Skip to content

[SYCL][DOC] Proposal to add number of HW threads per EU query #4876

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 1 commit into from
Nov 12, 2021
Merged

[SYCL][DOC] Proposal to add number of HW threads per EU query #4876

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
39 changes: 33 additions & 6 deletions sycl/doc/extensions/IntelGPU/IntelGPUDeviceInfo.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@ The Feature Test Macro SYCL\_EXT\_INTEL\_DEVICE\_INFO will be defined as one of
| ----- | ----------- |
| 1 | Initial extension version\. Base features are supported |
| 2 | Device UUID is supported |
| 3 | HW threads per EU device query is supported |


# Device UUID #
Expand Down Expand Up @@ -177,8 +178,6 @@ Then the number of EUs can be obtained using the standard get\_info() interface.

A new device descriptor will be added which will provide the number of slices on an Intel GPU. If the device is a subdevice, then the number of slices in the subdevice is returned.

This new device descriptor is only available for devices in the Level Zero platform, and the matching aspect is only true for those devices. The DPC++ default behavior is to expose GPU devices through the Level Zero platform.
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For reference: OpenCL extension for this: https://github.com/KhronosGroup/OpenCL-Docs/blob/master/extensions/cl_intel_device_attribute_query.asciidoc

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll update the implementation if we agree to remove this disclaimer.



## Version ##

Expand Down Expand Up @@ -214,8 +213,6 @@ Then the number of slices can be obtained using the standard get\_info() interfa

A new device descriptor will be added which will provide the number of subslices per slice on an Intel GPU. If the device is a subdevice, then the number of subslices per slice in the subdevice is returned.

This new device descriptor is only available for devices in the Level Zero platform, and the matching aspect is only true for those devices. The DPC++ default behavior is to expose GPU devices through the Level Zero platform.


## Version ##

Expand Down Expand Up @@ -251,8 +248,6 @@ Then the number of subslices per slice can be obtained using the standard get\_i

A new device descriptor will be added which will provide the number of EUs per subslice on an Intel GPU. If the device is a subdevice, then the number of EUs per subslice in the subdevice is returned.

This new device descriptor is only available for devices in the Level Zero platform, and the matching aspect is only true for those devices. The DPC++ default behavior is to expose GPU devices through the Level Zero platform.


## Version ##

Expand Down Expand Up @@ -283,6 +278,38 @@ Then the number of EUs per subslice can be obtained using the standard get\_info
auto euCount = dev.get_info<info::device::ext_intel_gpu_eu_count_per_subslice>();
}

# Intel GPU Number of hardware threads per EU #

A new device descriptor will be added which will provide the number of hardware threads per EU on an Intel GPU. If the device is a subdevice, then the number of hardware threads per EU in the subdevice is returned.


## Version ##

The extension supports this query in version 3 and later.

## Device Information Descriptors ##

| Device Descriptors | Return Type | Description |
| ------------------ | ----------- | ----------- |
| info\:\:device\:\:ext\_intel\_gpu\_hw\_threads\_per\_eu | uint32\_t| Returns the number of hardware threads in EU.|


## Aspects ##

A new aspect, ext\_intel\_gpu\_hw\_threads\_per\_eu, will be added.


## Error Condition ##

An invalid object runtime error will be thrown if the device does not support aspect\:\:ext\_intel\_gpu\_hw\_threads\_per\_eu.

## Example Usage ##

Then the number of hardware threads per EU can be obtained using the standard get\_info() interface.

if (dev.has(aspect::ext_intel_gpu_hw_threads_per_eu)) {
auto threadsCount = dev.get_info<info::device::ext_intel_gpu_hw_threads_per_eu>();
}

# Maximum Memory Bandwidth #

Expand Down