Iris Xe Graphics can't compute double-precision floating-point !?

[shane100]

When I use OpenCL call Iris Xe Graphics compute double-precision floating-point, OpenCL say"error: use of type 'double' requires cl_khr_fp64 extension to be enabled". but I use  NVIDIA GeForce MX350 compute not have any problem. therefore I inspection this device's extensions seem not have  cl_khr_fp64. So I can't use  Iris Xe Graphics compute double-precision floating-point!? 

[Alberto_Sykes]

shane100, Thank you for posting in the Intel® Communities Support.

In order for us to provide the most accurate assistance on this scenario, we just wanted to confirm a few details about your system:

What is the model of the motherboard?

If this is a laptop, what is the model of it?

Is this a new computer?

Was it working fine before?

When did the issue start?

Did you make any recent hardware/software changes?

Which Windows* version are you using?

Does the problem happen at home or in the work environment?

Please attach the SSU report so we can verify further details about the components in your platform, please check all the options in the report including the one that says "3rd party software logs":

https://downloadcenter.intel.com/download/25293/Intel-System-Support-Utility-for-Windows-?product=91600

Any questions, please let me know.

Regards,

Albert R.

Intel Customer Support Technician

[shane100]

Hi Alberto_R_Intel1,this my SSU.

This a new laptop.Used not over 6 month.This question is start in a few days ago,when I try to coding OpenCL.

Other responses are in the SSU report.

Thank you.

By the way,This is my Intel(R) Iris(R) Xe Graphics's extensions:

cl_khr_byte_addressable_store cl_khr_fp16 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_icd cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_intel_subgroups cl_intel_required_subgroup_size cl_intel_subgroups_short cl_khr_spir cl_intel_accelerator cl_intel_driver_diagnostics cl_khr_priority_hints cl_khr_throttle_hints cl_khr_create_command_queue cl_intel_subgroups_char cl_intel_subgroups_long cl_khr_il_program cl_intel_spirv_media_block_io cl_intel_spirv_subgroups cl_khr_spirv_no_integer_wrap_decoration cl_intel_unified_shared_memory_preview cl_khr_mipmap_image cl_khr_mipmap_image_writes cl_intel_planar_yuv cl_intel_packed_yuv cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_image2d_from_buffer cl_khr_depth_images cl_intel_media_block_io cl_khr_3d_image_writes cl_khr_gl_sharing cl_khr_gl_depth_images cl_khr_gl_event cl_khr_gl_msaa_sharing cl_intel_dx9_media_sharing cl_khr_dx9_media_sharing cl_khr_d3d10_sharing cl_khr_d3d11_sharing cl_intel_d3d11_nv12_media_sharing cl_intel_unified_sharing cl_intel_subgroup_local_block_io cl_intel_simultaneous_sharing

It look like support half-precision floating-point only ......

[Alberto_Sykes]

Hi shane100, You are very welcome, thank you very much for providing that information and the SSU report.

Based on the information showing in the SSU report, we can confirm that the current graphics driver installed on your system is version 27.20.100.8439, which is the driver tested and validated by Acer, that is actually the driver that we always recommend to install since that driver was customized by them to work with your specific platform.

So, we will do further research on this matter, as soon as I get any updates I will post all the details on this thread.

Regards,

Albert R.

Intel Customer Support Technician

[Alberto_Sykes]

Hello shane100, I just received an update on this matter.

I just moved your thread to the proper department so they can further assist you with this topic, they will provide additional assistance as soon as possible.

Regards,

Albert R.

Intel Customer Support Technician

[shane100]

I use my old laptop test the program(Intel® HD Graphics 620).It can compute.

[ArpitaP_Intel]

Hi,

Thanks for reaching out to us.

Please try setting up the below flag in your environment:

export OverrideDefaultFP64Settings=1 && export IGC_EnableDPEmulation=1

Let us know if the issue still persists.

Regards,

Arpita

[shane100]

Hi ArpitaP_Intel

I run this program in Windows , IDE is Visual Studio 2019.

I don't know how to set this.

[Ben_A_Intel]

Hello,

By default, the GPU OpenCL driver only exposes the double-precision floating point extension on devices where double precision is supported natively.  Because our latest client GPUs ("Ice Lake" and "Tiger Lake") do not support double precision natively, they do not advertise the extension by default.

If desired, for development purposes, we do support double-precision floating point emulation.  For more information about double-precision emulation, please see:

https://github.com/intel/compute-runtime/blob/master/opencl/doc/FAQ.md#feature-double-precision-emulation-fp64

[shane100]

Hello,Ben_A_Intel

So I  can not use double-precision floating point with Windows? 

The link say "FP64 emulation can only be enabled on Linux.".

[Ben_A_Intel]

So I  can not use double-precision floating point with Windows? 

---

I don't believe the debug variable mechanism used to enable this feature on Linux is enabled in Windows builds.  I'm checking to be sure, and also if there is an alternative mechanism that can enable the feature.

[shane100]

Hello,

Have any way can emulation fp64 in windows?

[ArpitaP_Intel]

Hi Shane100,

Please help us with the versions of your Visual studio, oneAPI and compiler.

We are checking internally, we will get back to you.

Regards,

Arpita

[shane100]

Hello ArpitaP_Intel,

My Visual Studio 2019's versions is 16.8.6 .

I don't know what is oneAPI, is this "intel_sdk_for_opencl_applications_2020.3.494"?

Thanks.

[Juan_G_1]

Hello There,

Only for curiosity, How many work-items does your Intel GPU show in each dimension?

[shane100]

Hi Juan_G_1,

This is not determine.

I use GPU to compute my homemade neural networks.

My currently set is:[ work_dim always 1 , global_work_size is the currently layer quantity ].

Originally like this,but iris do not support fp64.

So I currently is use Nvidia MX350 compute.

I initial is plan to 2 GPU compute this ,but ......

Intel always let me disappointed......

[Dunni_A_Intel · Accepted solution]

Hi Shane,

We have verified that double-precision floating point emulation is not available on Windows. 

[shane100]

OK,thanks