liuliu

58 ms·

781.
by liuliu 6y ago
I am not so sure about that. GPU has been slowly moving away from having a fixed graphics pipline at all in the past a few years. See the recent Nanite technology from Unreal 5: http://c0de517e.blogspot.com/2020/05&#x2F
782.
by liuliu 6y ago
There is no path for backing down for China. There is no clear actionable demand from the current U.S. administration. All players are waiting for the November result to pursuit a more strategically-oriented outcome.
783.
by liuliu 6y ago
Thanks. Taking on malloc and being competitive with much less code (mimalloc is around 6k, and it is the smallest I know that is still competitive) would be a great feat. Would be interesting to follow the development.
784.
by liuliu 6y ago
Do you have any extra readings on the rationale of building their own malloc rather than integrating mimalloc or jemalloc?
785.
by liuliu 6y ago
If you stick with 1 or 2 GPUs and common mainstream motherboards, more recent Linux distro (Ubuntu 20.04). You will be fine. The most problem I had is due to the "high-end" motherboard.
786.
by liuliu 6y ago
Updated to 3970x a few days ago. The performance has been good (llvm compiles around 5 minutes!) and it is stable (previous 2920x has soft lockups and have to workaround with idle=halt: https://bugzilla.kernel.org/show_bug.c
787.
by liuliu 7y ago
The second link is death per 100,000 people (not death of confirmed cases), that, for 18-45, is 14.21 per 100,000 (0.014%). There are confirmed cases of 1705.26 per 100,000 for 18-45. Thus, the death for 18-45 over confirmed cases is 0.83%,
788.
by liuliu 7y ago
Seems to be based on iup that supports both Windows and Linux? http://webserver2.tecgraf.puc-rio.br/iup/ The new crop of immediate mode cross platform GUI libraries are more interesting (such as nuklear and Dear Imgui)
789.
Early Thoughts on Reopening Our Family(drz.today)
1 points by liuliu 7y ago|0 comments
790.
by liuliu 7y ago
We don't have good GPUs that is as fast until 580 (or to some extents, the first Titan). Previous generations only about 2 to 5x faster depending on what types of CPU you compare against. IMHO, credit should always go to Alex Krizhevsk
791.
by liuliu 7y ago
What do you think of optimizers such as XLA, TVM or Glow that try to smooth out underlying hardware differences?
792.
by liuliu 7y ago
Dear imgui supports Windows, thus, there is no reason cannot be supported. I indeed will add that while progressing this project. Just at the moment I don't have access to any Windows dev machine.
793.
Write Cross-Platform GUI in Swift Like It Is 1998(liuliu.me)
10 points by liuliu 7y ago|3 comments
794.
by liuliu 7y ago
Thanks, that makes sense. I am not familiar with Zoom the product to know the connectors show up in participant list. I think even if proper e2e channel established, without authentication (Zoom just allows you to join any meeting with a to
795.
by liuliu 7y ago
it is unclear how key exchanges can be handled securely in these cases. The post seems to suggest Zoom cannot insert itself as middleman ("without showing on the participant list"). But that contradicts directly to how these &quot
796.
by liuliu 7y ago
Deficit is funded through debt (treasury notes). Fed buys treasury notes (through QE) and sets the interest rate on these debt. Lower interest rate in short-term reduces deficit (smaller coupon payment), but in longer-term encourages borrow
797.
by liuliu 7y ago
It can go negative, or Fed can do another round of QE, there are plenty of tools. The real question is whether the deficit can grow indefinitely without consequences, and if this is a moral hazard (e.g. no risk priced in borrowing).
798.
by liuliu 7y ago
I've encountered some latency issues with allreduce on transformer models due to vocabulary sizes when communication have to cross PCIe lanes. Increasing batch size helps a lot, but low-latency & high-throughput is universally more
799.
by liuliu 7y ago
Sorry, it is 2-month billing cycle. We have around 30 cents per kwh I think.
800.
by liuliu 7y ago
I've been playing with custom-built 2080 Ti workstation for a while: https://www.youtube.com/watch?v=OF3JYEIsjH8 Several issues: 1. electricity bill is still an issue, I've been paying anywhere between $500 to $10
801.
by liuliu 7y ago
Looks like what it does is similar to what Alex did a few years back with One Weird Trick paper: https://arxiv.org/abs/1404.5997 When attempting to train transformers, I do notice a lot of time spend on allreduce more
802.
by liuliu 7y ago
It jumped to 1900 again in the morning. That is just a reporting pipeline delay I think. Not diverging from "hyperbolic growth" as far as I can tell.
803.
by liuliu 7y ago
WAL mode doesn't really have "read-only" connection (if you try to open a read-only connection by marking as READ_ONLY with sqlite_open_v2, it will error out for WAL mode). A "supposedly" read-only connection can en
804.
by liuliu 7y ago
bcachefs should be heavily supported, it doesn't get nearly enough for what it supposes to do: https://www.patreon.com/bcachefs
805.
by liuliu 7y ago
Assassination: murder by sudden or secret attack often for political reasons https://www.merriam-webster.com/dictionary/assassination
806.
by liuliu 7y ago
I vaguely remember that at the dawn of the deep learning (2013 to 2014), there were talks and hopes that Xeon Phi would smash the performance of Nvidia GPUs. However, the sample people got are too late (I believe it is at the end of 2014) a
807.
by liuliu 7y ago
The rbtree implementation from jemalloc allows you to maintain the value semantics (it is an intrusive design as well). I believe it is also a left-leaning 2-3 red-black tree to be exact. The API is a bit harder to understand but once you g
808.
by liuliu 7y ago
Anyone who does too much C would know that manually implementing associative array (dictionary) is not fun. I started to use two implementations for my work, both worked pretty well for the things they are good at (and all of them are as &q
809.
Where Dispatch Block Keeps the QoS Information on iOS?(liuliu.me)
1 points by liuliu 7y ago|0 comments
810.
by liuliu 7y ago
> Languages like Swift and Objective-C hide the reference counting for you, so 'instruction pollution' is not really something I care about. The overhead at the instruction & memory level is fairly minimal for these langua

More ›