[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ci.rocm: invalid device function



Hi Christian,

On 2025-07-24 02:51, Christian Kastner wrote:
I wonder though whether --for-build should be dropped entirely, and its
functionality made the default:

   # In an unpacked src
   $ rocm-target-arch

   # Somewhere else
   $ rocm-target-arch
   ERROR

   # Somewhere else, with --dist -> "query database"
   $ rocm-target-arch --dist <dist>
   [list]


Thoughts, all?

I must admit that this discussion has revealed to me that I had misunderstood the --for-build option too. I'm not sure how I failed to understand what you were telling me, but I somehow thought it just changed the format of the output. I guess I never really thought about the experimental distribution.

Your proposed alternative seems reasonable to me. Of course, I'm also totally fine if you just want to keep the existing behaviour and rename --for-build to something else. Heck, I'm fine if you decide to make no code changes and focus instead on user education.

I'll just echo that there was confusion here and I support your actions to address it.

On 2025-07-24 00:22, Christian Kastner wrote:
Alternatively, the existing --sep option could infer/do this when use
with some special value, eg: --sep=offload-arch. Though that might
already be too clever?

IMO, that would be too clever.

Sincerely,
Cory Bloor


Reply to: