CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
401 lines
18 KiB
YAML
401 lines
18 KiB
YAML
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
|
|
# RUN: llc -run-pass=machine-combiner -mcpu=neoverse-n2 -mtriple=aarch64-none-linux-gnu -verify-machineinstrs %s -o - | FileCheck %s
|
|
|
|
---
|
|
name: split_loads_to_fpr128
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4
|
|
|
|
; CHECK-LABEL: name: split_loads_to_fpr128
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSroX [[COPY]], [[COPY1]], 0, 1
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG [[LD_i32]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, [[COPY2]]
|
|
; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr32 = LDRSui [[COPY3]], 0
|
|
; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD1_0]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i32 [[SECOND_REG]], 1, [[COPY4]]
|
|
; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_1]], [[LD1_1]]
|
|
; CHECK-NEXT: $q0 = COPY [[ZIP]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:fpr32 = LDRSroX %0, %1, 0, 1
|
|
%6:fpr128 = SUBREG_TO_REG %5, %subreg.ssub
|
|
%7:fpr128 = LD1i32 %6, 1, %2
|
|
%8:fpr128 = LD1i32 %7, 2, %3
|
|
%9:fpr128 = LD1i32 %8, 3, %4
|
|
$q0 = COPY %9
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: split_loads_to_fpr128_ui
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4
|
|
|
|
; CHECK-LABEL: name: split_loads_to_fpr128_ui
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSui [[COPY]], 0
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD_i32]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY1]]
|
|
; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr32 = LDRSui [[COPY2]], 0
|
|
; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD1_0]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i32 [[SECOND_REG]], 1, killed [[COPY3]]
|
|
; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_1]], [[LD1_1]]
|
|
; CHECK-NEXT: $q0 = COPY [[ZIP]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:fpr32 = LDRSui %0, 0
|
|
%6:fpr128 = SUBREG_TO_REG killed %5, %subreg.ssub
|
|
%7:fpr128 = LD1i32 %6, 1, killed %1
|
|
%8:fpr128 = LD1i32 %7, 2, killed %2
|
|
%9:fpr128 = LD1i32 %8, 3, killed %3
|
|
$q0 = COPY %9
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: split_loads_to_fpr128_i16
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8
|
|
|
|
; CHECK-LABEL: name: split_loads_to_fpr128_i16
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[COPY5:%[0-9]+]]:gpr64common = COPY $x5
|
|
; CHECK-NEXT: [[COPY6:%[0-9]+]]:gpr64common = COPY $x6
|
|
; CHECK-NEXT: [[COPY7:%[0-9]+]]:gpr64common = COPY $x7
|
|
; CHECK-NEXT: [[COPY8:%[0-9]+]]:gpr64common = COPY $x8
|
|
; CHECK-NEXT: [[LD_i16:%[0-9]+]]:fpr16 = LDRHroX [[COPY]], killed [[COPY1]], 0, 1
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD_i16]], %subreg.hsub
|
|
; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i16 [[FIRST_REG]], 1, killed [[COPY2]]
|
|
; CHECK-NEXT: [[LD0_2:%[0-9]+]]:fpr128 = LD1i16 [[LD0_1]], 2, killed [[COPY3]]
|
|
; CHECK-NEXT: [[LD0_3:%[0-9]+]]:fpr128 = LD1i16 [[LD0_2]], 3, killed [[COPY4]]
|
|
; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr16 = LDRHui [[COPY5]], 0
|
|
; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD1_0]], %subreg.hsub
|
|
; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i16 [[SECOND_REG]], 1, killed [[COPY6]]
|
|
; CHECK-NEXT: [[LD1_2:%[0-9]+]]:fpr128 = LD1i16 [[LD1_1]], 2, killed [[COPY7]]
|
|
; CHECK-NEXT: [[LD1_3:%[0-9]+]]:fpr128 = LD1i16 [[LD1_2]], 3, killed [[COPY8]]
|
|
; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_3]], [[LD1_3]]
|
|
; CHECK-NEXT: $q0 = COPY [[ZIP]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:gpr64common = COPY $x5
|
|
%6:gpr64common = COPY $x6
|
|
%7:gpr64common = COPY $x7
|
|
%8:gpr64common = COPY $x8
|
|
%9:fpr16 = LDRHroX %0, killed %1, 0, 1
|
|
%10:fpr128 = SUBREG_TO_REG killed %9, %subreg.hsub
|
|
%11:fpr128 = LD1i16 %10, 1, killed %2
|
|
%12:fpr128 = LD1i16 %11, 2, killed %3
|
|
%13:fpr128 = LD1i16 %12, 3, killed %4
|
|
%14:fpr128 = LD1i16 %13, 4, killed %5
|
|
%15:fpr128 = LD1i16 %14, 5, killed %6
|
|
%16:fpr128 = LD1i16 %15, 6, killed %7
|
|
%17:fpr128 = LD1i16 %16, 7, killed %8
|
|
$q0 = COPY %17
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: split_loads_to_fpr128_i16_ui
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8
|
|
|
|
; CHECK-LABEL: name: split_loads_to_fpr128_i16_ui
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[COPY5:%[0-9]+]]:gpr64common = COPY $x5
|
|
; CHECK-NEXT: [[COPY6:%[0-9]+]]:gpr64common = COPY $x6
|
|
; CHECK-NEXT: [[COPY7:%[0-9]+]]:gpr64common = COPY $x7
|
|
; CHECK-NEXT: [[COPY8:%[0-9]+]]:gpr64common = COPY $x8
|
|
; CHECK-NEXT: [[LD_i16:%[0-9]+]]:fpr16 = LDRHui [[COPY]], 0
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD_i16]], %subreg.hsub
|
|
; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i16 [[FIRST_REG]], 1, killed [[COPY1]]
|
|
; CHECK-NEXT: [[LD0_2:%[0-9]+]]:fpr128 = LD1i16 [[LD0_1]], 2, killed [[COPY2]]
|
|
; CHECK-NEXT: [[LD0_3:%[0-9]+]]:fpr128 = LD1i16 [[LD0_2]], 3, killed [[COPY3]]
|
|
; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr16 = LDRHui [[COPY4]], 0
|
|
; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD1_0]], %subreg.hsub
|
|
; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i16 [[SECOND_REG]], 1, killed [[COPY5]]
|
|
; CHECK-NEXT: [[LD1_2:%[0-9]+]]:fpr128 = LD1i16 [[LD1_1]], 2, killed [[COPY6]]
|
|
; CHECK-NEXT: [[LD1_3:%[0-9]+]]:fpr128 = LD1i16 [[LD1_2]], 3, killed [[COPY7]]
|
|
; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_3]], [[LD1_3]]
|
|
; CHECK-NEXT: $q0 = COPY [[ZIP]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:gpr64common = COPY $x5
|
|
%6:gpr64common = COPY $x6
|
|
%7:gpr64common = COPY $x7
|
|
%8:gpr64common = COPY $x8
|
|
%9:fpr16 = LDRHui %0, 0
|
|
%10:fpr128 = SUBREG_TO_REG killed %9, %subreg.hsub
|
|
%11:fpr128 = LD1i16 %10, 1, killed %1
|
|
%12:fpr128 = LD1i16 %11, 2, killed %2
|
|
%13:fpr128 = LD1i16 %12, 3, killed %3
|
|
%14:fpr128 = LD1i16 %13, 4, killed %4
|
|
%15:fpr128 = LD1i16 %14, 5, killed %5
|
|
%16:fpr128 = LD1i16 %15, 6, killed %6
|
|
%17:fpr128 = LD1i16 %16, 7, killed %7
|
|
$q0 = COPY %17
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: split_loads_to_fpr128_i8
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8, $x9, $x10, $x11, $x12, $x13, $x14, $x15, $x16
|
|
|
|
; CHECK-LABEL: name: split_loads_to_fpr128_i8
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[COPY5:%[0-9]+]]:gpr64common = COPY $x5
|
|
; CHECK-NEXT: [[COPY6:%[0-9]+]]:gpr64common = COPY $x6
|
|
; CHECK-NEXT: [[COPY7:%[0-9]+]]:gpr64common = COPY $x7
|
|
; CHECK-NEXT: [[COPY8:%[0-9]+]]:gpr64common = COPY $x8
|
|
; CHECK-NEXT: [[COPY9:%[0-9]+]]:gpr64common = COPY $x9
|
|
; CHECK-NEXT: [[COPY10:%[0-9]+]]:gpr64common = COPY $x10
|
|
; CHECK-NEXT: [[COPY11:%[0-9]+]]:gpr64common = COPY $x11
|
|
; CHECK-NEXT: [[COPY12:%[0-9]+]]:gpr64common = COPY $x12
|
|
; CHECK-NEXT: [[COPY13:%[0-9]+]]:gpr64common = COPY $x13
|
|
; CHECK-NEXT: [[COPY14:%[0-9]+]]:gpr64common = COPY $x14
|
|
; CHECK-NEXT: [[COPY15:%[0-9]+]]:gpr64common = COPY $x15
|
|
; CHECK-NEXT: [[COPY16:%[0-9]+]]:gpr64common = COPY $x16
|
|
; CHECK-NEXT: [[LD_i8:%[0-9]+]]:fpr8 = LDRBroX [[COPY]], killed [[COPY1]], 0, 0
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD_i8]], %subreg.bsub
|
|
; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i8 [[FIRST_REG]], 1, killed [[COPY2]]
|
|
; CHECK-NEXT: [[LD0_2:%[0-9]+]]:fpr128 = LD1i8 [[LD0_1]], 2, killed [[COPY3]]
|
|
; CHECK-NEXT: [[LD0_3:%[0-9]+]]:fpr128 = LD1i8 [[LD0_2]], 3, killed [[COPY4]]
|
|
; CHECK-NEXT: [[LD0_4:%[0-9]+]]:fpr128 = LD1i8 [[LD0_3]], 4, killed [[COPY5]]
|
|
; CHECK-NEXT: [[LD0_5:%[0-9]+]]:fpr128 = LD1i8 [[LD0_4]], 5, killed [[COPY6]]
|
|
; CHECK-NEXT: [[LD0_6:%[0-9]+]]:fpr128 = LD1i8 [[LD0_5]], 6, killed [[COPY7]]
|
|
; CHECK-NEXT: [[LD0_7:%[0-9]+]]:fpr128 = LD1i8 [[LD0_6]], 7, killed [[COPY8]]
|
|
; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr8 = LDRBui [[COPY9]], 0
|
|
; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD1_0]], %subreg.bsub
|
|
; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i8 [[SECOND_REG]], 1, killed [[COPY10]]
|
|
; CHECK-NEXT: [[LD1_2:%[0-9]+]]:fpr128 = LD1i8 [[LD1_1]], 2, killed [[COPY11]]
|
|
; CHECK-NEXT: [[LD1_3:%[0-9]+]]:fpr128 = LD1i8 [[LD1_2]], 3, killed [[COPY12]]
|
|
; CHECK-NEXT: [[LD1_4:%[0-9]+]]:fpr128 = LD1i8 [[LD1_3]], 4, killed [[COPY13]]
|
|
; CHECK-NEXT: [[LD1_5:%[0-9]+]]:fpr128 = LD1i8 [[LD1_4]], 5, killed [[COPY14]]
|
|
; CHECK-NEXT: [[LD1_6:%[0-9]+]]:fpr128 = LD1i8 [[LD1_5]], 6, killed [[COPY15]]
|
|
; CHECK-NEXT: [[LD1_7:%[0-9]+]]:fpr128 = LD1i8 [[LD1_6]], 7, killed [[COPY16]]
|
|
; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_7]], [[LD1_7]]
|
|
; CHECK-NEXT: $q0 = COPY [[ZIP]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:gpr64common = COPY $x5
|
|
%6:gpr64common = COPY $x6
|
|
%7:gpr64common = COPY $x7
|
|
%8:gpr64common = COPY $x8
|
|
%9:gpr64common = COPY $x9
|
|
%10:gpr64common = COPY $x10
|
|
%11:gpr64common = COPY $x11
|
|
%12:gpr64common = COPY $x12
|
|
%13:gpr64common = COPY $x13
|
|
%14:gpr64common = COPY $x14
|
|
%15:gpr64common = COPY $x15
|
|
%16:gpr64common = COPY $x16
|
|
%17:fpr8 = LDRBroX %0, killed %1, 0, 0
|
|
%18:fpr128 = SUBREG_TO_REG killed %17, %subreg.bsub
|
|
%19:fpr128 = LD1i8 %18, 1, killed %2
|
|
%20:fpr128 = LD1i8 %19, 2, killed %3
|
|
%21:fpr128 = LD1i8 %20, 3, killed %4
|
|
%22:fpr128 = LD1i8 %21, 4, killed %5
|
|
%23:fpr128 = LD1i8 %22, 5, killed %6
|
|
%24:fpr128 = LD1i8 %23, 6, killed %7
|
|
%25:fpr128 = LD1i8 %24, 7, killed %8
|
|
%26:fpr128 = LD1i8 %25, 8, killed %9
|
|
%27:fpr128 = LD1i8 %26, 9, killed %10
|
|
%28:fpr128 = LD1i8 %27, 10, killed %11
|
|
%29:fpr128 = LD1i8 %28, 11, killed %12
|
|
%30:fpr128 = LD1i8 %29, 12, killed %13
|
|
%31:fpr128 = LD1i8 %30, 13, killed %14
|
|
%32:fpr128 = LD1i8 %31, 14, killed %15
|
|
%33:fpr128 = LD1i8 %32, 15, killed %16
|
|
$q0 = COPY %33
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: negative_pattern_missing_lanes
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1
|
|
|
|
; CHECK-LABEL: name: negative_pattern_missing_lanes
|
|
; CHECK: [[LD1:%.*]]:fpr128 = LDRQui $x1, 0
|
|
; CHECK-NEXT: [[LD2:%.*]]:fpr128 = LD1i32 [[LD1]]
|
|
|
|
%0:gpr64common = COPY $x0
|
|
%1:fpr128 = LDRQui $x1, 0
|
|
%2:fpr128 = LD1i32 %1, 3, %0
|
|
$q0 = COPY %2
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: out_of_order_lanes
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4
|
|
|
|
; CHECK-LABEL: name: out_of_order_lanes
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSroX [[COPY]], killed [[COPY1]], 0, 1
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD_i32]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY3]]
|
|
; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr32 = LDRSui [[COPY2]], 0
|
|
; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD1_0]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i32 [[SECOND_REG]], 1, killed [[COPY4]]
|
|
; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_1]], [[LD1_1]]
|
|
; CHECK-NEXT: $q0 = COPY [[ZIP]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:fpr32 = LDRSroX %0, killed %1, 0, 1
|
|
%6:fpr128 = SUBREG_TO_REG killed %5, %subreg.ssub
|
|
%7:fpr128 = LD1i32 %6, 2, killed %2
|
|
%8:fpr128 = LD1i32 %7, 1, killed %3
|
|
%9:fpr128 = LD1i32 %8, 3, killed %4
|
|
$q0 = COPY %9
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: negative_pattern_no_subreg_to_reg
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3
|
|
|
|
; CHECK-LABEL: name: negative_pattern_no_subreg_to_reg
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[INITIAL_VEC:%[0-9]+]]:fpr128 = LDRQui [[COPY]], 0
|
|
; CHECK-NEXT: [[LD_LANE_1:%[0-9]+]]:fpr128 = LD1i32 [[INITIAL_VEC]], 1, killed [[COPY1]]
|
|
; CHECK-NEXT: [[LD_LANE_2:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_1]], 2, killed [[COPY2]]
|
|
; CHECK-NEXT: [[LD_LANE_3:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_2]], 3, killed [[COPY3]]
|
|
; CHECK-NEXT: $q0 = COPY [[LD_LANE_3]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:fpr128 = LDRQui %0, 0
|
|
%5:fpr128 = LD1i32 %4, 1, killed %1
|
|
%6:fpr128 = LD1i32 %5, 2, killed %2
|
|
%7:fpr128 = LD1i32 %6, 3, killed %3
|
|
$q0 = COPY %7
|
|
RET_ReallyLR implicit $q0
|
|
|
|
---
|
|
name: negative_pattern_multiple_users
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3, $x4
|
|
|
|
; CHECK-LABEL: name: negative_pattern_multiple_users
|
|
; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
|
|
; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSroX [[COPY]], killed [[COPY1]], 0, 1
|
|
; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LD_i32]], %subreg.ssub
|
|
; CHECK-NEXT: [[LD_LANE_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY2]]
|
|
; CHECK-NEXT: [[LD_LANE_2:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_1]], 2, killed [[COPY3]]
|
|
; CHECK-NEXT: [[LD_LANE_3:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_2]], 3, killed [[COPY4]]
|
|
; CHECK-NEXT: $q0 = COPY [[LD_LANE_3]]
|
|
; CHECK-NEXT: $q1 = COPY [[LD_LANE_2]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0, implicit $q1
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%4:gpr64common = COPY $x4
|
|
%5:fpr32 = LDRSroX %0, killed %1, 0, 1
|
|
%6:fpr128 = SUBREG_TO_REG killed %5, %subreg.ssub
|
|
%7:fpr128 = LD1i32 %6, 1, killed %2
|
|
%8:fpr128 = LD1i32 %7, 2, killed %3
|
|
%9:fpr128 = LD1i32 %8, 3, killed %4
|
|
$q0 = COPY %9
|
|
$q1 = COPY %8
|
|
RET_ReallyLR implicit $q0, implicit $q1
|
|
|
|
---
|
|
name: aliasing_store_between_vector_loads
|
|
alignment: 4
|
|
tracksRegLiveness: true
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $x0, $x1, $x2, $x3
|
|
|
|
; CHECK-LABEL: name: aliasing_store_between_vector_loads
|
|
; CHECK: [[BASE_PTR:%[0-9]+]]:gpr64common = COPY $x0
|
|
; CHECK-NEXT: [[OFFSET_PTR:%[0-9]+]]:gpr64common = COPY $x1
|
|
; CHECK-NEXT: [[ALIAS_ADDR:%[0-9]+]]:gpr64common = COPY $x2
|
|
; CHECK-NEXT: [[OTHER_ADDR:%[0-9]+]]:gpr64common = COPY $x3
|
|
; CHECK-NEXT: [[LOAD0:%[0-9]+]]:fpr32 = LDRSroX [[BASE_PTR]], killed [[OFFSET_PTR]], 0, 1
|
|
; CHECK-NEXT: [[VEC0:%[0-9]+]]:fpr128 = SUBREG_TO_REG killed [[LOAD0]], %subreg.ssub
|
|
; CHECK-NEXT: [[VEC1:%[0-9]+]]:fpr128 = LD1i32 [[VEC0]], 1, [[ALIAS_ADDR]]
|
|
; CHECK-NEXT: [[CONST:%[0-9]+]]:gpr32 = MOVi32imm 99
|
|
; CHECK-NEXT: STRWui [[CONST]], [[ALIAS_ADDR]], 0
|
|
; CHECK-NEXT: [[VEC2:%[0-9]+]]:fpr128 = LD1i32 [[VEC1]], 2, killed [[ALIAS_ADDR]]
|
|
; CHECK-NEXT: [[VEC3:%[0-9]+]]:fpr128 = LD1i32 [[VEC2]], 3, killed [[OTHER_ADDR]]
|
|
; CHECK-NEXT: $q0 = COPY [[VEC3]]
|
|
; CHECK-NEXT: RET_ReallyLR implicit $q0
|
|
%0:gpr64common = COPY $x0
|
|
%1:gpr64common = COPY $x1
|
|
%2:gpr64common = COPY $x2
|
|
%3:gpr64common = COPY $x3
|
|
%5:fpr32 = LDRSroX %0, killed %1, 0, 1
|
|
%6:fpr128 = SUBREG_TO_REG killed %5, %subreg.ssub
|
|
%7:fpr128 = LD1i32 %6, 1, %2
|
|
%10:gpr32 = MOVi32imm 99
|
|
STRWui %10, %2, 0
|
|
%8:fpr128 = LD1i32 %7, 2, killed %2
|
|
%9:fpr128 = LD1i32 %8, 3, killed %3
|
|
$q0 = COPY %9
|
|
RET_ReallyLR implicit $q0
|