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,SSE2
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE,SSE41
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
6 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
7 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX,AVX512,AVX512F
8 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX,AVX512,AVX512BW
10 declare i32 @llvm.sadd.sat.i32 (i32, i32)
11 declare i64 @llvm.sadd.sat.i64 (i64, i64)
12 declare <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16>, <8 x i16>)
14 ; fold (sadd_sat x, undef) -> -1
15 define i32 @combine_undef_i32(i32 %a0) {
16 ; CHECK-LABEL: combine_undef_i32:
18 ; CHECK-NEXT: movl $-1, %eax
20 %res = call i32 @llvm.sadd.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: pcmpeqd %xmm0, %xmm0
30 ; AVX-LABEL: combine_undef_v8i16:
32 ; AVX-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
34 %res = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> undef, <8 x i16> %a0)
38 ; fold (sadd_sat c1, c2) -> c3
39 define i32 @combine_constfold_i32() {
40 ; CHECK-LABEL: combine_constfold_i32:
42 ; CHECK-NEXT: movl $2147483647, %eax # imm = 0x7FFFFFFF
44 %res = call i32 @llvm.sadd.sat.i32(i32 2147483647, i32 100)
48 define <8 x i16> @combine_constfold_v8i16() {
49 ; SSE-LABEL: combine_constfold_v8i16:
51 ; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,0,256,65534,0,65280,32768,0]
54 ; AVX-LABEL: combine_constfold_v8i16:
56 ; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [1,0,256,65534,0,65280,32768,0]
58 %res = call <8 x i16> @llvm.sadd.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 = [65535,65535,65535,65534,0,65280,32768,0]
68 ; AVX-LABEL: combine_constfold_undef_v8i16:
70 ; AVX-NEXT: vmovaps {{.*#+}} xmm0 = [65535,65535,65535,65534,0,65280,32768,0]
72 %res = call <8 x i16> @llvm.sadd.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 (sadd_sat c, x) -> (sadd_sat x, c)
77 define i32 @combine_constant_i32(i32 %a0) {
78 ; CHECK-LABEL: combine_constant_i32:
80 ; CHECK-NEXT: xorl %eax, %eax
81 ; CHECK-NEXT: movl %edi, %ecx
82 ; CHECK-NEXT: incl %ecx
83 ; CHECK-NEXT: setns %al
84 ; CHECK-NEXT: addl $2147483647, %eax # imm = 0x7FFFFFFF
85 ; CHECK-NEXT: incl %edi
86 ; CHECK-NEXT: cmovnol %edi, %eax
88 %res = call i32 @llvm.sadd.sat.i32(i32 1, i32 %a0)
92 define <8 x i16> @combine_constant_v8i16(<8 x i16> %a0) {
93 ; SSE-LABEL: combine_constant_v8i16:
95 ; SSE-NEXT: paddsw {{.*}}(%rip), %xmm0
98 ; AVX-LABEL: combine_constant_v8i16:
100 ; AVX-NEXT: vpaddsw {{.*}}(%rip), %xmm0, %xmm0
102 %res = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, <8 x i16> %a0)
106 ; fold (sadd_sat c, 0) -> x
107 define i32 @combine_zero_i32(i32 %a0) {
108 ; CHECK-LABEL: combine_zero_i32:
110 ; CHECK-NEXT: movl %edi, %eax
112 %1 = call i32 @llvm.sadd.sat.i32(i32 %a0, i32 0)
116 define <8 x i16> @combine_zero_v8i16(<8 x i16> %a0) {
117 ; CHECK-LABEL: combine_zero_v8i16:
120 %1 = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %a0, <8 x i16> zeroinitializer)
124 ; fold (sadd_sat x, y) -> (add x, y) iff no overflow
125 define i32 @combine_no_overflow_i32(i32 %a0, i32 %a1) {
126 ; CHECK-LABEL: combine_no_overflow_i32:
128 ; CHECK-NEXT: sarl $16, %edi
129 ; CHECK-NEXT: shrl $16, %esi
130 ; CHECK-NEXT: xorl %eax, %eax
131 ; CHECK-NEXT: movl %edi, %ecx
132 ; CHECK-NEXT: addl %esi, %ecx
133 ; CHECK-NEXT: setns %al
134 ; CHECK-NEXT: addl $2147483647, %eax # imm = 0x7FFFFFFF
135 ; CHECK-NEXT: addl %edi, %esi
136 ; CHECK-NEXT: cmovnol %esi, %eax
138 %1 = ashr i32 %a0, 16
139 %2 = lshr i32 %a1, 16
140 %3 = call i32 @llvm.sadd.sat.i32(i32 %1, i32 %2)
144 define <8 x i16> @combine_no_overflow_v8i16(<8 x i16> %a0, <8 x i16> %a1) {
145 ; SSE-LABEL: combine_no_overflow_v8i16:
147 ; SSE-NEXT: psraw $10, %xmm0
148 ; SSE-NEXT: psrlw $10, %xmm1
149 ; SSE-NEXT: paddsw %xmm1, %xmm0
152 ; AVX-LABEL: combine_no_overflow_v8i16:
154 ; AVX-NEXT: vpsraw $10, %xmm0, %xmm0
155 ; AVX-NEXT: vpsrlw $10, %xmm1, %xmm1
156 ; AVX-NEXT: vpaddsw %xmm1, %xmm0, %xmm0
158 %1 = ashr <8 x i16> %a0, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
159 %2 = lshr <8 x i16> %a1, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
160 %3 = call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %1, <8 x i16> %2)