shaklee3

77 ms·

901.
by shaklee3 7y ago
The compression is a property of the table and done on the fly, transparently to the user. If the difference was compressing/decompressing as part of the user task, I'd agree. But this is something that comes for free by a few ext
902.
by shaklee3 7y ago
That's not really why they're slower. CPUs are significantly more complicated. Things like branch prediction, transactions, sophisticated prefetching all take up a lot of silicon.
903.
by shaklee3 7y ago
It's reading the specs wrong for apples to apples. OP is using the tensor core numbers, which is half precision and only for matrix multiplies. Operations that don't fit that will use the standard fp32/16 performance of the c
904.
by shaklee3 7y ago
DDR has a bandwidth of about 25GBps per channel . You can hit around 100-200GBps on Epyc processors if you're utilizing ram efficiently. GPUs tend to enforce programming models that ensure more sequential accesses, but CPU can do it
905.
by shaklee3 7y ago
Or just use a futex alone, since that mimics its behavior enough.
906.
by shaklee3 7y ago
Thanks! That's a great link for more than just Torvalds.
907.
No nuances, just buggy code (was: related to Spinlock implementation)(realworldtech.com)
540 points by shaklee3 7y ago|274 comments
908.
by shaklee3 7y ago
This is the normal use case for any DPDK software. I think anyone involved in HPC or high-speed networking knows that this is pretty common.
909.
by shaklee3 7y ago
This is also the reason DPDK uses a highly-efficient ring buffer. I believe (old benchmarks) it's faster than LMAX.
910.
by shaklee3 7y ago
With e-ink I was expecting a very long battery life, so only 1-3 days suggests the radio is still a huge battery drain.
911.
by shaklee3 7y ago
Do you know why the oss project is stagnant? It would seem if they wanted it to gain traction they would use that repo as the main one. Someone else mentioned service fabric was for stateful services, but at this point stateful services in
912.
by shaklee3 7y ago
It looks like this project is dead at this point, with the last commit 6 months ago.
913.
by shaklee3 7y ago
Brendan Burns has been leading the k8s/azure stuff for some time at Microsoft. I'm not sure internally how those two are competing, though.
914.
by shaklee3 7y ago
I think SpaceX, while cool now, will be either non-existent, or a shadow of what it is today due to costs going out of control.
915.
by shaklee3 7y ago
Microsoft already has a public kubernetes competitor that (comparatively) nobody uses: https://github.com/microsoft/service-fabric
916.
by shaklee3 7y ago
I have 2fa on my Android and this feature works fine on Google home.
917.
by shaklee3 7y ago
Not sure what you mean by a single bit for a spin lock. It's a cas operation, which doesn't operate on a single bit. Edit: I think I see what you're getting at, but the point from the article about cache invalidation for the
918.
by shaklee3 7y ago
Even with isolcpus, there are some kernel threads that will preempt user threads (kworker/rcuos/softirq) unless you set your priority higher than those with sched_fifo/rr
919.
by shaklee3 7y ago
As much as I like dpdk, XDP and eBPF can potentially hit the same rates as dpdk. But we're currently in a state where more driver support is needed.
920.
by shaklee3 7y ago
If we're talking about dedicating their life, that's likely to be Agner Fog: https://www.agner.org/optimize/ He's put out the most detailed third-party documentation on Intel and AMD processors that I&#x
921.
by shaklee3 7y ago
It seems like this can only be a con if other languages give you a choice. In other words, in C++ you can decide that your debug and release are the same, or you can make them different if you know what you're doing (like a good IDE do
922.
by shaklee3 7y ago
I think that works as long as you have no spouse or kids. If you do, that number will be very hard to be independent.
923.
by shaklee3 7y ago
I don't see how this is worse than most other things that hit the home page. It's interesting, is an in-depth article, is tech-related, and is something most people recognize. And the fact that it's not just cgi is interestin
924.
by shaklee3 7y ago
The link you sent says they're not profitable. They had their best quarter, and it was still a loss.
925.
by shaklee3 7y ago
Ubuntu has been officially supported on nvidia GPUs (cuda, really) for over 5 years. What did you need serviced? https://developer.nvidia.com/cuda-toolkit-60
926.
by shaklee3 7y ago
I remember him from the original Xbox hacking. I believe he was the one that originally figured out how to load custom software on it, and maybe even got sued?
927.
by shaklee3 7y ago
I know this is coming in late, but TMF just did an analysis of the economics: http://tmfassociates.com/blog/2019/12/12/reality-and-hype-in...
928.
by shaklee3 7y ago
Line rate doesn't imply small packets. But most userspace benchmarking uses 64B packets. That being said, the "imix" packet size, which is supposed to represent internet traffic, is around 500B.
929.
by shaklee3 7y ago
Correct me if I'm wrong, but I thought the mass estimates increased? either way, that should only affect how many that can launch at a given time. I didn't think that number had changed appreciably either.
930.
by shaklee3 7y ago
These are not guesses. These are calculations based on FCC filings with things like EIRP, PFD, etc, all included. SpaceX cannot simply transmit hotter than they filed for just because the filing is old. If anything, they're worse tha

More ›