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
88 lines
6.2 KiB
MLIR
88 lines
6.2 KiB
MLIR
// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
|
|
|
|
// CHECK-LABEL: @convert_f32x2_to_f16x2_rn
|
|
llvm.func @convert_f32x2_to_f16x2_rn(%srcA : f32, %srcB : f32) {
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rn(float %{{.*}}, float %{{.*}})
|
|
%res1 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rn.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res2 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>, sat = #nvvm.sat_mode<satfinite>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rn.relu(float %{{.*}}, float %{{.*}})
|
|
%res3 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>, relu = true} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rn.relu.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res4 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>, relu = true, sat = #nvvm.sat_mode<satfinite>} : vector<2xf16>
|
|
|
|
llvm.return
|
|
}
|
|
|
|
// CHECK-LABEL: @convert_f32x2_to_f16x2_rz
|
|
llvm.func @convert_f32x2_to_f16x2_rz(%srcA : f32, %srcB : f32) {
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rz(float %{{.*}}, float %{{.*}})
|
|
%res1 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rz.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res2 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>, sat = #nvvm.sat_mode<satfinite>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rz.relu(float %{{.*}}, float %{{.*}})
|
|
%res3 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>, relu = true} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rz.relu.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res4 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>, relu = true, sat = #nvvm.sat_mode<satfinite>} : vector<2xf16>
|
|
|
|
llvm.return
|
|
}
|
|
|
|
// CHECK-LABEL: @convert_f32x2_to_f16x2_rs_stochastic
|
|
llvm.func @convert_f32x2_to_f16x2_rs_stochastic(%srcA : f32, %srcB : f32, %rbits : i32) {
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rs(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res1 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB, %rbits {rnd = #nvvm.fp_rnd_mode<rs>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rs.relu(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res2 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB, %rbits {relu = true, rnd = #nvvm.fp_rnd_mode<rs>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rs.satfinite(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res3 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB, %rbits {rnd = #nvvm.fp_rnd_mode<rs>, sat = #nvvm.sat_mode<satfinite>} : vector<2xf16>
|
|
// CHECK: %{{.*}} = call <2 x half> @llvm.nvvm.ff2f16x2.rs.relu.satfinite(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res4 = nvvm.convert.f32x2.to.f16x2 %srcA, %srcB, %rbits {relu = true, rnd = #nvvm.fp_rnd_mode<rs>, sat = #nvvm.sat_mode<satfinite>} : vector<2xf16>
|
|
|
|
llvm.return
|
|
}
|
|
|
|
// -----
|
|
|
|
// CHECK-LABEL: @convert_f32x2_to_bf16x2_rn
|
|
llvm.func @convert_f32x2_to_bf16x2_rn(%srcA : f32, %srcB : f32) {
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn(float %{{.*}}, float %{{.*}})
|
|
%res1 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res2 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>, sat = #nvvm.sat_mode<satfinite>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.relu(float %{{.*}}, float %{{.*}})
|
|
%res3 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>, relu = true} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rn.relu.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res4 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rn>, relu = true, sat = #nvvm.sat_mode<satfinite>} : vector<2xbf16>
|
|
|
|
llvm.return
|
|
}
|
|
|
|
// CHECK-LABEL: @convert_f32x2_to_bf16x2_rz
|
|
llvm.func @convert_f32x2_to_bf16x2_rz(%srcA : f32, %srcB : f32) {
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz(float %{{.*}}, float %{{.*}})
|
|
%res1 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res2 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>, sat = #nvvm.sat_mode<satfinite>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.relu(float %{{.*}}, float %{{.*}})
|
|
%res3 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>, relu = true} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rz.relu.satfinite(float %{{.*}}, float %{{.*}})
|
|
%res4 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB {rnd = #nvvm.fp_rnd_mode<rz>, relu = true, sat = #nvvm.sat_mode<satfinite>} : vector<2xbf16>
|
|
|
|
llvm.return
|
|
}
|
|
|
|
// CHECK-LABEL: @convert_f32x2_to_bf16x2_rs_stochastic
|
|
llvm.func @convert_f32x2_to_bf16x2_rs_stochastic(%srcA : f32, %srcB : f32, %rbits : i32) {
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rs(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res1 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB, %rbits {rnd = #nvvm.fp_rnd_mode<rs>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rs.relu(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res2 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB, %rbits {relu = true, rnd = #nvvm.fp_rnd_mode<rs>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rs.satfinite(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res3 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB, %rbits {rnd = #nvvm.fp_rnd_mode<rs>, sat = #nvvm.sat_mode<satfinite>} : vector<2xbf16>
|
|
// CHECK: %{{.*}} = call <2 x bfloat> @llvm.nvvm.ff2bf16x2.rs.relu.satfinite(float %{{.*}}, float %{{.*}}, i32 %{{.*}})
|
|
%res4 = nvvm.convert.f32x2.to.bf16x2 %srcA, %srcB, %rbits {relu = true, rnd = #nvvm.fp_rnd_mode<rs>, sat = #nvvm.sat_mode<satfinite>} : vector<2xbf16>
|
|
|
|
llvm.return
|
|
}
|