Skip to content

vulkan: set all memory allocations to high priority #17624

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
0cc4m merged 2 commits into from
Dec 5, 2025
Merged

vulkan: set all memory allocations to high priority #17624

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b33c590
vulkan: set all memory allocations to high priority
jeffbolznv Nov 30, 2025
0ba4fee
gate by env var
jeffbolznv Dec 2, 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
22 changes: 21 additions & 1 deletion ggml/src/ggml-vulkan/ggml-vulkan.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -519,6 +519,7 @@ struct vk_device_struct {
bool fp16;
bool bf16;
bool pipeline_robustness;
bool memory_priority;
vk::Device device;
uint32_t vendor_id;
vk::DriverId driver_id;
Expand Down Expand Up @@ -2360,7 +2361,13 @@ static vk_buffer ggml_vk_create_buffer(vk_device& device, size_t size, const std

vk::PhysicalDeviceMemoryProperties mem_props = device->physical_device.getMemoryProperties();

const vk::MemoryAllocateFlagsInfo mem_flags_info { mem_flags };
const vk::MemoryPriorityAllocateInfoEXT mem_priority_info { 1.0f };

vk::MemoryAllocateFlagsInfo mem_flags_info { mem_flags };

if (device->memory_priority) {
mem_flags_info.setPNext(&mem_priority_info);
}

for (auto it = req_flags_list.begin(); it != req_flags_list.end(); it++) {
const auto & req_flags = *it;
Expand Down Expand Up @@ -4331,6 +4338,9 @@ static vk_device ggml_vk_get_device(size_t idx) {
#endif
} else if (strcmp("VK_KHR_pipeline_executable_properties", properties.extensionName) == 0) {
pipeline_executable_properties_support = true;
} else if (strcmp("VK_EXT_memory_priority", properties.extensionName) == 0 &&
getenv("GGML_VK_ENABLE_MEMORY_PRIORITY")) {
device->memory_priority = true;
}
}

Expand Down Expand Up @@ -4522,6 +4532,16 @@ static vk_device ggml_vk_get_device(size_t idx) {
device_extensions.push_back("VK_EXT_pipeline_robustness");
}

VkPhysicalDeviceMemoryPriorityFeaturesEXT memory_priority_features;
memory_priority_features.pNext = nullptr;
memory_priority_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_PRIORITY_FEATURES_EXT;
memory_priority_features.memoryPriority = VK_FALSE;
if (device->memory_priority) {
last_struct->pNext = (VkBaseOutStructure *)&memory_priority_features;
last_struct = (VkBaseOutStructure *)&memory_priority_features;
device_extensions.push_back("VK_EXT_memory_priority");
}

VkPhysicalDeviceSubgroupSizeControlFeaturesEXT subgroup_size_control_features;
subgroup_size_control_features.pNext = nullptr;
subgroup_size_control_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_SIZE_CONTROL_FEATURES_EXT;
Expand Down
Loading