1 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7 -mattr=+sse4.1 | FileCheck %s 2 3 4 ; Verify that we produce movss instead of blendvps when possible. 5 6 ;CHECK-LABEL: vsel_float: 7 ;CHECK-NOT: blend 8 ;CHECK: movss 9 ;CHECK: ret 10 define <4 x float> @vsel_float(<4 x float> %v1, <4 x float> %v2) { 11 %vsel = select <4 x i1> <i1 true, i1 false, i1 false, i1 false>, <4 x float> %v1, <4 x float> %v2 12 ret <4 x float> %vsel 13 } 14 15 ;CHECK-LABEL: vsel_4xi8: 16 ;CHECK-NOT: blend 17 ;CHECK: movss 18 ;CHECK: ret 19 define <4 x i8> @vsel_4xi8(<4 x i8> %v1, <4 x i8> %v2) { 20 %vsel = select <4 x i1> <i1 true, i1 false, i1 false, i1 false>, <4 x i8> %v1, <4 x i8> %v2 21 ret <4 x i8> %vsel 22 } 23 24 ;CHECK-LABEL: vsel_8xi16: 25 ; The select mask is 26 ; <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false> 27 ; which translates into the boolean mask (big endian representation): 28 ; 00010001 = 17. 29 ; '1' means takes the first argument, '0' means takes the second argument. 30 ; This is the opposite of the intel syntax, thus we expect 31 ; the inverted mask: 11101110 = 238. 32 ; According to the ABI: 33 ; v1 is in xmm0 => first argument is xmm0. 34 ; v2 is in xmm1 => second argument is xmm1. 35 ;CHECK: pblendw $238, %xmm1, %xmm0 36 ;CHECK: ret 37 define <8 x i16> @vsel_8xi16(<8 x i16> %v1, <8 x i16> %v2) { 38 %vsel = select <8 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false>, <8 x i16> %v1, <8 x i16> %v2 39 ret <8 x i16> %vsel 40 } 41
