From: Craig Topper Date: Fri, 13 Jan 2012 06:59:47 +0000 (+0000) Subject: Add patterns for v16i16 and v32i8 immAllZerosV to select VPXOR to match v4i64 and... X-Git-Url: http://plrg.eecs.uci.edu/git/?a=commitdiff_plain;h=c30432ab573a3755ccd15dfeb03d64e6c0b90ccb;p=oota-llvm.git Add patterns for v16i16 and v32i8 immAllZerosV to select VPXOR to match v4i64 and v8i32. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@148106 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td index 51cc7f384e5..d3cd6a82fdb 100644 --- a/lib/Target/X86/X86InstrSSE.td +++ b/lib/Target/X86/X86InstrSSE.td @@ -289,6 +289,14 @@ def AVX_SET0PDY : PDI<0x57, MRMInitReg, (outs VR256:$dst), (ins), "", // AVX has no support for 256-bit integer instructions, but since the 128-bit // VPXOR instruction writes zero to its upper part, it's safe build zeros. +def : Pat<(v32i8 immAllZerosV), (SUBREG_TO_REG (i8 0), (V_SET0), sub_xmm)>; +def : Pat<(bc_v32i8 (v8f32 immAllZerosV)), + (SUBREG_TO_REG (i8 0), (V_SET0), sub_xmm)>; + +def : Pat<(v16i16 immAllZerosV), (SUBREG_TO_REG (i16 0), (V_SET0), sub_xmm)>; +def : Pat<(bc_v16i16 (v8f32 immAllZerosV)), + (SUBREG_TO_REG (i16 0), (V_SET0), sub_xmm)>; + def : Pat<(v8i32 immAllZerosV), (SUBREG_TO_REG (i32 0), (V_SET0), sub_xmm)>; def : Pat<(bc_v8i32 (v8f32 immAllZerosV)), (SUBREG_TO_REG (i32 0), (V_SET0), sub_xmm)>; diff --git a/test/CodeGen/X86/avx2-shift.ll b/test/CodeGen/X86/avx2-shift.ll index b6cf54ebe8f..1f192a0d323 100644 --- a/test/CodeGen/X86/avx2-shift.ll +++ b/test/CodeGen/X86/avx2-shift.ll @@ -231,7 +231,7 @@ define <32 x i8> @sra_v32i8_7(<32 x i8> %A) nounwind { %B = ashr <32 x i8> %A, ret <32 x i8> %B ; CHECK: sra_v32i8_7: -; CHECK: vxorps +; CHECK: vpxor ; CHECK: vpcmpgtb ; CHECK: ret }