diff options
| author | Bruno Cardoso Lopes <bruno.cardoso@gmail.com> | 2011-08-09 03:04:29 +0000 |
|---|---|---|
| committer | Bruno Cardoso Lopes <bruno.cardoso@gmail.com> | 2011-08-09 03:04:29 +0000 |
| commit | 1155b1eafabcbba0ddc88bd3bcfbdb0a1f7ee107 (patch) | |
| tree | 2bccefd9366900f3f0eb8fad3fe977e0eab2a5bc | |
| parent | 0d0964d0998b8f291d74bb828ab017e5a3f9cee3 (diff) | |
| download | bcm5719-llvm-1155b1eafabcbba0ddc88bd3bcfbdb0a1f7ee107.tar.gz bcm5719-llvm-1155b1eafabcbba0ddc88bd3bcfbdb0a1f7ee107.zip | |
Add support for avx vector fextend
llvm-svn: 137105
| -rw-r--r-- | llvm/lib/Target/X86/X86InstrSSE.td | 7 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/avx-cvt.ll | 6 |
2 files changed, 12 insertions, 1 deletions
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td index 986e428b9ee..79d455a326c 100644 --- a/llvm/lib/Target/X86/X86InstrSSE.td +++ b/llvm/lib/Target/X86/X86InstrSSE.td @@ -1093,12 +1093,17 @@ def : Pat<(int_x86_avx_cvtt_ps2dq_256 VR256:$src), def : Pat<(int_x86_avx_cvtt_ps2dq_256 (memopv8f32 addr:$src)), (VCVTTPS2DQYrm addr:$src)>; -// Match fround for 128/256-bit conversions +// Match fround and fextend for 128/256-bit conversions def : Pat<(v4f32 (fround (v4f64 VR256:$src))), (VCVTPD2PSYrr VR256:$src)>; def : Pat<(v4f32 (fround (loadv4f64 addr:$src))), (VCVTPD2PSYrm addr:$src)>; +def : Pat<(v4f64 (fextend (v4f32 VR128:$src))), + (VCVTPS2PDYrr VR128:$src)>; +def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))), + (VCVTPS2PDYrm addr:$src)>; + //===----------------------------------------------------------------------===// // SSE 1 & 2 - Compare Instructions //===----------------------------------------------------------------------===// diff --git a/llvm/test/CodeGen/X86/avx-cvt.ll b/llvm/test/CodeGen/X86/avx-cvt.ll index 6da47af5ab5..e45010c903e 100644 --- a/llvm/test/CodeGen/X86/avx-cvt.ll +++ b/llvm/test/CodeGen/X86/avx-cvt.ll @@ -20,6 +20,12 @@ define <8 x float> @fptrunc00(<8 x double> %b) nounwind { ret <8 x float> %a } +; CHECK: vcvtps2pd %xmm +define <4 x double> @fpext00(<4 x float> %b) nounwind { + %a = fpext <4 x float> %b to <4 x double> + ret <4 x double> %a +} + ; CHECK: vcvtsi2sdq (% define double @funcA(i64* nocapture %e) nounwind uwtable readonly ssp { entry: |
