https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67839
Georg-Johann Lay <gjl at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Priority|P3 |P4
Status|UNCONFIRMED |NEW
Last reconfirmed| |2016-01-25
CC| |gjl at gcc dot gnu.org
Known to work| |4.9.2
Ever confirmed|0 |1
Known to fail| |5.2.1
--- Comment #4 from Georg-Johann Lay <gjl at gcc dot gnu.org> ---
Senthil, isn't this supposed to be fixed in 5.3?
Compiling the PR67839 code from the test suite with -Os -c -mmcu=attiny44
generates the following assembler code:
main:
sbi 0x1f,0
cbi 0x1f,0
sbi 0x20,0
cbi 0x20,0
in r24,__SREG__
out __SREG__,r24
ldi r30,lo8(96)
ldi r31,0
ld r24,Z
st Z,r24
ldi r24,0
ldi r25,0
ret
.size main, .-main
.ident "GCC: (GNU) 5.3.1 20160106"
BTW, why isn't 4.9 and older affected?
The *cbi and *sbi insn are usualy forged by combine. Where 4.9 rejects
Failed to match this instruction:
(set (mem/v:QI (const_int 64 [0x40]) [0 MEM[(volatile char *)64B]+0 S1 A8])
(ior:QI (mem/v:QI (const_int 64 [0x40]) [0 MEM[(volatile char *)64B]+0 S1
A8])
(const_int 1 [0x1])))
5.3 accepts the exact same insn:
Successfully matched this instruction:
(set (mem/v:QI (const_int 64 [0x40]) [0 MEM[(volatile char *)64B]+0 S1 A8])
(ior:QI (mem/v:QI (const_int 64 [0x40]) [0 MEM[(volatile char *)64B]+0 S1
A8])
(const_int 1 [0x1])))
Only difference is that the constraint has been changed from "n" to "i" which
looks very odd to me.
