derek/TheRock
1
0 Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
Files
1a182e2dac58879f215c0fad0f5f78c76dcafe56
TheRock/compiler/amd-llvm/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-opts-lifting.ll
T
derek.holloway 7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
Details
CodeQL / Analyze (python) (push) Canceled after 0s
Details
pre-commit / pre-commit (push) Canceled after 0s
Details
Multi-Arch CI / setup (push) Canceled after 0s
Details
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Details
Setup (Multi-Arch) / setup (push) Canceled after 0s
Details
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Details
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
Details
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Details
Multi-Arch CI / CI Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Details
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
Details
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

172 lines
9.1 KiB
LLVM
Raw Blame History

; RUN: opt -p lower-matrix-intrinsics -matrix-print-after-transpose-opt -disable-output -S %s 2>&1 | FileCheck %s
; REQUIRES: asserts
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
; FIXME: Lifted transpose dimensions are incorrect.
define <6 x double> @lift_through_add_matching_transpose_dimensions(<6 x double> %a, <6 x double> %b) {
; CHECK-LABEL: define <6 x double> @lift_through_add_matching_transpose_dimensions(<6 x double> %a, <6 x double> %b) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.+]] = fadd <6 x double> %a, %b
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[A]], i32 3, i32 2)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 3, i32 2)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 3, i32 2)
%add = fadd <6 x double> %a.t, %b.t
ret <6 x double> %add
}
define <6 x double> @lift_through_add_matching_transpose_dimensions_ops_also_have_shape_info(ptr %a.ptr, ptr %b.ptr) {
; CHECK-LABEL: define <6 x double> @lift_through_add_matching_transpose_dimensions_ops_also_have_shape_info(ptr %a.ptr, ptr %b.ptr)
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.+]] = load <6 x double>, ptr %a.ptr
; CHECK-NEXT: [[B:%.+]] = load <6 x double>, ptr %b.ptr
; CHECK-NEXT: [[ADD:%.+]] = fadd <6 x double> [[A]], [[B]]
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[ADD]], i32 3, i32 2)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a = load <6 x double>, ptr %a.ptr
%b = load <6 x double>, ptr %b.ptr
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 3, i32 2)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 3, i32 2)
%add = fadd <6 x double> %a.t, %b.t
ret <6 x double> %add
}
define <6 x double> @lift_through_add_mismatching_dimensions_1(<6 x double> %a, <6 x double> %b) {
; CHECK-LABEL: define <6 x double> @lift_through_add_mismatching_dimensions_1(<6 x double> %a, <6 x double> %b) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.+]] = fadd <6 x double> %a, %b
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[A]], i32 1, i32 6)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 1, i32 6)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 3, i32 2)
%add = fadd <6 x double> %a.t, %b.t
ret <6 x double> %add
}
define <6 x double> @lift_through_add_mismatching_dimensions_1_transpose_dimensions_ops_also_have_shape_info(ptr %a.ptr, ptr %b.ptr) {
; CHECK-LABEL: define <6 x double> @lift_through_add_mismatching_dimensions_1_transpose_dimensions_ops_also_have_shape_info(ptr %a.ptr, ptr %b.ptr)
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.+]] = load <6 x double>, ptr %a.ptr
; CHECK-NEXT: [[B:%.+]] = load <6 x double>, ptr %b.ptr
; CHECK-NEXT: [[ADD:%.+]] = fadd <6 x double> [[A]], [[B]]
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[ADD]], i32 1, i32 6)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a = load <6 x double>, ptr %a.ptr
%b = load <6 x double>, ptr %b.ptr
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 1, i32 6)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 3, i32 2)
%add = fadd <6 x double> %a.t, %b.t
ret <6 x double> %add
}
define <6 x double> @lift_through_add_mismatching_dimensions_2(<6 x double> %a, <6 x double> %b) {
; CHECK-LABEL: define <6 x double> @lift_through_add_mismatching_dimensions_2(<6 x double> %a, <6 x double> %b) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.+]] = fadd <6 x double> %a, %b
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[A]], i32 3, i32 2)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 3, i32 2)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 6, i32 1)
%add = fadd <6 x double> %a.t, %b.t
ret <6 x double> %add
}
define <6 x double> @lift_through_add_mismatching_dimensions_2_transpose_dimensions_ops_also_have_shape_info(ptr %a.ptr, ptr %b.ptr) {
; CHECK-LABEL: define <6 x double> @lift_through_add_mismatching_dimensions_2_transpose_dimensions_ops_also_have_shape_info(ptr %a.ptr, ptr %b.ptr)
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A:%.+]] = load <6 x double>, ptr %a.ptr
; CHECK-NEXT: [[B:%.+]] = load <6 x double>, ptr %b.ptr
; CHECK-NEXT: [[ADD:%.+]] = fadd <6 x double> [[A]], [[B]]
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[ADD]], i32 3, i32 2)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a = load <6 x double>, ptr %a.ptr
%b = load <6 x double>, ptr %b.ptr
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 3, i32 2)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 6, i32 1)
%add = fadd <6 x double> %a.t, %b.t
ret <6 x double> %add
}
define <9 x double> @lift_through_multiply(<6 x double> %a, <6 x double> %b) {
; CHECK-LABEL: define <9 x double> @lift_through_multiply(<6 x double> %a, <6 x double> %b) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[MUL:%.+]] = call <9 x double> @llvm.matrix.multiply.v9f64.v6f64.v6f64(<6 x double> %b, <6 x double> %a, i32 3, i32 2, i32 3)
; CHECK-NEXT: [[T:%.+]] = call <9 x double> @llvm.matrix.transpose.v9f64(<9 x double> [[MUL]], i32 3, i32 3)
; CHECK-NEXT: ret <9 x double> [[T]]
;
entry:
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 3, i32 2)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 2, i32 3)
%mul = call <9 x double> @llvm.matrix.multiply.v9f64.v6f64(<6 x double> %a.t, <6 x double> %b.t, i32 3, i32 2 , i32 3)
ret <9 x double> %mul
}
define <6 x double> @lift_through_multiply_2(<6 x double> %a, <4 x double> %b) {
; CHECK-LABEL: define <6 x double> @lift_through_multiply_2(<6 x double> %a, <4 x double> %b) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[MUL:%.+]] = call <6 x double> @llvm.matrix.multiply.v6f64.v4f64.v6f64(<4 x double> %b, <6 x double> %a, i32 2, i32 2, i32 3)
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[MUL]], i32 2, i32 3)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %a, i32 3, i32 2)
%b.t = call <4 x double> @llvm.matrix.transpose.v4f64(<4 x double> %b, i32 2, i32 2)
%mul = call <6 x double> @llvm.matrix.multiply.v6f64.v6f64.v4f64(<6 x double> %a.t, <4 x double> %b.t, i32 3, i32 2 , i32 2)
ret <6 x double> %mul
}
define <6 x double> @lift_through_multiply_3(<4 x double> %a, <6 x double> %b) {
; CHECK-LABEL: define <6 x double> @lift_through_multiply_3(<4 x double> %a, <6 x double> %b) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[MUL:%.+]] = call <6 x double> @llvm.matrix.multiply.v6f64.v6f64.v4f64(<6 x double> %b, <4 x double> %a, i32 3, i32 2, i32 2)
; CHECK-NEXT: [[T:%.+]] = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> [[MUL]], i32 3, i32 2)
; CHECK-NEXT: ret <6 x double> [[T]]
;
entry:
%a.t = call <4 x double> @llvm.matrix.transpose.v4f64(<4 x double> %a, i32 2, i32 2)
%b.t = call <6 x double> @llvm.matrix.transpose.v6f64(<6 x double> %b, i32 2, i32 3)
%mul = call <6 x double> @llvm.matrix.multiply.v6f64.v4f64.v6f64(<4 x double> %a.t, <6 x double> %b.t, i32 2, i32 2 , i32 3)
ret <6 x double> %mul
}
define void @test_remove_entries_from_shape_map(<3 x float> %a, <2 x float> %b, <6 x float> %c, ptr %dst) {
; CHECK-LABEL: define void @test_remove_entries_from_shape_map(
; CHECK-SAME: <3 x float> [[A:%.*]], <2 x float> [[B:%.*]], <6 x float> [[C:%.*]], ptr [[DST:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: [[TMP0:%.*]] = call <6 x float> @llvm.matrix.multiply.v6f32.v3f32.v2f32(<3 x float> [[A]], <2 x float> [[B]], i32 3, i32 1, i32 2)
; CHECK-NEXT: [[MFADD:%.*]] = fadd <6 x float> [[C]], [[TMP0]]
; CHECK-NEXT: [[MFADD_T:%.*]] = call <6 x float> @llvm.matrix.transpose.v6f32(<6 x float> [[MFADD]], i32 3, i32 2)
; CHECK-NEXT: store <6 x float> [[MFADD_T]], ptr [[DST]], align 4
; CHECK-NEXT: ret void
;
entry:
%m = tail call <6 x float> @llvm.matrix.multiply.v6f32.v3f32.v2f32(<3 x float> %a, <2 x float> %b, i32 3, i32 1, i32 2)
%add = fadd <6 x float> %c, %m
%t = tail call <6 x float> @llvm.matrix.transpose.v6f32(<6 x float> %add, i32 3, i32 2)
store <6 x float> %t, ptr %dst, align 4
ret void
}
declare <6 x double> @llvm.matrix.transpose.v6f64.v6f64(<6 x double>, i32, i32)
declare <4 x double> @llvm.matrix.transpose.v4f64.v4f64(<4 x double>, i32, i32)
declare <9 x double> @llvm.matrix.multiply.v9f64.v6f64(<6 x double>, <6 x double>, i32, i32, i32)
declare <6 x double> @llvm.matrix.multiply.v6f64.v6f64.v4f64(<6 x double>, <4 x double>, i32, i32, i32)
declare <6 x double> @llvm.matrix.multiply.v6f64.v6f64.v6f64(<6 x double>, <4 x double>, i32, i32, i32)
declare <6 x float> @llvm.matrix.transpose.v6f32(<6 x float>, i32 immarg, i32 immarg)
declare <6 x float> @llvm.matrix.multiply.v6f32.v3f32.v2f32(<3 x float>, <2 x float>, i32 immarg, i32 immarg, i32 immarg)
Reference in new issue View Git Blame Copy Permalink