Merge branch 'master' into preparation_for_size_specific_parameters

This commit is contained in:
Cedric Nugteren 2017-09-30 20:26:50 +02:00
commit 3b7371f81b
5 changed files with 16 additions and 6 deletions

View file

@ -58,8 +58,8 @@ build_script:
after_build:
- ps: pushd $env:CLBLAST_BUILD
- 7z a CLBlast-1.0.0-Windows-x64.zip .\install_dir\*
- ps: mv CLBlast-1.0.0-Windows-x64.zip $env:APPVEYOR_BUILD_FOLDER
- 7z a CLBlast-1.1.0-Windows-x64.zip .\install_dir\*
- ps: mv CLBlast-1.1.0-Windows-x64.zip $env:APPVEYOR_BUILD_FOLDER
artifacts:
- path: '*.zip'

View file

@ -21,7 +21,7 @@ matrix:
env:
global:
- CLBLAST_VERSION=1.0.0
- CLBLAST_VERSION=1.1.0
- CLBLAST_ROOT=${TRAVIS_BUILD_DIR}/bin/clblast
- CLBLAST_INSTALL=${TRAVIS_BUILD_DIR}/bin/CLBlast-${CLBLAST_VERSION}
- CLBLAST_TAR=CLBlast-${CLBLAST_VERSION}-${TRAVIS_OS_NAME}-x64.tar.gz

View file

@ -1,5 +1,5 @@
Development (next version)
Version 1.1.0
- The tuning database now has defaults per architecture (e.g. NVIDIA Kepler SM3.5, AMD Fiji)
- The tuning database now has a dictionary to translate vendor/device names to a common set
- The tuners can now distinguish between different AMD GPU board names of the same architecture

View file

@ -18,8 +18,8 @@ set(CMAKE_USER_MAKE_RULES_OVERRIDE_CXX ${CMAKE_CURRENT_SOURCE_DIR}/cmake/cxx_fla
# CMake project details
project("clblast" C CXX)
set(clblast_VERSION_MAJOR 1)
set(clblast_VERSION_MINOR 0)
set(clblast_VERSION_PATCH 1)
set(clblast_VERSION_MINOR 1)
set(clblast_VERSION_PATCH 0)
# Options and their default values
option(BUILD_SHARED_LIBS "Build a shared (ON) or static library (OFF)" ON)

View file

@ -316,6 +316,16 @@ Since there is no half-precision data-type in C or C++, OpenCL provides the `cl_
The `samples/haxpy.c` example shows how to use these convenience functions when calling the half-precision BLAS routine HAXPY.
Known issues
-------------
Known performance related issues:
* Severe performance issues with Beignet v1.3.0 due to missing support for local memory. Please downgrade to v1.2.1 or upgrade to v1.3.1 or newer.
* Performance issues on ARM Mali GPUs due to missing compiler for support for loop unrolling and array-to-register promotion.
Contributing
-------------