diff options
Diffstat (limited to 'examples/gpu/cuda-benchmark.zc')
| -rw-r--r-- | examples/gpu/cuda-benchmark.zc | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/examples/gpu/cuda-benchmark.zc b/examples/gpu/cuda-benchmark.zc index cea326e..8f4a85d 100644 --- a/examples/gpu/cuda-benchmark.zc +++ b/examples/gpu/cuda-benchmark.zc @@ -194,7 +194,7 @@ fn benchmark_matrix_multiply(N: int) { cuda_copy_to_device(d_B, h_B, size * sizeof(float)); // Configure grid - const BLOCK_SIZE = 16; + def BLOCK_SIZE = 16; let blocks_per_grid = (N + BLOCK_SIZE - 1) / BLOCK_SIZE; "-> Launching kernel: {blocks_per_grid}x{blocks_per_grid} blocks, {BLOCK_SIZE}x{BLOCK_SIZE} threads each"; @@ -232,8 +232,8 @@ fn benchmark_monte_carlo_pi(num_samples: u64) { "-> Estimating Pi with {num_samples} samples"; - const BLOCK_SIZE = 256; - const NUM_BLOCKS = 1024; + def BLOCK_SIZE = 256; + def NUM_BLOCKS = 1024; let total_threads = BLOCK_SIZE * NUM_BLOCKS; // Allocate memory @@ -342,7 +342,7 @@ fn benchmark_nbody(num_bodies: int, num_steps: int) { cuda_copy_to_device(d_vy, h_vy, num_bodies * sizeof(float)); cuda_copy_to_device(d_vz, h_vz, num_bodies * sizeof(float)); - const BLOCK_SIZE = 256; + def BLOCK_SIZE = 256; let num_blocks = (num_bodies + BLOCK_SIZE - 1) / BLOCK_SIZE; let dt = 0.01f; |