diff --git a/example/14_gemm_quantization/gemm_wmma_quantization_int8.cpp b/example/14_gemm_quantization/gemm_wmma_quantization_int8.cpp index cc5e3616ff..7437d0be9d 100644 --- a/example/14_gemm_quantization/gemm_wmma_quantization_int8.cpp +++ b/example/14_gemm_quantization/gemm_wmma_quantization_int8.cpp @@ -27,10 +27,11 @@ using ::ck::Tensor; template using S = ck::Sequence; -using I8 = int8_t; -using I32 = int32_t; -using Row = ck::tensor_layout::gemm::RowMajor; -using Col = ck::tensor_layout::gemm::ColumnMajor; +using I8 = int8_t; +using I32 = int32_t; +using Row = ck::tensor_layout::gemm::RowMajor; +using Col = ck::tensor_layout::gemm::ColumnMajor; +using Bypass = ck::tensor_layout::BypassLayoutVerification; using PassThrough = ck::tensor_operation::element_wise::PassThrough; using ActivationOp = PassThrough; @@ -125,11 +126,11 @@ int main(int /* argc */, char* /* argv */[]) if(std::is_same::value) { - return HostTensorDescriptor({row, col}, {stride, 1_uz}); + return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return HostTensorDescriptor({row, col}, {1_uz, stride}); + return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } }; diff --git a/example/65_gemm_multiply_multiply/gemm_add_add_wmma_fp16.cpp b/example/65_gemm_multiply_multiply/gemm_add_add_wmma_fp16.cpp index 24c58bb69a..1e3d946bad 100644 --- a/example/65_gemm_multiply_multiply/gemm_add_add_wmma_fp16.cpp +++ b/example/65_gemm_multiply_multiply/gemm_add_add_wmma_fp16.cpp @@ -31,8 +31,9 @@ using S = ck::Sequence; using F16 = ck::half_t; using F32 = float; -using Row = ck::tensor_layout::gemm::RowMajor; -using Col = ck::tensor_layout::gemm::ColumnMajor; +using Row = ck::tensor_layout::gemm::RowMajor; +using Col = ck::tensor_layout::gemm::ColumnMajor; +using Bypass = ck::tensor_layout::BypassLayoutVerification; using A0DataType = F16; using B0DataType = F16; @@ -139,11 +140,11 @@ int main(int argc, char* argv[]) if(std::is_same::value) { - return HostTensorDescriptor({row, col}, {stride, 1_uz}); + return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return HostTensorDescriptor({row, col}, {1_uz, stride}); + return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } }; diff --git a/example/65_gemm_multiply_multiply/run_gemm_multiply_multiply_wp_example.inc b/example/65_gemm_multiply_multiply/run_gemm_multiply_multiply_wp_example.inc index 2de3222380..10dce7fe64 100644 --- a/example/65_gemm_multiply_multiply/run_gemm_multiply_multiply_wp_example.inc +++ b/example/65_gemm_multiply_multiply/run_gemm_multiply_multiply_wp_example.inc @@ -5,6 +5,8 @@ int run_gemm_example(int argc, char* argv[]) { + using Bypass = ck::tensor_layout::BypassLayoutVerification; + bool do_verification = true; int init_method = 1; bool time_kernel = false; @@ -64,11 +66,11 @@ int run_gemm_example(int argc, char* argv[]) if(std::is_same::value) { - return ck::HostTensorDescriptor({row, col}, {stride, 1_uz}); + return ck::HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return ck::HostTensorDescriptor({row, col}, {1_uz, stride}); + return ck::HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } }; diff --git a/example/68_gemm_add/common.hpp b/example/68_gemm_add/common.hpp index 362dc2fff2..12d4b381b2 100644 --- a/example/68_gemm_add/common.hpp +++ b/example/68_gemm_add/common.hpp @@ -87,7 +87,7 @@ parse_cmd_args(int argc, char* argv[], ProblemSize& problem_size, ExecutionConfi config.init_method = std::stoi(argv[2]); config.time_kernel = std::stoi(argv[3]); } - else if(argc == 13) + else if(argc == 11) { config.do_verification = std::stoi(argv[1]); config.init_method = std::stoi(argv[2]); diff --git a/example/68_gemm_add/run_gemm_add_example_wmma.inc b/example/68_gemm_add/run_gemm_add_example_wmma.inc index ba15d03e07..0f2cc08edf 100644 --- a/example/68_gemm_add/run_gemm_add_example_wmma.inc +++ b/example/68_gemm_add/run_gemm_add_example_wmma.inc @@ -6,6 +6,7 @@ bool run_gemm_add(const ProblemSize& problem_size, const ExecutionConfig& config) { using namespace ck::literals; + using Bypass = ck::tensor_layout::BypassLayoutVerification; auto& [M, N, K, StrideA, StrideB, StrideD, StrideE] = problem_size; @@ -13,11 +14,11 @@ bool run_gemm_add(const ProblemSize& problem_size, const ExecutionConfig& config [](std::size_t row, std::size_t col, std::size_t stride, auto layout) { if(std::is_same::value) { - return HostTensorDescriptor({row, col}, {stride, 1_uz}); + return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return HostTensorDescriptor({row, col}, {1_uz, stride}); + return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } }; diff --git a/example/68_gemm_add/run_gemm_add_example_xdl.inc b/example/68_gemm_add/run_gemm_add_example_xdl.inc index da22230a4e..186423d32f 100644 --- a/example/68_gemm_add/run_gemm_add_example_xdl.inc +++ b/example/68_gemm_add/run_gemm_add_example_xdl.inc @@ -6,6 +6,7 @@ bool run_gemm_add(const ProblemSize& problem_size, const ExecutionConfig& config) { using namespace ck::literals; + using Bypass = ck::tensor_layout::BypassLayoutVerification; auto& [M, N, K, StrideA, StrideB, StrideD, StrideE] = problem_size; @@ -13,11 +14,11 @@ bool run_gemm_add(const ProblemSize& problem_size, const ExecutionConfig& config [](std::size_t row, std::size_t col, std::size_t stride, auto layout) { if(std::is_same::value) { - return HostTensorDescriptor({row, col}, {stride, 1_uz}); + return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return HostTensorDescriptor({row, col}, {1_uz, stride}); + return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } }; diff --git a/example/69_gemm_add_relu/common.hpp b/example/69_gemm_add_relu/common.hpp index e54c5317ae..de84d69a5e 100644 --- a/example/69_gemm_add_relu/common.hpp +++ b/example/69_gemm_add_relu/common.hpp @@ -87,7 +87,7 @@ parse_cmd_args(int argc, char* argv[], ProblemSize& problem_size, ExecutionConfi config.init_method = std::stoi(argv[2]); config.time_kernel = std::stoi(argv[3]); } - else if(argc == 13) + else if(argc == 11) { config.do_verification = std::stoi(argv[1]); config.init_method = std::stoi(argv[2]); diff --git a/example/69_gemm_add_relu/run_gemm_add_relu_example_wmma.inc b/example/69_gemm_add_relu/run_gemm_add_relu_example_wmma.inc index 8deac6dec8..c3cfd00ab3 100644 --- a/example/69_gemm_add_relu/run_gemm_add_relu_example_wmma.inc +++ b/example/69_gemm_add_relu/run_gemm_add_relu_example_wmma.inc @@ -6,6 +6,7 @@ bool run_gemm_add_relu(const ProblemSize& problem_size, const ExecutionConfig& config) { using namespace ck::literals; + using Bypass = ck::tensor_layout::BypassLayoutVerification; auto& [M, N, K, StrideA, StrideB, StrideD, StrideE] = problem_size; @@ -13,11 +14,11 @@ bool run_gemm_add_relu(const ProblemSize& problem_size, const ExecutionConfig& c [](std::size_t row, std::size_t col, std::size_t stride, auto layout) { if(std::is_same::value) { - return HostTensorDescriptor({row, col}, {stride, 1_uz}); + return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return HostTensorDescriptor({row, col}, {1_uz, stride}); + return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } }; diff --git a/example/69_gemm_add_relu/run_gemm_add_relu_example_xdl.inc b/example/69_gemm_add_relu/run_gemm_add_relu_example_xdl.inc index df7474bab5..cca85aa11c 100644 --- a/example/69_gemm_add_relu/run_gemm_add_relu_example_xdl.inc +++ b/example/69_gemm_add_relu/run_gemm_add_relu_example_xdl.inc @@ -6,6 +6,7 @@ bool run_gemm_add_relu(const ProblemSize& problem_size, const ExecutionConfig& config) { using namespace ck::literals; + using Bypass = ck::tensor_layout::BypassLayoutVerification; auto& [M, N, K, StrideA, StrideB, StrideD, StrideE] = problem_size; @@ -13,11 +14,11 @@ bool run_gemm_add_relu(const ProblemSize& problem_size, const ExecutionConfig& c [](std::size_t row, std::size_t col, std::size_t stride, auto layout) { if(std::is_same::value) { - return HostTensorDescriptor({row, col}, {stride, 1_uz}); + return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{}); } else { - return HostTensorDescriptor({row, col}, {1_uz, stride}); + return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{}); } };