1 ; Use CPU parameters to ensure that a CPU-specific attribute is not overriding the AVX definition. 2 3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s 4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=corei7-avx | FileCheck %s 5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=btver2 | FileCheck %s 6 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=-avx | FileCheck %s --check-prefix=SSE 7 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=corei7-avx -mattr=-avx | FileCheck %s --check-prefix=SSE 8 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=btver2 -mattr=-avx | FileCheck %s --check-prefix=SSE 9 10 ; No need to load unaligned operand from memory using an explicit instruction with AVX. 11 ; The operand should be folded into the AND instr. 12 13 ; With SSE, folding memory operands into math/logic ops requires 16-byte alignment 14 ; unless specially configured on some CPUs such as AMD Family 10H. 15 16 define <4 x i32> @test1(<4 x i32>* %p0, <4 x i32> %in1) nounwind { 17 %in0 = load <4 x i32>, <4 x i32>* %p0, align 2 18 %a = and <4 x i32> %in0, %in1 19 ret <4 x i32> %a 20 21 ; CHECK-LABEL: @test1 22 ; CHECK-NOT: vmovups 23 ; CHECK: vandps (%rdi), %xmm0, %xmm0 24 ; CHECK-NEXT: ret 25 26 ; SSE-LABEL: @test1 27 ; SSE: movups (%rdi), %xmm1 28 ; SSE-NEXT: andps %xmm1, %xmm0 29 ; SSE-NEXT: ret 30 } 31 32
