Parallella Community

Nick,

Internally, it's performing DWORD moves (for performance). So if your data isn't a multiple of 8 bytes or 8-byte aligned, you may have problems. Obviously, you found a bug that should be fixed. A workaround is to pad your data and round up the size of the transfer.

A new resources and download page is up:
http://www.browndeertechnology.com/reso ... prthr2.htm

There is now real documentation. And the first official Beta (version 20160701) is available now from the link above.

I haven't used it yet...

I got it going last night. It's looking good and having some documentation is great.

From what I have tested so far it seems to be on par with the last pre-release. I'm not sure what code improvements have been made given that the release notes are more of an installation guide.

I sorted the problem I had with sending data to and getting it from the epiphany. My mistake. I didn't call coprthr_memptr().

Browndeer has added some more examples of coprthr-2 programs.

Have a look here:

nick,

From my understanding, you shouldn't compile it like that. In the past, everything was compiled with clcc, but COPRTHR-2 is trying to improve interaction with the existing tools (like e-gcc).

Try compiling timer.c with e-gcc and linking the resulting binary (timer.o) with your passCL.c, compiled with coprcc.

Make sense?

jar,

I'm producing a .o file with e-gcc but I'm getting problems when I try and link it to the .e32 file.

Given that this would be of marginal value I think I'll leave it.

Again, thanks for your help.

nick