diff options
| author | Anton Korobeynikov <asl@math.spbu.ru> | 2012-08-14 23:36:01 +0000 |
|---|---|---|
| committer | Anton Korobeynikov <asl@math.spbu.ru> | 2012-08-14 23:36:01 +0000 |
| commit | c6d945b11abdab886c1e5da295f2b088bcd0c1af (patch) | |
| tree | 9a88bf91ca4574389ccd00589e7bf3fa5545811c /llvm/test/CodeGen/ARM/fp16.ll | |
| parent | 5f61a7498b43589d6fa6617d2e1793c6edc37bbf (diff) | |
| download | bcm5719-llvm-c6d945b11abdab886c1e5da295f2b088bcd0c1af.tar.gz bcm5719-llvm-c6d945b11abdab886c1e5da295f2b088bcd0c1af.zip | |
The names of VFP variants of half-to-float conversion instructions were
reversed. This leads to wrong codegen for float-to-half conversion
intrinsics which are used to support storage-only fp16 type.
NEON variants of same instructions are fine.
llvm-svn: 161907
Diffstat (limited to 'llvm/test/CodeGen/ARM/fp16.ll')
| -rw-r--r-- | llvm/test/CodeGen/ARM/fp16.ll | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/llvm/test/CodeGen/ARM/fp16.ll b/llvm/test/CodeGen/ARM/fp16.ll index c5583b94bef..1261ea50212 100644 --- a/llvm/test/CodeGen/ARM/fp16.ll +++ b/llvm/test/CodeGen/ARM/fp16.ll @@ -15,14 +15,14 @@ entry: %1 = load i16* @y, align 2 %2 = tail call float @llvm.convert.from.fp16(i16 %0) ; CHECK: __gnu_h2f_ieee -; CHECK-FP16: vcvtb.f16.f32 +; CHECK-FP16: vcvtb.f32.f16 %3 = tail call float @llvm.convert.from.fp16(i16 %1) ; CHECK: __gnu_h2f_ieee -; CHECK-FP16: vcvtb.f16.f32 +; CHECK-FP16: vcvtb.f32.f16 %4 = fadd float %2, %3 %5 = tail call i16 @llvm.convert.to.fp16(float %4) ; CHECK: __gnu_f2h_ieee -; CHECK-FP16: vcvtb.f32.f16 +; CHECK-FP16: vcvtb.f16.f32 store i16 %5, i16* @x, align 2 ret void } |
