RaisingSpear

10 ms·

61.
by RaisingSpear 3y ago
Not a downvoter, but something doesn't need to necessarily be a bottleneck to warrant speeding up. Put another way, why deliberately pick a slower method if a faster one exists? Many applications these days suffer from "death by
62.
by RaisingSpear 3y ago
Or simply: > report issue > issue is automatically closed 6 months later due to lack of activity ...with absolutely no change to or comment on the issue.
63.
by RaisingSpear 3y ago
> unless you are directly mmapping and munmapping (or DONTNEEDing or similar) each allocation You don't free up the memory you're not using? Sounds like you're doing it wrong then. > I bet a really good AVX-512 transco
64.
by RaisingSpear 3y ago
> wasting up to half the memory OSes generally implement overcommit - allocating a bunch of memory doesn't mean it's actually being used.
65.
by RaisingSpear 3y ago
Sorry, I don't quite understand you there - what is the "something" you speak of, and who does "you" refer to?
66.
by RaisingSpear 3y ago
There's nothing inherent with RVV or LMUL which makes this any easier/harder than other ISAs. Intel could design their little cores to break AVX-512 instructions into 4x128b parts if they felt like it, for example.
67.
by RaisingSpear 3y ago
> Your big cores can get wider execution units that can run LMUL=2 instructions natively while your little cores break them up into LMUL=.5 and processes them 4x slower RVV spec doesn't allow LMUL to vary across cores in a hybrid de
68.
by RaisingSpear 3y ago
> The only case I can currently fore see where using LMUL=1 and manually unrolling instead will likely be always beneficial is vrgather operations that don't need to cross between registers in a register group (e.g. byte swapping).
69.
by RaisingSpear 3y ago
I think if Discord thinks your IP is suspicious, they'll force you to enter in a mobile number when registering an account (this is obviously before joining any server). I'm guessing the captcha is the same - your device/IP
70.
by RaisingSpear 3y ago
> but I don't think it's reasonable to do that and then complain about it. "Entry into this establishment requires an exorbitant 20 minute procedure, or you can bypass that by showing nudie pics of yourself." Me: why
71.
by RaisingSpear 3y ago
> the process is exactly the same as it is for me on just about any other site/system when I log in on an untrusted device This is not my experience. Logging into Reddit/Twitter/Hacker News etc is nowhere near that leve
72.
by RaisingSpear 3y ago
Good to know. I refuse to let a bunch of things touch my phone (Discord doesn't exactly have a stellar privacy record), so have never tried, but for those that don't care so much about privacy, it sounds reasonable. But hopefully
73.
by RaisingSpear 3y ago
> Discord has been decent for a while. I suspect enshittification will commence soon Discord was enshittified years ago. It's just that nothing really changed recently, unlike Twitter/Reddit, so it's not making the news.
74.
by RaisingSpear 3y ago
Doesn't seem redundant to me. Even if you can multiplex requests, batches still have certain advantages, e.g. - compression is often more efficient with larger payloads - can reduce per-request overheads, e.g. do authentication once r
75.
by RaisingSpear 3y ago
> for reading any uncached data, the I/O stalls are unavoidable. Excuse me for a silly question, but whilst an I/O stall may be unavoidable, wouldn't a thread stall be avoidable if you're not using mmap? Assuming th
76.
by RaisingSpear 3y ago
AVX-512's VPCOMPRESS* instructions are very useful for binary-radix sorting though.
77.
by RaisingSpear 3y ago
Intel kinda messed up the ISA by requiring AVX-512F across all AVX-512 subsets. If AVX-512VL didn't depend on F, you could have a 256-bit only variant of "AVX-512". (and instead of making a new feature that allows VL without
78.
by RaisingSpear 3y ago
You might find this post interesting: https://bitmath.blogspot.com/2023/05/grevmul.html
79.
by RaisingSpear 3y ago
> I wonder if we'll see a similar instruction for 2 bits -> 2 bits? Could be useful! You don't even need to support all 16 functions, as a number are duplicates or could be implemented with others. The following list covers
80.
by RaisingSpear 3y ago
> So why not just statically link at this point? In many cases, you just can't. Other than glibc having issues with static linking, you can't use `dlopen` in a static binary. For example, if you're building an application
81.
by RaisingSpear 4y ago
...I mean, you could say the same thing about any other version of Windows, so I don't quite get the point you're making.
82.
by RaisingSpear 4y ago
> This totally ignores the concept of "zero-days", vulnerabilities discovered that are only known by bad actors or known but not patched yet. If we're talking zero-days, then XP offers no more or less protection than any o
83.
by RaisingSpear 4y ago
One thing to note is that DEFLATE (underlying algorithm that Gzip uses) doesn't indicate the length of blocks, or make it easy to figure out where they start/end without decoding everything proceeding it. This is likely why pigz c
84.
by RaisingSpear 4y ago
> but does so more efficiently from an algorithmic point of view. He is using a hash table rather than a linear search after all. Most other compression implementations also use a hash table, at least on the faster settings. They may swi
85.
by RaisingSpear 4y ago
> you haven't had to deal with a complex database setup from an operations perspective I've often found it to be the opposite actually. My experience is with RDS MySQL, and on that, RDS heavily restricts what you can do. Want
86.
by RaisingSpear 4y ago
The key difference is that on SVE, the hardware specifies the vector width, whilst with AVX, the programmer specifies the vector width. Of course, the actual hardware EU doesn't have to match either of these. There are various benefi
87.
ARM’s Scalable Vector Extensions: A Critical Look at SVE2 for Integer Workloads(gist.github.com)
2 points by RaisingSpear 5y ago|0 comments
88.
by RaisingSpear 5y ago
I see, thanks. I presume the end-user running the application will need to have up-to-date drivers to take advantage of these changes in the past 4 years?
89.
by RaisingSpear 5y ago
I see, thanks for explaining all that!
90.
by RaisingSpear 5y ago
Thanks! I guess I got confused with the information out there. Most examples seem to be using GLSL shaders for the kernel, but posts seem to indicate it uses SPIR-V as input [ https://community.khronos.org/t/is-a-vulkan

More ›