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
71 lines
2.8 KiB
LLVM
71 lines
2.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; RUN: opt < %s -passes=slsr -S | FileCheck %s
|
|
; RUN: llc < %s -march=nvptx64 -mcpu=sm_75 | FileCheck %s --check-prefix=PTX
|
|
|
|
target triple = "nvptx64-nvidia-cuda"
|
|
|
|
; Test SLSR can reuse the computation by complex variable delta.
|
|
; The original program needs 4 mul.wide.s32, after SLSR with
|
|
; variable-delta, it can reduce to 1 mul.wide.s32.
|
|
define void @foo(ptr %a, ptr %b, i32 %j) {
|
|
; PTX-LABEL: foo(
|
|
; PTX: {
|
|
; PTX-NEXT: .reg .b32 %r<4>;
|
|
; PTX-NEXT: .reg .b64 %rd<9>;
|
|
; PTX-EMPTY:
|
|
; PTX-NEXT: // %bb.0:
|
|
; PTX-NEXT: ld.param.b64 %rd1, [foo_param_0];
|
|
; PTX-NEXT: ld.b32 %r1, [%rd1];
|
|
; PTX-NEXT: ld.param.b64 %rd2, [foo_param_1];
|
|
; PTX-NEXT: ld.param.b32 %r2, [foo_param_2];
|
|
; PTX-NEXT: add.s32 %r3, %r1, %r2;
|
|
; PTX-NEXT: mul.wide.s32 %rd3, %r3, 4;
|
|
; PTX-NEXT: add.s64 %rd4, %rd2, %rd3;
|
|
; PTX-NEXT: st.b32 [%rd4], 0;
|
|
; PTX-NEXT: add.s64 %rd5, %rd4, %rd3;
|
|
; PTX-NEXT: st.b32 [%rd5], 1;
|
|
; PTX-NEXT: add.s64 %rd6, %rd5, 4;
|
|
; PTX-NEXT: st.b32 [%rd5+4], 2;
|
|
; PTX-NEXT: add.s64 %rd7, %rd6, %rd3;
|
|
; PTX-NEXT: st.b32 [%rd7], 3;
|
|
; PTX-NEXT: add.s64 %rd8, %rd7, %rd3;
|
|
; PTX-NEXT: st.b32 [%rd8], 4;
|
|
; PTX-NEXT: ret;
|
|
%i.0 = load i32, ptr %a, align 8
|
|
%i = add i32 %i.0, %j
|
|
; CHECK: [[L:%.*]] = load i32, ptr %a, align 8
|
|
; CHECK: [[I:%.*]] = add i32 [[L]], %j
|
|
%gep.24 = getelementptr float, ptr %b, i32 %i
|
|
; CHECK: [[GEP0:%.*]] = getelementptr float, ptr %b, i32 [[I]]
|
|
; CHECK: store i32 0, ptr [[GEP0]]
|
|
store i32 0, ptr %gep.24
|
|
%gep.24.sum1 = add i32 %i, %i
|
|
%gep.25 = getelementptr float, ptr %b, i32 %gep.24.sum1
|
|
; CHECK: [[EXT1:%.*]] = sext i32 [[I]] to i64
|
|
; CHECK: [[MUL1:%.*]] = shl i64 [[EXT1]], 2
|
|
; CHECK: [[GEP1:%.*]] = getelementptr i8, ptr [[GEP0]], i64 [[MUL1]]
|
|
; CHECK: store i32 1, ptr [[GEP1]]
|
|
store i32 1, ptr %gep.25
|
|
%gep.26.sum3 = add i32 1, %i
|
|
%gep.27.sum = add i32 %gep.26.sum3, %i
|
|
%gep.28 = getelementptr float, ptr %b, i32 %gep.27.sum
|
|
; CHECK: [[GEP2:%.*]] = getelementptr i8, ptr [[GEP1]], i64 4
|
|
; CHECK: store i32 2, ptr [[GEP2]]
|
|
store i32 2, ptr %gep.28
|
|
%gep.28.sum = add i32 %gep.27.sum, %i
|
|
%gep.29 = getelementptr float, ptr %b, i32 %gep.28.sum
|
|
; CHECK: [[EXT2:%.*]] = sext i32 [[I]] to i64
|
|
; CHECK: [[MUL2:%.*]] = shl i64 [[EXT2]], 2
|
|
; CHECK: [[GEP3:%.*]] = getelementptr i8, ptr [[GEP2]], i64 [[MUL2]]
|
|
; CHECK: store i32 3, ptr [[GEP3]]
|
|
store i32 3, ptr %gep.29
|
|
%gep.29.sum = add i32 %gep.28.sum, %i
|
|
%gep.30 = getelementptr float, ptr %b, i32 %gep.29.sum
|
|
; CHECK: [[EXT3:%.*]] = sext i32 [[I]] to i64
|
|
; CHECK: [[MUL3:%.*]] = shl i64 [[EXT3]], 2
|
|
; CHECK: [[GEP4:%.*]] = getelementptr i8, ptr [[GEP3]], i64 [[MUL3]]
|
|
; CHECK: store i32 4, ptr [[GEP4]]
|
|
store i32 4, ptr %gep.30
|
|
ret void
|
|
}
|