Reading data from FPGA on HPS with Linux - what's the best way to do it?

[jackhab]

I have a Cyclone V with Linux application on ARM.

The application accesses FPGA directly by mapping HW light- and heavy-weight bridges into virtual memory space (without kernel driver).

Currently, I have a working application which does all access via alt_read_word() and alt_write_word().

I recall from NIOS documentation that this is the proper way to access HW memory to avoid caching issues but I couldn't find anything about it with regards to ARM. So I'd like to ask the following:

1. Should FPGA memory be accessed by HPS only via SoCAL Memory Read/Write Utilities or it can be used as normal HPS RAM including calling memcopy() to move data between FPGA and HPS?
2. Is there a document for SW developers which describes how HPS should access FPGA memory space?

Thanks.

[Jeet14]

Hi,

If you use the HPS and FPFA on chip RAM, then this can be accessed using the mmap() function in linux. 

You can refer the Cyclone V GSRD on below link. https://www.rocketboards.org/foswiki/Documentation/CycloneVSoCGSRD

Cyclone V, FPGA-to-HPS Bridges Design Example

https://www.intel.com/content/www/us/en/support/programmable/support-resources/design-examples/horizontal/fpga-to-hps-bridges-design-example.html

Regards

Tiwari

[jackhab]

Hi Tiwari

I, indeed, use mmap() to gain access to FPGA memory space. But my question was about what happens during actual read and write.

Should I read the memory word-by-word using alt_read_word() function or reading big memory chunks using standard memcpy() also OK?

I am asking it because I'm not sure how ARM's data caching will behave during mempcpy() and I'm afraid caching can cause data corruption.

[Jeet14]

Hi,

Cross-compiler vendors generally include a precompiled set of standard class libraries, including a basic implementation of memcpy() . Unfortunately, since this same code must run on hardware with a variety of processors and memory architectures, it can't be optimized for any specific architecture. 

A variety of hardware and software factors might affect your decision about a memcpy() algorithm. These include the speed of your processor, the width of your memory bus, the availability and features of a data cache, and the size and alignment of the memory transfers your application will make.

An intimate knowledge of your target hardware and memory-transfer needs can help you write a much more efficient implementation of custom memcpy() functions.

You can also refer below link on memcpy and custom memcpy functions-

https://www.intel.com/content/www/us/en/developer/articles/technical/performance-optimization-of-memcpy-in-dpdk.html

I suggest you to use the data reading using driver APIs i.e. alt_read_word();

Regards

Jeetesh

[Jeet14]

Hi,

If you have any other question on the matter, please let me know.

Regards

Jeetesh