haraldwolff
/
CLBlast
mirror of https://github.com/CNugteren/CLBlast.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Changed back to cl_intel_subgroups as suggested

pull/357/head
Cedric Nugteren 2019-05-08 22:01:56 +02:00
parent c5a82f6978
commit 9cbffc9b7c
3 changed files with 1 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/kernels/level3/xgemm_part1.opencl
View File

@ -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

1
src/utilities/compile.cpp
View File

@ -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";

1
src/utilities/utilities.hpp
View File

@ -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;