77 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s --mattr=+complxnum,+sve,+fullfp16 -o - | FileCheck %s3 4target triple = "aarch64"5 6define <4 x double> @simple_symmetric_muladd2(<4 x double> %a, <4 x double> %b) {7; CHECK-LABEL: simple_symmetric_muladd2:8; CHECK: // %bb.0: // %entry9; CHECK-NEXT: mov x8, #-7378697629483820647 // =0x999999999999999910; CHECK-NEXT: movk x8, #3932211; CHECK-NEXT: movk x8, #16393, lsl #4812; CHECK-NEXT: dup v4.2d, x813; CHECK-NEXT: fmla v2.2d, v4.2d, v0.2d14; CHECK-NEXT: fmla v3.2d, v4.2d, v1.2d15; CHECK-NEXT: mov v0.16b, v2.16b16; CHECK-NEXT: mov v1.16b, v3.16b17; CHECK-NEXT: ret18entry:19 %ext00 = shufflevector <4 x double> %a, <4 x double> poison, <2 x i32> <i32 0, i32 2>20 %ext01 = shufflevector <4 x double> %a, <4 x double> poison, <2 x i32> <i32 1, i32 3>21 %fmul0 = fmul fast <2 x double> %ext00, splat (double 3.200000e+00)22 %ext10 = shufflevector <4 x double> %b, <4 x double> poison, <2 x i32> <i32 0, i32 2>23 %ext11 = shufflevector <4 x double> %b, <4 x double> poison, <2 x i32> <i32 1, i32 3>24 %fadd0 = fadd fast <2 x double> %ext10, %fmul025 %fmul1 = fmul fast <2 x double> %ext01, splat (double 3.200000e+00)26 %fadd1 = fadd fast <2 x double> %ext11, %fmul127 %interleaved.vec = shufflevector <2 x double> %fadd0, <2 x double> %fadd1, <4 x i32> <i32 0, i32 2, i32 1, i32 3>28 ret <4 x double> %interleaved.vec29}30 31define <8 x double> @simple_symmetric_muladd4(<8 x double> %a, <8 x double> %b) {32; CHECK-LABEL: simple_symmetric_muladd4:33; CHECK: // %bb.0: // %entry34; CHECK-NEXT: mov x8, #-7378697629483820647 // =0x999999999999999935; CHECK-NEXT: zip1 v16.2d, v0.2d, v2.2d36; CHECK-NEXT: zip2 v0.2d, v0.2d, v2.2d37; CHECK-NEXT: movk x8, #3932238; CHECK-NEXT: zip1 v2.2d, v1.2d, v3.2d39; CHECK-NEXT: zip2 v1.2d, v1.2d, v3.2d40; CHECK-NEXT: movk x8, #16393, lsl #4841; CHECK-NEXT: zip1 v3.2d, v4.2d, v6.2d42; CHECK-NEXT: zip2 v4.2d, v4.2d, v6.2d43; CHECK-NEXT: zip1 v17.2d, v5.2d, v7.2d44; CHECK-NEXT: zip2 v5.2d, v5.2d, v7.2d45; CHECK-NEXT: dup v6.2d, x846; CHECK-NEXT: fmla v3.2d, v6.2d, v16.2d47; CHECK-NEXT: fmla v4.2d, v6.2d, v0.2d48; CHECK-NEXT: fmla v17.2d, v6.2d, v2.2d49; CHECK-NEXT: fmla v5.2d, v6.2d, v1.2d50; CHECK-NEXT: zip1 v0.2d, v3.2d, v4.2d51; CHECK-NEXT: zip2 v2.2d, v3.2d, v4.2d52; CHECK-NEXT: zip1 v1.2d, v17.2d, v5.2d53; CHECK-NEXT: zip2 v3.2d, v17.2d, v5.2d54; CHECK-NEXT: ret55entry:56 %ext00 = shufflevector <8 x double> %a, <8 x double> poison, <2 x i32> <i32 0, i32 4>57 %ext01 = shufflevector <8 x double> %a, <8 x double> poison, <2 x i32> <i32 1, i32 5>58 %ext02 = shufflevector <8 x double> %a, <8 x double> poison, <2 x i32> <i32 2, i32 6>59 %ext03 = shufflevector <8 x double> %a, <8 x double> poison, <2 x i32> <i32 3, i32 7>60 %fmul0 = fmul fast <2 x double> %ext00, splat (double 3.200000e+00)61 %ext10 = shufflevector <8 x double> %b, <8 x double> poison, <2 x i32> <i32 0, i32 4>62 %ext11 = shufflevector <8 x double> %b, <8 x double> poison, <2 x i32> <i32 1, i32 5>63 %ext12 = shufflevector <8 x double> %b, <8 x double> poison, <2 x i32> <i32 2, i32 6>64 %ext13 = shufflevector <8 x double> %b, <8 x double> poison, <2 x i32> <i32 3, i32 7>65 %fadd0 = fadd fast <2 x double> %ext10, %fmul066 %fmul1 = fmul fast <2 x double> %ext01, splat (double 3.200000e+00)67 %fadd1 = fadd fast <2 x double> %ext11, %fmul168 %fmul2 = fmul fast <2 x double> %ext02, splat (double 3.200000e+00)69 %fadd2 = fadd fast <2 x double> %ext12, %fmul270 %fmul3 = fmul fast <2 x double> %ext03, splat (double 3.200000e+00)71 %fadd3 = fadd fast <2 x double> %ext13, %fmul372 %interleave.pt1 = shufflevector <2 x double> %fadd0, <2 x double> %fadd1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>73 %interleave.pt2 = shufflevector <2 x double> %fadd2, <2 x double> %fadd3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>74 %interleaved.vec = shufflevector <4 x double> %interleave.pt1, <4 x double> %interleave.pt2, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>75 ret <8 x double> %interleaved.vec76}77