♟ aartbik

Aart J.C. Bik received his PhD from the Leiden University in 1996. From 1998 to 2007, he worked at Intel, where he was the lead compiler architect of automatic vectorization in the Intel C++/Fortran compilers. In 2007, he moved to Google, where he has worked on compilers for ART and the Dart VM. He is now a member of the TensorFlow codegen team.

[mlir] X86Vector: Add AVX Rsqrt

[mlir] Use MCJIT to fix integration tests

aren't the AMX test affected by this as well then?

can you please rebase so we can continue the review in the context of the new dialect name?

[mlir] introduce "encoding" attribute to tensor type

In D99548#2681226, @mehdi_amini wrote:

To me there is nothing about property that makes it discardable, quite the contrary actually,

I am most likely not the right person to LG this, but, wow, thank you for adding this so quickly!!!

In D99548#2680605, @lattner wrote:

I'd recommend going with something active sounding, perhaps layout or encoding?

Uday, can you please verify that you are happy with the doc changes (autogen from td) that describe the format syntax.
Next are the interfaces (happy to include these here, but probably easier to review as next CL)....

In D99548#2675771, @bondhugula wrote:

This revision is missing an update to the tensor type documentation in mlir/doc/LangRef.md.

I don't think we have so many clients that renaming isn't possible, do we? We have done that before with much more heavily used files.
It's not a super big deal, but this way we lose the revision history of all copied files.

first round of comments

In D99548#2672931, @stellaraccident wrote:

figured someone else was more authoritative than me and would take this.

[mlir][sparse] support integral types i32,i16,i8 for *numerical* values

any early feedback on this? can I go ahead polishing this CL for review?

Very pleased to see this happen! This will also remove some artificial bail out in sparse compiler vectorization due to "index" as base type!

[mlir][sparse] move new test to right directory structure

[mlir][sparse] support for very narrow index and pointer types

pass python bindings tests

seems you were changing this in parallel, abandoning this one

please update integration tests too!

[mlir][cpu-runner] register all llvm ir dialects

[mlir][amx] regression test for tile-muli (all zero/sign-extension combinations)

fixed layout of parameters

[mlir][amx] reformatted examples

[mlir][amx] blocked tilezero integration test

[mlir][amx] Add Intel AMX dialect (architectural-specific vector dialect)

removed commented out code

better error messages, more doc on ops, new asserts, stride checks

merged LLVM IR AMX dialect with AMX dialect (other comments still to be addressed....)

[mlir][sparse] disable nonunit stride dense vectorization

In D98470#2622418, @dcaballe wrote:

Thanks for working on this, Aart! I think the progressive lowering approach that you are taking here is very on point! I’m not working with AMX but it would be great if you could add me as a reviewer to the related code reviews. It would be very educational for me since this approach is also applicable to similar internal problems we have.

This is a large drop, so I added some background on discourse:

aartbik (Aart Bik)
PhD

Projects

User Details

Recent Activity
View All

Today

Yesterday

Fri, Apr 9

Thu, Apr 8

Wed, Apr 7

Tue, Apr 6

Mon, Apr 5

Fri, Apr 2

Thu, Apr 1

Tue, Mar 30

Mon, Mar 29

Fri, Mar 19

Wed, Mar 17

Tue, Mar 16

Mon, Mar 15

Mar 12 2021

Mar 11 2021

Mar 8 2021

aartbik (Aart Bik)PhD