[Bug tree-optimization/83171] __builtin_popcountl ((unsigned long long)unsigned_var) is not being optimized to __builtin_popcount (unsigned_var)

[cvs-commit at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83171

--- Comment #4 from CVS Commits  ---
The master branch has been updated by Jakub Jelinek :

https://gcc.gnu.org/g:6dd4c703be17fa5dd56136d067e7fdc4a61584b3

commit r14-5557-g6dd4c703be17fa5dd56136d067e7fdc4a61584b3
Author: Jakub Jelinek 
Date:   Fri Nov 17 15:10:51 2023 +0100

match.pd: Optimize ctz/popcount/parity/ffs on extended argument [PR112566]

ctz(ext(X)) is the same as ctz(X) in the UB on zero case (or could be also
in the 2 argument case on large BITINT_TYPE by preserving the argument, not
implemented in this patch),
popcount(zext(X)) is the same as popcount(X),
parity(zext(X)) is the same as parity(X),
parity(sext(X)) is the same as parity(X) provided the bit difference
between
the extended and unextended types is even,
ffs(ext(X)) is the same as ffs(X).

The following patch optimizes those in match.pd if those are beneficial
(always in the large BITINT_TYPE case, or if the narrower type has optab
and the wider doesn't, or the wider is larger than word and narrower is
one of the standard argument sizes (tested just int and long long, as
long is on most targets same bitsize as one of those two).

Joseph in the PR mentioned that ctz(narrow(X)) is the same as ctz(X)
if UB on 0, but that can be handled incrementally (and would need different
decisions when it is profitable).
And clz(zext(X)) is clz(X) + bit_difference, but not sure we want to change
that in match.pd at all, perhaps during insn selection?

2023-11-17  Jakub Jelinek  

PR tree-optimization/112566
PR tree-optimization/83171
* match.pd (ctz(ext(X)) -> ctz(X), popcount(zext(X)) ->
popcount(X),
parity(ext(X)) -> parity(X), ffs(ext(X)) -> ffs(X)): New
simplifications.
( __builtin_ffs (X) == 0 -> X == 0): Use FFS rather than
BUILT_IN_FFS BUILT_IN_FFSL BUILT_IN_FFSLL BUILT_IN_FFSIMAX.

* gcc.dg/pr112566-1.c: New test.
* gcc.dg/pr112566-2.c: New test.
* gcc.target/i386/pr78057.c (foo): Pass another long long argument
and use it in __builtin_ia32_*zcnt_u64 instead of the int one.

[Bug tree-optimization/83171] __builtin_popcountl ((unsigned long long)unsigned_var) is not being optimized to __builtin_popcount (unsigned_var)

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83171

Andrew Pinski  changed:

   What|Removed |Added

   Last reconfirmed|2017-11-26 00:00:00 |2021-8-19

--- Comment #3 from Andrew Pinski  ---
>Clang (-O2, using libstdc++) and MSVC(/O2, Visual 2017) seem to handle this 
>correctly. 
clang is definitely able to handle it.
We should inline the popcount too if the target does not support popcount but
that is a different bug report I had saw before.

anyways here is a reduced testcase:
#include 
#include 

size_t foo1(uint32_t value)
{
  unsigned char t = value;
  return __builtin_popcount(t);
}
size_t foo2(uint32_t value)
{
  unsigned char t = value;
  unsigned int t1 = t;
  return __builtin_popcountl(t1);
}
- CUT 
Note for aarch64, it does not matter as we always zero_extend to 64bit, though
I suspect there might be a way to remove the addv if we are just doing one byte
popcount.

[Bug tree-optimization/83171] __builtin_popcountl ((unsigned long long)unsigned_var) is not being optimized to __builtin_popcount (unsigned_var)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83171

--- Comment #2 from Richard Biener  ---
A match.pd pattern would fix that I guess.  Similarly we can elide
conversions of args to ffs and ctz.