MCPcopy Create free account
hub / github.com/NVIDIA/TensorRT / cublasLtMatmul

Function cublasLtMatmul

plugin/fcPlugin/fcPlugin.h:303–326  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

301
302template <typename T>
303cublasStatus_t cublasLtMatmul(
304 LtContext& ctx, Gemm<T>& g, cublasLtMatmulAlgo_t algo, void* workspace, size_t workspaceSize, cudaStream_t stream)
305{
306 // clang-format off
307 return cublasLtMatmul(
308 ctx.cublas,
309 ctx.operationDesc,
310 &g.alpha,
311 g.A,
312 ctx.Adesc,
313 g.B,
314 ctx.Bdesc,
315 &g.beta,
316 g.C,
317 ctx.Cdesc,
318 g.C,
319 ctx.Cdesc,
320 &algo,
321 workspace,
322 workspaceSize,
323 stream
324 );
325 // clang-format on
326}
327
328// CAUTION : must match cublasLtMatmulTile_t
329char const* const matmulTileName[] = {

Callers 2

customMatmulRunFunction · 0.85
enqueueMethod · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected