Mfma matrix core timing

[mjkpolo]

A lookup table is used with values taken from the mi200 and mi300 ISA
specifying how long a SIMD's matrix core unit will be consumed by MFMAs.

This time can be adjusted for all MFMAs using the --mfma-scale parameter.

[mattsinc]

Needs to be merged in after #1693

[mattsinc]

Since Vishnu already integrated Nagendra's patches in #1693, please remove those from this patch before we review.

[mattsinc]

Few small things, but overall nice job!

[abmerop]

This is mostly good, thanks! I only have one additional comment on top of Matt's comments.

[erin-le]

@mjkpolo Hello, thank you for your contribution! Could you give an update on the status of this PR? Also, if you are able to make the changes requested within the next day or so/by the end of this week, we may be able to get this PR into the staging branch.

[mattsinc]

@mjkpolo Hello, thank you for your contribution! Could you give an update on the status of this PR? Also, if you are able to make the changes requested within the next day or so/by the end of this week, we may be able to get this PR into the staging branch.

@mjkpolo is working towards an ISCA submission, so my guess is he won't be able to address these comments in the timeline you provided.

[abmerop]

Should I just add a commit with the changes? It seems like we just want comments and an assert?

[mattsinc]

Should I just add a commit with the changes? It seems like we just want comments and an assert?

Thanks @abmerop ! I also fixed the print info.

[mattsinc]

Should I just add a commit with the changes? It seems like we just want comments and an assert?

Thanks @abmerop ! I also fixed the print info.

@abmerop since I pushed the print fix it won't let me approve ... can you?