Skip to content

vulkan: disable async for older Intel devices #17369

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Nov 21, 2025
Merged

vulkan: disable async for older Intel devices #17369

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
addc995
vulkan: disable async for older Intel devices
jeffbolznv Nov 18, 2025
7d68182
update detection logic
jeffbolznv Nov 20, 2025
703c61f
use name string for detection
jeffbolznv Nov 21, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 19 additions & 0 deletions ggml/src/ggml-vulkan/ggml-vulkan.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -513,6 +513,7 @@ struct vk_device_struct {
vk_queue compute_queue;
vk_queue transfer_queue;
bool single_queue;
bool support_async;
uint32_t subgroup_size;
uint32_t shader_core_count;
bool uma;
Expand Down Expand Up @@ -4222,6 +4223,16 @@ static vk_device ggml_vk_get_device(size_t idx) {
device->vendor_id = device->properties.vendorID;
device->driver_id = driver_props.driverID;

// Implementing the async backend interfaces seems broken on older Intel HW,
// see https://github.com/ggml-org/llama.cpp/issues/17302.
device->support_async = (device->vendor_id != VK_VENDOR_ID_INTEL ||
std::string(device->properties.deviceName.data()).find("(DG1)") == std::string::npos) &&
getenv("GGML_VK_DISABLE_ASYNC") == nullptr;

if (!device->support_async) {
GGML_LOG_DEBUG("ggml_vulkan: WARNING: Async execution disabled on certain Intel devices.\n");
}

const char* GGML_VK_FORCE_MAX_ALLOCATION_SIZE = getenv("GGML_VK_FORCE_MAX_ALLOCATION_SIZE");

if (GGML_VK_FORCE_MAX_ALLOCATION_SIZE != nullptr) {
Expand Down Expand Up @@ -12923,6 +12934,10 @@ static ggml_status ggml_backend_vk_graph_compute(ggml_backend_t backend, ggml_cg
ctx->device->perf_logger->print_timings();
}

if (!ctx->device->support_async) {
ggml_vk_synchronize(ctx);
}

return GGML_STATUS_SUCCESS;

UNUSED(backend);
Expand Down Expand Up @@ -13216,6 +13231,10 @@ ggml_backend_t ggml_backend_vk_init(size_t dev_num) {
/* .context = */ ctx,
};

if (!ctx->device->support_async) {
vk_backend->iface.get_tensor_async = nullptr;
}

return vk_backend;
}

Expand Down
Loading