Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-

In 2015....

Name: Anonymous 2015-01-25 6:17

DX arbitrary-precision integer arithmetic extensions Intel recently added
They are still working on making CPUs that can do arithmetic. Why not calculus already? I had a pretty hard time in my calculus class and I’m a human! Good luck with the integral calculus, CPUs, you're gonna need it.

Name: Anonymous 2015-01-25 11:35

Analog computers can do infinitesimal calculus easily.

Name: Anonymous 2015-01-25 11:58

>>2
False. Precision is limited by noise induced error. Even in an ideal setting with zero noise, you're limited by Planck scale quantization.

Name: Anonymous 2015-01-25 12:24

>>3
An analog computer with infinitely large components in a zero-noise setting can do infinitesimal calculus easily.

Name: Anonymous 2015-01-25 12:27

>>4
infinitely
Shalom!

Name: Anonymous 2015-01-25 12:29

Name: Anonymous 2015-01-25 12:34

>>3
That's not a problem as long as the user knows the SNR.

Name: Cudder !MhMRSATORI 2015-01-25 15:45

Those ADX extensions are pretty lame. They include

- multiplication without affecting the flags
- add-with-carry affecting only carry flag
- add-with-carry using overflow flag as a carry flag

It's not the awesomeness that a REP ADD/ADC/SUB/SBB/etc.S{B,W,D,Q} would be:

; esi = source bignum
; edi = destination bignum
; ecx = size in dords
rep addsd

That would be a real "arbitrary-precision integer arithmetic extension".

Name: Anonymous 2015-01-25 15:50

>>1
Why would you need arbitrary-precision, when 64 bits are enough for everything?

Name: Anonymous 2015-01-25 19:34

>>9
Why would you need 64 bits when 32 bits are enough for everything?

Name: Anonymous 2015-01-25 19:45

>>6
10 lines in scheme.

Name: Anonymous 2015-01-25 23:20

>>8
Handling rep prefixes on instructions that introduce dependencies on flags is a pain in the ass. It's not surprising they chose to implement another set of instructions that are easier for the existing OOO architecture to decode and retire.

Name: Anonymous 2015-01-25 23:37

>>8,12
Looks like they cared enough about killing dependencies between adds that they applied for a patent:

http://patentscope.wipo.int/search/en/WO2013095529

Seems like an obvious continuation of other work already done on Nehalem, but IANA patent lawyer.

Name: Anonymous 2015-01-25 23:55

Floating-point numbers are an invention of the kikes to bring down the White man.

Why hasn't anyone implemented arbitrary-precision numbers in hardware when doing the same is feasible in software? Because the Jews are the ones making the hardware.

Don't change these.
Name: Email:
Entire Thread Thread List