added architecture specific macros

1. Added __HIP_ARCH_GFXNUM__
2. Usage, -D__HIP_ARCH_GFX803__=1

Change-Id: I68b3a85d62cfab3a45d2b7a70cb3518ab2565236
This commit is contained in:
Aditya Atluri
2017-02-08 19:45:32 -06:00
parent f2bfc8a405
commit c6c90f6f5a
2 changed files with 10 additions and 2 deletions
@@ -25,6 +25,8 @@ THE SOFTWARE.
#define HALF_SIZE 64*sizeof(__half)
#define HALF2_SIZE 64*sizeof(__half2)
#if __HIP_ARCH_GFX803__ > 0
__global__ void __halfMath(hipLaunchParm lp, __half *A, __half *B, __half *C) {
int tx = hipThreadIdx_x;
__half a = A[tx];
@@ -56,6 +58,8 @@ __global__ void __half2Math(hipLaunchParm lp, __half2 *A, __half2 *B, __half2 *C
c = __hmul2_sat(b, c);
}
#endif
int main(){
__half *A, *B, *C;
hipMalloc(&A, HALF_SIZE);