Help Porting HLS Gzip Sample to OneAPI GPU on Tiber Devcloud

[wanyu]

Dear Intel DevCloud Community,

I am currently working on the gzip project from the Intel HLS samples (https://github.com/altera-fpga/hls-samples/tree/main/ReferenceDesigns/gzip ). I would like to run this project on the Tiber OneAPI GPU node instead of an FPGA.

Could you please advise if a GPU-compatible version of this gzip sample exists, or suggest the best practices for adapting it to a GPU target on Tiber?

Additionally, if a CPU version is available for debugging purposes, I’d appreciate any reference or guidance.

Thank you for your support!

Best regards,
Wanyu

[Luqman_Intel]

Hi wanyu,

We are currently looking into this matter and will get back to you shortly. We greatly appreciate your patience.

Regards,

Luqman

[Luqman_Intel]

Hi wanyu,

Thank you for your patience and interest in running the gzip project on Tiber AI Cloud. We understand that you were previously using FPGA, which is a free service. Please be aware that Tiber AI Cloud is a paid service. Would you still be interested in proceeding with this?

Regards,

Luqman

[wanyu]

Hi Luqman,

Thank you for the clarification and for your continued support.

Yes, I’m still interested in proceeding with the gzip project on Tiber AI Cloud. If a GPU version or CPU version is available for debugging purposes, that would be very helpful.

Thank you,

Wanyu

[Luqman_Intel]

Hi wanyu,

Thank you for your patience.

To answer your questions, a CPU-based implementation of gzip is available for debugging and validation purposes.The Altera FPGA gzip reference design primarily uses OpenCL for FPGA acceleration. However, for debugging and validation, a CPU-based gzip implementation can be utilized.

While a GPU-specific version of the gzip implementation isn't directly provided in the Altera reference design, the CPU-based version should suffice for debugging and validation purposes.

Regards,

Luqman