cuda : add error checking for cudaMemcpyAsync in argsort (llama/17599)

* cuda : add error checking for cudaMemcpyAsync in argsort (llama/12836)

* fix indentation

diff --git a/ggml/src/ggml-cuda/argsort.cu b/ggml/src/ggml-cuda/argsort.cu
index 3722cf3ab26ee7bc9e677636349f9c4bcd565e26..da9652c3be8c9d9ea0431b16e2803da2888f37d3 100644 (file)
--- a/ggml/src/ggml-cuda/argsort.cu
+++ b/ggml/src/ggml-cuda/argsort.cu
@@ -44,7 +44,7 @@ static void argsort_f32_i32_cuda_cub(ggml_cuda_pool & pool,
     const dim3 offset_grid((nrows + block_size - 1) / block_size);
     init_offsets<<<offset_grid, block_size, 0, stream>>>(d_offsets, ncols, nrows);
 
-    cudaMemcpyAsync(temp_keys, x, ncols * nrows * sizeof(float), cudaMemcpyDeviceToDevice, stream);
+    CUDA_CHECK(cudaMemcpyAsync(temp_keys, x, ncols * nrows * sizeof(float), cudaMemcpyDeviceToDevice, stream));
 
     size_t temp_storage_bytes = 0;