matrix basics in x86 assembly | SCHIZONE EP-10

Поделиться
HTML-код
  • Опубликовано: 9 сен 2024
  • Tenth video in a series on scientific computing from scratch using x86-64 assembly. This video describes matrix creation, manipulation & arithmetic.
    code: github.com/xmd...
    3 good resources:
    x86 instruction reference:
    www.felixclout...
    linux/freebsd syscall concordance:
    www.lurklurk.o...
    old draft of System-V ABI:
    refspecs.linux...
    secret discord link:
    xmdi.us/misc/d...

Комментарии • 8

  • @dipsiecamila
    @dipsiecamila Год назад

    I miss you!
    It would be good to create a video called SCHIZONE EP-0 where you say exactly where you want to go, so there would be more motivation for us. I know that where you want to go is very big and you want to surprise us, but not knowing where we are going can demotivate us. I would venture to say that you are a genius.

  • @L1Q
    @L1Q Год назад

    the ARRAY_OF_STRUCTS and functions for working on individual matrices do ring a small alarm bell for me. especially considering you went for option 3 - memory pointers. should work, but looks too pOOP from here, read: not every scalable with more data. it becomes important when we start talking graphics, that's what matrix is for, right?
    when implementing real code, consider using structs of arrays along with special function to handle arrays of matrices at once instead of accumulating function call overhead

    • @xmdi0
      @xmdi0  Год назад

      Yeah agreed, especially because of our very "conservative" calling convention (eccessive amount of push/pop), there is a bit of overhead if we were to call these functions billions of times as part of an algorithm. These particular functions may not be called quite as frequently as that, but we will eventually implement cross/dot product functions that will be called much more frequently, and that will really manifest what you are saying. Although performance isn't my main priority in this series (and in the grand scheme of things, we aren't losing much in terms of speed), we should definitely take a look at the performance benefit of inlining these simple operations wherever possible (most of these operations are just 1-2 nested loops).

  • @saikeerty820
    @saikeerty820 Год назад

    Absolutely love the content your putting out. I'm the CMO for CeruLearn, platform in development for online education. We're interested in partnering with you. Is there any way we can get in contact?

  • @artmcdargh9002
    @artmcdargh9002 Год назад

    *promo sm*

  • @darkexior
    @darkexior Месяц назад

    Σ