|
Members in Shoutbox
None.
Shoutbox Search
Shoutbox Commands
/w [name] > Whisper
/r > Reply to last whisper /me > Marks as action Shoutbox Information
Moderators may delete any and all shouts at will.
|
Global Shoutbox
Please log in to shout.
[2016-1-28. : 2:30 am]  O)FaRTy1billion[MM] -- x86 is cool, all the stupid primitive things are gone and you can pretty much use anything for anything ... unlike z80 that I was using where you can only add to and subtract from the a register, and with 16 bit registers you can only add to (but not subtract) the HL register.  [2016-1-28. : 2:26 am]  Dem0n -- Mini Moose 2707Mini Moose 2707 shouted: Oh, it's a subset of x86. Damn, I got excited. You ever done y86?[2016-1-28. : 2:26 am] O)FaRTy1billion[MM] -- the floats I made before weren't totally IEEE whatever standard, just because it was easier in the ASM I was using, but it'd be easy in C[2016-1-28. : 2:22 am] O)FaRTy1billion[MM] -- floats actually use a lot of integer math, and particularly with assembly (carry flags, add with carry instructions) it's trivial to extend math beyond the wordsize[2016-1-28. : 2:19 am] O)FaRTy1billion[MM] -- It's not hard to do software floats of arbitrary precision, though[2016-1-28. : 2:17 am] O)FaRTy1billion[MM] -- "Each x87 register, known as ST(0) through ST(7), is 80 bits wide and stores numbers in the IEEE floating-point standard double extended precision format."[2016-1-28. : 2:16 am] O)FaRTy1billion[MM] -- this says float registers are 80 bits and support doubles[2016-1-28. : 2:12 am] dumbducky -- I'd have to look into how super large floats are handled currently to see how they compare. Do they do them manually in the integer ALU or does the FP unit have the capability to do the large stuff?[2016-1-28. : 2:09 am] O)FaRTy1billion[MM] -- you can do it in multiple steps, but that's also part of why I'm curious [2016-1-28. : 2:07 am] dumbducky -- Hardware would be interesting. I think he explained how you can expand sub-32-bit unums (64bits as well, I assume) to fill up a whole register, but how are you going to fit larger unums into a fixed size multiplier/adder?[2016-1-28. : 2:02 am] O)FaRTy1billion[MM] -- is it a software implementation, or is it supposed to be a hardware implementation?[2016-1-28. : 2:02 am] dumbducky -- They've got a whole library set up in mathematica that implements it. I don't have mathematica so I can't really look at it[2016-1-28. : 1:58 am] dumbducky -- I think. Like I said, I was a little fuzzy on some of the details[2016-1-28. : 1:58 am] dumbducky -- He says you don't treat each one as variable, but rather as it's own data type. Compiler would determine what it needs and take care of all of that[2016-1-28. : 1:57 am] O)FaRTy1billion[MM] -- It just makes me curious how you address it in memory though if it is variable like I understood ...[2016-1-28. : 1:56 am] O)FaRTy1billion[MM] -- I got the impression he's just really hard trying to sell the idea Though with modern computing and hardware and stuff, I can see a place for more dynamic datatypes like that[2016-1-28. : 1:54 am] O)FaRTy1billion[MM] -- Also I guess some rounding/conversion errors sometimes make them not, but from what I read he basically just said there's a bit that indicates trailing digits, and that's his way of saying it's still associative in that regard[2016-1-28. : 1:53 am] O)FaRTy1billion[MM] -- they're not associative due to precision. If your number does not exceed the precision, they're perfectly associative[2016-1-28. : 1:53 am] dumbducky -- That's part of it. It also does things better than floats. Apparently floats aren't associative [(a+b)+c != a +(b+c)]. There's more but I watched it last night[2016-1-28. : 1:50 am] O)FaRTy1billion[MM] -- my understanding was just it's variable bits, so like if your number only needs, say, 6 bits of precision it's not using 52[2016-1-28. : 1:48 am] dumbducky -- It's crazy the sort of problems it outperforms floats on using far fewer bits[2016-1-28. : 1:47 am] O)FaRTy1billion[MM] -- I think the video would've shown it, since they were referring to slides and graphics. I looked it up on wikipedia when I got bored of reading them just rambling and it didn't really say anything[2016-1-28. : 1:45 am] Dem0n -- dumbduckydumbducky shouted: how much you squatting? lol ait was my first time squatting and first time doing legs in forever so i only did 55 a couple times[2016-1-28. : 1:43 am] O)FaRTy1billion[MM] -- Just sounds like variable-precision floats or something[2016-1-28. : 1:43 am] O)FaRTy1billion[MM] -- dumbduckydumbducky shouted: Also you nerds need to watch my video about unum computing I read most of the transcript[2016-1-28. : 1:40 am] O)FaRTy1billion[MM] -- Except now that I think about it, using the pointer class keeping the empty ones isn't necessary anymore.[2016-1-28. : 1:39 am] O)FaRTy1billion[MM] -- How I'm doing it lets me just directly copy the code that's already there in SC (though I don't mark the end of lists, since when it reaches the end it just adds more)[2016-1-28. : 1:38 am] O)FaRTy1billion[MM] -- I was just giving a general/vague example. I'm using a vector just to keep all the pointers together, but the pointers are all linked together manually[2016-1-28. : 1:37 am]  jjf28 -- well if you're using a standard container std::sort is good, if you're using a list you've built outta pointers do w/e [2016-1-28. : 1:37 am] O)FaRTy1billion[MM] -- I think -1 is before, 0 is doesn't matter, 1 is after. I'd have to look it up anyway, it's been a while [2016-1-28. : 1:37 am] Dem0n -- I only got almond milk because the store ran out of regular milk. And it's gross when you swallow balls or protein.[2016-1-28. : 1:36 am] O)FaRTy1billion[MM] -- I like qsort. just do return -1, return 0, or return 1 then qsort(mylist, &myfunction)[2016-1-28. : 1:35 am] Dem0n -- dumbduckydumbducky shouted: eat protein, wait two days, do it again my protein powder won't mix into this almond milk D:[2016-1-28. : 1:34 am] jjf28 -- std::sort would be preferred, that way you just need one extra function that tells std::sort which element comes before which |
boutta make a triple type and a quadruple type XDDDDD
Members Online: tokatta007