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/CodeGen/NVPTX/convert-sm80-sf.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

279 lines
10 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx81 | FileCheck %s
; RUN: %if ptxas-sm_80 && ptxas-isa-8.1 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx81 | %ptxas-verify -arch=sm_80 %}
; CHECK-LABEL: cvt_rna_satfinite_tf32_f32
define i32 @cvt_rna_satfinite_tf32_f32(float %f1) {
; CHECK-LABEL: cvt_rna_satfinite_tf32_f32(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<3>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rna_satfinite_tf32_f32_param_0];
; CHECK-NEXT: cvt.rna.satfinite.tf32.f32 %r2, %r1;
; CHECK-NEXT: st.param.b32 [func_retval0], %r2;
; CHECK-NEXT: ret;
%val = call i32 @llvm.nvvm.f2tf32.rna.satfinite(float %f1)
ret i32 %val
}
define <2 x bfloat> @cvt_rn_bf16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rn_bf16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_bf16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rn_bf16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rn.satfinite.bf16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.satfinite(float %f1, float %f2)
ret <2 x bfloat> %val
}
define <2 x bfloat> @cvt_rn_relu_bf16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rn_relu_bf16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_relu_bf16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rn_relu_bf16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rn.relu.satfinite.bf16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.relu.satfinite(float %f1, float %f2)
ret <2 x bfloat> %val
}
define <2 x bfloat> @cvt_rz_bf16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rz_bf16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_bf16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rz_bf16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rz.satfinite.bf16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.satfinite(float %f1, float %f2)
ret <2 x bfloat> %val
}
define <2 x bfloat> @cvt_rz_relu_bf16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rz_relu_bf16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_relu_bf16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rz_relu_bf16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rz.relu.satfinite.bf16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.relu.satfinite(float %f1, float %f2)
ret <2 x bfloat> %val
}
declare <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.satfinite(float, float)
declare <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.relu.satfinite(float, float)
declare <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.satfinite(float, float)
declare <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.relu.satfinite(float, float)
define <2 x half> @cvt_rn_f16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rn_f16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_f16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rn_f16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rn.satfinite.f16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x half> @llvm.nvvm.ff2f16x2.rn.satfinite(float %f1, float %f2)
ret <2 x half> %val
}
define <2 x half> @cvt_rn_relu_f16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rn_relu_f16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_relu_f16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rn_relu_f16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rn.relu.satfinite.f16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x half> @llvm.nvvm.ff2f16x2.rn.relu.satfinite(float %f1, float %f2)
ret <2 x half> %val
}
define <2 x half> @cvt_rz_f16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rz_f16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_f16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rz_f16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rz.satfinite.f16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x half> @llvm.nvvm.ff2f16x2.rz.satfinite(float %f1, float %f2)
ret <2 x half> %val
}
define <2 x half> @cvt_rz_relu_f16x2_f32_sf(float %f1, float %f2) {
; CHECK-LABEL: cvt_rz_relu_f16x2_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b32 %r<4>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_relu_f16x2_f32_sf_param_0];
; CHECK-NEXT: ld.param.b32 %r2, [cvt_rz_relu_f16x2_f32_sf_param_1];
; CHECK-NEXT: cvt.rz.relu.satfinite.f16x2.f32 %r3, %r1, %r2;
; CHECK-NEXT: st.param.b32 [func_retval0], %r3;
; CHECK-NEXT: ret;
%val = call <2 x half> @llvm.nvvm.ff2f16x2.rz.relu.satfinite(float %f1, float %f2)
ret <2 x half> %val
}
declare <2 x half> @llvm.nvvm.ff2f16x2.rn.satfinite(float, float)
declare <2 x half> @llvm.nvvm.ff2f16x2.rn.relu.satfinite(float, float)
declare <2 x half> @llvm.nvvm.ff2f16x2.rz.satfinite(float, float)
declare <2 x half> @llvm.nvvm.ff2f16x2.rz.relu.satfinite(float, float)
define bfloat @cvt_rn_bf16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rn_bf16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_bf16_f32_sf_param_0];
; CHECK-NEXT: cvt.rn.satfinite.bf16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call bfloat @llvm.nvvm.f2bf16.rn.satfinite(float %f1)
ret bfloat %val
}
define bfloat @cvt_rn_relu_bf16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rn_relu_bf16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_relu_bf16_f32_sf_param_0];
; CHECK-NEXT: cvt.rn.relu.satfinite.bf16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call bfloat @llvm.nvvm.f2bf16.rn.relu.satfinite(float %f1)
ret bfloat %val
}
define bfloat @cvt_rz_bf16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rz_bf16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_bf16_f32_sf_param_0];
; CHECK-NEXT: cvt.rz.satfinite.bf16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call bfloat @llvm.nvvm.f2bf16.rz.satfinite(float %f1)
ret bfloat %val
}
define bfloat @cvt_rz_relu_bf16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rz_relu_bf16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_relu_bf16_f32_sf_param_0];
; CHECK-NEXT: cvt.rz.relu.satfinite.bf16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call bfloat @llvm.nvvm.f2bf16.rz.relu.satfinite(float %f1)
ret bfloat %val
}
declare bfloat @llvm.nvvm.f2bf16.rn.satfinite(float)
declare bfloat @llvm.nvvm.f2bf16.rn.relu.satfinite(float)
declare bfloat @llvm.nvvm.f2bf16.rz.satfinite(float)
declare bfloat @llvm.nvvm.f2bf16.rz.relu.satfinite(float)
define half @cvt_rn_f16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rn_f16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_f16_f32_sf_param_0];
; CHECK-NEXT: cvt.rn.satfinite.f16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call half @llvm.nvvm.f2f16.rn.satfinite(float %f1)
ret half %val
}
define half @cvt_rn_relu_f16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rn_relu_f16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rn_relu_f16_f32_sf_param_0];
; CHECK-NEXT: cvt.rn.relu.satfinite.f16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call half @llvm.nvvm.f2f16.rn.relu.satfinite(float %f1)
ret half %val
}
define half @cvt_rz_f16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rz_f16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_f16_f32_sf_param_0];
; CHECK-NEXT: cvt.rz.satfinite.f16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call half @llvm.nvvm.f2f16.rz.satfinite(float %f1)
ret half %val
}
define half @cvt_rz_relu_f16_f32_sf(float %f1) {
; CHECK-LABEL: cvt_rz_relu_f16_f32_sf(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b32 %r1, [cvt_rz_relu_f16_f32_sf_param_0];
; CHECK-NEXT: cvt.rz.relu.satfinite.f16.f32 %rs1, %r1;
; CHECK-NEXT: st.param.b16 [func_retval0], %rs1;
; CHECK-NEXT: ret;
%val = call half @llvm.nvvm.f2f16.rz.relu.satfinite(float %f1)
ret half %val
}
declare half @llvm.nvvm.f2f16.rn.satfinite(float)
declare half @llvm.nvvm.f2f16.rn.relu.satfinite(float)
declare half @llvm.nvvm.f2f16.rz.satfinite(float)
declare half @llvm.nvvm.f2f16.rz.relu.satfinite(float)
Reference in new issue View Git Blame Copy Permalink