Ari

[Ari @AriWWU]

Our #ACM #TOPLAS paper (https://t.co/35TTnwnc0N) was presented at #PLDI24 (@PLDI) this week -- slides are available here: https://t.co/4HZkcT8Syk

We report speedups of >9x over #NVIDIA #cuBLAS & #cuDNN, #Intel #oneMKL & #oneDNN, and #TVM for #DeepLearning computations.

[Ari @AriWWU]

At @PLDI next week, we will present our #ACM #TOPLAS paper (https://t.co/35TTnwnc0N) about formally based code generation for #DeepLearning, with notable speedups over #NVIDIA #cuBLAS #cuDNN and #Intel #oneMKL #oneDNN.

#DL #CodeGeneration #GPU #CPU

[Ari @AriWWU]

In our new #TOPLAS paper (https://t.co/DvSKNNgrnp), we report speedups for our #MDH compiler -- for real-world #DeepLearning computations -- of up to 3.5x over #NVIDIA #cuDNN and #cuBLAS and speedups up to >9x over #Intel #oneDNN and #oneMKL.

[Ari @AriWWU]

We presented our #MLIR dialect "#MDH" at #EuroLLVM2024 and compared it to #Linalg dialect.

Slides are available here: https://t.co/Pk6L2jSkKn

More information about MDH is available here: https://t.co/Sz3DUiVOVW

[Ari @AriWWU]

Our MDH homepage is now online: https://t.co/5gGj2tPXDZ

Take a look if you are interested in GPU and CPU code generation for data-parallel computations (including #DeepLearning computations).

#NVIDIA #INTEL #GPU #CPU #AI #MLIR #MDH

[Ari @AriWWU]

We presented some of our latest results on code generation and optimization for AI algorithms at C4ML (https://t.co/ggQ0x304HB) this week. If you are interested - slides and posters are available here: https://t.co/5gGj2tPXDZ

#AI #DeepLearning #MLIR #NVIDIA #INTEL #GPU #CPU #MDH

[Ari @AriWWU]

*Auto-Tuning Framework (ATF)* is now open source (MIT license) and also available in #Python. You can find all information here: https://t.co/qVlFdCf9oJ

Please feel free to pass this announcement to other interested colleagues. Let's build a larger user community for ATF.

[Ari @AriWWU]

At #FCRC #PLDI, we presented our #MDH approach for #GPU & #CPU code generation.

If interested:

- slides available here: https://t.co/3gXeUgdxy0

- recording available here:
https://t.co/MJ8d7wXgCB

[Ari @AriWWU]

@MIT_CSAIL Math in UTF-8

[Ari @AriWWU]

If you are interested in GPU/CPU code optimizations and scheduling languages -- please check out our paper presented this week at @cc2020conf: https://t.co/E1mIcoSoL8

[Ari @AriWWU]

Interested in expressing parallelization and data movement optimizations systematically? Check out our WIP results presented at HiPar@SC’22

#DeepLearning #GPU #CPU #TVM #NVIDIA #cuBLAS #Intel #oneMKL #CUDA #OpenCL #OpenMP

[Ari @AriWWU]

Great talk given today by @luiginardi at AutoML about Hyperparameter Optimization !

Slides are available here: https://t.co/CfLHCfiEcg

[Ari @AriWWU]

Invited talk given today by my colleague Richard Schulze at AccML@#HiPEAC22 — Code Generation & Optimization for Deep Learning Computations based on our approach of Multi-Dimensional Homomorphisms (MDH)

Thanks to @jcanore, @ahcohen and other organizers for this great workshop!

[Ari @AriWWU]

A great week of interesting discussions on code generation and optimization is coming to an end. Many thanks to Cosmin Oancea for so warmly hosting our small meeting at @DIKU_Institut.

Anne C. Elster Mary Hall Richard Schulze Denys Shabalin

[Ari @AriWWU]

Our @lorentzcenter workshop last week was as enjoyable as it was productive. Many thanks to all participants!
https://t.co/MdP71YZm6y

[Lorentz Center @lorentzcenter]

Research groups working on auto-tuning and in high-level programming languages and compilers come together this week to work on a common interface to facilitate portability between tuning frameworks. https://t.co/sLoAQZSTB4 @eScienceCenter @ASTRON_NL

[Ari @AriWWU]

@lorentzcenter: Presenting #ATF (Auto-Tuning Framework) at first in-person event after 2 years of pandemic.

[Ari @AriWWU]

Our 3-year funded #DFG project on #CodeGeneration & #Optimization for #DeepLearning computations via our approaches of Multi-Dimensional Homomorphisms (#MDH) and the Auto-Tuning Framework (#ATF) started today. Stay tuned for interesting results :-)

#Google #MLIR #NVIDIA #GPU

[Ari @AriWWU]

We are happy to announce that we have started using our MDH code generation approach for deep learning computations.
Richard Schulze will present our work-in-progress results at #SC21 as research poster.

#DeepLearning #CodeGeneration #GPU #NVIDIA

[Ari @AriWWU]

@thoefler Is a recording available?