|
It's a marketing term, obviously.
But hiding behind that is a set of genuinely useful instructions, not exclusively for "deep learning" (whatever that is).
VPDPBUSD (base intrinsic: _mm_dpbusd_epi32), in the words of the official guide,
Multiply groups of 4 adjacent pairs of unsigned 8-bit integers in a with corresponding signed 8-bit integers in b, producing 4 intermediate signed 16-bit results. Sum these 4 results with the corresponding 32-bit integer in src, and store the packed 32-bit results in dst.
So it's like a slightly more elaborate version of the old PMADDUBSW (aka _mm_maddubs_epi16), summing groups of 4 instead of 2 and with an extra 32-bit addition at the end. There is also a version where the final sum is saturating. Since it sums together groups of 4, it has 32 bits to represent the sum of 4 products, and it is easy to avoid saturation. This is different from working with PMADDUBSW which sums together two adjacent 16-bit products with saturation, which in many cases limits the useful range of your scale factors.
It's also actually a fast instruction (CTRL+F for VNNI), at 2 operations per cycle. It's not inherently faster than VPMADDUBSW, but you get more range on your scale factors so the amount of useful work goes up a bit. The 512 bit version does not look super useful yet, possibly even harmful: it's twice as wide but can be done half as often so it does not increase the amount of work done per cycle, and since it is a 512bit instruction it may have funny side effects such as reducing clock speed and fusing ports 0 and 1, who knows, be careful.
Unfortunately there is still no VPMULLB or VPMULHUB or VPMULHB. Intel, AMD, please.
|
|
|
|
|
Good to know that there are still people interested in Assembler, the last time I tried it must have been 30 years ago 
|
|
|
|
|
You have got to be the most knowledgeable guy on this site!
I can't even imagine what such an instruction would be used for.
The difficult we do right away...
...the impossible takes slightly longer.
|
|
|
|
|
Lots of image processing stuff: convolutions, 4-way cross-fade, color space conversion, bilinear interpolation..
It does not seem as great for parsing decimal numbers, because the 4th scale factor won't fit in an sbyte. If you shuffle first you can make something happen but is it worth it? Maybe..
It could be used to turn blocks of 4 8-bit masks into a 4-bit combination of packed flags by setting the second operand to _mm_setr_epi8(-1, -2, -4, -8, -1, -2, -4, -8, -1, -2, -4, -8, -1, -2, -4, -8) (this puts junk in the upper 24 bits of the dwords but just discard that), sort of like a PMOVMSKB but with a vector result, I predict some synergy with PSHUFB (which can use those 4-bit groups to pick arbitrary 8-bit values out of a 16-entry lookup table).
|
|
|
|
|
I mean, I care about posix standards when I'm writing shell scripts.
But regular expressions? really?
*sigh*
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
How else would someone top 24x7 torture?
The sh*t I complain about
It's like there ain't a cloud in the sky and it's raining out - Eminem
~! Firewall !~
|
|
|
|
|
reality television? (either watching or participating, doesn't matter)
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
I hear that they're making Guantanamo prisoners do regular expression coding, and that's the REAL reason hillary and obama are panicking over the DoJ investigations.
".45 ACP - because shooting twice is just silly" - JSOP, 2010
-----
You can never have too much ammo - unless you're swimming, or on fire. - JSOP, 2010
-----
When you pry the gun from my cold dead hands, be careful - the barrel will be very hot. - JSOP, 2013
|
|
|
|
|
So the voices are getting stronger now. Hmmm. 
The sh*t I complain about
It's like there ain't a cloud in the sky and it's raining out - Eminem
~! Firewall !~
|
|
|
|
|
I've got a solution that's about 7 different significant projects (out of like 9)
Each one does its own thing, some are dependent on others, but each one can pretty much be explored and written about in its own right.
However, I can't reasonably segregate all the source to make it available project by project. There's just too many interdependencies to make that anything but confusing for people.
So I have to release the source for the whole thing even if i write 7 different articles - same codebase.
So my question is, should I write an article for each, just because each deserves the space and room to explore it, or should I write one really long article for the whole thing?
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
|
Quote: So my question is, should I write an article for each, just because each deserves the space and room to explore it, or should I write one really long article for the whole thing?
Go for a separate article, because I think 6+ projects are definitely going to require at least a 10k words article, which won't be a good experience for the readers.
I would surely break them into different articles, discussing each topic separately and linking them together with hyperlinks as needed.
The sh*t I complain about
It's like there ain't a cloud in the sky and it's raining out - Eminem
~! Firewall !~
|
|
|
|
|
THe problem I've run into with hyperlinking between articles is having to go back and add the links after each one is approved. When there's a lot of articles over time it gets hectic to remember them all.
Oh well, i still find myself agreeing with you. I think one article is too long.
It's sure too long for me to write in one sitting. 
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
You can put the links at the very beginning of the article like
Part 1 of X
Part 2 of X
Part 3 of X
...
As placeholders, then the article itself.
This way is easier to add the clicketies when the other article is done and approved. This way you only need to go to the latest of your serie and then click thorugh to the first one to add the newest one.
M.D.V. 
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
You can see all your published articles, and their links, in your personal pages
|
|
|
|
|
I realize that.
But I can't put a table of links at the top of each article until they all have been approved. Think about it: I don't get a link until after the article has been approved so i can't link to it, which means every time an article is approved i have to go back and edit all of the old ones.
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
And additionally as long as you have not the privileg "Create an article without requiring moderation" all the updated articles need to be approved again 
It does not solve my Problem, but it answers my question
modified 19-Jan-21 21:04pm.
|
|
|
|
|
yeah, i've been to this rodeo a few times already. LOL
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
|
There are quite a few article series floating round here. So how about:
Codewitch's parser stuff - Part 1 - overview
Codewitch's parser stuff - Part 2 - project A
Codewitch's parser stuff - Part 3 - project B
...
Cheers,
Peter
Software rusts. Simon Stephenson, ca 1994. So does this signature. me, 2012
|
|
|
|
|
It is best to make a series of it, starting with an overview and some introduction and some resumee at the end. You can link them too. Take a look at this outstanding article series.
Best is to prepare all articles so you and the reviewers are understanding the content, its flow and may give some tips.
Great and good luck! 
Press F1 for help or google it.
Greetings from Germany
|
|
|
|
|
I suggest you post this in the Article Writing Forum: [^].
«Where is the Life we have lost in living? Where is the wisdom we have lost in knowledge? Where is the knowledge we have lost in information?» T. S. Elliot
|
|
|
|
|
I create a multi-part article, and break up the narrative into meaningful chunks. As far as the downloadable source goes, I make one zip file, and add it to all of the articles in the set. That way, nothing (as far as the download is concerned) gets lost. I have a few multi-part article sets here.
".45 ACP - because shooting twice is just silly" - JSOP, 2010
-----
You can never have too much ammo - unless you're swimming, or on fire. - JSOP, 2010
-----
When you pry the gun from my cold dead hands, be careful - the barrel will be very hot. - JSOP, 2013
|
|
|
|
|
because I need it, I just made a Document Object Model for regular expressions.
So you can do things like
var expr = new RegexOrExpression(new RegexLiteralExpression('a'),new RegexLiteralExpression('b'));
to do "a|b" aka "[ab]"
in case you're feeling masochistic
but at least you'll understand them. And they work in VB, so you can all use them 
When I was growin' up, I was the smartest kid I knew. Maybe that was just because I didn't know that many kids. All I know is now I feel the opposite.
|
|
|
|
|
The or operator in regular expressions of my bane.
I usually wind up putting them in parentheses just to be sure it knows what I mean.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
