frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 12:54:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add device_id to XPU device properties (#156481)

Browse Source 
# Motivation

Some older Intel iGPUs may share the same device name across different hardware products.
(See [device name example](aaa01c06f9/shared/source/dll/devices/devices_base.inl (L190-L199)))
To help disambiguate which specific iGPU product is being used, we introduce the use of a
[device id](https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/supported/sycl_ext_intel_device_info.md#device-id). This device id corresponds to the Device ID in [official Intel product specification](https://www.intel.com/content/www/us/en/products/sku/232155/intel-core-i71360p-processor-18m-cache-up-to-5-00-ghz/specifications.html) and enables more accurate identification and troubleshooting for user issues.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/156481
Approved by: https://github.com/EikanWang, https://github.com/albanD
This commit is contained in:
Yu, Guangye
2025-06-29 06:17:25 +00:00
committed by PyTorch MergeBot
parent 7597988f1b
commit 5cc4e856fd
4 changed files with 22 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
c10/xpu/XPUDeviceProp.h
View File

@ -113,18 +113,21 @@ namespace c10::xpu {
_(native_vector_width_double) \
_(native_vector_width_half)
#define AT_FORALL_XPU_EXT_DEVICE_PROPERTIES(_) \
/* the number of EUs associated with the Intel GPU. */ \
_(gpu_eu_count, 512) \
\
/* the number of EUs in a subslice. */ \
_(gpu_eu_count_per_subslice, 8) \
\
/* the simd width of EU of GPU. */ \
_(gpu_eu_simd_width, 8) \
\
/* the number of hardware threads per EU of GPU. */ \
_(gpu_hw_threads_per_eu, 8)
#define AT_FORALL_XPU_EXT_DEVICE_PROPERTIES(_) \
/* the number of EUs associated with the Intel GPU. */ \
_(gpu_eu_count, 512) \
\
/* the number of EUs in a subslice. */ \
_(gpu_eu_count_per_subslice, 8) \
\
/* the simd width of EU of GPU. */ \
_(gpu_eu_simd_width, 8) \
\
/* the number of hardware threads per EU of GPU. */ \
_(gpu_hw_threads_per_eu, 8) \
\
/* the device identifier of the Intel GPU, also known as the product ID. */ \
_(device_id, 0)
#define AT_FORALL_XPU_DEVICE_ASPECT(_) \
/* sycl::half is supported on device. */ \

1
test/test_xpu.py
View File

@ -101,6 +101,7 @@ class TestXpu(TestCase):
self.assertEqual(device_name, torch.xpu.get_device_name())
device_capability = torch.xpu.get_device_capability(current_device)
self.assertTrue(device_capability["device_id"] > 0)
self.assertTrue(device_capability["max_work_group_size"] > 0)
self.assertTrue(device_capability["max_num_sub_groups"] > 0)
self.assertEqual(

1
torch/_C/__init__.pyi.in
View File

@ -2342,6 +2342,7 @@ class _XpuDeviceProperties:
name: str
platform_name: str
vendor: str
device_id: _int
driver_version: str
version: str
max_compute_units: _int

7
torch/csrc/xpu/Module.cpp
View File

@ -305,6 +305,7 @@ static void registerXpuDeviceProperties(PyObject* module) {
._(name) \
._(platform_name) \
._(vendor) \
._(device_id) \
._(driver_version) \
._(version) \
._(max_compute_units) \
@ -333,8 +334,10 @@ static void registerXpuDeviceProperties(PyObject* module) {
std::ostringstream stream;
stream << "_XpuDeviceProperties(name='" << prop.name
<< "', platform_name='" << prop.platform_name << "', type='"
<< get_device_type(prop) << "', driver_version='"
<< prop.driver_version << "', total_memory="
<< get_device_type(prop) << "', device_id=0x" << std::hex
<< std::uppercase << prop.device_id << std::dec
<< ", driver_version='" << prop.driver_version
<< "', total_memory="
<< prop.global_mem_size / (1024ull * 1024) << "MB"
<< ", max_compute_units=" << prop.max_compute_units
<< ", gpu_eu_count=" << prop.gpu_eu_count