|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 |
| 2 | +; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu \ |
| 3 | +; RUN: -mcpu=future -ppc-asm-full-reg-names \ |
| 4 | +; RUN: -ppc-vsr-nums-as-vr < %s | FileCheck %s |
| 5 | +; RUN: llc -verify-machineinstrs -mtriple=powerpc64-ibm-aix \ |
| 6 | +; RUN: -mcpu=future -ppc-asm-full-reg-names \ |
| 7 | +; RUN: -ppc-vsr-nums-as-vr < %s | FileCheck %s --check-prefix=CHECK-BE |
| 8 | + |
| 9 | +define void @tdmrz(ptr nocapture readonly %vp1, ptr nocapture %resp) { |
| 10 | +; CHECK-LABEL: tdmrz: |
| 11 | +; CHECK: # %bb.0: # %entry |
| 12 | +; CHECK-NEXT: dmsetdmrz dmr0 |
| 13 | +; CHECK-NEXT: dmxxextfdmr512 wacc0, vsp34, vsp36, 0 |
| 14 | +; CHECK-NEXT: stxvp vsp34, 96(r4) |
| 15 | +; CHECK-NEXT: stxvp vsp36, 64(r4) |
| 16 | +; CHECK-NEXT: dmxxextfdmr512 wacc_hi0, vsp34, vsp36, 1 |
| 17 | +; CHECK-NEXT: stxvp vsp34, 32(r4) |
| 18 | +; CHECK-NEXT: stxvp vsp36, 0(r4) |
| 19 | +; CHECK-NEXT: blr |
| 20 | +; |
| 21 | +; CHECK-BE-LABEL: tdmrz: |
| 22 | +; CHECK-BE: # %bb.0: # %entry |
| 23 | +; CHECK-BE-NEXT: dmsetdmrz dmr0 |
| 24 | +; CHECK-BE-NEXT: dmxxextfdmr512 wacc_hi0, vsp34, vsp36, 1 |
| 25 | +; CHECK-BE-NEXT: stxvp vsp36, 96(r4) |
| 26 | +; CHECK-BE-NEXT: stxvp vsp34, 64(r4) |
| 27 | +; CHECK-BE-NEXT: dmxxextfdmr512 wacc0, vsp34, vsp36, 0 |
| 28 | +; CHECK-BE-NEXT: stxvp vsp36, 32(r4) |
| 29 | +; CHECK-BE-NEXT: stxvp vsp34, 0(r4) |
| 30 | +; CHECK-BE-NEXT: blr |
| 31 | +entry: |
| 32 | + %z = call <1024 x i1> @llvm.ppc.mma.dmsetdmrz() |
| 33 | + store <1024 x i1> %z, ptr %resp, align 32 |
| 34 | + ret void |
| 35 | +} |
| 36 | + |
| 37 | +define void @tdmmr(ptr nocapture readonly %vp1, ptr nocapture %resp) { |
| 38 | +; CHECK-LABEL: tdmmr: |
| 39 | +; CHECK: # %bb.0: # %entry |
| 40 | +; CHECK-NEXT: lxvp vsp34, 0(r3) |
| 41 | +; CHECK-NEXT: lxvp vsp36, 32(r3) |
| 42 | +; CHECK-NEXT: dmxxinstfdmr512 wacc_hi0, vsp36, vsp34, 1 |
| 43 | +; CHECK-NEXT: lxvp vsp34, 64(r3) |
| 44 | +; CHECK-NEXT: lxvp vsp36, 96(r3) |
| 45 | +; CHECK-NEXT: dmxxinstfdmr512 wacc0, vsp36, vsp34, 0 |
| 46 | +; CHECK-NEXT: dmmr dmr0, dmr0 |
| 47 | +; CHECK-NEXT: dmxxextfdmr512 wacc0, vsp34, vsp36, 0 |
| 48 | +; CHECK-NEXT: stxvp vsp34, 96(r4) |
| 49 | +; CHECK-NEXT: stxvp vsp36, 64(r4) |
| 50 | +; CHECK-NEXT: dmxxextfdmr512 wacc_hi0, vsp34, vsp36, 1 |
| 51 | +; CHECK-NEXT: stxvp vsp34, 32(r4) |
| 52 | +; CHECK-NEXT: stxvp vsp36, 0(r4) |
| 53 | +; CHECK-NEXT: blr |
| 54 | +; |
| 55 | +; CHECK-BE-LABEL: tdmmr: |
| 56 | +; CHECK-BE: # %bb.0: # %entry |
| 57 | +; CHECK-BE-NEXT: lxvp vsp34, 96(r3) |
| 58 | +; CHECK-BE-NEXT: lxvp vsp36, 64(r3) |
| 59 | +; CHECK-BE-NEXT: dmxxinstfdmr512 wacc_hi0, vsp36, vsp34, 1 |
| 60 | +; CHECK-BE-NEXT: lxvp vsp34, 32(r3) |
| 61 | +; CHECK-BE-NEXT: lxvp vsp36, 0(r3) |
| 62 | +; CHECK-BE-NEXT: dmxxinstfdmr512 wacc0, vsp36, vsp34, 0 |
| 63 | +; CHECK-BE-NEXT: dmmr dmr0, dmr0 |
| 64 | +; CHECK-BE-NEXT: dmxxextfdmr512 wacc_hi0, vsp34, vsp36, 1 |
| 65 | +; CHECK-BE-NEXT: stxvp vsp36, 96(r4) |
| 66 | +; CHECK-BE-NEXT: stxvp vsp34, 64(r4) |
| 67 | +; CHECK-BE-NEXT: dmxxextfdmr512 wacc0, vsp34, vsp36, 0 |
| 68 | +; CHECK-BE-NEXT: stxvp vsp36, 32(r4) |
| 69 | +; CHECK-BE-NEXT: stxvp vsp34, 0(r4) |
| 70 | +; CHECK-BE-NEXT: blr |
| 71 | +entry: |
| 72 | + %l = load <1024 x i1>, ptr %vp1, align 32 |
| 73 | + %c = call <1024 x i1> @llvm.ppc.mma.dmmr(<1024 x i1> %l) |
| 74 | + store <1024 x i1> %c, ptr %resp, align 32 |
| 75 | + ret void |
| 76 | +} |
| 77 | + |
| 78 | +define void @tdmxor(ptr nocapture readonly %vp1, ptr %vp2, ptr nocapture %resp) { |
| 79 | +; CHECK-LABEL: tdmxor: |
| 80 | +; CHECK: # %bb.0: # %entry |
| 81 | +; CHECK-NEXT: lxvp vsp34, 0(r3) |
| 82 | +; CHECK-NEXT: lxvp vsp36, 32(r3) |
| 83 | +; CHECK-NEXT: dmxxinstfdmr512 wacc_hi0, vsp36, vsp34, 1 |
| 84 | +; CHECK-NEXT: lxvp vsp34, 64(r3) |
| 85 | +; CHECK-NEXT: lxvp vsp36, 96(r3) |
| 86 | +; CHECK-NEXT: dmxxinstfdmr512 wacc0, vsp36, vsp34, 0 |
| 87 | +; CHECK-NEXT: lxvp vsp34, 0(r4) |
| 88 | +; CHECK-NEXT: lxvp vsp36, 32(r4) |
| 89 | +; CHECK-NEXT: dmxxinstfdmr512 wacc_hi1, vsp36, vsp34, 1 |
| 90 | +; CHECK-NEXT: lxvp vsp34, 64(r4) |
| 91 | +; CHECK-NEXT: lxvp vsp36, 96(r4) |
| 92 | +; CHECK-NEXT: dmxxinstfdmr512 wacc1, vsp36, vsp34, 0 |
| 93 | +; CHECK-NEXT: dmxor dmr0, dmr1 |
| 94 | +; CHECK-NEXT: dmxxextfdmr512 wacc0, vsp34, vsp36, 0 |
| 95 | +; CHECK-NEXT: stxvp vsp34, 96(r5) |
| 96 | +; CHECK-NEXT: stxvp vsp36, 64(r5) |
| 97 | +; CHECK-NEXT: dmxxextfdmr512 wacc_hi0, vsp34, vsp36, 1 |
| 98 | +; CHECK-NEXT: stxvp vsp34, 32(r5) |
| 99 | +; CHECK-NEXT: stxvp vsp36, 0(r5) |
| 100 | +; CHECK-NEXT: blr |
| 101 | +; |
| 102 | +; CHECK-BE-LABEL: tdmxor: |
| 103 | +; CHECK-BE: # %bb.0: # %entry |
| 104 | +; CHECK-BE-NEXT: lxvp vsp34, 96(r3) |
| 105 | +; CHECK-BE-NEXT: lxvp vsp36, 64(r3) |
| 106 | +; CHECK-BE-NEXT: dmxxinstfdmr512 wacc_hi0, vsp36, vsp34, 1 |
| 107 | +; CHECK-BE-NEXT: lxvp vsp34, 32(r3) |
| 108 | +; CHECK-BE-NEXT: lxvp vsp36, 0(r3) |
| 109 | +; CHECK-BE-NEXT: dmxxinstfdmr512 wacc0, vsp36, vsp34, 0 |
| 110 | +; CHECK-BE-NEXT: lxvp vsp34, 96(r4) |
| 111 | +; CHECK-BE-NEXT: lxvp vsp36, 64(r4) |
| 112 | +; CHECK-BE-NEXT: dmxxinstfdmr512 wacc_hi1, vsp36, vsp34, 1 |
| 113 | +; CHECK-BE-NEXT: lxvp vsp34, 32(r4) |
| 114 | +; CHECK-BE-NEXT: lxvp vsp36, 0(r4) |
| 115 | +; CHECK-BE-NEXT: dmxxinstfdmr512 wacc1, vsp36, vsp34, 0 |
| 116 | +; CHECK-BE-NEXT: dmxor dmr0, dmr1 |
| 117 | +; CHECK-BE-NEXT: dmxxextfdmr512 wacc_hi0, vsp34, vsp36, 1 |
| 118 | +; CHECK-BE-NEXT: stxvp vsp36, 96(r5) |
| 119 | +; CHECK-BE-NEXT: stxvp vsp34, 64(r5) |
| 120 | +; CHECK-BE-NEXT: dmxxextfdmr512 wacc0, vsp34, vsp36, 0 |
| 121 | +; CHECK-BE-NEXT: stxvp vsp36, 32(r5) |
| 122 | +; CHECK-BE-NEXT: stxvp vsp34, 0(r5) |
| 123 | +; CHECK-BE-NEXT: blr |
| 124 | +entry: |
| 125 | + %l = load <1024 x i1>, ptr %vp1, align 32 |
| 126 | + %r = load <1024 x i1>, ptr %vp2, align 32 |
| 127 | + %x = call <1024 x i1> @llvm.ppc.mma.dmxor(<1024 x i1> %l, <1024 x i1> %r) |
| 128 | + store <1024 x i1> %x, ptr %resp, align 32 |
| 129 | + ret void |
| 130 | +} |
| 131 | + |
| 132 | +declare <1024 x i1> @llvm.ppc.mma.dmsetdmrz() |
| 133 | +declare <1024 x i1> @llvm.ppc.mma.dmmr(<1024 x i1>) |
| 134 | +declare <1024 x i1> @llvm.ppc.mma.dmxor(<1024 x i1>, <1024 x i1>) |
0 commit comments