| /external/gemmlowp/meta/ |
| D | multi_thread_gemm.h | 29 std::uint8_t* PrepareGemmTask(const Params& params, int kernel_m, int kernel_n, in PrepareGemmTask() argument
53 task, kernel_m, kernel_n, kernel_k); in PrepareGemmTask()
58 int kernel_m, int kernel_n, int kernel_k, in PrepareGemmTasks() argument
81 params, kernel_m, kernel_n, kernel_k, scratch, i * m_chunk, m_chunk, in PrepareGemmTasks()
85 PrepareGemmTask<Executor, Params>(params, kernel_m, kernel_n, kernel_k, in PrepareGemmTasks()
92 params, kernel_m, kernel_n, kernel_k, scratch, 0, params.m, 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,
110 Gemm<Executor, Params, kernel_m, kernel_n, kernel_k>(params); in Run()
119 int kernel_m, int kernel_n, int kernel_k>
[all …]
|
| D | single_thread_gemm.h | 24 template <typename Executor, typename Params, int kernel_m, int kernel_n,
31 static int EstimateScratchSize(const P& params, int kernel_m, int kernel_n, in EstimateScratchSize() argument
36 const int rhs_chunks = ((params.n + kernel_n - 1) / kernel_n); in EstimateScratchSize()
40 params.right_stream, kernel_n, kernel_k); in EstimateScratchSize()
220 static int EstimateScratchSize(const P& params, int kernel_m, int kernel_n, in EstimateScratchSize() argument
229 params.right_stream, kernel_n, kernel_k); in EstimateScratchSize()
441 static int EstimateScratchSize(const P& params, int kernel_m, int kernel_n, in EstimateScratchSize() argument
490 static int EstimateScratchSize(const P& params, int kernel_m, int kernel_n, in EstimateScratchSize() argument
676 template <typename Executor, typename Params, int kernel_m, int kernel_n,
679 internal::Dispatch3DStage1<Executor, Params, kernel_m, kernel_n, kernel_k,
[all …]
|
| D | base.h | 101 typename OutputStream, int kernel_m, int kernel_n, int pack_size>
|
| /external/gemmlowp/meta/generators/ |
| D | quantized_mul_kernels_common.py | 53 def Prepare(self, emitter, registers, kernel_m, kernel_n, lhs, rhs): argument
59 self.rhs_offsets = _ReadParams(emitter, registers, rhs, kernel_n, 4)
130 def Prepare(self, emitter, registers, kernel_m, kernel_n, lhs, rhs): argument
135 self.rhs_offsets = _ReadParams(emitter, registers, rhs, kernel_n, 4)
162 def Prepare(self, emitter, registers, kernel_m, kernel_n, lhs, rhs): argument
167 self.rhs_offsets = _ReadParams(emitter, registers, rhs, kernel_n, 4)
222 kernel_n): argument
227 emitter.EmitVStoreAE(data_type, kernel_n, datum, output, None)
410 def _GenerateNxMLoadMultiplyAggregate(emitter, registers, kernel_m, kernel_n, argument
422 rhs_load = [registers.DoubleRegister() for unused_i in range(kernel_n)]
[all …]
|
| D | common.py | 95 def SpecializeMulKernel(self, in_type, out_type, kernel_m, kernel_n, argument
100 kernel_n, pack_size
112 self.EmitMultiply(in_type, out_type, kernel_m, kernel_n, pack_size)
|