Modern CPUs Assign Registers To Speed Up Your Code - Computerphile

Comments · 160

• JB

  @jb31842 · 1 year ago

  - Matt Godbolt is a genius at understanding and explaining CPUs<br>- Can&apos;t get a webcam to focus right

  37

• ER

  @erichobbs4042 · 1 year ago

  I remember watching Hackers back in the 90&apos;s, and there&apos;s a scene where they are trying to get some files downloaded. You get a close-up of the keyboard and you can see K E R M I T typed out. I remember thinking, &quot;Boy I hope that&apos;s the username and not the protocol , otherwise this is going to take forever.&quot;

  9

• FL

  @flymypg · 1 year ago

  With apologies to Dire Straits: &quot;We got to move these register mappings. Custom program delivery!&quot;

  5

• ΝΊ

  @ΝίκοςΙστοσελίδα · 1 year ago

  God, aren&apos;t you on a roll! Malleable encryption, Matt Godbolt, the content is getting ever better

  22

• JW

  @jwmittag · 1 year ago

  And when you mispredict a branch and need to wipe the board clean, but you leave some smudges behind, that’s how you get transient execution vulnerabilities like the (in)famous Spectre and Meltdown.

  227

• JO

  @joehopfield · 1 year ago

  The 1970s PDP/11 instruction set was so elegant you could learn to read the machine code from octal in a few minutes.

  52

• VE

  @Veptis · 7 months ago

  The Godbolt videos are amazing. Will we get a season 4 where it&apos;s all of GPUs and their methods of parallelism?

• EI

  @Eidolon2003 · 1 year ago

  These videos have been absolutely fantastic Matt, thanks! And double thanks for your website :)

  29

• SH

  @shadamethyst1258 · 1 year ago

  Amazing presentation. Matt Godbolt really knows his stuff (not that surprising :p), but he also knows how to explain it in simple terms. I was currently learning this (alongside a few subjects), so that&apos;s gonna help!

  9

• VA

  @vadrif-draco · 1 year ago

  That was really useful and fun, thanks! This entire series has been great so far :)

  5

• HO

  @Howtheheckarehandleswit · 1 year ago

  This kinda makes me want to design a turbo-6502 that has the same interface as the classic 6502, but does these sorts of clever tricks to run old software incredibly fast.

  6

• MA

  @marcelopacheco2479 · 1 month ago

  An Intel CPU has 280 registers per core, but not per thread. With SMT (hyper threading) it shares the pipeline and cache between two tasks at a time, appearing like two physical CPUs to code executing on them. Pipeline stalls (due to register dependencies) are opportunities to give more attention to the other task.