[Bug target/115065] New: AVR clz is not always fast as can be

[Bug target/115065] New: AVR clz is not always fast as can be

[WolfgaNg.Hospital at arcor dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115065

            Bug ID: 115065
           Summary: AVR clz is not always fast as can be
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: WolfgaNg.Hospital at arcor dot de
  Target Milestone: ---

Created attachment 58186
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58186&action=edit
AVR assembly code for __clzhi2

For 8-bit AVRs, count leading zeroes is based on __clzhi2. (This is in contrast
to popcount using the single-register (qi) variant. Possibly due to the
convenience of having the MSB of the result cleared in just one place. No idea
if it was wiser to declare the result single-register, plenty even for
"double-width integer" arguments.)
Instruction order isn't optimal up to now, allowing to take one cycle off for
argument values 1-255. (This includes the worst case: 1.)

[Bug target/115065] AVR clz is not always fast as can be

[gjl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115065

--- Comment #1 from Georg-Johann Lay <gjl at gcc dot gnu.org> ---
IIUC, this is just about the timing of a branch, which in the general != 0 is
currently taken (takes 2 ticks), but it's better to only take it in the
non-common (= 0) case? So that the common case falls through and thus takes 1
cycle less.

[Bug target/115065] AVR clz is not always fast as can be

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115065

--- Comment #2 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Georg-Johann Lay <gjl@gcc.gnu.org>:

https://gcc.gnu.org/g:988838da722dea09bd81ee9d49800a6f24980372

commit r15-645-g988838da722dea09bd81ee9d49800a6f24980372
Author: Wolfgang Hospital <Wolfgang.Hospital@arcor.de>
Date:   Sat May 18 15:02:51 2024 +0200

    AVR: target/115065 - Tweak __clzhi2.

    The libgcc implementation of __clzhi2 can be tweaked by
    one cycle in some situations by re-arranging the instructions.
    It also reduces the WCET by 1 cycle.

    libgcc/
            PR target/115065
            * config/avr/lib1funcs.S (__clzhi2): Tweak.

[Bug target/115065] AVR clz is not always fast as can be

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115065

--- Comment #3 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-14 branch has been updated by Georg-Johann Lay
<gjl@gcc.gnu.org>:

https://gcc.gnu.org/g:3b88dade7ff8a07fd0843ac1281e095cfd94453e

commit r14-10217-g3b88dade7ff8a07fd0843ac1281e095cfd94453e
Author: Wolfgang Hospital <Wolfgang.Hospital@arcor.de>
Date:   Sat May 18 15:02:51 2024 +0200

    AVR: target/115065 - Tweak __clzhi2.

    The libgcc implementation of __clzhi2 can be tweaked by
    one cycle in some situations by re-arranging the instructions.
    It also reduces the WCET by 1 cycle.

    libgcc/
            PR target/115065
            * config/avr/lib1funcs.S (__clzhi2): Tweak.

    (cherry picked from commit 988838da722dea09bd81ee9d49800a6f24980372)

[Bug target/115065] AVR clz is not always fast as can be

[gjl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115065

Georg-Johann Lay <gjl at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P5
           Severity|normal                      |minor
   Target Milestone|---                         |14.2
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED

--- Comment #4 from Georg-Johann Lay <gjl at gcc dot gnu.org> ---
Fixed in v14.2+