make fftw platform agnostic

fftw.c

@@ -46,10 +46,9 @@ run_test
         }
 
         //  Copy data to device
-        hipMemcpy(A, hA, sizeof(*hA) * N, hipMemcpyHostToDevice);
-
+        __ASSERT(__PREFIX(Memcpy)(A, hA, sizeof(*hA) * N, __PREFIX(MemcpyHostToDevice)));
         // Create FFT plan
-	__FFTW_PLAN plan = NULL;
+	__FFTW_PLAN plan;
         size_t length = N;
 
 	char mes[128];
@@ -80,14 +79,16 @@ run_test
 #ifdef ROC
 		rocfft_execute(plan, (void**) &A, NULL, NULL);
 #elif CUDA
-	cufftExecC2C(plan,  A, A);
+	cufftExecC2C(plan,A , A, CUFFT_FORWARD);
 #endif
 		timer_stop(timer);
 
 		// Wait for execution to finish
 		sprintf(mes, "dim: %zu\tSynchronize round %d." ,N , r);
 		timer_start(timer, mes);
+#ifdef ROC
 		hipDeviceSynchronize();
+#endif
 		timer_stop(timer);
 	}
         // Destroy plan
@@ -96,7 +97,7 @@ run_test
         __DESTROY_PLAN(plan);
 	timer_stop(timer);
 
-        __ASSERT(__PREFIX(Memcpy)(hB, A, sizeof(*A) * N, hipMemcpyDeviceToHost));
+        __ASSERT(__PREFIX(Memcpy)(hB, A, sizeof(*A) * N, __PREFIX(MemcpyDeviceToHost)));
 
         __ASSERT(__PREFIX(Free)(A));
         free(hA);
@@ -111,7 +112,7 @@ main
 )
 {
 	int rep = 1;
-	int min_dim = 12;
+	int min_dim = 8;
 	int max_dim = 28;
 	int nofftws = 128;