Closed glennpj closed 1 year ago
We'd like to but currently don't have anyone on it. If you want to do it, we'd really appreciate the contribution!
One thing we've done for other compilers is implement proxies for x86_64_v2, x86_64_v3, x86_64_v4, etc. for older versions that do not support the names directly. If that's possible with oneapi and nvhpc that would be really helpful as well.
I expect the oneapi entries to be very similar to the intel entries. Looking at the man page, some of the options for features, such as -mfma
, -mbmi
, -mavx2
, and -mlzcnt
can only be turned on via -march=core-avx2
. That probably defeats the purpose of the x86_64_v[2-4]
profiles, which I guess is why there are no entries for intel for those.
@samcmill Would you, or anybody at NVIDIA, be interested in tackling this issue (for the nvhpc part, of course :smirk:)?
I guess I initiated this and then let it fall off of my radar. I will start working on the oneapi compilers.
@cparrott73 is looking at the nvhpc compilers.
I have changes queued up to support the nvhpc compilers, but I am just waiting for management approval before I can release them. Hopefully very soon.
I submitted #47 to cover the OneAPI part of this.
I have received management approval to contribute changes to support the nvhpc compilers to this project. What's the best way to contribute this? My changes consist of fairly straightforward modifications to the microarchitectures.json file. I can produce a diff/patch if that would be preferable.
@cparrott73 A pull request to the master
branch of this repository is easiest, since then the changes can be discussed in context easily if needed.
@cparrott73 A pull request to the
master
branch of this repository is easiest, since then the changes can be discussed in context easily if needed.
Okay. This was a bit of a learning curve for me, as I had never done this before. I believe I just successfully sent the pull request with my changes. Please take a look and let me know if you need anything else from me. Thanks!
This issue can be closed now that #67 is merged.
Are there any plans to add the Intel OneAPI and Nvidia NVHPC compilers to archspec-json? If this is not in the works I can take a stab at it.