Skip to content

[UR] Add a UR device info enum for checking native bfloat16 conversions #17372

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 12 commits into from
Apr 2, 2025
Merged

[UR] Add a UR device info enum for checking native bfloat16 conversions #17372

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
c7fda15
[UR] Add a UR device info enum for checking native bfloat16 conversions
jinge90 Mar 10, 2025
e6235b3
fix clang format
jinge90 Mar 10, 2025
ab891ad
address ur source check
jinge90 Mar 10, 2025
bd898ef
sync with latest syclos
jinge90 Mar 12, 2025
6fdbef4
Merge remote-tracking branch 'upstream/sycl' into ur_bf16_enum
jinge90 Mar 17, 2025
bb6395c
regenerate enum value for bfloat16_conv
jinge90 Mar 17, 2025
db6d7d3
regenerate ur enum for bf16 convert
jinge90 Mar 17, 2025
6d3a927
rename bf16 conversions intel to native
jinge90 Mar 25, 2025
0c5ca0f
add missing break
jinge90 Mar 25, 2025
985f363
Merge remote-tracking branch 'upstream/sycl' into ur_bf16_enum
jinge90 Mar 26, 2025
6a04a69
Merge remote-tracking branch 'upstream/sycl' into ur_bf16_enum
jinge90 Mar 31, 2025
11762ee
add test for bfloat16 conversions
jinge90 Mar 31, 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
2 changes: 2 additions & 0 deletions unified-runtime/include/ur_api.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

16 changes: 16 additions & 0 deletions unified-runtime/include/ur_print.hpp
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 2 additions & 0 deletions unified-runtime/scripts/core/device.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -449,6 +449,8 @@ etors:
desc: "[$x_bool_t] support the $xProgramSetSpecializationConstants entry point"
- name: USE_NATIVE_ASSERT
desc: "[$x_bool_t] return true if the device has a native assert implementation."
- name: BFLOAT16_CONVERSIONS_INTEL
Copy link
Contributor

@kbenzie kbenzie Mar 17, 2025

Choose a reason for hiding this comment

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

We should remove the _INTEL suffix here since we don't need to differentiate based on vendor for UR queries.

Copy link
Contributor Author

@jinge90 jinge90 Mar 25, 2025

Choose a reason for hiding this comment

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

Hi, @kbenzie
I changed the enum name to "UR_BFLOAT_CONVERSIONS_NATIVE".
Thanks very much.

desc: "[$x_bool_t] support for native bfloat16 conversions"
--- #--------------------------------------------------------------------------
type: function
desc: "Retrieves various information about device"
Expand Down
2 changes: 2 additions & 0 deletions unified-runtime/source/adapters/cuda/device.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -261,6 +261,8 @@ UR_APIEXPORT ur_result_t UR_APICALL urDeviceGetInfo(ur_device_handle_t hDevice,
bool BFloat16 = (Major >= 8) ? true : false;
return ReturnValue(BFloat16);
}
case UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL:
return ReturnValue(false);
case UR_DEVICE_INFO_SUB_GROUP_SIZES_INTEL: {
// NVIDIA devices only support one sub-group size (the warp size)
int WarpSize = 0;
Expand Down
2 changes: 2 additions & 0 deletions unified-runtime/source/adapters/hip/device.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1021,6 +1021,8 @@ UR_APIEXPORT ur_result_t UR_APICALL urDeviceGetInfo(ur_device_handle_t hDevice,
#endif
case UR_DEVICE_INFO_BFLOAT16:
return ReturnValue(true);
case UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL:
return ReturnValue(false);
case UR_DEVICE_INFO_ASYNC_BARRIER:
return ReturnValue(false);
case UR_DEVICE_INFO_IL_VERSION:
Expand Down
8 changes: 8 additions & 0 deletions unified-runtime/source/adapters/level_zero/device.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1038,6 +1038,14 @@ ur_result_t urDeviceGetInfo(
return ze2urResult(errc);
return ReturnValue(UrRootDev);
}
case UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL: {
bool Bfloat16ConversionSupport =
(Device->Platform->zeDriverExtensionMap.count(
ZE_BFLOAT16_CONVERSIONS_EXT_NAME)) ||
((Device->ZeDeviceProperties->deviceId & 0xfff) == 0x201 ||
(Device->ZeDeviceProperties->deviceId & 0xff0) == 0xbd0);
return ReturnValue(Bfloat16ConversionSupport);
}
case UR_DEVICE_INFO_COMMAND_BUFFER_SUPPORT_EXP:
return ReturnValue(true);
case UR_DEVICE_INFO_COMMAND_BUFFER_UPDATE_CAPABILITIES_EXP: {
Expand Down
2 changes: 2 additions & 0 deletions unified-runtime/source/adapters/native_cpu/device.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -346,6 +346,8 @@ UR_APIEXPORT ur_result_t UR_APICALL urDeviceGetInfo(ur_device_handle_t hDevice,
return ReturnValue(bool{1});
case UR_DEVICE_INFO_BFLOAT16:
return ReturnValue(bool{0});
case UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL:
return ReturnValue(bool{0});
case UR_DEVICE_INFO_MEM_CHANNEL_SUPPORT:
return ReturnValue(bool{0});
case UR_DEVICE_INFO_IMAGE_SRGB:
Expand Down
7 changes: 7 additions & 0 deletions unified-runtime/source/adapters/opencl/device.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1516,6 +1516,13 @@ UR_APIEXPORT ur_result_t UR_APICALL urDeviceGetInfo(ur_device_handle_t hDevice,
return ReturnValue(UR_EXP_DEVICE_2D_BLOCK_ARRAY_CAPABILITY_FLAG_LOAD |
UR_EXP_DEVICE_2D_BLOCK_ARRAY_CAPABILITY_FLAG_STORE);
}
case UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL: {
bool Supported = false;
UR_RETURN_ON_FAILURE(cl_adapter::checkDeviceExtensions(
cl_adapter::cast<cl_device_id>(hDevice),
{"cl_intel_bfloat16_conversions"}, Supported));
return ReturnValue(Supported);
}
case UR_DEVICE_INFO_COMMAND_BUFFER_SUPPORT_EXP: {
cl_device_id Dev = cl_adapter::cast<cl_device_id>(hDevice);
size_t ExtSize = 0;
Expand Down
2 changes: 1 addition & 1 deletion unified-runtime/test/conformance/device/urDeviceGetInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1758,7 +1758,7 @@ TEST_P(urDeviceGetInfoTest, SuccessAtomicFenceScopeCapabilities) {
ASSERT_EQ(property_value & UR_MEMORY_SCOPE_CAPABILITY_FLAGS_MASK, 0);
}

TEST_P(urDeviceGetInfoTest, SuccessBFloat64) {
TEST_P(urDeviceGetInfoTest, SuccessBFloat16) {
size_t property_size = 0;
const ur_device_info_t property_name = UR_DEVICE_INFO_BFLOAT16;
Copy link
Contributor

@aarongreig aarongreig Mar 17, 2025

Choose a reason for hiding this comment

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

this is for the old deprecated bfloat16 query, there's a pr open to remove it https://github.com/intel/llvm/pull/17053/files#diff-4ad6c68f3c3e27ec6c124dab1d1c2cd28dc6bba6f4e0960f44eec79b3c0c1c98

for testing if you could just copy this function but rename the test and update property_name that'd be ideal, like this:

TEST_P(urDeviceGetInfoTest, SuccessBFloat16Conversions) {
  size_t property_size = 0;
  const ur_device_info_t property_name = UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL;


Expand Down
6 changes: 6 additions & 0 deletions unified-runtime/test/conformance/testing/source/utils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -629,6 +629,12 @@ ur_result_t GetDeviceBFloat16Support(ur_device_handle_t device, bool &support) {
return GetDeviceInfo<bool>(device, UR_DEVICE_INFO_BFLOAT16, support);
}

ur_result_t GetDeviceBFloat16ConversionsIntelSupport(ur_device_handle_t device,
bool &support) {
return GetDeviceInfo<bool>(device, UR_DEVICE_INFO_BFLOAT16_CONVERSIONS_INTEL,
support);
}

ur_result_t GetDeviceMaxComputeQueueIndices(ur_device_handle_t device,
uint32_t &max_indices) {
return GetDeviceInfo<uint32_t>(
Expand Down
3 changes: 3 additions & 0 deletions unified-runtime/tools/urinfo/urinfo.hpp
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.