|
Cybernetic Vermin posted:decode to run on a suitably narrow unit (afaik what amd does even for avx-256, and almost certainly for forthcoming -512) zen 1 had two 128 bit adders and two 128 bit multipliers and split larger instructions zen 2 widened them all to 256 unknown how avx-512 support will work
|
#
?
Aug 23, 2021 03:17
|
|
|
|
| # ? Jun 8, 2024 17:05 |
|
|
Nomnom Cookie posted:there are so many people just panting for intel to be competitive in hpc/gpu that every time intel puts out a statement like "our next batch of igp will definitely be good" it just gets reported everywhere amd everyone is like oh yeah intel put those hot fragments in my raster hole and then it launches and won't even run without a hacked bios on a particular mobo lmao yeah could have sworn ive heard the whole "this next intel gpu will be competitive" for about a decade now. they've gotten a lot better but still nowhere near nvidia/amd perf.
|
|
#
?
Aug 23, 2021 04:26
|
|
|
did any intel c-levels get fired for missing the gpu train? by all measures they were ready for it. they just snoozed.
|
|
#
?
Aug 23, 2021 09:00
|
|
|
Workaday Wizard posted:did any intel c-levels get fired for missing the gpu train? by all measures they were ready for it. they just snoozed. if i remember correctly when the gpu train was taking off no one weren't buying intel so i imagine they got big fat bonuses
|
|
#
?
Aug 23, 2021 09:06
|
|
|
Sagacity posted:hang on, i was doing software rendering in the late 90s, right before 3d cards came along and rendered it obsolete. mmx was really needs suiting for that Yeah world+dog was adding similar stuff at the time, remember the 90s was the time of ~multimedia~ so the idea was also for things like playing videos. AMD had 3dnow! Even, like, PA-RISC added MAX. Most of them weren't dumb enough to alias the FPU registers though.
|
|
#
?
Aug 23, 2021 16:04
|
|
|
yeah everyone started adding vector registers after PowerPC made it a thing
|
|
#
?
Aug 24, 2021 08:40
|
|
|
Amazing how PowerPPC was able to do that when it introduced AltiVec after MMX, 3DNow!, and SSE.
|
|
#
?
Aug 24, 2021 15:35
|
|
|
it's just so powerful that linear time doesn't apply to it
|
|
#
?
Aug 24, 2021 19:30
|
|
|
.
Joe Chip fucked around with this message at 20:50 on Aug 24, 2021 |
|
#
?
Aug 24, 2021 20:21
|
|
|
oh right, it was SSE that came out after and was similar to AltiVec/VMX, what with separate vector registers, floating point, etc. of course all that stuff originally was done by Cray… and “array processors” were popular peripherals for mainframes, minicomputers, and workstations starting in the 1970s a friend has a file cabinet sized array processor peripheral from like 1988 or so that uses SCSI so you can attach it to any workstation of the era, and transfer data to and from it at a whopping 5-20 megabytes per second; it renders a Mandelbrot set pretty quickly Symbolics made an array processor board set for their 3600 series of lisp machines that had a bunch of Weitek FPUs on it, evidently about the only things that used it were the neural network product, MACSYMA (which is what Stephen Wolfram shamelessly paid other people to knock off for Mathematica, and then took credit for), and S-Graphics (which eventually became Mirai) at least in their case you could write fairly generic vector code and run it without the array proc too
|
|
#
?
Aug 25, 2021 05:01
|
|
|
SSE predates AltiVec though.
|
|
#
?
Aug 25, 2021 15:21
|
|
|
only by 6 months though, which was within reach of the time travel technology of the time
|
|
#
?
Aug 25, 2021 16:35
|
|
|
eschaton posted:(which is what Stephen Wolfram shamelessly ... took credit for), wouldn't expect anything else
|
|
#
?
Aug 25, 2021 16:42
|
|
|
they were obviously designed in parallel, it’s not like somebody slapped their head and said “vectors, why didn’t we think of that” and then added a vector unit to their upcoming chip design in a couple months
|
|
#
?
Aug 25, 2021 17:28
|
|
|
rjmccall posted:they were obviously designed in parallel
|
|
#
?
Aug 25, 2021 17:43
|
|
|
lol
|
|
#
?
Aug 25, 2021 19:46
|
|
|
lol
|
|
#
?
Aug 26, 2021 19:28
|
|
|
is the discord broke for anybody else anyway lol: quote:someone at Microsoft who wields great power has made the decision that features such as hot reload cannot be given away for free as part of the open source .NET SDK anymore. These features must be reserved to proprietary commercial products such as Visual Studio. In fact, there is a bigger internal strategy being formed at Microsoft to make Visual Studio the main IDE for .NET again, because some people at Microsoft are annoyed that Visual Studio Code and other third party tools have been undermining Visual Studio for way too long now. As a result the hot reload feature was ripped out of the SDK in a last minute effort, breaking all promises and conventions of Microsoft’s release candidate policies and announcing that hot reload will be a Visual Studio only feature going forward.
|
|
#
?
Oct 23, 2021 15:00
|
|
|
do we know that is what is happening? some microsoft person claimed they just cut the feature from 6.0 to get it out the door. it seems kind of fantastic that hot code reload would be the distinguishing commercial feature, and that the way they'd make this strategic move is cutting it in release candidate 2
|
|
#
?
Oct 23, 2021 15:21
|
|
|
we don't, it's that person's speculation. however imo it's worth making a note of the prediction at the very least for future reference
|
|
#
?
Oct 23, 2021 15:24
|
|
|
It's complete speculation but the decision appears to have been made higher up in what is apparently the only issue that isn't publicly viewable github repo with no public explanation so if the reason WASN'T intentionally crippling the open source tooling they have done an amazingly bad job communicating it and lmao at them for causing the entire .net community to start freaking out for absolutely no reason. Perhaps people also wouldn't have jumped to conclusions if it weren't for the other kerfuffle recently and I suspect that the fact that this blew up indicates that people are already primed to worry that Microsoft is about due to gently caress up .net somehow. Ideally it will result in Microsoft apologizing and explaining that it was just because they decided it wasn't ready or something (although people were already using it because microsoft doesn't usually pull features when they have been in release candidates that are described as being basically ready for production so that would be dumb anyway). This does highlight the issue with .net, though, which is that even though it's open source now and theoretically someone else could take it over, the development and ecosystem are still completely centered around Microsoft and decisions Microsoft makes can still easily kill it overnight. mystes fucked around with this message at 16:15 on Oct 23, 2021 |
|
#
?
Oct 23, 2021 16:08
|
|
|
my first thought when I heard they were pulling it so late was that they’d run into a patent somewhere and hadn’t been able to settle it in time
|
|
#
?
Oct 23, 2021 16:16
|
|
|
I mean literally they just need to provide any explanation at this point and it's surprising that they haven't said something already. People are still going to be pissed that they pulled it but that's much better than the entire .net community imagining the worst possible scenario that microsoft is getting ready to kill the open source version and that it's time to bail. It is funny because it is totally microsoft's fault that that's the first thing everyone assumes though. The attention to .net is clearly worth it in terms of getting people to make stuff that works on windows/azure but it is easy to imagine some manager saying "no you have to sell more copies of vs" and completely ruining it to squeeze a few more bucks out of people.
|
|
#
?
Oct 23, 2021 16:21
|
|
|
iirc from the early days of .NET CORE like beta8 or 9 dotnet watch was a tool developed by someone independently of dotnet, maybe a msft employer but still not as part of his job. this poo poo makes me really sad because I would probably soon have the chance to leave frontend land for a while and do some c# again and with this I don’t wanna…
|
|
#
?
Oct 23, 2021 18:33
|
|
|
if statements like “there is a bigger internal strategy being formed at Microsoft to make Visual Studio the main IDE for .NET again, because some people at Microsoft are annoyed that Visual Studio Code and other third party tools have been undermining Visual Studio” have not been reported at all, i.e. have no supporting basis from investigation and are just total speculation, then this article is utter trash journalistically to the point of opening itself up to legal liability if anyone cared to pursue it
|
|
#
?
Oct 23, 2021 19:19
|
|
|
rjmccall posted:if statements like “there is a bigger internal strategy being formed at Microsoft to make Visual Studio the main IDE for .NET again, because some people at Microsoft are annoyed that Visual Studio Code and other third party tools have been undermining Visual Studio” have not been reported at all, i.e. have no supporting basis from investigation and are just total speculation, then this article is utter trash journalistically to the point of opening itself up to legal liability if anyone cared to pursue it the choice to identify a specific MS employee and link to their linkedin profile seems totally hosed to me. something you’d do to encourage harassment is very much how that reads to me the article basically claims “we’ve gotten leaks, other outlets reporting on this have gotten leaks, and if you sit an informed MS employee down for coffee you too could have a leak that will tell you how it is” which is, uh, interesting?
|
|
#
?
Oct 23, 2021 19:43
|
|
|
hm, watch/hot reload always seemed kinda like fluff to me, until I (unfortunately) started using webpack'ed javascript, where it just feels like The Community decided making 'watch' work was easier than making 'incremental builds' work.
|
|
#
?
Oct 23, 2021 20:05
|
|
|
it seems hard to imagine that there aren't internal tensions at ms between the free-but-good ide team and the commercial-but-bad ide team, with teams like dotnet getting caught in the crossfire
|
|
#
?
Oct 23, 2021 20:27
|
|
|
Zlodo posted:it seems hard to imagine that there aren't internal tensions at ms between the free-but-good ide team and the commercial-but-bad ide team, with teams like dotnet getting caught in the crossfire quite possibly, but i really think this is way too random a feature with far too likely a different interpretation (bug got reported in it, they're trying to ship the release candidate with everyone already loaded down with other bugs, just axe the feature for now) to make it very likely. if someone wanted to protect visual studio and had the power to get things dropped i think they'd hollow out vscode, at least the .net support, just saying "for .net proper visual studio will suit you better". but fully speculation on my end as well, microsoft certainly has a history of being stupid and inconsistent enough to quite possibly do this sort of thing.
|
|
#
?
Oct 23, 2021 20:59
|
|
|
https://devblogs.microsoft.com/dotnet/net-hot-reload-support-via-cli/quote:.NET Hot Reload Support via CLI Edit: I had never seen the actual announcement (https://devblogs.microsoft.com/dotnet/update-on-net-hot-reload-progress-and-visual-studio-2022-highlights/) so I thought more of this was speculation than actually was; I hadn't realized it but they had officially said that it would be VS only from now on. Based on this, I'm not sure this response really seems insufficient to be honest. They're apologizing for actually deleting it from the code base, but it doesn't sound like they're even saying that they're necessarily committed to continuing to developing it in the open source sdk. It kind of sounds like the speculation was spot on. mystes fucked around with this message at 22:52 on Oct 23, 2021 |
|
#
?
Oct 23, 2021 22:45
|
|
|
quote:In our effort to scope, we inadvertently ended up deleting the source code instead of just not invoking that code path. imagine having to type that sentence, knowing that other people will read it the ifdef and git rm buttons are not that close together on most keyboards
|
|
#
?
Oct 24, 2021 00:04
|
|
|
it's amazing how these companies will jump up and down to blame an employee for outages and promising everyone to reinforce procedures and controls to make sure it never happens again (which is not a good solution anyway but that's besides the point), whereas in this case "woops we mistakenly deleted code rather than adding a conditional" and all of a sudden nothing needs to change and we're sorry you were upset by our actions.
|
|
#
?
Oct 24, 2021 00:53
|
|
|
The problem is it looks like the people in charge right now really don't understand what they're doing and I suspect they really were just trying to sell more copies of visual studio or something completely stupid like that. The blog post mentions the majority of .net developers using visual studio, but realistically the majority of .net developers are still using .net framework and hate change and would be happier if .net core / .net 5/6 just went away (heck they probably wish they were still using legacy vb). If you ignore those people, I would expect that among people who are actually interested in using modern versions of .net, at least 50% are using vs code. Pissing off exactly the people who are excited about .net 6 and effectively telling them they don't matter because Microsoft only cares about people using legacy versions of .net out of inertia is really not a great strategy in the long term if they want people to continue to use .net going forward.
|
|
#
?
Oct 24, 2021 01:03
|
|
|
On a totally unrelated note I want to take a moment to laugh at what a completely stupid pointless clusterfuck ReasonML was.
|
|
#
?
Oct 24, 2021 01:08
|
|
|
visual studio and vscode seem like another incarnation of the whole GM vs Saturn thing, although i am not a car person proposal: gm proper is too full of grifting bureaucrats, make a separate fast-moving division to create a successful product in the market consequence: the fast-moving division is briefly successful, but it turns out the grifting bureaucrats are a lot better at playing corporate politics than the people who actually get poo poo done
|
|
#
?
Oct 24, 2021 01:13
|
|
|
mystes posted:On a totally unrelated note I want to take a moment to laugh at what a completely stupid pointless clusterfuck ReasonML was. i remember being utterly flabbergasted and disgusted when it was announced 'we took one of the cleanest, most elegant languages ever designed and made it deliberately more noisy and ugly. this has absolutely no advantage whatsoever but our hordes of javascript monkeys feel cold without a nice blanket of bullshit ({[;]}) crap around them' it felt like facebook didn't want to leave any stone unturned in its quest to destroy every form of beauty and peace in the world compared to that rob pike keeping his language dumb for the sake of junior googlers was the act of a brave visionary
|
|
#
?
Oct 24, 2021 01:25
|
|
|
NihilCredo posted:it felt like facebook didn't want to leave any stone unturned in its quest to destroy every form of beauty and peace in the world I think it was just a passion project of Jordan Walke’s that got some internal momentum and he’s one of the open source guys so out it goes
|
|
#
?
Oct 24, 2021 01:48
|
|
|
it's been a while since facebook put any effort into any open source, isn't it? react native seems to have lost its internal investment, at the very least. it's kinda just react/jest left and i'm not sure how much jest work comes from facebook (i know facebook still has an actual react team, but i don't know if any of these other projects do) are hack/hhvm dead these days
|
|
#
?
Oct 24, 2021 01:58
|
|
|
|
| # ? Jun 8, 2024 17:05 |
|
|
abraham linksys posted:it's been a while since facebook put any effort into any open source, isn't it? react native seems to have lost its internal investment, at the very least. it's kinda just react/jest left and i'm not sure how much jest work comes from facebook (i know facebook still has an actual react team, but i don't know if any of these other projects do) jens axboe, the buy behind io_uring who made io apis impossibly fast, works for facebook so 
|
|
#
?
Oct 24, 2021 02:14
|
|