[1/*] I wanted to post a brief illustration of how hard it is to use CLANG when you're trying to write anything that needs to be carefully optimized. I am not cherry-picking this - this happens to me literally all the time with CLANG.
I do like the idea of ISPC (although I've never used it myself), but most of the time these LLVM disasters really don't have much to do with vectorization. So it's kind of a separate problem. C++ is definitely terrible at expressing vectorization, though, that is true!
-
-
It's actually the lazy/dynamic dispatching I like the most about ISPC - i.e. your binary can hold versions of a function for various instruction sets and the best (widest) will be selected for the system at run-time.
-
For me, ISPC generated slow code (20+% slower than naive intrinsics) when compiling even trivial stuff - granted this was 12 months ago, so perhaps things have improved incredibly. The dynamic dispatch was a huge overhead for our code too. Maybe I need to revisit it...
- Show replies
New conversation -
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.