1 # RUN: llc -mtriple=x86_64-linux-gnu -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NO_AVX512VL --check-prefix=NO_AVX512F --check-prefix=SSE
2 # RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NO_AVX512VL --check-prefix=NO_AVX512F --check-prefix=AVX
3 # RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=NO_AVX512VL --check-prefix=AVX512ALL --check-prefix=AVX512F
4 # RUN: llc -mtriple=x86_64-linux-gnu -mattr=+avx512f -mattr=+avx512vl -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=ALL --check-prefix=AVX512ALL --check-prefix=AVX512VL
7 define <4 x i32> @test_load_v4i32_noalign(<4 x i32>* %p1) {
8 %r = load <4 x i32>, <4 x i32>* %p1, align 1
12 define <4 x i32> @test_load_v4i32_align(<4 x i32>* %p1) {
13 %r = load <4 x i32>, <4 x i32>* %p1, align 16
17 define <4 x i32>* @test_store_v4i32_align(<4 x i32> %val, <4 x i32>* %p1) {
18 store <4 x i32> %val, <4 x i32>* %p1, align 16
22 define <4 x i32>* @test_store_v4i32_noalign(<4 x i32> %val, <4 x i32>* %p1) {
23 store <4 x i32> %val, <4 x i32>* %p1, align 1
29 # ALL-LABEL: name: test_load_v4i32_noalign
30 name: test_load_v4i32_noalign
35 - { id: 0, class: gpr }
36 - { id: 1, class: vecr }
37 # ALL: %0:gr64 = COPY $rdi
38 # SSE: %1:vr128 = MOVUPSrm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1, align 1)
39 # AVX: %1:vr128 = VMOVUPSrm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1, align 1)
40 # AVX512F: %1:vr128x = VMOVUPSZ128rm_NOVLX %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1, align 1)
41 # AVX512VL: %1:vr128x = VMOVUPSZ128rm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1, align 1)
42 # ALL: $xmm0 = COPY %1
48 %1(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.p1, align 1)
49 $xmm0 = COPY %1(<4 x s32>)
54 # ALL-LABEL: name: test_load_v4i32_align
55 name: test_load_v4i32_align
60 - { id: 0, class: gpr }
61 - { id: 1, class: vecr }
62 # ALL: %0:gr64 = COPY $rdi
63 # SSE: %1:vr128 = MOVAPSrm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1)
64 # AVX: %1:vr128 = VMOVAPSrm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1)
65 # AVX512F: %1:vr128x = VMOVAPSZ128rm_NOVLX %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1)
66 # AVX512VL: %1:vr128x = VMOVAPSZ128rm %0, 1, $noreg, 0, $noreg :: (load 16 from %ir.p1)
67 # ALL: $xmm0 = COPY %1
73 %1(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.p1)
74 $xmm0 = COPY %1(<4 x s32>)
79 # ALL-LABEL: name: test_store_v4i32_align
80 name: test_store_v4i32_align
85 - { id: 0, class: vecr }
86 - { id: 1, class: gpr }
87 # NO_AVX512F: %0:vr128 = COPY $xmm0
88 # AVX512ALL: %0:vr128x = COPY $xmm0
89 # ALL: %1:gr64 = COPY $rdi
90 # SSE: MOVAPSmr %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1)
91 # AVX: VMOVAPSmr %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1)
92 # AVX512F: VMOVAPSZ128mr_NOVLX %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1)
93 # AVX512VL: VMOVAPSZ128mr %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1)
99 %0(<4 x s32>) = COPY $xmm0
101 G_STORE %0(<4 x s32>), %1(p0) :: (store 16 into %ir.p1, align 16)
107 # ALL-LABEL: name: test_store_v4i32_noalign
108 name: test_store_v4i32_noalign
111 regBankSelected: true
113 - { id: 0, class: vecr }
114 - { id: 1, class: gpr }
115 # NO_AVX512F: %0:vr128 = COPY $xmm0
116 # AVX512ALL: %0:vr128x = COPY $xmm0
117 # ALL: %1:gr64 = COPY $rdi
118 # SSE: MOVUPSmr %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1, align 1)
119 # AVX: VMOVUPSmr %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1, align 1)
120 # AVX512F: VMOVUPSZ128mr_NOVLX %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1, align 1)
121 # AVX512VL: VMOVUPSZ128mr %1, 1, $noreg, 0, $noreg, %0 :: (store 16 into %ir.p1, align 1)
122 # ALL: $rax = COPY %1
127 %0(<4 x s32>) = COPY $xmm0
129 G_STORE %0(<4 x s32>), %1(p0) :: (store 16 into %ir.p1, align 1)