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
158 lines
4.8 KiB
LLVM
158 lines
4.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
|
|
|
|
; i256 shift left with variable amount
|
|
define i256 @shl_i256(i256 %a, i256 %amt) nounwind {
|
|
; CHECK-LABEL: shl_i256:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: sub sp, sp, #64
|
|
; CHECK-NEXT: movi v0.2d, #0000000000000000
|
|
; CHECK-NEXT: lsr x8, x4, #3
|
|
; CHECK-NEXT: mov x9, sp
|
|
; CHECK-NEXT: add x9, x9, #32
|
|
; CHECK-NEXT: stp x2, x3, [sp, #48]
|
|
; CHECK-NEXT: mvn w14, w4
|
|
; CHECK-NEXT: and x8, x8, #0x18
|
|
; CHECK-NEXT: stp x0, x1, [sp, #32]
|
|
; CHECK-NEXT: and x11, x4, #0x3f
|
|
; CHECK-NEXT: sub x8, x9, x8
|
|
; CHECK-NEXT: eor x11, x11, #0x3f
|
|
; CHECK-NEXT: stp q0, q0, [sp]
|
|
; CHECK-NEXT: ldp x10, x9, [x8]
|
|
; CHECK-NEXT: ldp x12, x8, [x8, #16]
|
|
; CHECK-NEXT: lsr x13, x9, #1
|
|
; CHECK-NEXT: lsr x15, x10, #1
|
|
; CHECK-NEXT: lsl x9, x9, x4
|
|
; CHECK-NEXT: lsl x8, x8, x4
|
|
; CHECK-NEXT: lsl x0, x10, x4
|
|
; CHECK-NEXT: lsr x13, x13, x14
|
|
; CHECK-NEXT: lsr x14, x12, #1
|
|
; CHECK-NEXT: lsr x15, x15, x11
|
|
; CHECK-NEXT: lsl x12, x12, x4
|
|
; CHECK-NEXT: lsr x11, x14, x11
|
|
; CHECK-NEXT: orr x1, x9, x15
|
|
; CHECK-NEXT: orr x2, x12, x13
|
|
; CHECK-NEXT: orr x3, x8, x11
|
|
; CHECK-NEXT: add sp, sp, #64
|
|
; CHECK-NEXT: ret
|
|
%r = shl i256 %a, %amt
|
|
ret i256 %r
|
|
}
|
|
|
|
; i256 logical shift right with variable amount
|
|
define i256 @lshr_i256(i256 %a, i256 %amt) nounwind {
|
|
; CHECK-LABEL: lshr_i256:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: sub sp, sp, #64
|
|
; CHECK-NEXT: movi v0.2d, #0000000000000000
|
|
; CHECK-NEXT: lsr x8, x4, #3
|
|
; CHECK-NEXT: mov x9, sp
|
|
; CHECK-NEXT: stp x2, x3, [sp, #16]
|
|
; CHECK-NEXT: and x11, x4, #0x3f
|
|
; CHECK-NEXT: mvn w13, w4
|
|
; CHECK-NEXT: and x8, x8, #0x18
|
|
; CHECK-NEXT: stp x0, x1, [sp]
|
|
; CHECK-NEXT: eor x11, x11, #0x3f
|
|
; CHECK-NEXT: add x8, x9, x8
|
|
; CHECK-NEXT: stp q0, q0, [sp, #32]
|
|
; CHECK-NEXT: ldp x9, x10, [x8]
|
|
; CHECK-NEXT: ldp x12, x8, [x8, #16]
|
|
; CHECK-NEXT: lsl x14, x10, #1
|
|
; CHECK-NEXT: lsr x9, x9, x4
|
|
; CHECK-NEXT: lsr x10, x10, x4
|
|
; CHECK-NEXT: lsl x15, x12, #1
|
|
; CHECK-NEXT: lsl x16, x8, #1
|
|
; CHECK-NEXT: lsr x12, x12, x4
|
|
; CHECK-NEXT: lsl x14, x14, x11
|
|
; CHECK-NEXT: lsr x3, x8, x4
|
|
; CHECK-NEXT: lsl x8, x15, x13
|
|
; CHECK-NEXT: lsl x11, x16, x11
|
|
; CHECK-NEXT: orr x0, x14, x9
|
|
; CHECK-NEXT: orr x1, x10, x8
|
|
; CHECK-NEXT: orr x2, x11, x12
|
|
; CHECK-NEXT: add sp, sp, #64
|
|
; CHECK-NEXT: ret
|
|
%r = lshr i256 %a, %amt
|
|
ret i256 %r
|
|
}
|
|
|
|
; i256 arithmetic shift right with variable amount
|
|
define i256 @ashr_i256(i256 %a, i256 %amt) nounwind {
|
|
; CHECK-LABEL: ashr_i256:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: sub sp, sp, #64
|
|
; CHECK-NEXT: lsr x9, x4, #3
|
|
; CHECK-NEXT: asr x8, x3, #63
|
|
; CHECK-NEXT: mov x10, sp
|
|
; CHECK-NEXT: stp x2, x3, [sp, #16]
|
|
; CHECK-NEXT: mvn w14, w4
|
|
; CHECK-NEXT: and x13, x4, #0x3f
|
|
; CHECK-NEXT: and x9, x9, #0x18
|
|
; CHECK-NEXT: stp x0, x1, [sp]
|
|
; CHECK-NEXT: eor x13, x13, #0x3f
|
|
; CHECK-NEXT: stp x8, x8, [sp, #48]
|
|
; CHECK-NEXT: add x9, x10, x9
|
|
; CHECK-NEXT: stp x8, x8, [sp, #32]
|
|
; CHECK-NEXT: ldp x10, x8, [x9, #8]
|
|
; CHECK-NEXT: ldr x11, [x9]
|
|
; CHECK-NEXT: ldr x9, [x9, #24]
|
|
; CHECK-NEXT: lsr x11, x11, x4
|
|
; CHECK-NEXT: lsl x12, x8, #1
|
|
; CHECK-NEXT: lsl x15, x10, #1
|
|
; CHECK-NEXT: lsr x10, x10, x4
|
|
; CHECK-NEXT: lsr x8, x8, x4
|
|
; CHECK-NEXT: asr x3, x9, x4
|
|
; CHECK-NEXT: lsl x12, x12, x14
|
|
; CHECK-NEXT: lsl x14, x9, #1
|
|
; CHECK-NEXT: lsl x15, x15, x13
|
|
; CHECK-NEXT: lsl x13, x14, x13
|
|
; CHECK-NEXT: orr x0, x15, x11
|
|
; CHECK-NEXT: orr x1, x10, x12
|
|
; CHECK-NEXT: orr x2, x13, x8
|
|
; CHECK-NEXT: add sp, sp, #64
|
|
; CHECK-NEXT: ret
|
|
%r = ashr i256 %a, %amt
|
|
ret i256 %r
|
|
}
|
|
|
|
; i256 shift left by constant
|
|
define i256 @shl_i256_const(i256 %a) nounwind {
|
|
; CHECK-LABEL: shl_i256_const:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: extr x8, x2, x1, #47
|
|
; CHECK-NEXT: lsl x9, x0, #17
|
|
; CHECK-NEXT: extr x1, x1, x0, #47
|
|
; CHECK-NEXT: extr x3, x3, x2, #47
|
|
; CHECK-NEXT: mov x0, x9
|
|
; CHECK-NEXT: mov x2, x8
|
|
; CHECK-NEXT: ret
|
|
%r = shl i256 %a, 17
|
|
ret i256 %r
|
|
}
|
|
|
|
; i256 logical shift right by constant
|
|
define i256 @lshr_i256_const(i256 %a) nounwind {
|
|
; CHECK-LABEL: lshr_i256_const:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: extr x0, x1, x0, #17
|
|
; CHECK-NEXT: extr x1, x2, x1, #17
|
|
; CHECK-NEXT: extr x2, x3, x2, #17
|
|
; CHECK-NEXT: lsr x3, x3, #17
|
|
; CHECK-NEXT: ret
|
|
%r = lshr i256 %a, 17
|
|
ret i256 %r
|
|
}
|
|
|
|
; i256 arithmetic shift right by constant
|
|
define i256 @ashr_i256_const(i256 %a) nounwind {
|
|
; CHECK-LABEL: ashr_i256_const:
|
|
; CHECK: // %bb.0:
|
|
; CHECK-NEXT: extr x0, x1, x0, #17
|
|
; CHECK-NEXT: extr x1, x2, x1, #17
|
|
; CHECK-NEXT: extr x2, x3, x2, #17
|
|
; CHECK-NEXT: asr x3, x3, #17
|
|
; CHECK-NEXT: ret
|
|
%r = ashr i256 %a, 17
|
|
ret i256 %r
|
|
}
|