[CK][Examples] Fixing stride issues in ck examples 14/65/68/69 by workaround - Bypassing hostTensor validation

AMD-mkulikow · AMD-mkulikow · commit f696f557f5a2 · 2026-01-15T12:35:24.000+01:00
-Fixing args num in ck examples 68/69

Signed-off-by: Michal Kulikowski &lt;Michal.Kulikowski@amd.com&gt;
diff --git 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 <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
 
-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<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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
@@ -31,8 +31,9 @@ using S = ck::Sequence<Is...>;
 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<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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
@@ -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<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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
@@ -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
@@ -6,18 +6,19 @@
 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;
 
     auto f_host_tensor_descriptor =
         [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
             if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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
@@ -6,18 +6,19 @@
 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;
 
     auto f_host_tensor_descriptor =
         [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
             if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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
@@ -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
@@ -6,18 +6,19 @@
 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;
 
     auto f_host_tensor_descriptor =
         [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
             if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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
@@ -6,18 +6,19 @@
 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;
 
     auto f_host_tensor_descriptor =
         [](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
             if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::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{});
             }
         };
 

Original file line number	Diff line number	Diff line change
`@@ -31,8 +31,9 @@ using S = ck::Sequence<Is...>;`
`31`	`31`	`using F16 = ck::half_t;`
`32`	`32`	`using F32 = float;`
`33`	`33`
`34`		`-using Row = ck::tensor_layout::gemm::RowMajor;`
`35`		`-using Col = ck::tensor_layout::gemm::ColumnMajor;`
	`34`	`+using Row = ck::tensor_layout::gemm::RowMajor;`
	`35`	`+using Col = ck::tensor_layout::gemm::ColumnMajor;`
	`36`	`+using Bypass = ck::tensor_layout::BypassLayoutVerification;`
`36`	`37`
`37`	`38`	`using A0DataType = F16;`
`38`	`39`	`using B0DataType = F16;`
`@@ -139,11 +140,11 @@ int main(int argc, char* argv[])`
`139`	`140`
`140`	`141`	`if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)`
`141`	`142`	`{`
`142`		`- return HostTensorDescriptor({row, col}, {stride, 1_uz});`
	`143`	`+ return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{});`
`143`	`144`	`}`
`144`	`145`	`else`
`145`	`146`	`{`
`146`		`- return HostTensorDescriptor({row, col}, {1_uz, stride});`
	`147`	`+ return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{});`
`147`	`148`	`}`
`148`	`149`	`};`
`149`	`150`
Original file line number	Diff line number	Diff line change
`@@ -5,6 +5,8 @@`
`5`	`5`
`6`	`6`	`int run_gemm_example(int argc, char* argv[])`
`7`	`7`	`{`
	`8`	`+ using Bypass = ck::tensor_layout::BypassLayoutVerification;`
	`9`	`+`
`8`	`10`	`bool do_verification = true;`
`9`	`11`	`int init_method = 1;`
`10`	`12`	`bool time_kernel = false;`
`@@ -64,11 +66,11 @@ int run_gemm_example(int argc, char* argv[])`
`64`	`66`
`65`	`67`	`if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)`
`66`	`68`	`{`
`67`		`- return ck::HostTensorDescriptor({row, col}, {stride, 1_uz});`
	`69`	`+ return ck::HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{});`
`68`	`70`	`}`
`69`	`71`	`else`
`70`	`72`	`{`
`71`		`- return ck::HostTensorDescriptor({row, col}, {1_uz, stride});`
	`73`	`+ return ck::HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{});`
`72`	`74`	`}`
`73`	`75`	`};`
`74`	`76`
Original file line number	Diff line number	Diff line change
`@@ -87,7 +87,7 @@ parse_cmd_args(int argc, char* argv[], ProblemSize& problem_size, ExecutionConfi`
`87`	`87`	`config.init_method = std::stoi(argv[2]);`
`88`	`88`	`config.time_kernel = std::stoi(argv[3]);`
`89`	`89`	`}`
`90`		`- else if(argc == 13)`
	`90`	`+ else if(argc == 11)`
`91`	`91`	`{`
`92`	`92`	`config.do_verification = std::stoi(argv[1]);`
`93`	`93`	`config.init_method = std::stoi(argv[2]);`
Original file line number	Diff line number	Diff line change
`@@ -6,18 +6,19 @@`
`6`	`6`	`bool run_gemm_add(const ProblemSize& problem_size, const ExecutionConfig& config)`
`7`	`7`	`{`
`8`	`8`	`using namespace ck::literals;`
	`9`	`+ using Bypass = ck::tensor_layout::BypassLayoutVerification;`
`9`	`10`
`10`	`11`	`auto& [M, N, K, StrideA, StrideB, StrideD, StrideE] = problem_size;`
`11`	`12`
`12`	`13`	`auto f_host_tensor_descriptor =`
`13`	`14`	`[](std::size_t row, std::size_t col, std::size_t stride, auto layout) {`
`14`	`15`	`if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)`
`15`	`16`	`{`
`16`		`- return HostTensorDescriptor({row, col}, {stride, 1_uz});`
	`17`	`+ return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{});`
`17`	`18`	`}`
`18`	`19`	`else`
`19`	`20`	`{`
`20`		`- return HostTensorDescriptor({row, col}, {1_uz, stride});`
	`21`	`+ return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{});`
`21`	`22`	`}`
`22`	`23`	`};`
`23`	`24`
Original file line number	Diff line number	Diff line change
`@@ -6,18 +6,19 @@`
`6`	`6`	`bool run_gemm_add_relu(const ProblemSize& problem_size, const ExecutionConfig& config)`
`7`	`7`	`{`
`8`	`8`	`using namespace ck::literals;`
	`9`	`+ using Bypass = ck::tensor_layout::BypassLayoutVerification;`
`9`	`10`
`10`	`11`	`auto& [M, N, K, StrideA, StrideB, StrideD, StrideE] = problem_size;`
`11`	`12`
`12`	`13`	`auto f_host_tensor_descriptor =`
`13`	`14`	`[](std::size_t row, std::size_t col, std::size_t stride, auto layout) {`
`14`	`15`	`if(std::is_same<decltype(layout), ck::tensor_layout::gemm::RowMajor>::value)`
`15`	`16`	`{`
`16`		`- return HostTensorDescriptor({row, col}, {stride, 1_uz});`
	`17`	`+ return HostTensorDescriptor({row, col}, {stride, 1_uz}, Bypass{});`
`17`	`18`	`}`
`18`	`19`	`else`
`19`	`20`	`{`
`20`		`- return HostTensorDescriptor({row, col}, {1_uz, stride});`
	`21`	`+ return HostTensorDescriptor({row, col}, {1_uz, stride}, Bypass{});`
`21`	`22`	`}`
`22`	`23`	`};`
`23`	`24`