Can vtune profiling compute shader of dx11

[chris_pn]

Hi,

    Can vtune profiling hlsl code of dx11? For example, I want to find out which line of code cost most of time, is it possible? (Or I should use intel gpa for hlsl profiling?) Thanks!

[Egor_S_Intel1]

Hi! Now you should use Intel GPA for it. You can find additional information in this article: Desktop API: Optimizing Shaders

[chris_pn]

Hi Egor,

    Thanks for your reply. I also notice one metrics in vtune "Typed Memory Read Bandwidth":

https://software.intel.com/content/www/us/en/develop/documentation/vtune-help/top/reference/gpu-metrics-reference/typed-memory-read-bandwidth-gb-sec.html

http://agroda.gzu.edu.cn:9999/lib/intel/vtune_amplifier_xe_2015.3.0.403110/documentation/en/help/GUID-B4F6C59E-CC1F-11E2-9387-E02A82CB6E13.htm

    If vtune can't profiling directx hlsl, what is this metrics used for (applicable to which type of opencl memory)?

    Thanks!

[AthiraM_Intel]

Hi,

Thanks for reaching out to us.

We are working on it internally. We will get back to you soon.

Thanks

[Johny_P_Intel · Accepted solution]

Hi Chris,

The "Typed Memory Read Bandwidth" is a metric that appears to be strictly for graphics rendering, not OpenCL offload.

As Egor mentioned above, Intel GPA is the appropriate tool for profiling your application.

Regards,

Johny.

[chris_pn]

Got it. Thanks!