Add cudaver to docs

[dentalfloss1]

Adds a link to the docs where a table of test results of the CUDA toolkit live. The table has been added to the docs, but of course, we'll also need to update the hosted version of the docs.

[jaycedowell]

Two questions:

• What does "likely working" mean?
• The comments in CUDA 11.8 + bifrost.fir = error #201 imply that CUDA 11.8 doesn't actually work but I also see this listed as likely working in the table. Does the test suite pass?

[dentalfloss1]

Perhaps we can just change "likely working" to "working." The card I was using sometimes puts values just out of spec for the tests, so technically no tests would fully pass. It's weird and confusing. Let's just upgrade all "likely working" to "working."
That bug is interesting. All the tests passed I believe on 11.8. But, if we have an existing issue with 11.8 perhaps we should just say "not working" to be safe.
What do you think?

[jaycedowell]

Yes and yes. I'll update the table.

[jaycedowell]

What do you think about dropping the "CUDA_ERROR_UNSUPPORTED_PTX_VERSION" entries? From what I've read those come from mis-matched kernel driver and toolkit versions (one is too old for the other). It's not really a CUDA version thing or a Bifrost thing but more of an installation thing.

[dentalfloss1]

Sounds good. Maybe we can just add a footnote that errors can occur from mismatched drivers?

…

Sent from my iPhone On Jul 22, 2024, at 1:12 PM, jaycedowell ***@***.***> wrote:  [EXTERNAL] What do you think about dropping the "CUDA_ERROR_UNSUPPORTED_PTX_VERSION" entries? From what I've read those come from mis-matched kernel driver and toolkit versions (one is too old for the other). It's not really a CUDA version thing or a Bifrost thing but more of an installation thing. — Reply to this email directly, view it on GitHub<#244 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AF35E67X2LE7ZOONBVEPRN3ZNVKQ3AVCNFSM6AAAAABLDLR27KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENBTGYZTSMZZGE>. You are receiving this because you authored the thread.Message ID: ***@***.***>