150 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s --mattr=+sve2 -o - | FileCheck %s3 4target triple = "aarch64"5 6; Expected to not transform as the type's minimum size is less than 128 bits.7define <vscale x 4 x i16> @complex_mul_v4i16(<vscale x 4 x i16> %a, <vscale x 4 x i16> %b) {8; CHECK-LABEL: complex_mul_v4i16:9; CHECK: // %bb.0: // %entry10; CHECK-NEXT: uunpkhi z2.d, z0.s11; CHECK-NEXT: uunpklo z0.d, z0.s12; CHECK-NEXT: uunpkhi z3.d, z1.s13; CHECK-NEXT: uunpklo z1.d, z1.s14; CHECK-NEXT: ptrue p0.d15; CHECK-NEXT: uzp1 z4.d, z0.d, z2.d16; CHECK-NEXT: uzp2 z0.d, z0.d, z2.d17; CHECK-NEXT: uzp1 z2.d, z1.d, z3.d18; CHECK-NEXT: uzp2 z1.d, z1.d, z3.d19; CHECK-NEXT: mul z5.d, z2.d, z0.d20; CHECK-NEXT: mul z2.d, z2.d, z4.d21; CHECK-NEXT: mad z4.d, p0/m, z1.d, z5.d22; CHECK-NEXT: msb z0.d, p0/m, z1.d, z2.d23; CHECK-NEXT: zip2 z1.d, z0.d, z4.d24; CHECK-NEXT: zip1 z0.d, z0.d, z4.d25; CHECK-NEXT: uzp1 z0.s, z0.s, z1.s26; CHECK-NEXT: ret27entry:28 %a.deinterleaved = tail call { <vscale x 2 x i16>, <vscale x 2 x i16> } @llvm.vector.deinterleave2.nxv4i16(<vscale x 4 x i16> %a)29 %a.real = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %a.deinterleaved, 030 %a.imag = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %a.deinterleaved, 131 %b.deinterleaved = tail call { <vscale x 2 x i16>, <vscale x 2 x i16> } @llvm.vector.deinterleave2.nxv4i16(<vscale x 4 x i16> %b)32 %b.real = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %b.deinterleaved, 033 %b.imag = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %b.deinterleaved, 134 %0 = mul <vscale x 2 x i16> %b.imag, %a.real35 %1 = mul <vscale x 2 x i16> %b.real, %a.imag36 %2 = add <vscale x 2 x i16> %1, %037 %3 = mul <vscale x 2 x i16> %b.real, %a.real38 %4 = mul <vscale x 2 x i16> %a.imag, %b.imag39 %5 = sub <vscale x 2 x i16> %3, %440 %interleaved.vec = tail call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> %5, <vscale x 2 x i16> %2)41 ret <vscale x 4 x i16> %interleaved.vec42}43 44; Expected to transform45define <vscale x 8 x i16> @complex_mul_v8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {46; CHECK-LABEL: complex_mul_v8i16:47; CHECK: // %bb.0: // %entry48; CHECK-NEXT: movi v2.2d, #000000000000000049; CHECK-NEXT: cmla z2.h, z1.h, z0.h, #050; CHECK-NEXT: cmla z2.h, z1.h, z0.h, #9051; CHECK-NEXT: mov z0.d, z2.d52; CHECK-NEXT: ret53entry:54 %a.deinterleaved = tail call { <vscale x 4 x i16>, <vscale x 4 x i16> } @llvm.vector.deinterleave2.nxv8i16(<vscale x 8 x i16> %a)55 %a.real = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %a.deinterleaved, 056 %a.imag = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %a.deinterleaved, 157 %b.deinterleaved = tail call { <vscale x 4 x i16>, <vscale x 4 x i16> } @llvm.vector.deinterleave2.nxv8i16(<vscale x 8 x i16> %b)58 %b.real = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %b.deinterleaved, 059 %b.imag = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %b.deinterleaved, 160 %0 = mul <vscale x 4 x i16> %b.imag, %a.real61 %1 = mul <vscale x 4 x i16> %b.real, %a.imag62 %2 = add <vscale x 4 x i16> %1, %063 %3 = mul <vscale x 4 x i16> %b.real, %a.real64 %4 = mul <vscale x 4 x i16> %a.imag, %b.imag65 %5 = sub <vscale x 4 x i16> %3, %466 %interleaved.vec = tail call <vscale x 8 x i16> @llvm.vector.interleave2.nxv8i16(<vscale x 4 x i16> %5, <vscale x 4 x i16> %2)67 ret <vscale x 8 x i16> %interleaved.vec68}69; Expected to transform70define <vscale x 16 x i16> @complex_mul_v16i16(<vscale x 16 x i16> %a, <vscale x 16 x i16> %b) {71; CHECK-LABEL: complex_mul_v16i16:72; CHECK: // %bb.0: // %entry73; CHECK-NEXT: movi v4.2d, #000000000000000074; CHECK-NEXT: movi v5.2d, #000000000000000075; CHECK-NEXT: cmla z5.h, z2.h, z0.h, #076; CHECK-NEXT: cmla z4.h, z3.h, z1.h, #077; CHECK-NEXT: cmla z5.h, z2.h, z0.h, #9078; CHECK-NEXT: cmla z4.h, z3.h, z1.h, #9079; CHECK-NEXT: mov z0.d, z5.d80; CHECK-NEXT: mov z1.d, z4.d81; CHECK-NEXT: ret82entry:83 %a.deinterleaved = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.vector.deinterleave2.nxv16i16(<vscale x 16 x i16> %a)84 %a.real = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %a.deinterleaved, 085 %a.imag = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %a.deinterleaved, 186 %b.deinterleaved = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.vector.deinterleave2.nxv16i16(<vscale x 16 x i16> %b)87 %b.real = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %b.deinterleaved, 088 %b.imag = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %b.deinterleaved, 189 %0 = mul <vscale x 8 x i16> %b.imag, %a.real90 %1 = mul <vscale x 8 x i16> %b.real, %a.imag91 %2 = add <vscale x 8 x i16> %1, %092 %3 = mul <vscale x 8 x i16> %b.real, %a.real93 %4 = mul <vscale x 8 x i16> %a.imag, %b.imag94 %5 = sub <vscale x 8 x i16> %3, %495 %interleaved.vec = tail call <vscale x 16 x i16> @llvm.vector.interleave2.nxv16i16(<vscale x 8 x i16> %5, <vscale x 8 x i16> %2)96 ret <vscale x 16 x i16> %interleaved.vec97}98 99; Expected to transform100define <vscale x 32 x i16> @complex_mul_v32i16(<vscale x 32 x i16> %a, <vscale x 32 x i16> %b) {101; CHECK-LABEL: complex_mul_v32i16:102; CHECK: // %bb.0: // %entry103; CHECK-NEXT: movi v24.2d, #0000000000000000104; CHECK-NEXT: movi v25.2d, #0000000000000000105; CHECK-NEXT: movi v26.2d, #0000000000000000106; CHECK-NEXT: movi v27.2d, #0000000000000000107; CHECK-NEXT: cmla z24.h, z4.h, z0.h, #0108; CHECK-NEXT: cmla z25.h, z5.h, z1.h, #0109; CHECK-NEXT: cmla z27.h, z6.h, z2.h, #0110; CHECK-NEXT: cmla z26.h, z7.h, z3.h, #0111; CHECK-NEXT: cmla z24.h, z4.h, z0.h, #90112; CHECK-NEXT: cmla z25.h, z5.h, z1.h, #90113; CHECK-NEXT: cmla z27.h, z6.h, z2.h, #90114; CHECK-NEXT: cmla z26.h, z7.h, z3.h, #90115; CHECK-NEXT: mov z0.d, z24.d116; CHECK-NEXT: mov z1.d, z25.d117; CHECK-NEXT: mov z2.d, z27.d118; CHECK-NEXT: mov z3.d, z26.d119; CHECK-NEXT: ret120entry:121 %a.deinterleaved = tail call { <vscale x 16 x i16>, <vscale x 16 x i16> } @llvm.vector.deinterleave2.nxv32i16(<vscale x 32 x i16> %a)122 %a.real = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %a.deinterleaved, 0123 %a.imag = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %a.deinterleaved, 1124 %b.deinterleaved = tail call { <vscale x 16 x i16>, <vscale x 16 x i16> } @llvm.vector.deinterleave2.nxv32i16(<vscale x 32 x i16> %b)125 %b.real = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %b.deinterleaved, 0126 %b.imag = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %b.deinterleaved, 1127 %0 = mul <vscale x 16 x i16> %b.imag, %a.real128 %1 = mul <vscale x 16 x i16> %b.real, %a.imag129 %2 = add <vscale x 16 x i16> %1, %0130 %3 = mul <vscale x 16 x i16> %b.real, %a.real131 %4 = mul <vscale x 16 x i16> %a.imag, %b.imag132 %5 = sub <vscale x 16 x i16> %3, %4133 %interleaved.vec = tail call <vscale x 32 x i16> @llvm.vector.interleave2.nxv32i16(<vscale x 16 x i16> %5, <vscale x 16 x i16> %2)134 ret <vscale x 32 x i16> %interleaved.vec135}136 137declare { <vscale x 2 x i16>, <vscale x 2 x i16> } @llvm.vector.deinterleave2.nxv4i16(<vscale x 4 x i16>)138declare <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16>, <vscale x 2 x i16>)139 140declare { <vscale x 4 x i16>, <vscale x 4 x i16> } @llvm.vector.deinterleave2.nxv8i16(<vscale x 8 x i16>)141declare <vscale x 8 x i16> @llvm.vector.interleave2.nxv8i16(<vscale x 4 x i16>, <vscale x 4 x i16>)142 143declare { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.vector.deinterleave2.nxv16i16(<vscale x 16 x i16>)144declare <vscale x 16 x i16> @llvm.vector.interleave2.nxv16i16(<vscale x 8 x i16>, <vscale x 8 x i16>)145 146declare { <vscale x 16 x i16>, <vscale x 16 x i16> } @llvm.vector.deinterleave2.nxv32i16(<vscale x 32 x i16>)147declare <vscale x 32 x i16> @llvm.vector.interleave2.nxv32i16(<vscale x 16 x i16>, <vscale x 16 x i16>)148 149 150