1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=avx2 -show-mc-encoding | FileCheck %s --check-prefix=AVX2 3; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx512f,+avx512bw,+avx512vl,+avx512dq -show-mc-encoding | FileCheck %s --check-prefix=AVX512VL 4; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=avx2 -show-mc-encoding | FileCheck %s --check-prefix=AVX2 5; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx512f,+avx512bw,+avx512vl,+avx512dq -show-mc-encoding | FileCheck %s --check-prefix=AVX512VL 6 7; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/sse2-builtins.c 8 9define <32 x i8> @test_x86_avx2_paddus_b(<32 x i8> %a0, <32 x i8> %a1) { 10; AVX2-LABEL: test_x86_avx2_paddus_b: 11; AVX2: ## %bb.0: 12; AVX2-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 ## encoding: [0xc5,0xfd,0xdc,0xc1] 13; AVX2-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 14; 15; AVX512VL-LABEL: test_x86_avx2_paddus_b: 16; AVX512VL: ## %bb.0: 17; AVX512VL-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0xc1] 18; AVX512VL-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 19 %1 = add <32 x i8> %a0, %a1 20 %2 = icmp ugt <32 x i8> %a0, %1 21 %3 = select <32 x i1> %2, <32 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <32 x i8> %1 22 ret <32 x i8> %3 23} 24 25define <16 x i16> @test_x86_avx2_paddus_w(<16 x i16> %a0, <16 x i16> %a1) { 26; AVX2-LABEL: test_x86_avx2_paddus_w: 27; AVX2: ## %bb.0: 28; AVX2-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 ## encoding: [0xc5,0xfd,0xdd,0xc1] 29; AVX2-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 30; 31; AVX512VL-LABEL: test_x86_avx2_paddus_w: 32; AVX512VL: ## %bb.0: 33; AVX512VL-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0xc1] 34; AVX512VL-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 35 %1 = add <16 x i16> %a0, %a1 36 %2 = icmp ugt <16 x i16> %a0, %1 37 %3 = select <16 x i1> %2, <16 x i16> <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>, <16 x i16> %1 38 ret <16 x i16> %3 39} 40 41define <32 x i8> @test_x86_avx2_psubus_b(<32 x i8> %a0, <32 x i8> %a1) { 42; AVX2-LABEL: test_x86_avx2_psubus_b: 43; AVX2: ## %bb.0: 44; AVX2-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 ## encoding: [0xc5,0xfd,0xd8,0xc1] 45; AVX2-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 46; 47; AVX512VL-LABEL: test_x86_avx2_psubus_b: 48; AVX512VL: ## %bb.0: 49; AVX512VL-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0xc1] 50; AVX512VL-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 51 %cmp = icmp ugt <32 x i8> %a0, %a1 52 %sel = select <32 x i1> %cmp, <32 x i8> %a0, <32 x i8> %a1 53 %sub = sub <32 x i8> %sel, %a1 54 ret <32 x i8> %sub 55} 56 57define <16 x i16> @test_x86_avx2_psubus_w(<16 x i16> %a0, <16 x i16> %a1) { 58; AVX2-LABEL: test_x86_avx2_psubus_w: 59; AVX2: ## %bb.0: 60; AVX2-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 ## encoding: [0xc5,0xfd,0xd9,0xc1] 61; AVX2-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 62; 63; AVX512VL-LABEL: test_x86_avx2_psubus_w: 64; AVX512VL: ## %bb.0: 65; AVX512VL-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0xc1] 66; AVX512VL-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 67 %cmp = icmp ugt <16 x i16> %a0, %a1 68 %sel = select <16 x i1> %cmp, <16 x i16> %a0, <16 x i16> %a1 69 %sub = sub <16 x i16> %sel, %a1 70 ret <16 x i16> %sub 71} 72 73define <32 x i16> @test_x86_avx2_paddus_w_512(<32 x i16> %a, <32 x i16> %b) { 74; AVX2-LABEL: test_x86_avx2_paddus_w_512: 75; AVX2: ## %bb.0: 76; AVX2-NEXT: vpaddusw %ymm2, %ymm0, %ymm0 ## encoding: [0xc5,0xfd,0xdd,0xc2] 77; AVX2-NEXT: vpaddusw %ymm3, %ymm1, %ymm1 ## encoding: [0xc5,0xf5,0xdd,0xcb] 78; AVX2-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 79; 80; AVX512VL-LABEL: test_x86_avx2_paddus_w_512: 81; AVX512VL: ## %bb.0: 82; AVX512VL-NEXT: vpaddusw %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xdd,0xc1] 83; AVX512VL-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 84 %1 = add <32 x i16> %a, %b 85 %2 = icmp ugt <32 x i16> %a, %1 86 %3 = select <32 x i1> %2, <32 x i16> <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>, <32 x i16> %1 87 ret <32 x i16> %3 88} 89 90define <32 x i16> @test_x86_avx2_psubus_w_512(<32 x i16> %a, <32 x i16> %b) { 91; AVX2-LABEL: test_x86_avx2_psubus_w_512: 92; AVX2: ## %bb.0: 93; AVX2-NEXT: vpsubusw %ymm2, %ymm0, %ymm0 ## encoding: [0xc5,0xfd,0xd9,0xc2] 94; AVX2-NEXT: vpsubusw %ymm3, %ymm1, %ymm1 ## encoding: [0xc5,0xf5,0xd9,0xcb] 95; AVX2-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 96; 97; AVX512VL-LABEL: test_x86_avx2_psubus_w_512: 98; AVX512VL: ## %bb.0: 99; AVX512VL-NEXT: vpsubusw %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xd9,0xc1] 100; AVX512VL-NEXT: ret{{[l|q]}} ## encoding: [0xc3] 101 %cmp = icmp ugt <32 x i16> %a, %b 102 %sel = select <32 x i1> %cmp, <32 x i16> %a, <32 x i16> %b 103 %sub = sub <32 x i16> %sel, %b 104 ret <32 x i16> %sub 105} 106