Changed back to cl_intel_subgroups as suggested
parent
c5a82f6978
commit
9cbffc9b7c
|
@ -127,7 +127,7 @@ R"(
|
|||
|
||||
// Intel subgroups (https://www.khronos.org/registry/OpenCL/extensions/intel/cl_intel_subgroups.html)
|
||||
#if USE_SUBGROUP_SHUFFLING == 1 && SUBGROUP_SHUFFLING_INTEL == 1
|
||||
#pragma OPENCL EXTENSION cl_intel_device_side_avc_motion_estimation: enable
|
||||
#pragma OPENCL EXTENSION cl_intel_subgroups: enable
|
||||
#define SUBGROUP_SIZE 8 // Assumes subgroup size is always 8 on Intel GPUs
|
||||
#endif
|
||||
|
||||
|
|
|
@ -60,7 +60,6 @@ std::shared_ptr<Program> CompileFromSource(
|
|||
|
||||
// For Intel GPUs with subgroup support, use subgroup shuffling.
|
||||
if (device.IsGPU() && device.HasExtension(kKhronosIntelSubgroups) &&
|
||||
device.HasExtension(kKhronosIntelAvcMotionEstimation) &&
|
||||
(precision == Precision::kSingle || precision == Precision::kHalf)) {
|
||||
header_string += "#define USE_SUBGROUP_SHUFFLING 1\n";
|
||||
header_string += "#define SUBGROUP_SHUFFLING_INTEL 1\n";
|
||||
|
|
|
@ -48,7 +48,6 @@ using double2 = std::complex<double>;
|
|||
const std::string kKhronosAttributesAMD = "cl_amd_device_attribute_query";
|
||||
const std::string kKhronosAttributesNVIDIA = "cl_nv_device_attribute_query";
|
||||
const std::string kKhronosIntelSubgroups = "cl_intel_subgroups";
|
||||
const std::string kKhronosIntelAvcMotionEstimation = "cl_intel_device_side_avc_motion_estimation";
|
||||
|
||||
// Catched an unknown error
|
||||
constexpr auto kUnknownError = -999;
|
||||
|
|
Loading…
Reference in New Issue