state of a typical C compiler, 1998

[pervognsen@mastodon.social]

@joe I love that the experience of reading that table is like "sure, I can see how that would happen" and then you get to the TWENTY MINUTE COMPILE TIME. The "server crash" one is fun too.

[atax1a@infosec.exchange]

@regehr @joe ceci nes pas une peephole optimization

[ratsnakegames@mastodon.social]

@joe the heck is -!0

[ratsnakegames@mastodon.social]

@joe if you put that in your code, you deserve a compiler error

[rotopenguin@mastodon.social]

@atax1a @joe

[joe@f.duriansoftware.com]

@ratsnakegames not much, what’s -! With you

[regehr@mastodon.social]

@joe "Twenty-minute compile time" is so perfect and understated

[regehr@mastodon.social]

@joe of course with modern OOO processors this would be like 0.2 seconds and would go unnoticed

[cr1901@mastodon.social]

@regehr @joe "Server crash" is doing some work here :D

[regehr@mastodon.social]

@cr1901 @joe it's so good because it implicates the OS as an accomplice

[rotopenguin@mastodon.social]

@regehr @joe good thing it was on the best OO processor of its day, the iAPX432

[rotopenguin@mastodon.social]

@cr1901 singularity (1 million parentheses)

[aburka@hachyderm.io]

@joe "several compiler crashes"

[aburka@hachyderm.io]

@joe crashing multiple times in a single run implies some advanced error handling

[rygorous@mastodon.gamedev.place]

@regehr @joe we had widespread OoO by 1998, John! :P

[tay@transfem.social]

@joe@f.duriansoftware.com okay I'm curious about the original source for this and the reason as to why these happen. Some are obvious but how does if (sizeof(char)+d) f(d) cause an illegal instruction to be emitted?

[rygorous@mastodon.gamedev.place]

@regehr @joe

wild to think how 1989-1999 went from 33MHz pipelined in-order scalar to 500MHz out-of-order superscalar, 1999 to 2004 went to like 3GHz from there, and then 2004 to 2014 was "oh boy, hair shirt time"

[dotstdy@mastodon.social]

@regehr @joe thankfully with modern computers we also have 64 bits so we can now shift by UINT64_MAX

[rygorous@mastodon.gamedev.place]

@regehr @joe or I guess, more accurately, waiting for the cooling people to catch up.

[joe@f.duriansoftware.com]

@tay I put a link to the paper in the alt text: https://www.cs.swarthmore.edu/~bylvisa1/cs97/f13/Papers/DifferentialTestingForSoftware.pdf