ExternalVendorCode/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-04-08 03:44:46 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Vulkan: Default to 1GB allocations instead of 4GB to avoid fragmentation and driver issues (llama/12434)

Browse Source
This commit is contained in:
0cc4m 2025-03-18 07:21:40 +01:00 committed by Georgi Gerganov
parent 6c15539c54
commit fa72479cfb
1 changed files with 1 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@ -2524,13 +2524,9 @@ static vk_device ggml_vk_get_device(size_t idx) {
if (GGML_VK_SUBALLOCATION_BLOCK_SIZE != nullptr) {
device->suballocation_block_size = std::stoul(GGML_VK_SUBALLOCATION_BLOCK_SIZE);
#if defined(_WIN32)
} else if (device->vendor_id == VK_VENDOR_ID_NVIDIA) {
} else {
// Limit batching of allocations to 1GB by default to avoid fragmentation issues
device->suballocation_block_size = 1024*1024*1024;
#endif
} else {
device->suballocation_block_size = device->max_memory_allocation_size;
}
device->suballocation_block_size = std::min(device->suballocation_block_size, device->max_memory_allocation_size);