adding bwd data v4r1 (multiple kernel launch)

driver/include/device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_deprecated.hpp

@@ -205,7 +205,8 @@ void device_convolution_implicit_gemm_v4r4_nchw_kcyx_nkhw_deprecated(InDesc,
 
     for(index_t i = 0; i < nrepeat; ++i)
     {
-        float time = launch_kernel(run_gridwise_convolution_kernel<decltype(gridwise_conv), T>,
+        float time =
+            launch_and_time_kernel(run_gridwise_convolution_kernel<decltype(gridwise_conv), T>,
                                    dim3(GridSize),
                                    dim3(BlockSize),
                                    0,

driver/include/device_direct_convolution_2_vectorized_nchw_kcyx_nkhw.hpp

@@ -178,7 +178,7 @@ void device_direct_convolution_2_vectorized_nchw_kcyx_nkhw(InDesc,
 
     for(index_t i = 0; i < nrepeat; ++i)
     {
-        float time = launch_kernel(
+        float time = launch_and_time_kernel(
             gridwise_direct_convolution_2_vectorized_nchw_kcyx_nkhw<TInWei,
                                                                     TOut,
                                                                     accum_t,

driver/src/conv_bwd_data_driver.cpp

@@ -17,18 +17,19 @@
 #include "device_convolution_backward_data_implicit_gemm_v1r2_nchw_kcyx_nkhw.hpp"
 #include "device_convolution_backward_data_implicit_gemm_v2r1_nchw_kcyx_nkhw.hpp"
 #include "device_convolution_backward_data_implicit_gemm_v3r1_nchw_kcyx_nkhw.hpp"
+#include "device_convolution_backward_data_implicit_gemm_v4r1_nchw_kcyx_nkhw.hpp"
 
 int main(int argc, char* argv[])
 {
     using namespace ck;
 
-#if 0
+#if 1
     // 3x3 filter, 2x2 stride, 35x35 input
     constexpr index_t N  = 128;
     constexpr index_t C  = 1024;
     constexpr index_t HI = 35;
     constexpr index_t WI = 35;
-    constexpr index_t K  = 1024;
+    constexpr index_t K  = 128;
     constexpr index_t Y  = 3;
     constexpr index_t X  = 3;
 
@@ -251,8 +252,10 @@ int main(int argc, char* argv[])
     device_convolution_backward_data_implicit_gemm_v1r2_nchw_kcyx_nkhw
 #elif 0
     device_convolution_backward_data_implicit_gemm_v2r1_nchw_kcyx_nkhw
+#elif 0
+    device_convolution_backward_data_implicit_gemm_v2r1_nchw_kcyx_nkhw
 #else
-    device_convolution_backward_data_implicit_gemm_v3r1_nchw_kcyx_nkhw
+    device_convolution_backward_data_implicit_gemm_v4r1_nchw_kcyx_nkhw
 #endif
     (in_nchw_desc,
      in_nchw_device,