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

Function LtGemmSearch

plugin/fcPlugin/fcPlugin.h:186–215  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

184// clang-format on
185template <typename T>
186void LtGemmSearch(cublasLtHandle_t ltHandle, Gemm<T> const& g, void* workSpace, size_t workSpaceSize,
187 std::vector<customMatmulPerf_t>& perfResults)
188{
189 // clang-format off
190 LtGemmSearch(
191 ltHandle,
192 g.opA,
193 g.opB,
194 g.m,
195 g.n,
196 g.k,
197 &g.alpha,
198 g.A,
199 g.ldA,
200 g.B,
201 g.ldB,
202 &g.beta,
203 g.C,
204 g.ldC,
205 workSpace,
206 workSpaceSize,
207 Gemm<T>::Types::cudaTypeCom,
208 Gemm<T>::Types::cudaTypeS,
209 Gemm<T>::Types::cudaTypeI,
210 Gemm<T>::Types::cudaTypeI,
211 Gemm<T>::Types::cudaTypeO,
212 perfResults
213 );
214 // clang-format on
215}
216
217struct LtContext
218{

Callers 1

gemmSearchFunction · 0.70

Calls

no outgoing calls

Tested by

no test coverage detected