[PATCH] D99152: [AMX] Prototype for vector and amx bitcast.

Tue Mar 23 04:37:29 PDT 2021

lebedev.ri added a comment.

In D99152#2644171 <https://reviews.llvm.org/D99152#2644171>, @LuoYuanke wrote:

> In D99152#2644017 <https://reviews.llvm.org/D99152#2644017>, @lebedev.ri wrote:
>
>> load <https://llvm.org/docs/LangRef.html#load-instruction> instruction loads contigious bytes.
>> If that is not what is AMX is trying to use it for, then it is being used incorrectly.
>
> Isn't it a reason that we can't cast "load <256 x i32>*" to "load x86_amx*"? The load memory for <256 x i32> is contiguous, and load memory for x86_amx is not contiguous.

To be honest i don't really understand why `x86_amx` type is even there.
It seems to me that if you just directly used `@llvm.x86.tileloadd64.internal` / `@llvm.x86.tilestored64.internal`,
and `s/x86_amx/<256 x i32>/`, none of these problems would be here.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99152/new/

https://reviews.llvm.org/D99152