CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: fix DEBUG_CUDA_MALLOC (#5729)

Browse Source
This commit is contained in:
Johannes Gäßler
2024-02-26 15:36:38 +01:00
committed by GitHub
parent c4d7f81786
commit 47bb7b48c7
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
ggml-cuda.cu
View File

@@ -8079,8 +8079,8 @@ static void * ggml_cuda_pool_malloc_leg(int device, size_t size, size_t * actual
*actual_size = look_ahead_size;
g_cuda_pool_size[device] += look_ahead_size;
#ifdef DEBUG_CUDA_MALLOC
fprintf(stderr, "%s[%d]: %d buffers, max_size = %u MB, pool_size = %u MB, requested %u MB\n", __func__, id, nnz,
(uint32_t)(max_size/1024/1024), (uint32_t)(g_cuda_pool_size[id]/1024/1024), (uint32_t)(size/1024/1024));
fprintf(stderr, "%s[%d]: %d buffers, max_size = %u MB, pool_size = %u MB, requested %u MB\n", __func__, device, nnz,
(uint32_t)(max_size/1024/1024), (uint32_t)(g_cuda_pool_size[device]/1024/1024), (uint32_t)(size/1024/1024));
#endif
return ptr;
}
@@ -8166,7 +8166,7 @@ static void * ggml_cuda_pool_malloc_vmm(int device, size_t size, size_t * actual
g_cuda_pool_used[device] += size;
#ifdef DEBUG_CUDA_MALLOC
printf("cuda pool[%d]: allocated %llu bytes at %llx [%s]\n", id, (unsigned long long) size, ptr);
printf("cuda pool[%d]: allocated %llu bytes at %llx\n", device, (unsigned long long) size, ptr);
#endif
return ptr;
@@ -8176,7 +8176,7 @@ static void ggml_cuda_pool_free_vmm(int device, void * ptr, size_t size) {
scoped_spin_lock lock(g_cuda_pool_lock);
#ifdef DEBUG_CUDA_MALLOC
printf("cuda pool[%d]: freed %llu bytes at %llx\n", id, (unsigned long long) size, ptr);
printf("cuda pool[%d]: freed %llu bytes at %llx\n", device, (unsigned long long) size, ptr);
#endif
g_cuda_pool_used[device] -= size;