SWAR Demos - stage of converting eight ASCII bytes to int and string length, including AVX2 implementation. #72
Conversation
…swar benchmarks to share between Google benchmark and Catch2
Co-authored-by: thecppzoo <thecppzoo@gmail.com>
SWAR Demos on ARM
thecppzoo
changed the title
| @@ -23,7 +29,7 @@ uint32_t parse_eight_digits_swar(const char *chars) { | |||
|
|
|||
| // Note: eight digits can represent from 0 to (10^9) - 1, the logarithm base 2 | |||
| // of 10^9 is slightly less than 30, thus, only 30 bits are needed. | |||
| auto lemire_as_zoo_swar(const char *chars) { | |||
| uint32_t lemire_as_zoo_swar(const char *chars) { | |||
There was a problem hiding this comment.
Truly astonishing to see you return a concrete type. :D
There was a problem hiding this comment.
It is an exported symbol, I have to manually set the type in the header, I may as well set it manually in the implementation. Good to see that you noticed
| namespace zoo { | ||
|
|
||
| //constexpr |
There was a problem hiding this comment.
Raises the question of why it isn't constexpr
There was a problem hiding this comment.
It would not be exportable.
Also, all of SIMD support is un-usable in constexpr. I should delete the comment, but that's the reason why we have to dial down constexpr quite a lot in many parts of the library.
| // speculative, it might read bytes outside of the actual string. | ||
| // It is safe to read within the page where the string occurs, and to | ||
| // guarantee that, simply make aligned reads because the size of the SWAR | ||
| // base size will always divide the memory page size |
There was a problem hiding this comment.
Should this comment go in a SWAR-library level document?
There was a problem hiding this comment.
Yes. I have not yet created such a document. Because this is introductory code we can keep it like this, right? once we have a body of improvements we can populate other parts of the library such as design documentation.
The most critical piece that only lives as a comment is the one describing associative iteration.
There was a problem hiding this comment.
Would it be valuable to set up a Doxygen generation from this and host it somewhere?
Starting to have some documentation infra could motivate more and better documentation.
There was a problem hiding this comment.
Once in a blue moon I try Doxygen. Sadly, it just does not understand my code. I still use doxygen comments in the code.
| // AVX does not offer a practical way to generate a mask of all ones in | ||
| // the least significant positions, thus we cant invoke adjustFor_strlen. | ||
| // We will do the first iteration as a special case to explicitly take into | ||
| // account misalignment |
There was a problem hiding this comment.
Were any of the comments generated by ChatGPT? I think that could be good to know too.
There was a problem hiding this comment.
I think I opened a can of worms by crediting ChatGPT 4.
Yes, some comments are from the tool, the others mine, and I forgot which is who's.
What's your opinion about crediting the AI? I think it is sufficient to indicate that a tool was involved and the result is hybrid.
What I wanted was to let the reader know I did not have to do the heavy lifting of selecting the intrinsics.
The ARM/NEON flavor was a different experience, though, I did not manage to get usable code from the tool, I had to decode the GLIBC assembler to learn the counter-intuitive technique of, to narrow bytes to nibbles, to start by making byte-pairs to shift them by 4 bits (a nibble) so the upper and lower nibbles straddle the byte boundary, then reduce to bytes, which in reality is a pair of nibbles.
If I indicate what comments were mine, sure, but when does it stop? do I also indicate my prompts? what about my user configuration, so that people can replicate exactly? then I'd also have to indicate the day when I had the session, since the tool can change radically one day to the next.
ARM Strlen from Godbolt
No description provided.