public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug rtl-optimization/112268] New: AVR-GCC generates suboptimal code for bit shifts
@ 2023-10-28 20:26 anton at tchekov dot net
0 siblings, 0 replies; only message in thread
From: anton at tchekov dot net @ 2023-10-28 20:26 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112268
Bug ID: 112268
Summary: AVR-GCC generates suboptimal code for bit shifts
Product: gcc
Version: 13.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: anton at tchekov dot net
Target Milestone: ---
This function:
uint8_t extract1(uint32_t val)
{
return val >> 26;
}
generates the following assembly code, which shifts all four registers 26 times
in a loop.
It is exactly the same on optimization levels O2, O3 and Os. (The exact shift
amounts are not important)
extract1:
mov r27,r25
mov r26,r24
mov r25,r23
mov r24,r22
ldi r18,26
1:
lsr r27
ror r26
ror r25
ror r24
dec r18
brne 1b
ret
It is possible to do a lot better with this workaround which uses only 3
instructions:
(and is exactly equivalent)
uint8_t extract2(uint32_t val)
{
uint8_t tmp = val >> 24;
return tmp >> 2;
}
extract2:
mov r24,r25
lsr r24
lsr r24
ret
The "shift loop" only happens with 32-bit integers, but not with 16-bit, where
the optimization opportunity is recognized:
uint8_t extract3(uint16_t val)
{
return val >> 9;
}
extract3:
mov r24,r25
lsr r24
ret
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-10-28 20:26 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-10-28 20:26 [Bug rtl-optimization/112268] New: AVR-GCC generates suboptimal code for bit shifts anton at tchekov dot net
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).