Raw byte io and type punning are areas where programmers always say "it's so simple, just DWIM!" but it's highly non-simple if you want \
If you read the standards (even old versions) & rationale it's clear this was just due to bad impls.
-
-
I've referred to different versions of ISO C & C++ for several years, and I do not immediately recall any \
-
particular passage that makes this clear.
-
But regardless of the intentions of "The Founding Fathers," a very popular behavior was established, and then it was broken.
-
And it didn't need to be broken. E.g. if Clang had -fno-strict-aliasing set by default, it would still be a conforming impl.
-
A conforming implementation that never performs vectorization unless perhaps you use the restrict keyword.
-
This is understood, and this was the old behavior, and this was preferable to having shit break unexpectedly.
-
The old behavior was having people write thousands of lines of SSE asm or intrinsics by hand, all non-portable & full of bugs..
-
I accept that this is a real problem. I suggest that better solutions need to be found, \
- 5 more replies
New conversation -
-
-
Even 3 decades ago there was significant thought & attention put into C as a HLL, HL optimizations.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.