| AArch64 | |
| ppc64le | |
| s390x | |
| x86-64 |
- Update to version 1.1.2: * [cmake] Bump patch version * [snap] fix yaml for latest version of snapcraft * [snap] move to core20
- Update to version 1.1.1: * Add non-blocking read/write transfers. * Added 24bit fast MAD Integer Compute test. * Updated results with various Intel, AMD and NVidia hardware. * Use opencl.hpp instead of cl.hpp * Fix icd build issue.
- Update to upstream release 1.1.0
* All the kernels are now de-unrolled
* Prior to this, all mad or fetch operations were manually
unrolled which created register pressure on low-end devices.
Now the operations are put in a for loop and left for the
compiler to unroll
* General improvements
- Update to version 1.0+git.20191206: * enqueueWriteBuffer: Initialize host buffer to obtain accurate measurement * Move jetson result to pocl folder * [snap] use new cmake file for versions * Add AMD Ryzen 7 3700X * Bump rc * [travis] remove explicit branch * [src] auto format cpp files * Bump version to rc1 * [android] cosmetic changes * [travis] add deploy section * [travis] use cmake to invoke cpack * [travis] explicitly specify configuration to cpack * Update travis file * [cmake] fix a string in set * Add cpack * Add suffix to version * remove int kernel restriction for apple * [cmake] fix extra libs for static version * [travis] add arm64 * skip exit code for no platform error * [travis] call ctest * [cmake] add a test run * Create Jeston Nano (Tegra X1) * Undo disabling of int kernels for intel * Reduce register usage in kernels * [android] update cmake to include version * Create Intel_UHD_Graphics_620.log * Update README.md * Print half-precision with a capital letter * Add results for Radeon RX 570 * Add version info * [travis] use generator instead of platform type for windows * Revert "[cmake] pass CMAKE_GENERATOR_PLATFORM to external project build" * Revert "[cmake] fix error in prev commit" * [cmake] fix error in prev commit * [cmake] pass CMAKE_GENERATOR_PLATFORM to external project build * [travis] try other method of forcing win64 * [travis] force x64 on windows * Update readme * [travis] trim os list * [travis] add more os in matrix * [travis] specify xenial * [yml] Rename to yml * [travis] add initial yaml file * [cmake] install LICENSE file * Delete STATUS file * [cmake] fix typo in prev commit * [cmake] windows specific changes * [cmake] Build ICD if not present * [results] remove bin symbols * Add results for GeForce RTX 2080 Ti * Add snapcraft file * [cmake] add an install target * Remove exception for snapdragon platform * trim fn: check before call * [android] upgrade to latest sdk * Place max transfer bw size in common * Place max global bw size in common * Add a seperate multiplier for DP * some more warn fixes * [cmake] throw warning on conversion loss * Use higher size datatypes for memory size * Add an unused macro * Disable depcration warn on windows * Flush the queue after every run iteration * Disable integer kernel in intel only on windows * Revert "Allow int tests on intel" * fix global bw test to use right multiplier * Allow int tests on intel * Global size now is multiple of local size * Trim strings returned by cl APIs * fix warnings * set OPENCL target version * Stop using cl.hpp from deps * Add a gitignore * [cmake] OPENCL_ROOT is not required * Use ulong for alloc vars * Reduce load on integer kernel * Use FindOpenCL rather than custom OpenCL finders * Added result for RTX2080
- Update to 1.0+git.20170625 to fix build on AArch64: * Remove march native flag * Always define OS_NAME even for unknown architectures * Add AMD_Ryzen_5_1600.log
- Update to version 1.0+git.20170525: * Add GTX 1080 Ti results * Remove non printable chars from log file * Add One Plus 3T results * Macbook pro results submitted by Harry Mallon * Rename Macbook_Pro_2013.log to Mac_Pro_Late_2013_.log * Not passing half as kernel argument * Fix uninitialized variable * Ignore ignored-attributes warning in g++, clang++ * Enable support for FreeBSD.
- Update dependencies
- Update to version 1.0+git.20160805: * Add Galaxy S7 (Adreno 530) results * migrate to android studio * add missing x86 dummy folder * cosmetic ui layout changes * fix jumpdown list index * gpus only for snapdragon * update android.mk for new half-precision * add dummy folders for release * update android build files - Change naming scheme - Some small spec file cleanups
- Update to version 0.0.0.git1422696808.9a39c0c: + Update README.md + Handle mem alloc failures gracefully + Add results of Hawaii given by Schleimer Ben
- Update to version 0.0.0.git1419703962.3df45ca: + Mali T628 results shared by Sebastian Schaetz + Added compute unit, clock frequency and clock frequency units to XML file + CMakeLists: Remove -march=native options in case of cross compilation + src/common: Implement roundtoPowOf2 with binary logic + Add AMD A10-7850K and GeForce GTX 750 + Add Celeron N2930 and Core i5-4570 + Support clang in cmake Add armv8 in android jni + Preserve isComputeInt and isComputeDP across platforms + Add integer compute (GIOPS) to A10-7850K results + Add results from monster GTX 980 submitted by Gosuke yosita + Link statically to libopencl-stub + Add missing symbol in logger_android Fix bug in roundToPowOf2 - minor specfile cleanup
- initial package, version 0.0+git.1406390327.78b5cde: + 1. android logger interface. Renamed few files 2. Call stub reset api if required + Decrease work-group size to 128 for qualcomm platforms, with 256 kernel fails to launch + Update README.md + Update README with proper formatting + Build program per device. Otherwise qualcomm platform fails to compile for 2 devices at 1 shot! + Adding march native in cmake for linux. Required for x86 Adding results of Apple Iris Pro & other devices submitted by users + Fix win32 warning + Android GUI app initial commit + Avoid OsLib setenv. It is not implemented in 4.2.2 Create a jni function to call setenv + Remove trailing white-spaces Update README. Android app available + Nexus-5 & Moto-G results