/external/gemmlowp/meta/ |
H A D | legacy_operations_common.h | 20 Quantized8BitOperation(std::int32_t lhs_offset, std::int32_t rhs_offset, argument 24 rhs_offset(rhs_offset), 31 std::int32_t rhs_offset; member in class:Quantized8BitOperation 39 FloatOperation(std::int32_t lhs_offset, std::int32_t rhs_offset, argument 42 rhs_offset(rhs_offset), 47 std::int32_t rhs_offset; member in class:FloatOperation 53 Int32Operation(std::int32_t lhs_offset, std::int32_t rhs_offset) argument 54 : lhs_offset(lhs_offset), rhs_offset(rhs_offse 58 std::int32_t rhs_offset; member in class:Int32Operation [all...] |
H A D | legacy_multi_thread_gemv.h | 33 GemvQuantized8BitOperation(std::int32_t lhs_offset, std::int32_t rhs_offset, argument 36 : Quantized8BitOperation(lhs_offset, rhs_offset, sum_offset, multiplier, 43 gemv_q8(scratch, lhs, rhs, n, k, lhs_offset, rhs_offset, sum_offset, 55 GemvFloatOperation(std::int32_t lhs_offset, std::int32_t rhs_offset, argument 57 : FloatOperation(lhs_offset, rhs_offset, result_offset) {} 63 gemv_f(scratch, lhs, rhs, n, k, lhs_offset, rhs_offset, result_offset, 75 GemvInt32Operation(std::int32_t lhs_offset, std::int32_t rhs_offset) argument 76 : Int32Operation(lhs_offset, rhs_offset) {} 82 gemv_i32(scratch, lhs, rhs, n, k, lhs_offset, rhs_offset, result); 103 std::int32_t rhs_offset, st 99 multi_thread_gemv_q8(gemmlowp::WorkersPool* pool, std::int32_t max_threads, std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t sum_offset, std::int32_t multiplier, std::int32_t shift, std::uint8_t* result) argument 123 multi_thread_gemv_f(gemmlowp::WorkersPool* pool, std::int32_t max_threads, std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, float result_offset, float* result) argument 145 multi_thread_gemv_i32(gemmlowp::WorkersPool* pool, std::int32_t max_threads, std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t* result) argument [all...] |
H A D | legacy_multi_thread_gemm.h | 61 GemmQuantized8BitOperation(std::int32_t lhs_offset, std::int32_t rhs_offset, argument 64 : Quantized8BitOperation(lhs_offset, rhs_offset, sum_offset, multiplier, 81 gemm_q8_strided(scratch, lhs, rhs, m, n, k, lhs_offset, rhs_offset, 93 GemmFloatOperation(std::int32_t lhs_offset, std::int32_t rhs_offset, argument 95 : FloatOperation(lhs_offset, rhs_offset, result_offset) {} 111 gemm_f_strided(scratch, lhs, rhs, m, n, k, lhs_offset, rhs_offset, 123 GemmInt32Operation(std::int32_t lhs_offset, std::int32_t rhs_offset) argument 124 : Int32Operation(lhs_offset, rhs_offset) {} 140 gemm_i32_strided(scratch, lhs, rhs, m, n, k, lhs_offset, rhs_offset, result, 162 std::int32_t lhs_offset, std::int32_t rhs_offset, 158 multi_thread_gemm_q8(gemmlowp::WorkersPool* pool, std::int32_t max_threads, std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t m, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t sum_offset, std::int32_t multiplier, std::int32_t shift, std::uint8_t* result) argument 193 multi_thread_gemm_f(gemmlowp::WorkersPool* pool, std::int32_t max_threads, std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t m, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, float result_offset, float* result) argument 226 multi_thread_gemm_i32(gemmlowp::WorkersPool* pool, std::int32_t max_threads, std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t m, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t* result) argument [all...] |
H A D | legacy_single_thread_gemm.h | 32 std::int32_t rhs_offset, std::int32_t result_offset, 56 params.left_stream.multiplicative_sum_offset = rhs_offset; 58 result_offset + k * lhs_offset * rhs_offset; 76 std::int32_t lhs_offset, std::int32_t rhs_offset, 100 params.left_stream.multiplicative_sum_offset = rhs_offset; 102 result_offset + k * lhs_offset * rhs_offset; 125 std::int32_t rhs_offset, std::int32_t* result, 149 params.left_stream.multiplicative_sum_offset = rhs_offset; 150 params.left_stream.additive_sum_offset = k * lhs_offset * rhs_offset; 165 std::int32_t lhs_offset, std::int32_t rhs_offset, 29 gemm_q8_strided(std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t m, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t result_offset, std::int32_t multiplicative_offset, std::int32_t shift, std::uint8_t* result, std::int32_t result_stride) argument 74 gemv_q8(std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t result_offset, std::int32_t multiplicative_offset, std::int32_t shift, std::uint8_t* result) argument 122 gemm_i32_strided(std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t m, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t* result, std::int32_t result_stride) argument 163 gemv_i32(std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t* result) argument 207 gemm_f_strided(std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t m, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, float result_offset, float* result, std::int32_t result_stride) argument 248 gemv_f(std::uint8_t* scratch, const std::uint8_t* lhs, const std::uint8_t* rhs, std::int32_t n, std::int32_t k, std::int32_t lhs_offset, std::int32_t rhs_offset, float result_offset, float* result) argument [all...] |
/external/gemmlowp/public/ |
H A D | gemmlowp.h | 39 const RhsOffset& rhs_offset, 42 context, lhs, rhs, result, lhs_offset, rhs_offset, output_pipeline); 57 int lhs_offset, int rhs_offset, 62 const OffsetRowDup rhs_offset_vector(rhs_offset, rhs.cols()); 78 int rhs_offset, int result_offset, int result_mult_int, 81 context, lhs, rhs, result, lhs_offset, rhs_offset, 34 GemmWithOutputPipelinePC(GemmContextType* context, const MatrixMap<const InputScalar, LhsOrder>& lhs, const MatrixMap<const InputScalar, RhsOrder>& rhs, MatrixMap<OutputScalar, ResultOrder>* result, const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, const OutputPipelineType& output_pipeline) argument 53 GemmWithOutputPipeline(GemmContextType* context, const MatrixMap<const InputScalar, LhsOrder>& lhs, const MatrixMap<const InputScalar, RhsOrder>& rhs, MatrixMap<OutputScalar, ResultOrder>* result, int lhs_offset, int rhs_offset, const OutputPipelineType& output_pipeline) argument 74 Gemm(GemmContextType* context, const MatrixMap<const Scalar, LhsOrder>& lhs, const MatrixMap<const Scalar, RhsOrder>& rhs, MatrixMap<Scalar, ResultOrder>* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift) argument
|
/external/gemmlowp/internal/ |
H A D | unpack.h | 98 const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, 115 LoadForBroadcasting<RegisterBlockType>(rhs_offset, src_col); 134 const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, 188 rhs_offset, depth, r, c, global_row, global_col, 0, cx); 201 rhs_offset, depth, r, c, global_row, global_col, global_row, 213 rhs_offset, depth, r, c, global_row, global_col, global_row, 229 rhs_sums_of_each_slice, lhs_offset, rhs_offset, depth, r, c, 236 rhs_sums_of_each_slice, lhs_offset, rhs_offset, depth, r, c, 243 rhs_sums_of_each_slice, lhs_offset, rhs_offset, depth, r, c, 256 rhs_sums_of_each_slice, lhs_offset, rhs_offset, dept 92 UnpackResultBlock(const SrcMapType& src, const OutputPipelineExecutorType& executor, DstType* dst, const VectorMap<const std::int32_t, VectorShape::Col>& lhs_sums_of_each_slice, const VectorMap<const std::int32_t, VectorShape::Row>& rhs_sums_of_each_slice, const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, int depth, int src_row, int src_col, int src_global_row, int src_global_col, int dst_row, int dst_col) argument 130 UnpackResult(ResultBlockType* dst, const MatrixBlockBounds& dst_block, const PackedResultType& src, int depth, const std::int32_t* lhs_sums_of_each_slice_ptr, const std::int32_t* rhs_sums_of_each_slice_ptr, const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, const OutputPipelineType& output_pipeline) argument [all...] |
H A D | single_thread_gemm.h | 71 const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, 148 lhs_offset.block(r, rs), rhs_offset.block(c, cs), output_pipeline); 66 SingleThreadGemm(SingleThreadGemmContext* context, const KernelBase& kernel, const MatrixMap<const InputScalar, LhsOrder>& lhs, const MatrixMap<const InputScalar, RhsOrder>& rhs, MatrixMap<OutputScalar, ResultOrder>* result, const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, const OutputPipelineType& output_pipeline) argument
|
H A D | multi_thread_gemm.h | 457 rhs_offset(_rhs_offset), 491 rhs_offset.block(curr_result_block.start_col, cs), output_pipeline); 505 const RhsOffset& rhs_offset; member in struct:gemmlowp::GemmWithPackedRhsTask 612 const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, 635 lhs_offset, rhs_offset, 682 lhs_offset, rhs_offset, block_params, output_pipeline)); 608 MultiThreadGemm(GemmContextType* context, const KernelBase& kernel, const MatrixMap<const InputScalar, LhsOrder>& lhs, const MatrixMap<const InputScalar, RhsOrder>& rhs, MatrixMap<OutputScalar, ResultOrder>* result, const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, const OutputPipelineType& output_pipeline) argument
|
H A D | dispatch_gemm_shape.h | 159 const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, 177 Transpose(rhs_offset), Transpose(lhs_offset), 184 lhs_offset, rhs_offset, output_pipeline); 155 DispatchGemmShape(GemmContextType* context, const MatrixMap<const InputScalar, LhsOrder>& lhs, const MatrixMap<const InputScalar, RhsOrder>& rhs, MatrixMap<OutputScalar, ResultOrder>* result, const LhsOffset& lhs_offset, const RhsOffset& rhs_offset, const OutputPipelineType& output_pipeline) argument
|
/external/gemmlowp/eight_bit_int_gemm/ |
H A D | eight_bit_int_gemm.cc | 79 const int rhs_offset = b_offset; local 99 context, lhs, rhs, &result, lhs_offset, rhs_offset, result_offset, \ 117 const int rhs_offset = b_offset; local 136 context, lhs, rhs, &result, lhs_offset, rhs_offset, empty_pipeline); \ 254 std::int32_t lhs_offset, std::int32_t rhs_offset, 265 rhs_offset, sum_offset, multiplicative_offset, 270 scratch->buffer(), rhs, lhs, n, m, k, rhs_offset, 280 std::int32_t lhs_offset, std::int32_t rhs_offset, 289 rhs_offset, result_offset, result); 293 scratch->buffer(), rhs, lhs, n, m, k, rhs_offset, 252 MetaGemmQuantized8Bit(GemmContext* context, const std::uint8_t* lhs, const std::uint8_t* rhs, int m, int n, int k, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t sum_offset, std::int32_t multiplicative_offset, std::int32_t shift, bool result_transpose, std::int32_t result_stride, std::uint8_t* result) argument 278 MetaGemmFloat(GemmContext* context, const std::uint8_t* lhs, const std::uint8_t* rhs, int m, int n, int k, std::int32_t lhs_offset, std::int32_t rhs_offset, float result_offset, bool result_transpose, std::int32_t result_stride, float* result) argument [all...] |
/external/gemmlowp/test/ |
H A D | test.cc | 134 int rhs_offset, int result_offset, int result_mult_int, 145 const OffsetRowDup rhs_offset_vector(rhs_offset, cols); 174 int rhs_offset, int result_offset, int result_mult_int, 186 const OffsetRowDup rhs_offset_vector(rhs_offset, cols); 211 int rhs_offset, int result_offset, int result_mult_int, 216 rhs_offset, result_offset, result_mult_int, 245 int rhs_offset, int result_offset, int result_mult_int, 254 rhs_offset, rhs.stride(), result->data(), result_offset, 271 int rhs_offset, int result_offset, int result_mult_int, 276 lhs.stride(), rhs.data(), rhs_offset, rh 130 Gemm(Context* context, const MatrixMap<const Scalar, LhsOrder>& lhs, const MatrixMap<const Scalar, RhsOrder>& rhs, MatrixMap<Scalar, ResultOrder>* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift) argument 170 Gemm(Context* context, const MatrixMap<const Scalar, LhsOrder>& lhs, const MatrixMap<const Scalar, RhsOrder>& rhs, MatrixMap<Scalar, ResultOrder>* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift) argument 207 Gemm(Context* context, const MatrixMap<const Scalar, LhsOrder>& lhs, const MatrixMap<const Scalar, RhsOrder>& rhs, MatrixMap<Scalar, ResultOrder>* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift) argument 242 Gemm(Context*, const MatrixMap<const Scalar, LhsOrder>& lhs, const MatrixMap<const Scalar, RhsOrder>& rhs, MatrixMap<Scalar, ResultOrder>* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift) argument 267 Gemm(bool transpose_a, bool transpose_b, bool transpose_c, const MatrixMap<const Scalar, LhsOrder>& lhs, const MatrixMap<const Scalar, RhsOrder>& rhs, MatrixMap<Scalar, ResultOrder>* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift) argument 401 test_gemm_impl(typename GemmWrapper::Context* context, const LhsType& lhs, const RhsType& rhs, ResultType* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int, int result_shift_min, int result_shift_max) argument 494 test_gemm(typename GemmWrapper::Context* context, const LhsType& lhs, const RhsType& rhs, ResultType* result, int lhs_offset, int rhs_offset, int result_offset, int result_mult_int) argument 1209 const int rhs_offset = -34; local [all...] |
H A D | correctness_meta_gemm.cc | 64 std::int32_t lhs_offset, std::int32_t rhs_offset, 75 (static_cast<std::int32_t>(right[depth * j + k]) + rhs_offset); 107 std::int32_t lhs_offset, std::int32_t rhs_offset, 116 (static_cast<std::int32_t>(right[depth * j + k]) + rhs_offset); 141 std::int32_t lhs_offset, std::int32_t rhs_offset) { 149 (static_cast<std::int32_t>(right[depth * j + k]) + rhs_offset); 62 check_result(std::uint8_t* left, std::uint8_t* right, std::uint8_t* result, std::int32_t rows, std::int32_t cols, std::int32_t depth, std::int32_t lhs_offset, std::int32_t rhs_offset, std::int32_t sum_offset, std::int32_t mul_offset, std::int32_t shift) argument 105 check_result_f(std::uint8_t* left, std::uint8_t* right, float* result, std::int32_t rows, std::int32_t cols, std::int32_t depth, std::int32_t lhs_offset, std::int32_t rhs_offset, float result_offset) argument 138 check_result_i32(std::uint8_t* left, std::uint8_t* right, std::int32_t* result, std::int32_t rows, std::int32_t cols, std::int32_t depth, std::int32_t lhs_offset, std::int32_t rhs_offset) argument
|
/external/gemmlowp/doc/ |
H A D | quantization_example.cc | 333 const int rhs_offset = -rhs_qparams.zero_point; 364 &actual_uint8_result_map, lhs_offset, rhs_offset, output_pipeline);
|