HackerTrans
TopNewTrendsCommentsPastAskShowJobs

qualifiedeephd

no profile record

comments

qualifiedeephd
·vorig jaar·discuss
Serious researchers use HPC clusters not desktop workstations. Currently the biggest HPC cluster in North America has AMD GPUs. I think it'll be fine.
qualifiedeephd
·4 jaar geleden·discuss
qualifiedeephd
·4 jaar geleden·discuss
qualifiedeephd
·4 jaar geleden·discuss
maybe someone will see this (since the OP is about inlining) and give me good ideas: i would like to do loop carried dependency analysis through functions calls e.g., `foo` has a loop nest that calls `bar` that has a loop nest that modifies state in `foo` that affects foo's loop nest. my naive, off-top-of-my-head, idea was to inline `bar` and perform the analysis.

now i'm sure this goes wrong for real code in a billion different ways, but is there something i'm not considering regarding the principle (i.e. something at the intersection of inlining and dependency analysis)?
qualifiedeephd
·4 jaar geleden·discuss
>Facebook on the surface looks to be the least interesting.

lol. i work on compilers at FB. it's like being a kid in a candy store.

edit: and in comparison to google (not sure about amazon and ms), it's a much more open culture regarding technical reporting.

maybe it's surprising, maybe it's not, but internal communications are done over workplace i.e. FB (so FB for FB). that means i can see all manner of technical writeups from all parts of the company, and i can immediately reach out to people (chat) and quickly learn about new things. just imagine, thousands and thousands of experts in their area, working on high tech, a chat away. and i'm not even just meaning SWE type people - i've chatted at length with some of the machinists at that support prototyping for oculus.

it's not utopian (often times you can't get an answer from someone when you need it) but it's pretty great. when/if i leave for another FAANG i will definitely miss that.
qualifiedeephd
·4 jaar geleden·discuss
>A fully qualified EE understands circuits at a deep mathematical level.

this is so funny. i have a buddy who's finishing up his phd in ee that works on SDR and who lead a team that won the darpa spectrum challenge a few years ago. he has no idea (absolutely no idea) how odes model LRC systems. i'm closer to CS than EE so he was explaining SDR to me and at some point i said "oh like the ode representation <something something>" and he had no clue what i was talking about. sure he took the class where that was covered but he completely ignored it (probably barely passed).

just one data point.