mirror/gcc
2
0
Fork 0
mirror of git://gcc.gnu.org/git/gcc.git synced 2025-04-10 19:31:17 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix PR target/99184: Wrong cast from double to 16-bit and 32-bit ints

Browse Source 
this patch fixed PR target/99184 which incorrectly rounded during 64-bit
(long) double to 16-bit and 32-bit integers.

The patch just removes the respective roundings from
libf7-asm.sx::to_integer and ::to_unsigned.  Luckily, LibF7 does nowhere
use respective functions internally, the only user is in libf7.c::f7_exp

which reads

   f7_round (qq, qq);
   int16_t q = f7_get_s16 (qq);

so that f7_get_s16() operates on an already rounded value, and therefore
this code works unaltered with or without rounding in to_integer.

	PR target/99184
libgcc/config/avr/libf7/
	* libf7-asm.sx (to_integer, to_unsigned): Don't round 16-bit
	and 32-bit integers.
This commit is contained in:
Georg-Johann Lay 2022-09-19 09:46:58 +02:00 committed by Richard Biener
parent 0990a77837
commit 0b5b8ac5cb
1 changed files with 0 additions and 50 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
libgcc/config/avr/libf7/libf7-asm.sx
View File

@ -601,9 +601,6 @@ DEFUN to_integer
tst C6
brmi .Lsaturate.T ; > INTxx_MAX => saturate
rcall .Lround
brmi .Lsaturate.T ; > INTxx_MAX => saturate
brtc 9f ; >= 0 => return
sbrc Mask, 5
.global __negdi2
@ -658,30 +655,6 @@ DEFUN to_integer
.global __clr_8
XJMP __clr_8
.Lround:
;; C6.7 is known to be 0 here.
;; Return N = 1 iff we have to saturate.
cpi Mask, 0xf
breq .Lround16
cpi Mask, 0x1f
breq .Lround32
;; For now, no rounding in the 64-bit case. This rounding
;; would have to be integrated into the right-shift.
cln
ret
.Lround32:
rol C2
adc C3, ZERO
adc C4, ZERO
rjmp 2f
.Lround16:
rol C4
2: adc C5, ZERO
adc C6, ZERO
ret
ENDF to_integer
#endif /* F7MOD_to_integer_ */
@ -725,29 +698,6 @@ DEFUN to_unsigned
clr CA
F7call lshrdi3
POP r16
;; Rounding
;; ??? C6.7 is known to be 0 here.
cpi Mask, 0xf
breq .Lround16
cpi Mask, 0x1f
breq .Lround32
;; For now, no rounding in the 64-bit case. This rounding
;; would have to be integrated into the right-shift.
ret
.Lround32:
rol C2
adc C3, ZERO
adc C4, ZERO
rjmp 2f
.Lround16:
rol C4
2: adc C5, ZERO
adc C6, ZERO
brcs .Lset_0xffff ; Rounding overflow => saturate
ret
.Lset_0xffff: