116 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_add_v4i16(<vscale x 4 x i16> %a, <vscale x 4 x i16> %b) {8; CHECK-LABEL: complex_add_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: uzp1 z4.d, z0.d, z2.d15; CHECK-NEXT: uzp2 z0.d, z0.d, z2.d16; CHECK-NEXT: uzp2 z2.d, z1.d, z3.d17; CHECK-NEXT: uzp1 z1.d, z1.d, z3.d18; CHECK-NEXT: sub z0.d, z1.d, z0.d19; CHECK-NEXT: add z1.d, z2.d, z4.d20; CHECK-NEXT: zip2 z2.d, z0.d, z1.d21; CHECK-NEXT: zip1 z0.d, z0.d, z1.d22; CHECK-NEXT: uzp1 z0.s, z0.s, z2.s23; CHECK-NEXT: ret24entry:25 %a.deinterleaved = tail call { <vscale x 2 x i16>, <vscale x 2 x i16> } @llvm.vector.deinterleave2.nxv4i16(<vscale x 4 x i16> %a)26 %a.real = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %a.deinterleaved, 027 %a.imag = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %a.deinterleaved, 128 %b.deinterleaved = tail call { <vscale x 2 x i16>, <vscale x 2 x i16> } @llvm.vector.deinterleave2.nxv4i16(<vscale x 4 x i16> %b)29 %b.real = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %b.deinterleaved, 030 %b.imag = extractvalue { <vscale x 2 x i16>, <vscale x 2 x i16> } %b.deinterleaved, 131 %0 = sub <vscale x 2 x i16> %b.real, %a.imag32 %1 = add <vscale x 2 x i16> %b.imag, %a.real33 %interleaved.vec = tail call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> %0, <vscale x 2 x i16> %1)34 ret <vscale x 4 x i16> %interleaved.vec35}36 37; Expected to transform38define <vscale x 8 x i16> @complex_add_v8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {39; CHECK-LABEL: complex_add_v8i16:40; CHECK: // %bb.0: // %entry41; CHECK-NEXT: cadd z1.h, z1.h, z0.h, #9042; CHECK-NEXT: mov z0.d, z1.d43; CHECK-NEXT: ret44entry:45 %a.deinterleaved = tail call { <vscale x 4 x i16>, <vscale x 4 x i16> } @llvm.vector.deinterleave2.nxv8i16(<vscale x 8 x i16> %a)46 %a.real = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %a.deinterleaved, 047 %a.imag = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %a.deinterleaved, 148 %b.deinterleaved = tail call { <vscale x 4 x i16>, <vscale x 4 x i16> } @llvm.vector.deinterleave2.nxv8i16(<vscale x 8 x i16> %b)49 %b.real = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %b.deinterleaved, 050 %b.imag = extractvalue { <vscale x 4 x i16>, <vscale x 4 x i16> } %b.deinterleaved, 151 %0 = sub <vscale x 4 x i16> %b.real, %a.imag52 %1 = add <vscale x 4 x i16> %b.imag, %a.real53 %interleaved.vec = tail call <vscale x 8 x i16> @llvm.vector.interleave2.nxv8i16(<vscale x 4 x i16> %0, <vscale x 4 x i16> %1)54 ret <vscale x 8 x i16> %interleaved.vec55}56 57; Expected to transform58define <vscale x 16 x i16> @complex_add_v16i16(<vscale x 16 x i16> %a, <vscale x 16 x i16> %b) {59; CHECK-LABEL: complex_add_v16i16:60; CHECK: // %bb.0: // %entry61; CHECK-NEXT: cadd z3.h, z3.h, z1.h, #9062; CHECK-NEXT: cadd z2.h, z2.h, z0.h, #9063; CHECK-NEXT: mov z0.d, z2.d64; CHECK-NEXT: mov z1.d, z3.d65; CHECK-NEXT: ret66entry:67 %a.deinterleaved = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.vector.deinterleave2.nxv16i16(<vscale x 16 x i16> %a)68 %a.real = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %a.deinterleaved, 069 %a.imag = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %a.deinterleaved, 170 %b.deinterleaved = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.vector.deinterleave2.nxv16i16(<vscale x 16 x i16> %b)71 %b.real = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %b.deinterleaved, 072 %b.imag = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } %b.deinterleaved, 173 %0 = sub <vscale x 8 x i16> %b.real, %a.imag74 %1 = add <vscale x 8 x i16> %b.imag, %a.real75 %interleaved.vec = tail call <vscale x 16 x i16> @llvm.vector.interleave2.nxv16i16(<vscale x 8 x i16> %0, <vscale x 8 x i16> %1)76 ret <vscale x 16 x i16> %interleaved.vec77}78 79; Expected to transform80define <vscale x 32 x i16> @complex_add_v32i16(<vscale x 32 x i16> %a, <vscale x 32 x i16> %b) {81; CHECK-LABEL: complex_add_v32i16:82; CHECK: // %bb.0: // %entry83; CHECK-NEXT: cadd z6.h, z6.h, z2.h, #9084; CHECK-NEXT: cadd z4.h, z4.h, z0.h, #9085; CHECK-NEXT: cadd z5.h, z5.h, z1.h, #9086; CHECK-NEXT: cadd z7.h, z7.h, z3.h, #9087; CHECK-NEXT: mov z0.d, z4.d88; CHECK-NEXT: mov z1.d, z5.d89; CHECK-NEXT: mov z2.d, z6.d90; CHECK-NEXT: mov z3.d, z7.d91; CHECK-NEXT: ret92entry:93 %a.deinterleaved = tail call { <vscale x 16 x i16>, <vscale x 16 x i16> } @llvm.vector.deinterleave2.nxv32i16(<vscale x 32 x i16> %a)94 %a.real = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %a.deinterleaved, 095 %a.imag = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %a.deinterleaved, 196 %b.deinterleaved = tail call { <vscale x 16 x i16>, <vscale x 16 x i16> } @llvm.vector.deinterleave2.nxv32i16(<vscale x 32 x i16> %b)97 %b.real = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %b.deinterleaved, 098 %b.imag = extractvalue { <vscale x 16 x i16>, <vscale x 16 x i16> } %b.deinterleaved, 199 %0 = sub <vscale x 16 x i16> %b.real, %a.imag100 %1 = add <vscale x 16 x i16> %b.imag, %a.real101 %interleaved.vec = tail call <vscale x 32 x i16> @llvm.vector.interleave2.nxv32i16(<vscale x 16 x i16> %0, <vscale x 16 x i16> %1)102 ret <vscale x 32 x i16> %interleaved.vec103}104 105declare { <vscale x 2 x i16>, <vscale x 2 x i16> } @llvm.vector.deinterleave2.nxv4i16(<vscale x 4 x i16>)106declare <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16>, <vscale x 2 x i16>)107 108declare { <vscale x 4 x i16>, <vscale x 4 x i16> } @llvm.vector.deinterleave2.nxv8i16(<vscale x 8 x i16>)109declare <vscale x 8 x i16> @llvm.vector.interleave2.nxv8i16(<vscale x 4 x i16>, <vscale x 4 x i16>)110 111declare { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.vector.deinterleave2.nxv16i16(<vscale x 16 x i16>)112declare <vscale x 16 x i16> @llvm.vector.interleave2.nxv16i16(<vscale x 8 x i16>, <vscale x 8 x i16>)113 114declare { <vscale x 16 x i16>, <vscale x 16 x i16> } @llvm.vector.deinterleave2.nxv32i16(<vscale x 32 x i16>)115declare <vscale x 32 x i16> @llvm.vector.interleave2.nxv32i16(<vscale x 16 x i16>, <vscale x 16 x i16>)116