28 template <typename Executor, typename Params>
29 std::uint8_t* PrepareGemmTask(const Params& params, int kernel_m, int kernel_n,  in PrepareGemmTask()
32                               std::vector<Params>* tasks) {  in PrepareGemmTask()
34   Params& task = tasks->back();  in PrepareGemmTask()
39       StreamUtil<typename Params::InType, typename Params::LeftStream>::Offset(  in PrepareGemmTask()
44       StreamUtil<typename Params::InType, typename Params::RightStream>::Offset(  in PrepareGemmTask()
48       StreamUtil<typename Params::OutType, typename Params::OutputStream>::  in PrepareGemmTask()
52   return scratch + Executor::template EstimateScratchSize<Params>(  in PrepareGemmTask()
56 template <typename MultiThreadingContext, typename Executor, typename Params>
57 bool PrepareGemmTasks(MultiThreadingContext* context, const Params& params,  in PrepareGemmTasks()
59                       std::vector<Params>* task_params) {  in PrepareGemmTasks()
80       scratch = PrepareGemmTask<Executor, Params>(  in PrepareGemmTasks()
85     PrepareGemmTask<Executor, Params>(params, kernel_m, kernel_n, kernel_k,  in PrepareGemmTasks()
91       scratch = PrepareGemmTask<Executor, Params>(  in PrepareGemmTasks()
96     PrepareGemmTask<Executor, Params>(params, kernel_m, kernel_n, kernel_k,  in PrepareGemmTasks()
104 template <typename Executor, typename Params, int kernel_m, int kernel_n,
107   GemmTaskRunner(const Params& params) : params(params) {}  in GemmTaskRunner()
110     Gemm<Executor, Params, kernel_m, kernel_n, kernel_k>(params);  in Run()
113   Params params;
118 template <typename MultiThreadingContext, typename Executor, typename Params,
121                             const Params& params) {  in MultiThreadGemm()
122   typedef internal::GemmTaskRunner<Executor, Params, kernel_m, kernel_n,  in MultiThreadGemm()
126   std::vector<Params> task_params;  in MultiThreadGemm()
127   if (!internal::PrepareGemmTasks<MultiThreadingContext, Executor, Params>(  in MultiThreadGemm()
129     Gemm<Executor, Params, kernel_m, kernel_n, kernel_k>(params);  in MultiThreadGemm()