Remove obsolete max/max_int32 LLKs.

[jasondavies]

AFAIK, these are not used anywhere other than the obsolete tt-budabackend; max/min is supported via binary max min in tt-metal.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update

Checklist

• All post commit CI passes
• Blackhole Post commit CI passes (if applicable)

[fvranicTT]

Please check #1008.
There's an idea to revamp max. FYI @ldjurovicTT

[jasondavies]

Please check #1008. There's an idea to revamp max. FYI @ldjurovicTT

Looks like there's also a "unary" max too in tt-metal, but with two operands. All a bit confusing. I suspect the binary min/max are the more recent versions and these others are legacy.

[jasondavies]

Further investigation:

• confirmed that the LLKs removed in this PR are not used by tt-metal at the moment
• there is a max_tile in tt-metal, which uses a similar calculate_max in tt-metal, but I believe this is obsolete and should be removed in favour of binary_max_tile: max_tile is obsolete, along with calculate_{min,max} tt-metal#35502
• the unary min/max APIs in tt-metal refer to computing min/max(tensor, scalar)

[ldjurovicTT]

@jasondavies I am modifying ckernel_sfpu_max.h and testing max rewritten to TTI_ in LLK test infra. Also I have separated init and calculate functions, and I plan to add some SDPA specific functions to this file. So if we decide to remove the file please lmk so I can move my work to metal. @fvranicTT is this ok plan?

[jasondavies]

@ldjurovicTT FYI, I made a proposal to remove max_tile and associated LLKs here: tenstorrent/tt-metal#35502 - binary_max_tile is more flexible. Hopefully this doesn't conflict with your work. I have also proposed a revamp of max/min to @cmaryanTT (including fixing bugs with int32 handling) but I'm still waiting to hear back.