1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=CHECK --check-prefix=SKX 3 4 declare void @f() 5 define <4 x i1> @test_4i1(<4 x i32> %a, <4 x i32> %b) { 6 ; CHECK-LABEL: test_4i1: 7 ; CHECK: ## BB#0: 8 ; CHECK-NEXT: pushq %rax 9 ; CHECK-NEXT: Ltmp0: 10 ; CHECK-NEXT: .cfi_def_cfa_offset 16 11 ; CHECK-NEXT: vpcmpnleud %xmm1, %xmm0, %k0 12 ; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Folded Spill 13 ; CHECK-NEXT: vpcmpgtd %xmm1, %xmm0, %k0 14 ; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Folded Spill 15 ; CHECK-NEXT: callq _f 16 ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k0 ## 2-byte Folded Reload 17 ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k1 ## 2-byte Folded Reload 18 ; CHECK-NEXT: korw %k1, %k0, %k0 19 ; CHECK-NEXT: vpmovm2d %k0, %xmm0 20 ; CHECK-NEXT: popq %rax 21 ; CHECK-NEXT: retq 22 23 %cmp_res = icmp ugt <4 x i32> %a, %b 24 %cmp_res2 = icmp sgt <4 x i32> %a, %b 25 call void @f() 26 %res = or <4 x i1> %cmp_res, %cmp_res2 27 ret <4 x i1> %res 28 } 29 30 define <8 x i1> @test_8i1(<8 x i32> %a, <8 x i32> %b) { 31 ; CHECK-LABEL: test_8i1: 32 ; CHECK: ## BB#0: 33 ; CHECK-NEXT: pushq %rax 34 ; CHECK-NEXT: Ltmp1: 35 ; CHECK-NEXT: .cfi_def_cfa_offset 16 36 ; CHECK-NEXT: vpcmpnleud %ymm1, %ymm0, %k0 37 ; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Folded Spill 38 ; CHECK-NEXT: vpcmpgtd %ymm1, %ymm0, %k0 39 ; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Folded Spill 40 ; CHECK-NEXT: callq _f 41 ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k0 ## 2-byte Folded Reload 42 ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k1 ## 2-byte Folded Reload 43 ; CHECK-NEXT: korb %k1, %k0, %k0 44 ; CHECK-NEXT: vpmovm2w %k0, %xmm0 45 ; CHECK-NEXT: popq %rax 46 ; CHECK-NEXT: retq 47 48 %cmp_res = icmp ugt <8 x i32> %a, %b 49 %cmp_res2 = icmp sgt <8 x i32> %a, %b 50 call void @f() 51 %res = or <8 x i1> %cmp_res, %cmp_res2 52 ret <8 x i1> %res 53 } 54 55 define <16 x i1> @test_16i1(<16 x i32> %a, <16 x i32> %b) { 56 ; CHECK-LABEL: test_16i1: 57 ; CHECK: ## BB#0: 58 ; CHECK-NEXT: pushq %rax 59 ; CHECK-NEXT: Ltmp2: 60 ; CHECK-NEXT: .cfi_def_cfa_offset 16 61 ; CHECK-NEXT: vpcmpnleud %zmm1, %zmm0, %k0 62 ; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Folded Spill 63 ; CHECK-NEXT: vpcmpgtd %zmm1, %zmm0, %k0 64 ; CHECK-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) ## 2-byte Folded Spill 65 ; CHECK-NEXT: callq _f 66 ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k0 ## 2-byte Folded Reload 67 ; CHECK-NEXT: kmovw {{[0-9]+}}(%rsp), %k1 ## 2-byte Folded Reload 68 ; CHECK-NEXT: korw %k1, %k0, %k0 69 ; CHECK-NEXT: vpmovm2b %k0, %xmm0 70 ; CHECK-NEXT: popq %rax 71 ; CHECK-NEXT: retq 72 %cmp_res = icmp ugt <16 x i32> %a, %b 73 %cmp_res2 = icmp sgt <16 x i32> %a, %b 74 call void @f() 75 %res = or <16 x i1> %cmp_res, %cmp_res2 76 ret <16 x i1> %res 77 } 78 79 define <32 x i1> @test_32i1(<32 x i16> %a, <32 x i16> %b) { 80 ; CHECK-LABEL: test_32i1: 81 ; CHECK: ## BB#0: 82 ; CHECK-NEXT: pushq %rax 83 ; CHECK-NEXT: Ltmp3: 84 ; CHECK-NEXT: .cfi_def_cfa_offset 16 85 ; CHECK-NEXT: vpcmpnleuw %zmm1, %zmm0, %k0 86 ; CHECK-NEXT: kmovd %k0, {{[0-9]+}}(%rsp) ## 4-byte Folded Spill 87 ; CHECK-NEXT: vpcmpgtw %zmm1, %zmm0, %k0 88 ; CHECK-NEXT: kmovd %k0, (%rsp) ## 4-byte Folded Spill 89 ; CHECK-NEXT: callq _f 90 ; CHECK-NEXT: kmovd {{[0-9]+}}(%rsp), %k0 ## 4-byte Folded Reload 91 ; CHECK-NEXT: kmovd (%rsp), %k1 ## 4-byte Folded Reload 92 ; CHECK-NEXT: kord %k1, %k0, %k0 93 ; CHECK-NEXT: vpmovm2b %k0, %ymm0 94 ; CHECK-NEXT: popq %rax 95 ; CHECK-NEXT: retq 96 %cmp_res = icmp ugt <32 x i16> %a, %b 97 %cmp_res2 = icmp sgt <32 x i16> %a, %b 98 call void @f() 99 %res = or <32 x i1> %cmp_res, %cmp_res2 100 ret <32 x i1> %res 101 } 102 103 define <64 x i1> @test_64i1(<64 x i8> %a, <64 x i8> %b) { 104 ; CHECK-LABEL: test_64i1: 105 ; CHECK: ## BB#0: 106 ; CHECK-NEXT: subq $24, %rsp 107 ; CHECK-NEXT: Ltmp4: 108 ; CHECK-NEXT: .cfi_def_cfa_offset 32 109 ; CHECK-NEXT: vpcmpnleub %zmm1, %zmm0, %k0 110 ; CHECK-NEXT: kmovq %k0, {{[0-9]+}}(%rsp) ## 8-byte Folded Spill 111 ; CHECK-NEXT: vpcmpgtb %zmm1, %zmm0, %k0 112 ; CHECK-NEXT: kmovq %k0, {{[0-9]+}}(%rsp) ## 8-byte Folded Spill 113 ; CHECK-NEXT: callq _f 114 ; CHECK-NEXT: kmovq {{[0-9]+}}(%rsp), %k0 ## 8-byte Folded Reload 115 ; CHECK-NEXT: kmovq {{[0-9]+}}(%rsp), %k1 ## 8-byte Folded Reload 116 ; CHECK-NEXT: korq %k1, %k0, %k0 117 ; CHECK-NEXT: vpmovm2b %k0, %zmm0 118 ; CHECK-NEXT: addq $24, %rsp 119 ; CHECK-NEXT: retq 120 121 %cmp_res = icmp ugt <64 x i8> %a, %b 122 %cmp_res2 = icmp sgt <64 x i8> %a, %b 123 call void @f() 124 %res = or <64 x i1> %cmp_res, %cmp_res2 125 ret <64 x i1> %res 126 } 127
