revert workaround for square sample and update doc on GGL

Change-Id: I731c68ca4111e7dc2e45bef51c4cad2c23fc81f8


[ROCm/clr commit: 14f8c7cf74]

projects/clr/hipamd/docs/markdown/hip_faq.md

@@ -243,10 +243,4 @@ $HIP/include/hip/hcc_detail/hip_runtime_api.h
 $HIP/include/hip/hcc_detail/host_defines.h
 Or pass "-DGENERIC_GRID_LAUNCH=1" to hipcc at application compilation time.
 
-There are some limitation/assumptions of GGL implementation right now:
-1. GGL was only tested with Ubuntu16.04, assuming HCC and HIP only link against libstdc++ but not libc++.
-2. GGL currently requires templated kernel fucntions passed to hipLaunchKernel to be specialized. e.g.:
-
-```
-hipLaunchKernel(vector_square<float>, dim3(blocks), dim3(threadsPerBlock), 0, nullptr, C_d, A_d, N);
-```
+GGL was only tested with Ubuntu16.04, assuming HCC and HIP only link against libstdc++ but not libc++.

projects/clr/hipamd/include/hip/hcc_detail/grid_launch_v2.hpp

@@ -323,7 +323,6 @@ namespace glo_tests
             make_lambda_wrapper(kernel_name, __VA_ARGS__),                     \
             ##__VA_ARGS__);                                                    \
     }
-#warning "GGL hipLaunchKernel Reached"
     #define hipLaunchKernel(                                                 \
         kernel_name,                                                           \
         num_blocks,                                                            \

projects/clr/hipamd/include/hip/hcc_detail/hip_runtime_api.h

@@ -32,7 +32,7 @@ THE SOFTWARE.
 #include <iostream>
 
 #ifndef GENERIC_GRID_LAUNCH
-#define GENERIC_GRID_LAUNCH 0
+#define GENERIC_GRID_LAUNCH 0 
 #endif
 
 #include <hip/hcc_detail/host_defines.h>

projects/clr/hipamd/include/hip/hcc_detail/host_defines.h

@@ -32,7 +32,7 @@ THE SOFTWARE.
 
 // Add guard to Generic Grid Launch method
 #ifndef GENERIC_GRID_LAUNCH
-#define GENERIC_GRID_LAUNCH 0
+#define GENERIC_GRID_LAUNCH 0 
 #endif
 
 #ifdef __HCC__

projects/clr/hipamd/samples/0_Intro/square/square.hipref.cpp

@@ -83,7 +83,7 @@ int main(int argc, char *argv[])
     const unsigned threadsPerBlock = 256;
 
     printf ("info: launch 'vector_square' kernel\n");
-    hipLaunchKernel(vector_square<float>, dim3(blocks), dim3(threadsPerBlock), 0, nullptr, C_d, A_d, N);
+    hipLaunchKernel(vector_square, dim3(blocks), dim3(threadsPerBlock), 0, nullptr, C_d, A_d, N);
 
     printf ("info: copy Device2Host\n");
     CHECK ( hipMemcpy(C_h, C_d, Nbytes, hipMemcpyDeviceToHost));