hipBLASLt provider operation support

hipBLASLt provider operation support#

2026-03-31

1 min read time

Applies to Linux and Windows

The hipBLASLt provider plugin integrates with the ROCm hipBLASLt library that provides optimized GEMM operations.

Operation support#

The hipBLASLt provider plugin supports standalone Matmul (GEMM, general matrix multiplication) operations with these features and constraints:

  • Input and output data types:

    • FP32: Single-precision floating point (32-bit)

    • FP16: Half-precision floating point (16-bit)

    • BFP16: Brain floating point (16-bit).

  • Compute data type: FP32.

  • Transposed inputs: Supported.

  • Batched matmuls: Only equal batch sizes are supported, along with broadcasting when one input has a single batch (batch=1).

  • Fused operations: Matmul supports fused bias, forward activation (ReLU, clamp, GELU with tanh approximation, and Swish with unit beta), and fused bias + forward activation (same supported activations).