Draft slide deck for AFDS 2012
May 23, 2012
Posted by on
I’ve been working on my presentation for AMD Fusion Developer Summit 2012. Here’s the draft slide deck: http://golem5.org/bucket/afds2012/PL-4158_AFTER_REVIEW_DRAFT_052312.pdf . I’m in with the Programming Languages and Models track.
The AFDS 2012 session catalog looks interesting. There’s a lot of stuff on other tracks. But I know from experience that conference organizers do a good job of sorting speakers. If you hang around with your assigned track, you are likely to learn the most.
My experience of presenting is that I learn what I’m doing when trying to explain it to an audience. I’ve heard teachers say this, that they learn as much as students do. The narrative of this talk for AFDS 2012 turned out differently than I expected. It feels much more like product marketing. The audience isn’t compiler engineers so much as developers who need to solve problems with GPGPU.
Also, I didn’t understand how close a workalike clone of PeakStream this is until I saw the code samples in the slide deck. Chai doesn’t have everything PeakStream did. However, PeakStream didn’t have everything Chai does now.
The obvious work is in the JIT middle-end. I cheated and skipped a lot of safety logic so the language platform could be filled out first. Now that this is done, I have to go back and rework the JIT.
My big question, one I hope to answer from attending AFDS 2012, is if anyone cares. The world is changing. HPC and GPGPU… is this still relevant for an economy built around smart phones connected to social networks in the cloud?