Validate examples in CI (#233)

* validate examples in ctest runs

* format

* fix usage of check_err

* amend

* add example codes to custom target 'check'
Co-authored-by: Chao Liu <chao.liu2@amd.com>

example/15_grouped_gemm/grouped_gemm_xdl_fp16.cpp

@@ -211,6 +211,7 @@ int main(int argc, char* argv[])
     std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
               << gemm.GetTypeString() << std::endl;
 
+    bool pass = true;
     if(do_verification)
     {
         for(std::size_t i = 0; i < gemm_shapes.size(); i++)
@@ -227,9 +228,9 @@ int main(int argc, char* argv[])
                                                       c_element_op);
 
             ref_invoker.Run(ref_argument);
-            ck::utils::check_err(c_device_tensors[i].mData, c_host_tensors[i].mData);
+            pass &= ck::utils::check_err(c_device_tensors[i].mData, c_host_tensors[i].mData);
         }
     }
 
-    return 0;
+    return pass ? 0 : 1;
 }

example/16_gemm_reduce/gemm_reduce_xdl_fp16.cpp

@@ -4,6 +4,7 @@
 #include <cstdlib>
 #include <stdlib.h>
 #include <half.hpp>
+#include "check_err.hpp"
 #include "config.hpp"
 #include "device.hpp"
 #include "host_tensor.hpp"
@@ -211,6 +212,7 @@ int main(int argc, char* argv[])
     std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
               << gemm.GetTypeString() << std::endl;
 
+    bool pass = true;
     if(do_verification)
     {
         c_device_buf.FromDevice(c_m_n_device_result.mData.data());
@@ -247,10 +249,19 @@ int main(int argc, char* argv[])
             d1_m_host_result(m) = ck::type_convert<DDataType>(d1_acc);
         }
 
-        check_error(c_m_n_host_result, c_m_n_device_result);
-        check_error(d0_m_host_result, d0_m_device_result);
-        check_error(d1_m_host_result, d1_m_device_result);
+        pass &= ck::utils::check_err(
+            c_m_n_device_result.mData, c_m_n_host_result.mData, "Error: Incorrect results c");
+        pass &= ck::utils::check_err(d0_m_device_result.mData,
+                                     d0_m_host_result.mData,
+                                     "Error: Incorrect results d0",
+                                     1e-3,
+                                     1e-3);
+        pass &= ck::utils::check_err(d1_m_device_result.mData,
+                                     d1_m_host_result.mData,
+                                     "Error: Incorrect results d1",
+                                     1e-3,
+                                     1e-3);
     }
 
-    return 0;
+    return pass ? 0 : 1;
 }

example/17_convnd_bwd_data_xdl/convnd_bwd_data_xdl.cpp

@@ -322,7 +322,10 @@ int main(int argc, char* argv[])
 
             in_device_buf.FromDevice(in_n_c_hi_wi_device_result.mData.data());
 
-            check_error(in_n_c_hi_wi_host_result, in_n_c_hi_wi_device_result);
+            return ck::utils::check_err(in_n_c_hi_wi_device_result.mData,
+                                        in_n_c_hi_wi_host_result.mData)
+                       ? 0
+                       : 1;
         };
 
         switch(num_dim_spatial)
@@ -347,4 +350,5 @@ int main(int argc, char* argv[])
         }
         }
     }
+    return 0;
 }

example/18_batched_gemm_reduce/batched_gemm_reduce_xdl_fp16.cpp

@@ -4,6 +4,7 @@
 #include <cstdlib>
 #include <stdlib.h>
 #include <half.hpp>
+#include "check_err.hpp"
 #include "config.hpp"
 #include "device.hpp"
 #include "host_tensor.hpp"
@@ -62,13 +63,13 @@ int main(int argc, char* argv[])
     bool time_kernel     = false;
 
     // GEMM shape
-    ck::index_t M = 3840;
-    ck::index_t N = 4096;
-    ck::index_t K = 4096;
+    ck::index_t M = 2048;
+    ck::index_t N = 1920;
+    ck::index_t K = 2048;
 
-    ck::index_t StrideA = 4096;
-    ck::index_t StrideB = 4096;
-    ck::index_t StrideC = 4096;
+    ck::index_t StrideA = 2048;
+    ck::index_t StrideB = 2048;
+    ck::index_t StrideC = 1920;
 
     ck::index_t BatchCount = 4;
 
@@ -96,7 +97,7 @@ int main(int argc, char* argv[])
         StrideB = std::stoi(argv[8]);
         StrideC = std::stoi(argv[9]);
 
-        BatchCount = std::stoi(argv[9]);
+        BatchCount = std::stoi(argv[10]);
     }
     else
     {
@@ -224,6 +225,7 @@ int main(int argc, char* argv[])
     std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
               << batched_gemm.GetTypeString() << std::endl;
 
+    bool pass = true;
     if(do_verification)
     {
         c_device_buf.FromDevice(c_g_m_n_device_result.mData.data());
@@ -247,7 +249,7 @@ int main(int argc, char* argv[])
 
                 for(int n = 0; n < N; ++n)
                 {
-                    float d0_val = ck::type_convert<float>(c_g_m_n_host_result(m, n));
+                    float d0_val = ck::type_convert<float>(c_g_m_n_host_result(batch, m, n));
                     float d1_val;
 
                     d1_element_op(d1_val, d0_val);
@@ -260,10 +262,18 @@ int main(int argc, char* argv[])
             }
         }
 
-        check_error(c_g_m_n_host_result, c_g_m_n_device_result);
-        check_error(d0_g_m_host_result, d0_g_m_device_result);
-        check_error(d1_g_m_host_result, d1_g_m_device_result);
+        pass &= ck::utils::check_err(c_g_m_n_host_result.mData, c_g_m_n_device_result.mData);
+        pass &= ck::utils::check_err(d0_g_m_device_result.mData,
+                                     d0_g_m_host_result.mData,
+                                     "Error: Incorrect results! D0",
+                                     1e-3,
+                                     1e-3);
+        pass &= ck::utils::check_err(d1_g_m_device_result.mData,
+                                     d1_g_m_host_result.mData,
+                                     "Error: Incorrect results! D1",
+                                     1e-3,
+                                     1e-3);
     }
 
-    return 0;
+    return pass ? 0 : 1;
 }

example/CMakeLists.txt

@@ -19,9 +19,18 @@ include_directories(BEFORE
 
 add_custom_target(examples)
 
-function(add_example_executable EXAMPLE_NAME)
+function(add_example_executable EXAMPLE_NAME FILE_NAME)
     message("adding example ${EXAMPLE_NAME}")
-    add_executable(${EXAMPLE_NAME} ${ARGN})
+    add_executable(${EXAMPLE_NAME} ${FILE_NAME})
+    target_link_libraries(${EXAMPLE_NAME} PRIVATE host_tensor)
+    add_test(NAME ${EXAMPLE_NAME} COMMAND $<TARGET_FILE:${EXAMPLE_NAME}> ${ARGN})
+    add_dependencies(examples ${EXAMPLE_NAME})
+    add_dependencies(check ${EXAMPLE_NAME})
+endfunction(add_example_executable EXAMPLE_NAME)
+
+function(add_example_executable_no_testing EXAMPLE_NAME FILE_NAME)
+    message("adding example ${EXAMPLE_NAME}")
+    add_executable(${EXAMPLE_NAME} ${FILE_NAME})
     target_link_libraries(${EXAMPLE_NAME} PRIVATE host_tensor)
     add_dependencies(examples ${EXAMPLE_NAME})
 endfunction(add_example_executable EXAMPLE_NAME)

test/CMakeLists.txt

@@ -24,7 +24,6 @@ include_directories(BEFORE
 
 include(googletest)
 
-add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND} --output-on-failure -C ${CMAKE_CFG_INTDIR})
 add_custom_target(tests)