Removed some old hack I used for benchmarking a while back

This commit is contained in:
jpekkila
2019-07-10 14:34:05 +03:00
parent 9af7193ffb
commit 866ec8a192
2 changed files with 5 additions and 4 deletions

View File

@@ -55,7 +55,6 @@ extern "C" {
#define REGISTERS_PER_THREAD (255)
#define MAX_REGISTERS_PER_BLOCK (65536)
#define MAX_THREADS_PER_BLOCK (1024)
#define NUM_ITERATIONS (10)
#define WARP_SIZE (32)
/*
* =============================================================================

View File

@@ -466,7 +466,8 @@ autoOptimize(const Device device)
cudaEventRecord(tstart); // ---------------------------------------- Timing start
for (int i = 0; i < NUM_ITERATIONS; ++i)
const int num_iterations = 10;
for (int i = 0; i < num_iterations; ++i)
solve<2><<<bpg, tpb>>>(start, end, device->vba, FLT_EPSILON);
cudaEventRecord(tstop); // ----------------------------------------- Timing end
@@ -483,8 +484,9 @@ autoOptimize(const Device device)
}
}
#if VERBOSE_PRINTING
printf("Auto-optimization done. The best threadblock dimensions for rkStep: (%d, %d, %d) %f ms\n", best_dims.x, best_dims.y, best_dims.z,
double(best_time) / NUM_ITERATIONS);
printf(
"Auto-optimization done. The best threadblock dimensions for rkStep: (%d, %d, %d) %f ms\n",
best_dims.x, best_dims.y, best_dims.z, double(best_time) / num_iterations);
#endif
/*
FILE* fp = fopen("../config/rk3_tbdims.cuh", "w");