diff options
5 files changed, 36 insertions, 18 deletions
diff --git a/examples/gemm_tuner/cl_gemm_reshaped.cpp b/examples/gemm_tuner/cl_gemm_reshaped.cpp index bbc42a7e55..59044477bf 100644 --- a/examples/gemm_tuner/cl_gemm_reshaped.cpp +++ b/examples/gemm_tuner/cl_gemm_reshaped.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2021 Arm Limited. + * Copyright (c) 2019-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -256,6 +256,11 @@ public: kernel_info.broadcast_bias = true; kernel_info.activation_info = act_info; + if(rhs_info.h0 == 0) + { + rhs_info.h0 = std::max(kernel_info.n / rhs_info.n0, 1U); + } + // Initialise lhs_reshaped tensor info lhs_reshaped.allocator()->init(TensorInfo(compute_lhs_reshaped_shape(*lhs.info(), lhs_info), 1, params.data_type)); diff --git a/examples/gemm_tuner/cl_gemm_reshaped_rhs_only.cpp b/examples/gemm_tuner/cl_gemm_reshaped_rhs_only.cpp index 79f251ae7d..0ad2a65dc2 100644 --- a/examples/gemm_tuner/cl_gemm_reshaped_rhs_only.cpp +++ b/examples/gemm_tuner/cl_gemm_reshaped_rhs_only.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019-2021 Arm Limited. + * Copyright (c) 2019-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -225,6 +225,11 @@ public: kernel_info.broadcast_bias = true; kernel_info.activation_info = act_info; + if(rhs_info.h0 == 0) + { + rhs_info.h0 = std::max(kernel_info.n / rhs_info.n0, 1U); + } + // Initialise rhs_reshaped tensor info rhs_reshaped.allocator()->init(TensorInfo(compute_rhs_reshaped_shape(*rhs.info(), rhs_info), 1, params.data_type)); diff --git a/examples/gemm_tuner/cl_gemmlowp_reshaped.cpp b/examples/gemm_tuner/cl_gemmlowp_reshaped.cpp index 115ce6ba23..9cf9c9fed0 100644 --- a/examples/gemm_tuner/cl_gemmlowp_reshaped.cpp +++ b/examples/gemm_tuner/cl_gemmlowp_reshaped.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2021 Arm Limited. + * Copyright (c) 2020-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -240,6 +240,11 @@ public: rhs_info.transpose = configs.transpose_rhs; rhs_info.export_to_cl_image = false; // CL image not supported for quantized cases yet + if(rhs_info.h0 == 0) + { + rhs_info.h0 = std::max(static_cast<unsigned int>(params.N) / rhs_info.n0, 1U); + } + lhs_reshaped.allocator()->init(TensorInfo(compute_lhs_reshaped_shape(*lhs.info(), lhs_info), 1, params.data_type)); rhs_reshaped.allocator()->init(TensorInfo(compute_rhs_reshaped_shape(*rhs.info(), rhs_info), 1, params.data_type)); lhs_reshaped.info()->set_quantization_info(q_info); diff --git a/examples/gemm_tuner/cl_gemmlowp_reshaped_rhs_only_fused_output_stage_fixedpoint.cpp b/examples/gemm_tuner/cl_gemmlowp_reshaped_rhs_only_fused_output_stage_fixedpoint.cpp index 0e0917daa9..94f3c93166 100644 --- a/examples/gemm_tuner/cl_gemmlowp_reshaped_rhs_only_fused_output_stage_fixedpoint.cpp +++ b/examples/gemm_tuner/cl_gemmlowp_reshaped_rhs_only_fused_output_stage_fixedpoint.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2021 Arm Limited. + * Copyright (c) 2020-2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -220,6 +220,11 @@ public: rhs_info.transpose = configs.transpose_rhs; rhs_info.export_to_cl_image = false; // CL image not supported for quantized cases yet + if(rhs_info.h0 == 0) + { + rhs_info.h0 = std::max(static_cast<unsigned int>(params.N) / rhs_info.n0, 1U); + } + rhs_reshaped.allocator()->init(TensorInfo(compute_rhs_reshaped_shape(*rhs.info(), rhs_info), 1, params.data_type)); rhs_reshaped.info()->set_quantization_info(q_info); if(rhs_info.export_to_cl_image) diff --git a/src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp b/src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp index 76551b076a..d08bf84c72 100644 --- a/src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp +++ b/src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp @@ -562,21 +562,19 @@ std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOn if(m == 1) { - if(r_mn <= 0.0045f) + const GeMMConfigsMatrix configs_mnkb_best = { - if(workload <= 278.7000f) - { - return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 8, 0, 0, 0, 1, 1); - } - else - { - return configure_lhs_rhs_info(m, n, 1, 4, 8, 1, 32, 0, 0, 1, 0, 0); - } - } - else - { - return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 8, 0, 0, 1, 0, 0); - } + { 1, 8984, 640, 1, 1, 4, 2, 1, 0, 1, 0, 1, 1, 0 }, + { 1, 420, 392, 1, 1, 2, 4, 1, 0, 1, 0, 1, 0, 0 }, + { 1, 644, 5288, 1, 1, 2, 4, 1, 0, 1, 0, 1, 0, 0 }, + { 1, 6512, 6404, 1, 1, 2, 2, 1, 0, 1, 0, 1, 1, 0 }, + { 1, 5304, 640, 1, 1, 2, 2, 1, 0, 1, 0, 1, 0, 0 }, + { 1, 1352, 1520, 1, 1, 2, 4, 1, 0, 1, 0, 1, 0, 0 }, + { 1, 4096, 25088, 1, 1, 2, 4, 1, 0, 1, 0, 1, 0, 0 }, + { 1, 732, 8988, 1, 1, 2, 4, 1, 0, 1, 0, 1, 0, 0 } + }; + + return find_lhs_rhs_info(configs_mnkb_best, m, n, k, b); } else { |