enable hc_am example in hcc_ddialects example

Change-Id: Iec2f9eb05f95cb025c157fee8fd284aab844d1a2


[ROCm/hip commit: cd3a0a2d61]

projects/hip/samples/0_Intro/hcc_dialects/Makefile

@@ -1,8 +1,9 @@
 HCC_HOME?=/opt/rocm/hcc
 HCC = $(HCC_HOME)/bin/hcc
 
-HCC_CFLAGS= `$(HCC_HOME)/bin/hcc-config --cxxflags` 
-HCC_LDFLAGS= `$(HCC_HOME)/bin/hcc-config --ldflags`
+OPT=-O2
+HCC_CFLAGS= `$(HCC_HOME)/bin/hcc-config --cxxflags` ${OPT}
+HCC_LDFLAGS= `$(HCC_HOME)/bin/hcc-config --ldflags` ${OPT}
 
 CPPAMP_CFLAGS= -std=c++amp -stdlib=libc++ -I$(HCC_HOME)/include
 CPPAMP_LDFLAGS=  -std=c++amp -L$(HCC_HOME)/lib -Wl,--rpath=$(HCC_HOME)/lib -lc++ -lc++abi -ldl -lpthread -Wl,--whole-archive -lmcwamp -Wl,--no-whole-archive 
@@ -19,7 +20,7 @@ $(error hcc_dialects requires hcc compiler and only runs on hcc platform)
 endif
 
 
-TARGETS=vadd_hc_arrayview vadd_hc_array vadd_amp_arrayview  vadd_hip  
+TARGETS=vadd_hc_arrayview vadd_hc_array vadd_hc_am vadd_amp_arrayview  vadd_hip  
 
 all: $(TARGETS)
 
@@ -51,7 +52,7 @@ vadd_hc_array: vadd_hc_array.o
 vadd_hc_am.o: vadd_hc_am.cpp
 	$(HCC) $(HCC_CFLAGS) -c $< -o $@
 vadd_hc_am: vadd_hc_am.o
-	$(HCC) $(HCC_LDFLAGS) $< -o $@
+	$(HCC) $(HCC_LDFLAGS) -lhc_am $< -o $@
 
 
 

projects/hip/samples/0_Intro/hcc_dialects/vadd_hc_am.cpp

@@ -32,23 +32,30 @@ int main(int argc, char *argv[])
     for (int i=0; i<sizeElements; i++) {
         A_h[i] = 1.618f * i; 
         B_h[i] = 3.142f * i;
+        C_h[i] = 0;
     }
 
-    av.copy(A_h, A_d); // C++ copy H2D
-    av.copy(B_h, B_d); //C++  copy H2D
+    av.copy(A_h, A_d, sizeBytes); // C++ copy H2D
+    av.copy(B_h, B_d, sizeBytes); // C++ copy H2D
 
     // Launch kernel onto AV.  
     // Because the kernel PFE and the copies are submitted to same AV, they will execute in order
     // and we don't need additional synchronization to ensure the copies complete before the PFE begins.
+#if 1
+    hc::completion_future cf=
     hc::parallel_for_each(av,  hc::extent<1> (sizeElements),
       [&] (hc::index<1> idx) [[hc]] { 
         int i = idx[0];
-        C_d[i] = A_d[i] + B_d[i];
+        //C_d[i] = A_d[i] + B_d[i];
+        C_d[0] = A_d[1] + B_d[2];
     });
+    cf.wait();
+#endif
+
 
    
     // This copy is in same AV as the kernel and thus will wait for the kernel to finish before executing.
-    av.copy(C_d, C_h); // C++ copy D2H
+    av.copy(C_d, C_h, sizeBytes); // C++ copy D2H
 
 
     for (int i=0; i<sizeElements; i++) {