Files
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
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

223 lines
8.1 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc < %s -mtriple=nvptx -mcpu=sm_61 | FileCheck %s
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_61 | FileCheck %s
target triple = "nvptx-nvidia-cuda"
declare i32 @llvm.nvvm.idp4a.s.s(i32, i32, i32)
declare i32 @llvm.nvvm.idp4a.s.u(i32, i32, i32)
declare i32 @llvm.nvvm.idp4a.u.s(i32, i32, i32)
declare i32 @llvm.nvvm.idp4a.u.u(i32, i32, i32)
define i32 @test_dp4a_u32_u32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp4a_u32_u32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp4a_u32_u32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp4a_u32_u32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp4a_u32_u32_param_2];
; CHECK-NEXT: dp4a.u32.u32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp4a.u.u(i32 %a, i32 %b, i32 %c)
ret i32 %call
}
define i32 @test_dp4a_u32imm_u32imm(i32 %c) {
; CHECK-LABEL: test_dp4a_u32imm_u32imm(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp4a_u32imm_u32imm_param_0];
; CHECK-NEXT: mov.b32 %r2, 0;
; CHECK-NEXT: dp4a.u32.u32 %r3, %r2, %r2, %r1;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp4a.u.u(i32 0, i32 0, i32 %c)
ret i32 %call
}
define i32 @test_dp4a_u32_s32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp4a_u32_s32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp4a_u32_s32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp4a_u32_s32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp4a_u32_s32_param_2];
; CHECK-NEXT: dp4a.u32.s32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp4a.u.s(i32 %a, i32 %b, i32 %c)
ret i32 %call
}
define i32 @test_dp4a_s32_u32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp4a_s32_u32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp4a_s32_u32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp4a_s32_u32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp4a_s32_u32_param_2];
; CHECK-NEXT: dp4a.s32.u32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp4a.s.u(i32 %a, i32 %b, i32 %c)
ret i32 %call
}
define i32 @test_dp4a_s32_s32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp4a_s32_s32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp4a_s32_s32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp4a_s32_s32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp4a_s32_s32_param_2];
; CHECK-NEXT: dp4a.s32.s32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp4a.s.s(i32 %a, i32 %b, i32 %c)
ret i32 %call
}
declare i32 @llvm.nvvm.idp2a.s.s(i32, i32, i1 immarg, i32)
declare i32 @llvm.nvvm.idp2a.s.u(i32, i32, i1 immarg, i32)
declare i32 @llvm.nvvm.idp2a.u.s(i32, i32, i1 immarg, i32)
declare i32 @llvm.nvvm.idp2a.u.u(i32, i32, i1 immarg, i32)
define i32 @test_dp2a_lo_u32_u32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_lo_u32_u32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_lo_u32_u32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_lo_u32_u32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_lo_u32_u32_param_2];
; CHECK-NEXT: dp2a.lo.u32.u32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.u.u(i32 %a, i32 %b, i1 0, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_lo_u32_s32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_lo_u32_s32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_lo_u32_s32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_lo_u32_s32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_lo_u32_s32_param_2];
; CHECK-NEXT: dp2a.lo.u32.s32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.u.s(i32 %a, i32 %b, i1 0, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_lo_s32_u32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_lo_s32_u32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_lo_s32_u32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_lo_s32_u32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_lo_s32_u32_param_2];
; CHECK-NEXT: dp2a.lo.s32.u32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.s.u(i32 %a, i32 %b, i1 0, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_lo_s32_s32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_lo_s32_s32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_lo_s32_s32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_lo_s32_s32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_lo_s32_s32_param_2];
; CHECK-NEXT: dp2a.lo.s32.s32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.s.s(i32 %a, i32 %b, i1 0, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_hi_u32_u32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_hi_u32_u32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_hi_u32_u32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_hi_u32_u32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_hi_u32_u32_param_2];
; CHECK-NEXT: dp2a.hi.u32.u32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.u.u(i32 %a, i32 %b, i1 1, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_hi_u32_s32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_hi_u32_s32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_hi_u32_s32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_hi_u32_s32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_hi_u32_s32_param_2];
; CHECK-NEXT: dp2a.hi.u32.s32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.u.s(i32 %a, i32 %b, i1 1, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_hi_s32_u32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_hi_s32_u32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_hi_s32_u32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_hi_s32_u32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_hi_s32_u32_param_2];
; CHECK-NEXT: dp2a.hi.s32.u32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.s.u(i32 %a, i32 %b, i1 1, i32 %c)
ret i32 %call
}
define i32 @test_dp2a_hi_s32_s32(i32 %a, i32 %b, i32 %c) {
; CHECK-LABEL: test_dp2a_hi_s32_s32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [test_dp2a_hi_s32_s32_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [test_dp2a_hi_s32_s32_param_1];
; CHECK-NEXT: ld.param.b32 %r3, [test_dp2a_hi_s32_s32_param_2];
; CHECK-NEXT: dp2a.hi.s32.s32 %r4, %r1, %r2, %r3;
; CHECK-NEXT: st.param.b32 [func_retval0], %r4;
; CHECK-NEXT: ret;
%call = call i32 @llvm.nvvm.idp2a.s.s(i32 %a, i32 %b, i1 1, i32 %c)
ret i32 %call
}