1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX
6 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX
7 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX
8 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX
10 declare i32 @llvm.ssub.sat.i32 (i32, i32)
11 declare i64 @llvm.ssub.sat.i64 (i64, i64)
12 declare <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16>, <8 x i16>)
14 ; fold (ssub_sat x, undef) -> 0
15 define i32 @combine_undef_i32(i32 %a0) {
16 ; CHECK-LABEL: combine_undef_i32:
18 ; CHECK-NEXT: xorl %eax, %eax
20 %res = call i32 @llvm.ssub.sat.i32(i32 %a0, i32 undef)
24 define <8 x i16> @combine_undef_v8i16(<8 x i16> %a0) {
25 ; SSE-LABEL: combine_undef_v8i16:
27 ; SSE-NEXT: xorps %xmm0, %xmm0
30 ; AVX-LABEL: combine_undef_v8i16:
32 ; AVX-NEXT: vxorps %xmm0, %xmm0, %xmm0
34 %res = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> undef, <8 x i16> %a0)
38 ; fold (ssub_sat c1, c2) -> c3
39 define i32 @combine_constfold_i32() {
40 ; CHECK-LABEL: combine_constfold_i32:
42 ; CHECK-NEXT: movl $-2147483547, %eax # imm = 0x80000065
44 %res = call i32 @llvm.ssub.sat.i32(i32 100, i32 2147483647)
48 define <8 x i16> @combine_constfold_v8i16() {
49 ; SSE-LABEL: combine_constfold_v8i16:
51 ; SSE-NEXT: movaps {{.*#+}} xmm0 = [65535,2,254,0,65534,65282,32786,2]
54 ; AVX-LABEL: combine_constfold_v8i16:
56 ; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [65535,2,254,0,65534,65282,32786,2]
58 %res = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> <i16 0, i16 1, i16 255, i16 65535, i16 -1, i16 -255, i16 -32760, i16 1>, <8 x i16> <i16 1, i16 65535, i16 1, i16 65535, i16 1, i16 65535, i16 -10, i16 65535>)
62 define <8 x i16> @combine_constfold_undef_v8i16() {
63 ; SSE-LABEL: combine_constfold_undef_v8i16:
65 ; SSE-NEXT: movaps {{.*#+}} xmm0 = [0,0,0,0,65534,65282,32786,2]
68 ; AVX-LABEL: combine_constfold_undef_v8i16:
70 ; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [0,0,0,0,65534,65282,32786,2]
72 %res = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> <i16 undef, i16 1, i16 undef, i16 65535, i16 -1, i16 -255, i16 -32760, i16 1>, <8 x i16> <i16 1, i16 undef, i16 undef, i16 65535, i16 1, i16 65535, i16 -10, i16 65535>)
76 ; fold (ssub_sat x, 0) -> x
77 define i32 @combine_zero_i32(i32 %a0) {
78 ; CHECK-LABEL: combine_zero_i32:
80 ; CHECK-NEXT: movl %edi, %eax
82 %1 = call i32 @llvm.ssub.sat.i32(i32 %a0, i32 0)
86 define <8 x i16> @combine_zero_v8i16(<8 x i16> %a0) {
87 ; CHECK-LABEL: combine_zero_v8i16:
90 %1 = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %a0, <8 x i16> zeroinitializer)
94 ; fold (ssub_sat x, x) -> 0
95 define i32 @combine_self_i32(i32 %a0) {
96 ; CHECK-LABEL: combine_self_i32:
98 ; CHECK-NEXT: xorl %eax, %eax
100 %1 = call i32 @llvm.ssub.sat.i32(i32 %a0, i32 %a0)
104 define <8 x i16> @combine_self_v8i16(<8 x i16> %a0) {
105 ; SSE-LABEL: combine_self_v8i16:
107 ; SSE-NEXT: xorps %xmm0, %xmm0
110 ; AVX-LABEL: combine_self_v8i16:
112 ; AVX-NEXT: vxorps %xmm0, %xmm0, %xmm0
114 %1 = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %a0, <8 x i16> %a0)
118 ; fold (ssub_sat x, y) -> (sub x, y) iff no overflow
119 define i32 @combine_no_overflow_i32(i32 %a0, i32 %a1) {
120 ; CHECK-LABEL: combine_no_overflow_i32:
122 ; CHECK-NEXT: movl %edi, %eax
123 ; CHECK-NEXT: sarl $16, %eax
124 ; CHECK-NEXT: shrl $16, %esi
125 ; CHECK-NEXT: subl %esi, %eax
127 %1 = ashr i32 %a0, 16
128 %2 = lshr i32 %a1, 16
129 %3 = call i32 @llvm.ssub.sat.i32(i32 %1, i32 %2)
133 define <8 x i16> @combine_no_overflow_v8i16(<8 x i16> %a0, <8 x i16> %a1) {
134 ; SSE-LABEL: combine_no_overflow_v8i16:
136 ; SSE-NEXT: psraw $10, %xmm0
137 ; SSE-NEXT: psrlw $10, %xmm1
138 ; SSE-NEXT: psubw %xmm1, %xmm0
141 ; AVX-LABEL: combine_no_overflow_v8i16:
143 ; AVX-NEXT: vpsraw $10, %xmm0, %xmm0
144 ; AVX-NEXT: vpsrlw $10, %xmm1, %xmm1
145 ; AVX-NEXT: vpsubw %xmm1, %xmm0, %xmm0
147 %1 = ashr <8 x i16> %a0, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
148 %2 = lshr <8 x i16> %a1, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
149 %3 = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %1, <8 x i16> %2)
153 ; fold (ssub_sat (shuffle x, u, m), (shuffle y, u, m)) -> (shuffle (ssub_sat x, y), u, m)
154 define <8 x i16> @combine_shuffle_shuffle_v8i16(<8 x i16> %x0, <8 x i16> %y0) {
155 ; SSE-LABEL: combine_shuffle_shuffle_v8i16:
157 ; SSE-NEXT: psubsw %xmm1, %xmm0
158 ; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
161 ; AVX-LABEL: combine_shuffle_shuffle_v8i16:
163 ; AVX-NEXT: vpsubsw %xmm1, %xmm0, %xmm0
164 ; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
166 %x1= shufflevector <8 x i16> %x0, <8 x i16> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
167 %y1 = shufflevector <8 x i16> %y0, <8 x i16> poison, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
168 %res = tail call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %x1, <8 x i16> %y1)