This is an old version of ROCm documentation. Read the latest ROCm release documentation to stay informed of all our developments.

Welcome to hipBLASLt’s documentation!#


General Description#

hipBLASLt is created using the HIP programming language and optimized for AMD’s latest discrete GPUs. hipBLASLt is an equivalent math library to CUDA’s cuBLASLt which is a lightweight library dedicated to GEneral Matrix-to-matrix Multiply (GEMM) operations with a new flexible API. This library adds flexibility in matrix data layouts, input types, compute types, and also in choosing the algorithmic implementations and heuristics through parameter programmability. After a set of options for the intended GEMM operation are identified by the user, these options can be used repeatedly for different inputs. It is designed to be used from C and C++ code. The GEMM operation of hipBLASLt is performed by hipblasLtMatmul(). The equation is listed here:

\[D = Activation(alpha \cdot op(A) \cdot op(B) + beta \cdot op(C) + bias)\]

where \(op(A)/op(B)\) refers to in-place operations such as transpose/non-transpose, and alpha, beta are scalars.

The code is open and hosted here: ROCmSoftwarePlatform/hipBLASLt

hipBLASLt exports an interface that does not require the client to change, regardless of the chosen backend. hipBLASLt currently support rocBLASLt (core implementation of hipBLASLt) as backend. hipBLASLt will support cuBLASLt as backends in the coming release.

hipBLASLt Logging#

The hipBLASLt logging mechanism can be enabled by setting the following environment variables before launching the target application:

HIPBLASLT_LOG_LEVEL=<level> - while level is one of the following levels:

“0” - Off - logging is disabled (default)

“1” - Error - only errors will be logged

“2” - Trace - API calls that launch HIP kernels will log their parameters and important information

“3” - Hints - hints that can potentially improve the application’s performance

“4” - Info - provides general information about the library execution, may contain details about heuristic status

“5” - API Trace - API calls will log their parameter and important information

HIPLASLT_LOG_MASK=<mask> - while mask is a combination of the following masks:

“0” - Off

“1” - Error

“2” - Trace

“4” - Hints

“8” - Info

“16” - API Trace

HIPBLASLT_LOG_FILE=<file_name> - while file name is a path to a logging file. File name may contain %i, that will be replaced with the process ID. For example, “<file_name>_%i.log”. If HIPBLASLT_LOG_FILE is not defined, the log messages are printed to stdout.

Heuristics Cache#

hipBLASLt uses heuristics to pick the most suitable matmul kernel for execution based on the problem sizes, GPU configuration, and other parameters. This requires performing some computations on the host CPU, which could take tens of microseconds. To overcome this overhead, it is recommended to query the heuristics once using hipblasLtMatmulAlgoGetHeuristic() and then reuse the result for subsequent computations using hipblasLtMatmul().