Goals for this week, as written, are
- Go hard on Arxiv (what does it mean?)
- Make a final push for LATTE
- Finish Python notebook (notebook is private) (see output here) for the RTML demo. Namely, (get generated C code working)
I just wrote this bit for a Google application; I thought I should log it here:
My focus is in making the accelerator design process accessible. I believe that hardware—especially accelerators—are hardware manifestations of the software workloads that run on them, with a heavy serving of institutional architectural knowledge (e.g. how a computation should be optimally pipelined; how to design the network; how to set up the memory hierarchy). Copying the process of software compilation, I believe it should be possible to progressively lower a workload or set of workloads to an accelerator design, using lowering steps which encode institutional architectural knowledge and information about the target hardware platform.
Monday pomodoros, writing LATTE: XXXXX
Tuesday pomodoros, writing LATTE: XXXX
Wednesday pomodoros, writing LATTE: XXXX
Thursday I also worked the whole day on Arxiv.
Reflection on Goals
I certainly put a lot of time into Arxiv. That’s going fine!
I made a final push for LATTE, but didn’t make it. I need to turn those thoughts into a blog post.
I also finished up the codegen for the notebook.