Installation and Building for Windows

Installation and Building for Windows#

Prerequisites#

If using the rocBLAS backend on an AMD machine:
- An AMD HIP SDK enabled platform. Find more information on the System requirements (Windows) page.
- A compatible version of hipblas-common.
- A compatible version of rocBLAS.
- A compatible version of rocSOLVER and its dependencies, rocSPARSE and rocPRIM, for full functionality.
- hipBLAS is supported on the same Windows versions and toolchains that are supported by the HIP SDK.
hipBLAS does not currently support using the cuBLAS backend on Windows.

Installing pre-built packages#

hipBLAS can be installed on Windows 11 or Windows 10 using the AMD HIP SDK installer.

The simplest way to use hipBLAS in your code would be using CMake for which you would add the SDK installation location to your CMAKE_PREFIX_PATH in your CMake configure step.

-DCMAKE_PREFIX_PATH="C:\hipSDK"

Then in your CMakeLists.txt use:

find_package(hipblas)
target_link_libraries( your_exe PRIVATE roc::hipblas )

Building and Installing hipBLAS#

For most users, building from source is not necessary, as hipBLAS can be used after installing the pre-built packages as described above. However, users can use the following instructions to build hipBLAS from source if necessary.

Download hipBLAS Source Code#

The hipBLAS source code is available at the hipBLAS github page. The version of the ROCm HIP SDK may be shown in the path of default installation, but you can run the HIP SDK compiler to report the versino from the bin/ folder with:

hipcc --version

The HIP version has major, minor, and patch fields, possibly followed by a build specific identifier. For example, HIP version could be 5.4.22880-135e1ab4; this corresponds to major = 5, minor = 4, patch = 22880, build identifier 135e1ab4. There are GitHub branches at the hipBLAS site with names release/rocm-rel-major.minor where major and minor are the same as in the HIP version. For example for you can use the following to download rocBLAS:

git clone -b release/rocm-rel-x.y ROCm/hipBLAS.git

Replace x.y in the above command with the version of the HIP SDK installed on your machine. For example, if you have HIP 6.2 installed, then use -b release/rocm-rel-6.2. You can add the SDK tools to your path with an entry such as:

%HIP_PATH%\bin

Building#

The root of this repository has a helper python script rmake.py to build and install hipBLAS with a single command. It does take a lot of options and hard-coded configuration that can be specified through invoking cmake directly, but it’s a great way to get started quickly and can serve as an example of how to build and install. A few commands in the script need sudo access so it may prompt you for a password.

Typical uses of rmake.py to build (library dependencies + library) are in the table below.

Command	Description
`python3 rmake.py -h`	Help information.
`python3 rmake.py -d`	Build library dependencies and library in your local directory. The -d flag only needs to be used once. For subsequent invocations of rmake.py it is not necessary to rebuild the dependencies.
`python3 rmake.py`	Build library in your local directory. It is assumed dependencies have been built.
`python3 rmake.py -i`	Build library, then build and install hipBLAS package in C:\hipSDK. You will be prompted for sudo access. This will install for all users. If you want to keep hipBLAS in your local directory, you do not need the -i flag.
`python3 rmake.py -n`	Build library without functionality provided by rocSOLVER. rocSOLVER, rocSPARSE, and rocPRIM dependencies will not be needed. This flag has no effect when building with cuda backend.

Build library dependencies + client dependencies + library + client#

The client contains executables in the table below.

Executable name	Description
hipblas-test	runs Google Tests to test the library
hipblas-bench	executable to benchmark or test individual functions
hipblas-example-*	various examples showing hipblas usage

Common uses of rmake.py to build (dependencies + library + client) are in the table below.

Command	Description
`python3 rmake.py -dc`	Build library dependencies, client dependencies, library, and client in your local directory. The -d flag only needs to be used once. For subsequent invocations of rmake.py it is not necessary to rebuild the dependencies.
`python3 rmake.py -c`	Build library and client in your local directory. It is assumed the dependencies have been built.
`python3 rmake.py -idc`	Build library dependencies, client dependencies, library, client, then build and install the hipBLAS package. You will be prompted for sudo access. It is expected that if you want to install for all users you use the -i flag. If you want to keep hipBLAS in your local directory, you do not need the -i flag.
`python3 rmake.py -ic`	Build and install hipBLAS package, and build the client. You will be prompted for sudo access. This will install for all users. If you want to keep hipBLAS in your local directory, you do not need the -i flag.

Dependencies for building library#

Use rmake.py with -d option to install dependencies required to build the library. This will not install the hipblas-common, rocBLAS, rocSOLVER, rocSPARSE, and rocPRIM dependencies. When building hipBLAS it is important to note version dependencies of other libraries. The hipblas-common, rocBLAS and rocSOLVER versions needed for an AMD backend build are listed in the top level CMakeLists.txt file. rocSPARSE and rocPRIM are currently dependencies of rocSOLVER. To build these libraries from source, please visit the rocBLAS Documentation, rocSOLVER Documentation, rocSPARSE Documentation, and rocPRIM Documentation.

CMake has a minimum version requirement which is currently 3.16.8. See --cmake_install flag in rmake.py to upgrade automatically.

For the test and benchmark clients’ host reference functions you must manually download and install AMD’s ILP64 version of the AOCL libraries, version 4.2, from https://www.amd.com/en/developer/aocl.html. If you download and run the full Windows AOCL installer into the default location (C:\Program Files\AMD\AOCL-Windows\) then the AOCL reference BLAS (amd-blis) should be found by the clients’ CMakeLists.txt.

Note, if you only use the rmake.py -d dependency script and change the default CMake option LINK_BLIS=ON, you may experience hipblas-test stress test failures due to 32-bit integer overflow on the host unless you exclude the stress tests via command line argument --gtest_filter=-*stress*.