Parsa (@ParsaIsBack) header
Parsa (@ParsaIsBack) profile picture

Building a JavaScript Optimization Engine (Joe) | Probably writing code | often qti3e | πŸ¦€πŸ³οΈβ€πŸŒˆ

San Francisco, CA
Joined November 2022
Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)Photo by Parsa (@ParsaIsBack)
Parsa retweeted
a comment from 3blue1brown's most recent video
308
1,492
266
19,241
1,098,117
it's crazy how the AI advancements revealed how much we resent each other. im not good at math or drawing but i never resented the math people or the artists for their ability to do that stuff
79
987
31
11,091
134,337
Parsa retweeted
Everyone's trying to make CPUs faster. I’m trying to make them worse. New project: CPU deoptimization. Searching for the slowest possible machine instructions x86 single instruction record: 198,002,498,236 cycles, 62 seconds The assembly hall of shame: github.com/xoreaxeaxeax/asm-…
92
348
43
4,185
292,193
Parsa retweeted
Pentium 4 propaganda! Branch predictor hints (the 0x2e/3e prefix) on most modern CPUs don't have any effect, the important thing is about putting the likely paths near and making the unlikely case a far jump. This improves the performance by putting less pressure on the frontend and icache for instruction that you're likely to not run. C's original design, and to great extent still true to this day, is a line-by-line compiled language so to create that same effect what you do is you put the unlikely cases at the end of the function and use "goto" statements. So C actually had the most manual branch hint tooling (goto), later compilers and languages gave likely/unlikely/#[cold] so they could do that same basic block linearization for you instead forcing you to write the code in its assembly instruction order. And there is always inline asm xD
1
6
172
I left K3 running autonomously for 3 days with a goal to get OpenBSD working on my reMarkable Paper Pro. It reverse engineered the boot flow by doing lots of HW experiments and got the kernel booting just now!
2
1
3
952
as someone who literally reverse engineered the XZ attack, this is how you get more XZ attacks. suuure let’s just bundle already compiled blobs into other compiled blobs. it’s real fun to audit compressed garbage.
This is exactly right. Source code is on the verge of becoming like assembly. The next step is getting rid of β€œsource code” entirely and just making an efficient binary directly with AI.
143
243
17
4,598
191,902
Parsa retweeted
Replying to @melqtx
I get offended when someone misgenders my favorite model.
1
2
117
completely false. the real programmer's model is the actual microarchitecture.
As a programmer, if you think about the CPU as circuits and electricity.... You might be getting it all wrong. As a programmer this is what you need to know about the CPU is - the programmer's model. It is an abstraction that lets you visualise the CPU as a state-machine having certain essential elements to it, primarily to read instructions and manipulate data. Once you have a clear idea of what it is, you will be able to reason clearly about the programs, their performance and the limitations. I ported over yet another chapter from my book to the web, here's the link - code.pyjamacafe.com/courses/…
19
45
6
733
60,539
I’m convinced that a large % of programmers don’t actually like computers. As a side effect, are also perfectly happy to throw away their reasoning to a model as soon as they can. I don’t get it, at ALL. Don’t you *LIKE* understanding the magic of the machine? You do realize hand-programming (I hate that I even have to specify hand now) is fun…right?
1,000
716
271
8,950
970,807
ADHD isn't a disorder. It's a nervous system designed for a world that DOES NOT EXIST anymore. Let me explain:
152
1,017
183
13,795
3,495,337
i use nixos btw
8
1
43
1,554
Parsa retweeted
Replying to @tiboadriana
Did grandmasters stop playing chess because of Stockfish?
1
1
73
Unpopular opinion - I think the current AI β€œagents” theme is mid curve and took off because nothing else was going on to amuse the chronic gamblers. (Ducks out of the room)
731
149
175
4,055
849,883
Introducing Willow, our new state-of-the-art quantum computing chip with a breakthrough that can reduce errors exponentially as we scale up using more qubits, cracking a 30-year challenge in the field. In benchmark tests, Willow solved a standard computation in <5 mins that would take a leading supercomputer over 10^25 years, far beyond the age of the universe(!).
2,811
11,805
5,622
74,370
19,274,655
Docker is stupid. β€œSorry, we can’t deploy a single statically linked executable of size 10 MB, so LET’S JUST SHIP GIGABYTES OF DATA CONTAINING AN ENTIRE OS WITH ALL DEPENDENCIES FOR EVERY SINGLE SERVICE”
535
359
129
8,092
910,864
Parsa retweeted
If you didn't notice this week, we did a lil' UI revamp ⚑ We've introduced a new vertical sidebar navigation and enhanced our breadcrumb accessibility, making navigation throughout the app easier. This is just the beginningβ€”more improvements are on the way πŸš€
1
3
1
29
1,513
Parsa retweeted
πŸ’›
1
1
1
19
2,417
Next.js be like
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
5
5
95
15,028
Parsa retweeted
1
1
13
1,143
⚑ Fleek Network Phase {3}: Preview ⚑ The Fleek Foundation is excited to release the details for testnet Phase {3}, starting January 31st. This phase will showcase the first service running on the network: Edge Compute πŸ‘‡
5
13
4
88
8,786