MCPcopy Create free account
hub / github.com/ROCm/AMDMIGraphX / run

Method run

src/targets/gpu/gemm_impl.cpp:261–281  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

259 }
260
261 void run(context& ctx, const std::vector<argument>& input_args, int32_t solution_idx = 0) const
262 {
263 if(strided_batched)
264 {
265 auto common_args = create_strided_batched_args_common(ctx, compute_type, input_args);
266 rocblas_invoke(&rocblas_gemm_strided_batched_ex,
267 common_args,
268 rocblas_gemm_algo_solution_index,
269 solution_idx,
270 gemm_flags);
271 }
272 else
273 {
274 auto common_args = create_gemm_ex_args_common(ctx, compute_type, input_args);
275 rocblas_invoke(&rocblas_gemm_ex,
276 common_args,
277 rocblas_gemm_algo_solution_index,
278 solution_idx,
279 gemm_flags);
280 }
281 }
282
283#ifdef MIGRAPHX_USE_ROCBLAS_TUNING_API
284 auto validate(context& ctx, const std::vector<shape>& input_shapes, int32_t solution_idx) const

Callers 1

gemm_computeFunction · 0.45

Calls 1

rocblas_invokeFunction · 0.85

Tested by

no test coverage detected