt -S -loop-unroll -codegenprepare < %s | FileCheck %s + +target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64" +target triple = "armv7--linux-gnueabihf" + +; CHECK-LABEL: @f +define i32 @f(i32 %a) #0 { +; CHECK: call i32 @llvm.bitreverse.i32 +entry: + br label %for.body + +for.cond.cleanup: ; preds = %for.body + ret i32 %or + +for.body: ; preds = %for.body, %entry + %i.08 = phi i32 [ 0, %entry ], [ %inc, %for.body ] + %b.07 = phi i32 [ 0, %entry ], [ %or, %for.body ] + %shr = lshr i32 %a, %i.08 + %and = and i32 %shr, 1 + %sub = sub nuw nsw i32 31, %i.08 + %shl = shl i32 %and, %sub + %or = or i32 %shl, %b.07 + %inc = add nuw nsw i32 %i.08, 1 + %exitcond = icmp eq i32 %inc, 32 + br i1 %exitcond, label %for.cond.cleanup, label %for.body, !llvm.loop !3 +} + +attributes #0 = { norecurse nounwind readnone "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="cortex-a8" "target-features"="+dsp,+neon,+vfp3" "unsafe-fp-math"="false" "use-soft-float"="false" } + +!llvm.module.flags = !{!0, !1} +!llvm.ident = !{!2} + +!0 = !{i32 1, !"wchar_size", i32 4} +!1 = !{i32 1, !"min_enum_size", i32 4} +!2 = !{!"clang version 3.8.0 (http://llvm.org/git/clang.git b7441a0f42c43a8eea9e3e706be187252db747fa)"} +!3 = distinct !{!3, !4} +!4 = !{!"llvm.loop.unroll.full"} diff --git a/llvm/test/Transforms/CodeGenPrepare/ARM/lit.local.cfg b/llvm/test/Transforms/CodeGenPrepare/ARM/lit.local.cfg new file mode 100644 index 00000000000..98c6700c209 --- /dev/null +++ b/llvm/test/Transforms/CodeGenPrepare/ARM/lit.local.cfg @@ -0,0 +1,3 @@ +if not 'ARM' in config.root.targets: + config.unsupported = True + diff --git a/llvm/test/Transforms/CodeGenPrepare/bitreverse-hang.ll b/llvm/test/Transforms/CodeGenPrepare/bitreverse-hang.ll new file mode 100644 index 00000000000..c81dcc15cae --- /dev/null +++ b/llvm/test/Transforms/CodeGenPrepare/bitreverse-hang.ll @@ -0,0 +1,53 @@ +; RUN: opt < %s -loop-unroll -codegenprepare -S | FileCheck %s + +; This test is a worst-case scenario for bitreversal/byteswap detection. +; After loop unrolling (the unrolled loop is unreadably large so it has been kept +; rolled here), we have a binary tree of OR operands (as bitreversal detection +; looks straight through shifts): +; +; OR +; | \ +; | LSHR +; | / +; OR +; | \ +; | LSHR +; | / +; OR +; +; This results in exponential runtime. The loop here is 32 iterations which will +; totally hang if we don't deal with this case cleverly. + +@b = common global i32 0, align 4 + +; CHECK: define i32 @fn1 +define i32 @fn1() #0 { +entry: + %b.promoted = load i32, i32* @b, align 4, !tbaa !2 + br label %for.body + +for.body: ; preds = %for.body, %entry + %or4 = phi i32 [ %b.promoted, %entry ], [ %or, %for.body ] + %i.03 = phi i32 [ 0, %entry ], [ %inc, %for.body ] + %shr = lshr i32 %or4, 1 + %or = or i32 %shr, %or4 + %inc = add nuw nsw i32 %i.03, 1 + %exitcond = icmp eq i32 %inc, 32 + br i1 %exitcond, label %for.end, label %for.body + +for.end: ; preds = %for.body + store i32 %or, i32* @b, align 4, !tbaa !2 + ret i32 undef +} + +attributes #0 = { norecurse nounwind ssp uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="core2" "target-features"="+cx16,+fxsr,+mmx,+sse,+sse2,+sse3,+ssse3" "unsafe-fp-math"="false" "use-soft-float"="false" } + +!llvm.module.flags = !{!0} +!llvm.ident = !{!1} + +!0 = !{i32 1, !"PIC Level", i32 2} +!1 = !{!"clang version 3.8.0 (http://llvm.org/git/clang.git eb70f4e9cc9a4dc3dd57b032fb858d56b4b64a0e)"} +!2 = !{!3, !3, i64 0} +!3 = !{!"int", !4, i64 0} +!4 = !{!"omnipotent char", !5, i64 0} +!5 = !{!"Simple C/C++ TBAA"} diff --git a/llvm/test/Transforms/InstCombine/bitreverse-hang.ll b/llvm/test/Transforms/InstCombine/bitreverse-hang.ll new file mode 100644 index 00000000000..6823bd0ed65 --- /dev/null +++ b/llvm/test/Transforms/InstCombine/bitreverse-hang.ll @@ -0,0 +1,53 @@ +; RUN: opt < %s -loop-unroll -instcombine -S | FileCheck %s + +; This test is a worst-case scenario for bitreversal/byteswap detection. +; After loop unrolling (the unrolled loop is unreadably large so it has been kept +; rolled here), we have a binary tree of OR operands (as bitreversal detection +; looks straight through shifts): +; +; OR +; | \ +; | LSHR +; | / +; OR +; | \ +; | LSHR +; | / +; OR +; +; This results in exponential runtime. The loop here is 32 iterations which will +; totally hang if we don't deal with this case cleverly. + +@b = common global i32 0, align 4 + +; CHECK: define i32 @fn1 +define i32 @fn1() #0 { +entry: + %b.promoted = load i32, i32* @b, align 4, !tbaa !2 + br label %for.body + +for.body: ; preds = %for.body, %entry + %or4 = phi i32 [ %b.promoted, %entry ], [ %or, %for.body ] + %i.03 = phi i32 [ 0, %entry ], [ %inc, %for.body ] + %shr = lshr i32 %or4, 1 + %or = or i32 %shr, %or4 + %inc = add nuw nsw i32 %i.03, 1 + %exitcond = icmp eq i32 %inc, 32 + br i1 %exitcond, label %for.end, label %for.body + +for.end: ; preds = %for.body + store i32 %or, i32* @b, align 4, !tbaa !2 + ret i32 undef +} + +attributes #0 = { norecurse nounwind ssp uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="core2" "target-features"="+cx16,+fxsr,+mmx,+sse,+sse2,+sse3,+ssse3" "unsafe-fp-math"="false" "use-soft-float"="false" } + +!llvm.module.flags = !{!0} +!llvm.ident = !{!1} + +!0 = !{i32 1, !"PIC Level", i32 2} +!1 = !{!"clang version 3.8.0 (http://llvm.org/git/clang.git eb70f4e9cc9a4dc3dd57b032fb858d56b4b64a0e)"} +!2 = !{!3, !3, i64 0} +!3 = !{!"int", !4, i64 0} +!4 = !{!"omnipotent char", !5, i64 0} +!5 = !{!"Simple C/C++ TBAA"} diff --git a/llvm/test/Transforms/InstCombine/bitreverse-recognize.ll b/llvm/test/Transforms/InstCombine/bitreverse-recognize.ll deleted file mode 100644 index fbd5cb6d139..00000000000 --- a/llvm/test/Transforms/InstCombine/bitreverse-recognize.ll +++ /dev/null @@ -1,114 +0,0 @@ -; RUN: opt < %s -instcombine -S | FileCheck %s - -target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128" -target triple = "x86_64-apple-macosx10.10.0" - -define zeroext i8 @f_u8(i8 zeroext %a) { -; CHECK-LABEL: @f_u8 -; CHECK-NEXT: %[[A:.*]] = call i8 @llvm.bitreverse.i8(i8 %a) -; CHECK-NEXT: ret i8 %[[A]] - %1 = shl i8 %a, 7 - %2 = shl i8 %a, 5 - %3 = and i8 %2, 64 - %4 = shl i8 %a, 3 - %5 = and i8 %4, 32 - %6 = shl i8 %a, 1 - %7 = and i8 %6, 16 - %8 = lshr i8 %a, 1 - %9 = and i8 %8, 8 - %10 = lshr i8 %a, 3 - %11 = and i8 %10, 4 - %12 = lshr i8 %a, 5 - %13 = and i8 %12, 2 - %14 = lshr i8 %a, 7 - %15 = or i8 %14, %1 - %16 = or i8 %15, %3 - %17 = or i8 %16, %5 - %18 = or i8 %17, %7 - %19 = or i8 %18, %9 - %20 = or i8 %19, %11 - %21 = or i8 %20, %13 - ret i8 %21 -} - -; The ANDs with 32 and 64 have been swapped here, so the sequence does not -; completely match a bitreverse. -define zeroext i8 @f_u8_fail(i8 zeroext %a) { -; CHECK-LABEL: @f_u8_fail -; CHECK-NOT: call -; CHECK: ret i8 - %1 = shl i8 %a, 7 - %2 = shl i8 %a, 5 - %3 = and i8 %2, 32 - %4 = shl i8 %a, 3 - %5 = and i8 %4, 64 - %6 = shl i8 %a, 1 - %7 = and i8 %6, 16 - %8 = lshr i8 %a, 1 - %9 = and i8 %8, 8 - %10 = lshr i8 %a, 3 - %11 = and i8 %10, 4 - %12 = lshr i8 %a, 5 - %13 = and i8 %12, 2 - %14 = lshr i8 %a, 7 - %15 = or i8 %14, %1 - %16 = or i8 %15, %3 - %17 = or i8 %16, %5 - %18 = or i8 %17, %7 - %19 = or i8 %18, %9 - %20 = or i8 %19, %11 - %21 = or i8 %20, %13 - ret i8 %21 -} - -define zeroext i16 @f_u16(i16 zeroext %a) { -; CHECK-LABEL: @f_u16 -; CHECK-NEXT: %[[A:.*]] = call i16 @llvm.bitreverse.i16(i16 %a) -; CHECK-NEXT: ret i16 %[[A]] - %1 = shl i16 %a, 15 - %2 = shl i16 %a, 13 - %3 = and i16 %2, 16384 - %4 = shl i16 %a, 11 - %5 = and i16 %4, 8192 - %6 = shl i16 %a, 9 - %7 = and i16 %6, 4096 - %8 = shl i16 %a, 7 - %9 = and i16 %8, 2048 - %10 = shl i16 %a, 5 - %11 = and i16 %10, 1024 - %12 = shl i16 %a, 3 - %13 = and i16 %12, 512 - %14 = shl i16 %a, 1 - %15 = and i16 %14, 256 - %16 = lshr i16 %a, 1 - %17 = and i16 %16, 128 - %18 = lshr i16 %a, 3 - %19 = and i16 %18, 64 - %20 = lshr i16 %a, 5 - %21 = and i16 %20, 32 - %22 = lshr i16 %a, 7 - %23 = and i16 %22, 16 - %24 = lshr i16 %a, 9 - %25 = and i16 %24, 8 - %26 = lshr i16 %a, 11 - %27 = and i16 %26, 4 - %28 = lshr i16 %a, 13 - %29 = and i16 %28, 2 - %30 = lshr i16 %a, 15 - %31 = or i16 %30, %1 - %32 = or i16 %31, %3 - %33 = or i16 %32, %5 - %34 = or i16 %33, %7 - %35 = or i16 %34, %9 - %36 = or i16 %35, %11 - %37 = or i16 %36, %13 - %38 = or i16 %37, %15 - %39 = or i16 %38, %17 - %40 = or i16 %39, %19 - %41 = or i16 %40, %21 - %42 = or i16 %41, %23 - %43 = or i16 %42, %25 - %44 = or i16 %43, %27 - %45 = or i16 %44, %29 - ret i16 %45 -}
\ No newline at end of file |
