derek.holloway
7502bf64a4
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
430 lines
17 KiB
LLVM
430 lines
17 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
|
|
; RUN: opt < %s -passes=instcombine -S | FileCheck %s
|
|
|
|
target triple = "aarch64-unknown-linux-gnu"
|
|
|
|
declare <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32>, <4 x i32>)
|
|
declare <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32>, <4 x i32>)
|
|
declare <8 x i16> @llvm.aarch64.neon.sshl.v8i16(<8 x i16>, <8 x i16>)
|
|
declare <8 x i16> @llvm.aarch64.neon.ushl.v8i16(<8 x i16>, <8 x i16>)
|
|
declare <16 x i8> @llvm.aarch64.neon.sshl.v16i8(<16 x i8>, <16 x i8>)
|
|
declare <16 x i8> @llvm.aarch64.neon.ushl.v16i8(<16 x i8>, <16 x i8>)
|
|
declare i64 @llvm.aarch64.neon.sshl.i64(i64, i64)
|
|
declare i64 @llvm.aarch64.neon.ushl.i64(i64, i64)
|
|
declare i32 @llvm.aarch64.neon.sshl.i32(i32, i32)
|
|
declare i32 @llvm.aarch64.neon.ushl.i32(i32, i32)
|
|
|
|
declare <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32>, <2 x i32>)
|
|
declare <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32>, <2 x i32>)
|
|
declare <4 x i16> @llvm.arm.neon.vshifts.v4i16(<4 x i16>, <4 x i16>)
|
|
declare <4 x i16> @llvm.arm.neon.vshiftu.v4i16(<4 x i16>, <4 x i16>)
|
|
declare <8 x i8> @llvm.arm.neon.vshifts.v8i8(<8 x i8>, <8 x i8>)
|
|
declare <8 x i8> @llvm.arm.neon.vshiftu.v8i8(<8 x i8>, <8 x i8>)
|
|
|
|
define <4 x i32> @sshl_positive_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_positive_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <4 x i32> [[A:%.*]], splat (i32 3)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> <i32 3, i32 3, i32 3, i32 3>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_positive_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_positive_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <4 x i32> [[A:%.*]], splat (i32 7)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> <i32 7, i32 7, i32 7, i32 7>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
; Positive non-splat constants
|
|
define <4 x i32> @ushl_positive_non_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_positive_non_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <4 x i32> [[A:%.*]], <i32 0, i32 1, i32 2, i32 3>
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> <i32 0, i32 1, i32 2, i32 3>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_zero(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_zero(
|
|
; CHECK-NEXT: ret <4 x i32> [[A:%.*]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> zeroinitializer)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <8 x i16> @sshl_i16_positive(<8 x i16> %a) {
|
|
; CHECK-LABEL: @sshl_i16_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <8 x i16> [[A:%.*]], splat (i16 4)
|
|
; CHECK-NEXT: ret <8 x i16> [[RESULT]]
|
|
;
|
|
%result = call <8 x i16> @llvm.aarch64.neon.sshl.v8i16(<8 x i16> %a, <8 x i16> splat (i16 4))
|
|
ret <8 x i16> %result
|
|
}
|
|
|
|
define <16 x i8> @ushl_i8_positive(<16 x i8> %a) {
|
|
; CHECK-LABEL: @ushl_i8_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <16 x i8> [[A:%.*]], splat (i8 2)
|
|
; CHECK-NEXT: ret <16 x i8> [[RESULT]]
|
|
;
|
|
%result = call <16 x i8> @llvm.aarch64.neon.ushl.v16i8(<16 x i8> %a, <16 x i8> splat (i8 2))
|
|
ret <16 x i8> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_negative_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_negative_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr <4 x i32> [[A:%.*]], splat (i32 5)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> splat (i32 -5))
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_negative_non_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_negative_non_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr <4 x i32> [[A:%.*]], <i32 1, i32 2, i32 3, i32 4>
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> <i32 -1, i32 -2, i32 -3, i32 -4>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_negative_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_negative_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <4 x i32> [[A:%.*]], splat (i32 2)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> splat (i32 -2))
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_negative_non_splat(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_negative_non_splat(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <4 x i32> [[A:%.*]], <i32 1, i32 2, i32 3, i32 4>
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> <i32 -1, i32 -2, i32 -3, i32 -4>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <8 x i16> @sshl_i16_negative(<8 x i16> %a) {
|
|
; CHECK-LABEL: @sshl_i16_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr <8 x i16> [[A:%.*]], splat (i16 3)
|
|
; CHECK-NEXT: ret <8 x i16> [[RESULT]]
|
|
;
|
|
%result = call <8 x i16> @llvm.aarch64.neon.sshl.v8i16(<8 x i16> %a, <8 x i16> splat (i16 -3))
|
|
ret <8 x i16> %result
|
|
}
|
|
|
|
define <16 x i8> @ushl_i8_negative(<16 x i8> %a) {
|
|
; CHECK-LABEL: @ushl_i8_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <16 x i8> [[A:%.*]], splat (i8 1)
|
|
; CHECK-NEXT: ret <16 x i8> [[RESULT]]
|
|
;
|
|
%result = call <16 x i8> @llvm.aarch64.neon.ushl.v16i8(<16 x i8> %a, <16 x i8> splat (i8 -1))
|
|
ret <16 x i8> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_i32_max_left(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_i32_max_left(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <4 x i32> [[A:%.*]], splat (i32 31)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> splat (i32 31))
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <8 x i16> @ushl_i16_max_left(<8 x i16> %a) {
|
|
; CHECK-LABEL: @ushl_i16_max_left(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <8 x i16> [[A:%.*]], splat (i16 15)
|
|
; CHECK-NEXT: ret <8 x i16> [[RESULT]]
|
|
;
|
|
%result = call <8 x i16> @llvm.aarch64.neon.ushl.v8i16(<8 x i16> %a, <8 x i16> splat (i16 15))
|
|
ret <8 x i16> %result
|
|
}
|
|
|
|
define <16 x i8> @sshl_i8_max_left(<16 x i8> %a) {
|
|
; CHECK-LABEL: @sshl_i8_max_left(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <16 x i8> [[A:%.*]], splat (i8 7)
|
|
; CHECK-NEXT: ret <16 x i8> [[RESULT]]
|
|
;
|
|
%result = call <16 x i8> @llvm.aarch64.neon.sshl.v16i8(<16 x i8> %a, <16 x i8> splat (i8 7))
|
|
ret <16 x i8> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_i32_max_right(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_i32_max_right(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <4 x i32> [[A:%.*]], splat (i32 31)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> splat (i32 -31))
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <8 x i16> @sshl_i16_max_right(<8 x i16> %a) {
|
|
; CHECK-LABEL: @sshl_i16_max_right(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr <8 x i16> [[A:%.*]], splat (i16 15)
|
|
; CHECK-NEXT: ret <8 x i16> [[RESULT]]
|
|
;
|
|
%result = call <8 x i16> @llvm.aarch64.neon.sshl.v8i16(<8 x i16> %a, <8 x i16> splat (i16 -15))
|
|
ret <8 x i16> %result
|
|
}
|
|
|
|
define <16 x i8> @ushl_i8_max_right(<16 x i8> %a) {
|
|
; CHECK-LABEL: @ushl_i8_max_right(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <16 x i8> [[A:%.*]], splat (i8 7)
|
|
; CHECK-NEXT: ret <16 x i8> [[RESULT]]
|
|
;
|
|
%result = call <16 x i8> @llvm.aarch64.neon.ushl.v16i8(<16 x i8> %a, <16 x i8> splat (i8 -7))
|
|
ret <16 x i8> %result
|
|
}
|
|
|
|
define i64 @sshl_scalar_positive(i64 %a) {
|
|
; CHECK-LABEL: @sshl_scalar_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl i64 [[A:%.*]], 5
|
|
; CHECK-NEXT: ret i64 [[RESULT]]
|
|
;
|
|
%result = call i64 @llvm.aarch64.neon.sshl.i64(i64 %a, i64 5)
|
|
ret i64 %result
|
|
}
|
|
|
|
define i64 @ushl_scalar_negative(i64 %a) {
|
|
; CHECK-LABEL: @ushl_scalar_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr i64 [[A:%.*]], 7
|
|
; CHECK-NEXT: ret i64 [[RESULT]]
|
|
;
|
|
%result = call i64 @llvm.aarch64.neon.ushl.i64(i64 %a, i64 -7)
|
|
ret i64 %result
|
|
}
|
|
|
|
define i32 @sshl_i32_scalar_negative(i32 %a) {
|
|
; CHECK-LABEL: @sshl_i32_scalar_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr i32 [[A:%.*]], 10
|
|
; CHECK-NEXT: ret i32 [[RESULT]]
|
|
;
|
|
%result = call i32 @llvm.aarch64.neon.sshl.i32(i32 %a, i32 -10)
|
|
ret i32 %result
|
|
}
|
|
|
|
define i32 @ushl_i32_scalar_positive(i32 %a) {
|
|
; CHECK-LABEL: @ushl_i32_scalar_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl i32 [[A:%.*]], 8
|
|
; CHECK-NEXT: ret i32 [[RESULT]]
|
|
;
|
|
%result = call i32 @llvm.aarch64.neon.ushl.i32(i32 %a, i32 8)
|
|
ret i32 %result
|
|
}
|
|
|
|
;; Cases that should NOT be transformed
|
|
|
|
define <4 x i32> @sshl_out_of_range_positive(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_out_of_range_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> splat (i32 32))
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> splat (i32 32))
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_out_of_range_negative(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_out_of_range_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> splat (i32 -32))
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> splat (i32 -32))
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <8 x i16> @sshl_i16_out_of_range(<8 x i16> %a) {
|
|
; CHECK-LABEL: @sshl_i16_out_of_range(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <8 x i16> @llvm.aarch64.neon.sshl.v8i16(<8 x i16> [[A:%.*]], <8 x i16> splat (i16 16))
|
|
; CHECK-NEXT: ret <8 x i16> [[RESULT]]
|
|
;
|
|
%result = call <8 x i16> @llvm.aarch64.neon.sshl.v8i16(<8 x i16> %a, <8 x i16> splat (i16 16))
|
|
ret <8 x i16> %result
|
|
}
|
|
|
|
define <16 x i8> @ushl_i8_out_of_range(<16 x i8> %a) {
|
|
; CHECK-LABEL: @ushl_i8_out_of_range(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <16 x i8> @llvm.aarch64.neon.ushl.v16i8(<16 x i8> [[A:%.*]], <16 x i8> splat (i8 -8))
|
|
; CHECK-NEXT: ret <16 x i8> [[RESULT]]
|
|
;
|
|
%result = call <16 x i8> @llvm.aarch64.neon.ushl.v16i8(<16 x i8> %a, <16 x i8> splat (i8 -8))
|
|
ret <16 x i8> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_mixed_signs(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_mixed_signs(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> <i32 1, i32 -1, i32 2, i32 -2>)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> <i32 1, i32 -1, i32 2, i32 -2>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_mixed_signs(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_mixed_signs(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> <i32 3, i32 -1, i32 4, i32 -2>)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> <i32 3, i32 -1, i32 4, i32 -2>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_zero_and_negative(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_zero_and_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> <i32 0, i32 -1, i32 -2, i32 -3>)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> <i32 0, i32 -1, i32 -2, i32 -3>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_variable(<4 x i32> %a, <4 x i32> %shift) {
|
|
; CHECK-LABEL: @sshl_variable(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[SHIFT:%.*]])
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> %shift)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_variable(<4 x i32> %a, <4 x i32> %shift) {
|
|
; CHECK-LABEL: @ushl_variable(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[SHIFT:%.*]])
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> %shift)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_mixed_valid_invalid(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_mixed_valid_invalid(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> <i32 1, i32 2, i32 31, i32 32>)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> <i32 1, i32 2, i32 31, i32 32>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <4 x i32> @ushl_mostly_negative_invalid(<4 x i32> %a) {
|
|
; CHECK-LABEL: @ushl_mostly_negative_invalid(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> [[A:%.*]], <4 x i32> <i32 -1, i32 -2, i32 -3, i32 -32>)
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.ushl.v4i32(<4 x i32> %a, <4 x i32> <i32 -1, i32 -2, i32 -3, i32 -32>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define i64 @sshl_scalar_out_of_range(i64 %a) {
|
|
; CHECK-LABEL: @sshl_scalar_out_of_range(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call i64 @llvm.aarch64.neon.sshl.i64(i64 [[A:%.*]], i64 64)
|
|
; CHECK-NEXT: ret i64 [[RESULT]]
|
|
;
|
|
%result = call i64 @llvm.aarch64.neon.sshl.i64(i64 %a, i64 64)
|
|
ret i64 %result
|
|
}
|
|
|
|
define i32 @ushl_scalar_out_of_range(i32 %a) {
|
|
; CHECK-LABEL: @ushl_scalar_out_of_range(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call i32 @llvm.aarch64.neon.ushl.i32(i32 [[A:%.*]], i32 -33)
|
|
; CHECK-NEXT: ret i32 [[RESULT]]
|
|
;
|
|
%result = call i32 @llvm.aarch64.neon.ushl.i32(i32 %a, i32 -33)
|
|
ret i32 %result
|
|
}
|
|
|
|
define <4 x i32> @sshl_positive_with_zero(<4 x i32> %a) {
|
|
; CHECK-LABEL: @sshl_positive_with_zero(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <4 x i32> [[A:%.*]], <i32 0, i32 1, i32 2, i32 3>
|
|
; CHECK-NEXT: ret <4 x i32> [[RESULT]]
|
|
;
|
|
%result = call <4 x i32> @llvm.aarch64.neon.sshl.v4i32(<4 x i32> %a, <4 x i32> <i32 0, i32 1, i32 2, i32 3>)
|
|
ret <4 x i32> %result
|
|
}
|
|
|
|
define <2 x i32> @vshifts_positive(<2 x i32> %a) {
|
|
; CHECK-LABEL: @vshifts_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <2 x i32> [[A:%.*]], splat (i32 4)
|
|
; CHECK-NEXT: ret <2 x i32> [[RESULT]]
|
|
;
|
|
%result = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> %a, <2 x i32> <i32 4, i32 4>)
|
|
ret <2 x i32> %result
|
|
}
|
|
|
|
define <2 x i32> @vshiftu_positive(<2 x i32> %a) {
|
|
; CHECK-LABEL: @vshiftu_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <2 x i32> [[A:%.*]], splat (i32 3)
|
|
; CHECK-NEXT: ret <2 x i32> [[RESULT]]
|
|
;
|
|
%result = call <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32> %a, <2 x i32> <i32 3, i32 3>)
|
|
ret <2 x i32> %result
|
|
}
|
|
|
|
define <2 x i32> @vshifts_negative(<2 x i32> %a) {
|
|
; CHECK-LABEL: @vshifts_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr <2 x i32> [[A:%.*]], splat (i32 2)
|
|
; CHECK-NEXT: ret <2 x i32> [[RESULT]]
|
|
;
|
|
%result = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> %a, <2 x i32> <i32 -2, i32 -2>)
|
|
ret <2 x i32> %result
|
|
}
|
|
|
|
define <2 x i32> @vshiftu_negative(<2 x i32> %a) {
|
|
; CHECK-LABEL: @vshiftu_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <2 x i32> [[A:%.*]], splat (i32 5)
|
|
; CHECK-NEXT: ret <2 x i32> [[RESULT]]
|
|
;
|
|
%result = call <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32> %a, <2 x i32> <i32 -5, i32 -5>)
|
|
ret <2 x i32> %result
|
|
}
|
|
|
|
define <4 x i16> @vshifts_i16_positive(<4 x i16> %a) {
|
|
; CHECK-LABEL: @vshifts_i16_positive(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = shl <4 x i16> [[A:%.*]], splat (i16 6)
|
|
; CHECK-NEXT: ret <4 x i16> [[RESULT]]
|
|
;
|
|
%result = call <4 x i16> @llvm.arm.neon.vshifts.v4i16(<4 x i16> %a, <4 x i16> <i16 6, i16 6, i16 6, i16 6>)
|
|
ret <4 x i16> %result
|
|
}
|
|
|
|
define <4 x i16> @vshiftu_i16_negative(<4 x i16> %a) {
|
|
; CHECK-LABEL: @vshiftu_i16_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = lshr <4 x i16> [[A:%.*]], splat (i16 8)
|
|
; CHECK-NEXT: ret <4 x i16> [[RESULT]]
|
|
;
|
|
%result = call <4 x i16> @llvm.arm.neon.vshiftu.v4i16(<4 x i16> %a, <4 x i16> <i16 -8, i16 -8, i16 -8, i16 -8>)
|
|
ret <4 x i16> %result
|
|
}
|
|
|
|
define <8 x i8> @vshifts_i8_negative(<8 x i8> %a) {
|
|
; CHECK-LABEL: @vshifts_i8_negative(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = ashr <8 x i8> [[A:%.*]], splat (i8 3)
|
|
; CHECK-NEXT: ret <8 x i8> [[RESULT]]
|
|
;
|
|
%result = call <8 x i8> @llvm.arm.neon.vshifts.v8i8(<8 x i8> %a, <8 x i8> <i8 -3, i8 -3, i8 -3, i8 -3, i8 -3, i8 -3, i8 -3, i8 -3>)
|
|
ret <8 x i8> %result
|
|
}
|
|
|
|
define <2 x i32> @vshifts_out_of_range(<2 x i32> %a) {
|
|
; CHECK-LABEL: @vshifts_out_of_range(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> [[A:%.*]], <2 x i32> splat (i32 32))
|
|
; CHECK-NEXT: ret <2 x i32> [[RESULT]]
|
|
;
|
|
%result = call <2 x i32> @llvm.arm.neon.vshifts.v2i32(<2 x i32> %a, <2 x i32> <i32 32, i32 32>)
|
|
ret <2 x i32> %result
|
|
}
|
|
|
|
define <2 x i32> @vshiftu_mixed_signs(<2 x i32> %a) {
|
|
; CHECK-LABEL: @vshiftu_mixed_signs(
|
|
; CHECK-NEXT: [[RESULT:%.*]] = call <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32> [[A:%.*]], <2 x i32> <i32 2, i32 -3>)
|
|
; CHECK-NEXT: ret <2 x i32> [[RESULT]]
|
|
;
|
|
%result = call <2 x i32> @llvm.arm.neon.vshiftu.v2i32(<2 x i32> %a, <2 x i32> <i32 2, i32 -3>)
|
|
ret <2 x i32> %result
|
|
}
|