Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix OpenCL positive and negative INF constants. #8266

Merged
merged 3 commits into from
Jun 20, 2024
Merged

Fix OpenCL positive and negative INF constants. #8266

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
944639a
Fix OpenCL positive and negative INF constants.
alexreinking Jun 5, 2024
91b4418
Use the float16_t::make_* functions instead.
alexreinking Jun 6, 2024
44fe3f3
Enable gpu_f16_intrinsics on OpenCL+CLHalf
alexreinking Jun 7, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 8 additions & 3 deletions src/CodeGen_OpenCL_Dev.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@
#include "EliminateBoolVectors.h"
#include "EmulateFloat16Math.h"
#include "ExprUsesVar.h"
#include "Float16.h"
#include "IRMutator.h"
#include "IROperator.h"
#include "Simplify.h"
Expand Down Expand Up @@ -1183,11 +1184,15 @@ void CodeGen_OpenCL_Dev::init_module() {
}

if (target.has_feature(Target::CLHalf)) {
const uint16_t nan_f16 = float16_t::make_nan().to_bits();
const uint16_t neg_inf_f16 = float16_t::make_negative_infinity().to_bits();
const uint16_t inf_f16 = float16_t::make_infinity().to_bits();

src_stream << "#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n"
<< "inline half half_from_bits(unsigned short x) {return __builtin_astype(x, half);}\n"
<< "inline half nan_f16() { return half_from_bits(32767); }\n"
<< "inline half neg_inf_f16() { return half_from_bits(31744); }\n"
<< "inline half inf_f16() { return half_from_bits(64512); }\n"
<< "inline half nan_f16() { return half_from_bits(" << nan_f16 << "); }\n"
<< "inline half neg_inf_f16() { return half_from_bits(" << neg_inf_f16 << "); }\n"
<< "inline half inf_f16() { return half_from_bits(" << inf_f16 << "); }\n"
<< "inline bool is_nan_f16(half x) {return isnan(x); }\n"
<< "inline bool is_inf_f16(half x) {return isinf(x); }\n"
<< "inline bool is_finite_f16(half x) {return isfinite(x); }\n"
Expand Down
5 changes: 3 additions & 2 deletions test/correctness/gpu_f16_intrinsics.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,8 +4,9 @@ using namespace Halide;
int main(int argc, char *argv[]) {

auto target = get_jit_target_from_environment();
if (!target.has_feature(Target::Metal)) {
printf("[SKIP] No metal target enabled.\n");
if (!target.has_feature(Target::Metal) &&
!target.features_all_of({Target::OpenCL, Target::CLHalf})) {
printf("[SKIP] Test only applies to Metal and OpenCL+CLHalf.\n");
return 0;
}

Expand Down
Loading