Seems we’re in an irrational Trough of Disillusionment re. GPUs, where folks get wrongly ridiculed as naive for even suggesting using them.
-
-
Replying to @pcwalton
Yes, there are tons of limitations and difficulties. But CPU perf has stalled. The alternative to parallelism is NOT MAKING ANYTHING FASTER.
1 reply 4 retweets 16 likes -
Replying to @pcwalton
Most software has a lot of room to get faster on a single core - all the excess cycles got used up to optimize developer time. (Rightly so!)
1 reply 0 retweets 0 likes -
Replying to @BrandonBloom @pcwalton
sort of. so much of the unused wins in CPU land is data parallelism (AVX, cache lines, ...)... so might as well do GPU
3 replies 3 retweets 6 likes -
The percentage of cycles eaten by electron far outweighs those that can be recovered by data parallelism for typical apps.
2 replies 0 retweets 0 likes -
I'm not trying to say "JS=bad", I've built an electron app teasingly happily, in saying that there is a lot of low hanging fruit
2 replies 0 retweets 0 likes
Oh, I’m talking about algorithms already optimized for single core, including choosing the fastest language, etc.
-
-
Replying to @pcwalton @BrandonBloom
likewise, web is easier to parallelize than weird native frameworks due to structure. either way - i do beyond-native: we <3 server GPUs :)
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.