MLIR Open Design meeting (6/25): Basic Compute Algorithms for the GPU Dialect

[Mehdi AMINI]

Hi all,

Tomorrow (10am PT), Alexander Meißner will talk about "Basic Compute Algorithms" for the GPU Dialect, and depending on the time left we will also discuss the modeling of the types for the LLVM dialect.

 

As usual, the agenda is https://docs.google.com/document/d/1y_9f1AbfgcoVdJh4_aM6-BaSHvrHl8zuA5G4jv_94K8/edit

The information for joining the meeting are usually unchanged from week to week:

https://meet.google.com/aue-vgas-egu

‪+1 218-301-8485‬ PIN: ‪255 745‬#

Cheers,

-- 

Mehdi

[Mehdi AMINI]

If we have time, we can also talk about the RFC on the MLIR Test Case Reducer Tool

-- 

Mehdi

[Peter Balazovic]

Hello Guys,

it seems to be an interesting topic  - is this lecture  "Basic Compute Algorithms"  recorded (archived)? BTW how to get the notification (the invitation into the MS Outlook) for https://meet.google.com/aue-vgas-egu?

Thank you.

Peter

[Jacques Pienaar]

Hey,

All the open design meetings are recorded (https://mlir.llvm.org/talks/) and available afterwards. I'm not sure how best to get the notification in Outlook unfortunately (I believe joining this group should result in a calendar entry appearing in Google Calendar for the recurring meeting).

Best,

Jacques

--
You received this message because you are subscribed to the Google Groups "MLIR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mlir+uns...@tensorflow.org.
To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/mlir/70f2ee38-fcde-49de-8f3a-9e5292e84587o%40tensorflow.org.

[Madhur Amilkanthwar]

Hey guys,

We ran out of time so I couldn't ask these questions. May be dumb questions.

1. How does MLIR plan to implement these ops? Is it like as a part of progressive lowering we "decompose" the ops like "scan" into micro-ops in descendent dialect which eventually would realize the op?

2. How do we target libraries in MLIR? I want to utilize scan implementation from a specialized library, say. How would that work, it we intend to?

To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/mlir/CAM4W%2BYcGvWWZD9H9QM7kVX4U7XVtV73AUOukfgawwtBJ-5fVBQ%40mail.gmail.com.

--

Disclaimer: Views, concerns, thoughts, questions, ideas expressed in this mail are of my own and my employer has no take in it. 

Thank You.
Madhur D. Amilkanthwar

[Mehdi AMINI]

Hi Madhur,

I would follow-up in the thread on Discourse: "Basic Compute Algorithms" for this topic.

Thanks,

-- 

Mehdi

[Madhur Amilkanthwar]

Cool, thanks!