Archives Discussions

[ryta1203]

Is there a way to get a 1D array size larger than 8192 so that I can use scatter with an array greater than 8192?

[MicahVillmow]

ryta, the scatter sample in the SDK allows sizes larger than 8192.

[ryta1203]

Yes, it appears this way; however, is there a reason I keep getting Brook.dll runtime errors (symbols not loaded)? Can you think of a reason this would be happening? Much appreciated.

It will stop on either the streamReads or the kernel call. Kinda lost here, couldn't find anything in documentation about this and my scatter seems to be ok (follows the scatter SDK example).

[ryta1203]

I also get incorrect values whenever I send an "int" as a parameter and then use that as an array index.

[MicahVillmow]

ryta, can you send me the generated IL so I can examine it to see what is going on in a PM?

[ryta1203]

Originally posted by: MicahVillmow

ryta, can you send me the generated IL so I can examine it to see what is going on in a PM?

Did you get my PMs?

[MicahVillmow]

Some of the code got cut off, so I couldn't see the export section. However the code that I did look at seemed fine to me, but can't say for certain until I manage to see the end section of the IL code.

[ryta1203]

Micah,

I emailed you the code, both IL and Brook level kernel code along with the wrapper functions after I got your PM about the code getting cut off, sorry about that. They are attachments in the email so it shouldn't be a problem.

[ryta1203]

Micah,

My apologies, it seems as though there was a streamRead(ss, s) that was taking s[] of size 9 and putting it into ss<> of size gx*gy (where gx=gy and are the first two size numbers I have mentioned above, aka 32*32 or 64*64).

That said, I still think it's odd that I didn't have any problems doing a streamRead(32*32, 9) sizes but I do (64*64, 9) sizes? Do you know why this might be?