1; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx --show-mc-encoding | FileCheck %s 2 3define i32 @f256(<8 x float> %A, <8 x float> %AA, i8* %B, <4 x double> %C, <4 x double> %CC, i32 %D, <4 x i64> %E, <4 x i64> %EE, i32* %loadptr) { 4; CHECK: vmovntps %ymm{{.*}} ## EVEX TO VEX Compression encoding: [0xc5 5 %v0 = load i32, i32* %loadptr, align 1 6 %cast = bitcast i8* %B to <8 x float>* 7 %A2 = fadd <8 x float> %A, %AA 8 store <8 x float> %A2, <8 x float>* %cast, align 64, !nontemporal !0 9; CHECK: vmovntdq %ymm{{.*}} ## EVEX TO VEX Compression encoding: [0xc5 10 %v1 = load i32, i32* %loadptr, align 1 11 %cast1 = bitcast i8* %B to <4 x i64>* 12 %E2 = add <4 x i64> %E, %EE 13 store <4 x i64> %E2, <4 x i64>* %cast1, align 64, !nontemporal !0 14; CHECK: vmovntpd %ymm{{.*}} ## EVEX TO VEX Compression encoding: [0xc5 15 %v2 = load i32, i32* %loadptr, align 1 16 %cast2 = bitcast i8* %B to <4 x double>* 17 %C2 = fadd <4 x double> %C, %CC 18 store <4 x double> %C2, <4 x double>* %cast2, align 64, !nontemporal !0 19 %v3 = load i32, i32* %loadptr, align 1 20 %sum1 = add i32 %v0, %v1 21 %sum2 = add i32 %sum1, %v2 22 %sum3 = add i32 %sum2, %v3 23 ret i32 %sum3 24} 25 26define i32 @f128(<4 x float> %A, <4 x float> %AA, i8* %B, <2 x double> %C, <2 x double> %CC, i32 %D, <2 x i64> %E, <2 x i64> %EE, i32* %loadptr) { 27 %v0 = load i32, i32* %loadptr, align 1 28; CHECK: vmovntps %xmm{{.*}} ## EVEX TO VEX Compression encoding: [0xc5 29 %cast = bitcast i8* %B to <4 x float>* 30 %A2 = fadd <4 x float> %A, %AA 31 store <4 x float> %A2, <4 x float>* %cast, align 64, !nontemporal !0 32; CHECK: vmovntdq %xmm{{.*}} ## EVEX TO VEX Compression encoding: [0xc5 33 %v1 = load i32, i32* %loadptr, align 1 34 %cast1 = bitcast i8* %B to <2 x i64>* 35 %E2 = add <2 x i64> %E, %EE 36 store <2 x i64> %E2, <2 x i64>* %cast1, align 64, !nontemporal !0 37; CHECK: vmovntpd %xmm{{.*}} ## EVEX TO VEX Compression encoding: [0xc5 38 %v2 = load i32, i32* %loadptr, align 1 39 %cast2 = bitcast i8* %B to <2 x double>* 40 %C2 = fadd <2 x double> %C, %CC 41 store <2 x double> %C2, <2 x double>* %cast2, align 64, !nontemporal !0 42 %v3 = load i32, i32* %loadptr, align 1 43 %sum1 = add i32 %v0, %v1 44 %sum2 = add i32 %sum1, %v2 45 %sum3 = add i32 %sum2, %v3 46 ret i32 %sum3 47} 48!0 = !{i32 1} 49