Conversation

Notices

1. Embed this notice
   John Regehr (regehr@mastodon.social)'s status on Tuesday, 14-Jan-2025 13:35:38 JST John Regehr

   Computer Systems: A Programmer’s Perspective, a textbook by Bryant and O’Halloran, is in its third edition and also its third decade of teaching students falsehoods about C

   • Embed this notice
     John Regehr (regehr@mastodon.social)'s status on Tuesday, 14-Jan-2025 13:36:54 JST John Regehr

     in reply to

     switching over to codegen for overflow checking for addition, we don't appear to get optimal code for any of the obvious ways to do it, from latest GCC or LLVM, unless we use the intrinsic

     https://gcc.godbolt.org/z/9bcPdzYcY

   • Embed this notice
     John Regehr (regehr@mastodon.social)'s status on Tuesday, 14-Jan-2025 13:36:54 JST John Regehr

     in reply to

     • Martin Uecker

     @uecker reminded me that this C23 feature is a better choice than the intrinsic, and also results in optimal codegen https://gcc.godbolt.org/z/xP3reE9n8

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Tuesday, 14-Jan-2025 13:36:54 JST Rich Felker

     in reply to

     • Martin Uecker

     @regehr @uecker And it can be provided by a dropin if you don't have a C23 compiler.

   • Embed this notice
     John Regehr (regehr@mastodon.social)'s status on Tuesday, 14-Jan-2025 13:36:55 JST John Regehr

     in reply to

     absolutely indefensible code. I mailed them about this years and years ago

     https://gcc.godbolt.org/z/W3jdEdnn4

   • Embed this notice
     Andrew Kelley (andrewrk@mastodon.social)'s status on Tuesday, 14-Jan-2025 15:49:08 JST Andrew Kelley

     in reply to

     @regehr I almost didn't graduate because of crap like this, it really sucked the credibility out of the entire university system. Especially since I was barely scraping by, paying my own way through college, and then what is my money going towards? Between being told that every C++ instance has its executable code copied for every method, being assigned UML diagrams by profs that obviously had never written a line of code in their life... ASU CS program was a shit show

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Dan Cross (cross@discuss.systems)'s status on Tuesday, 14-Jan-2025 15:49:58 JST Dan Cross

     in reply to

     • Andrew Kelley

     @andrewrk @regehr in one of the few programming courses I took, I had an instructor take off points on a program because I used a function-like macro and they couldn't find the function that I was invoking.

     A few years ago, I volunteers was one of the "industry" types who review student code for MIT's 6.172 "Performance Engineering" course. It was a maddening experience; often the grad students teaching the course would give advice that was a) wrong, and b) contradicted the industry people. Many times, the reviews were extra difficult because the code from the instructors was incorrect.

     We started begging them to give us their code (the assignment frameworks and so on) to review _before_ they gave it to the students, but they kept telling us they were too busy with research and couldn't do it. Finally I had to ask, "if you're too busy to do the bare minimum to avoid wasting _my_ time, why should I continue volunteering?"

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     CF Bolz-Tereick (cfbolz@mastodon.social)'s status on Wednesday, 15-Jan-2025 21:04:42 JST CF Bolz-Tereick

     in reply to

     • Rich Felker
     • Martin Uecker

     @dalias @regehr @uecker some quick googling turned up this code here: https://github.com/jart/jtckdint uses the C23 intrinsics if available, otherwise has a dropin. is this what the youths call a polyfill?