public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content
@ 2011-10-27 22:59 hjl.tools at gmail dot com
2011-10-27 23:11 ` [Bug rtl-optimization/50891] " hjl.tools at gmail dot com
` (4 more replies)
0 siblings, 5 replies; 6+ messages in thread
From: hjl.tools at gmail dot com @ 2011-10-27 22:59 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50891
Bug #: 50891
Summary: move2add_note_store fails to properly track register
content
Classification: Unclassified
Product: gcc
Version: 4.7.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
AssignedTo: unassigned@gcc.gnu.org
ReportedBy: hjl.tools@gmail.com
There are
(insn 18 323 318 2 (set (reg:DI 2 cx [orig:110 MEM[(c_char * {ref-all})&A.0]+8
] [110])
(mem/s/u/c:DI (const:SI (plus:SI (symbol_ref:SI ("A.0.1724") [flags
0x2] <var_decl 0x7ffff1435460 A.0>)
(const_int 8 [0x8]))) [0 MEM[(c_char * {ref-all})&A.0]+8 S8
A64])) x.f90:7 62 {*movdi_internal_rex64}
(expr_list:REG_EQUIV (mem/s/u/c:DI (const:SI (plus:SI (symbol_ref:SI
("A.0.1724") [flags 0x2] <var_decl 0x7ffff1435460 A.0>)
(const_int 8 [0x8]))) [0 MEM[(c_char * {ref-all})&A.0]+8 S8
A64])
(nil)))
....
(insn 73 71 74 11 (set (reg:SI 2 cx [125])
(const_int 16 [0x10])) x.f90:12 64 {*movsi_internal}
(expr_list:REG_EQUAL (const_int 16 [0x10])
(nil)))
For some reason, move2add_use_add2_insn replaces it with
(insn 73 71 74 11 (set (strict_low_part (reg:QI 2 cx))
(const_int 16 [0x10])) x.f90:12 81 {*movstrictqi_1}
(expr_list:REG_EQUAL (const_int 16 [0x10])
(nil)))
The problem is the upper 48bits of CX isn't zero.
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug rtl-optimization/50891] move2add_note_store fails to properly track register content
2011-10-27 22:59 [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content hjl.tools at gmail dot com
@ 2011-10-27 23:11 ` hjl.tools at gmail dot com
2011-10-27 23:21 ` hjl.tools at gmail dot com
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: hjl.tools at gmail dot com @ 2011-10-27 23:11 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50891
--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-27 23:10:45 UTC ---
reload_cse_move2add treats
(insn 61 60 62 8 (set (reg:CCZ 17 flags)
(compare:CCZ (reg:SI 2 cx)
(const_int 3 [0x3]))) x.f90:9 6 {*cmpsi_1}
(nil))
(jump_insn 62 61 63 8 (set (pc)
(if_then_else (ne (reg:CCZ 17 flags)
(const_int 0 [0]))
(label_ref 231)
(pc))) x.f90:9 589 {*jcc_1}
(expr_list:REG_BR_PROB (const_int 1400 [0x578])
(nil))
-> 231)
with
(gdb) call debug_rtx (implicit_set)
(set (reg:SI 2 cx)
(const_int 3 [0x3]))
(gdb)
at
2046 move2add_note_store (SET_DEST (implicit_set), implicit_set,
insn);
It is wrong.
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug rtl-optimization/50891] move2add_note_store fails to properly track register content
2011-10-27 22:59 [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content hjl.tools at gmail dot com
2011-10-27 23:11 ` [Bug rtl-optimization/50891] " hjl.tools at gmail dot com
@ 2011-10-27 23:21 ` hjl.tools at gmail dot com
2011-10-28 21:20 ` hjl.tools at gmail dot com
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: hjl.tools at gmail dot com @ 2011-10-27 23:21 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50891
--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-27 23:20:47 UTC ---
Does this:
---
diff --git a/gcc/postreload.c b/gcc/postreload.c
index 0e50d85..526015b 100644
--- a/gcc/postreload.c
+++ b/gcc/postreload.c
@@ -2039,6 +2039,10 @@ reload_cse_move2add (rtx first)
allocation if possible. */
&& SCALAR_INT_MODE_P (GET_MODE (XEXP (cnd, 0)))
&& hard_regno_nregs[REGNO (XEXP (cnd, 0))][GET_MODE (XEXP (cnd, 0))]
== 1
+ /* Use implicit set only if the new mode is wider than
+ the previous mode since register isn't really set. */
+ && (GET_MODE_SIZE (GET_MODE (XEXP (cnd, 0)))
+ >= GET_MODE_SIZE (reg_mode[REGNO (XEXP (cnd, 0))]))
&& CONST_INT_P (XEXP (cnd, 1)))
{
rtx implicit_set =
---
make any senses?
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug rtl-optimization/50891] move2add_note_store fails to properly track register content
2011-10-27 22:59 [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content hjl.tools at gmail dot com
2011-10-27 23:11 ` [Bug rtl-optimization/50891] " hjl.tools at gmail dot com
2011-10-27 23:21 ` hjl.tools at gmail dot com
@ 2011-10-28 21:20 ` hjl.tools at gmail dot com
2021-09-13 5:56 ` pinskia at gcc dot gnu.org
2021-09-13 11:40 ` hjl.tools at gmail dot com
4 siblings, 0 replies; 6+ messages in thread
From: hjl.tools at gmail dot com @ 2011-10-28 21:20 UTC (permalink / raw)
To: gcc-bugs
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50891
--- Comment #3 from H.J. Lu <hjl.tools at gmail dot com> 2011-10-28 21:20:09 UTC ---
We have
(set (reg:DI 2) (mem))
...
(set (reg:HI 2) (const_int))
...
(set (reg:SI 2) (const_int))
Since we don't if know
(set (reg:HI 2) (const_int))
will update the whole register, move2add_note_store should
stop if the new mode is narrower than the previous mode.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [Bug rtl-optimization/50891] move2add_note_store fails to properly track register content
2011-10-27 22:59 [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content hjl.tools at gmail dot com
` (2 preceding siblings ...)
2011-10-28 21:20 ` hjl.tools at gmail dot com
@ 2021-09-13 5:56 ` pinskia at gcc dot gnu.org
2021-09-13 11:40 ` hjl.tools at gmail dot com
4 siblings, 0 replies; 6+ messages in thread
From: pinskia at gcc dot gnu.org @ 2021-09-13 5:56 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50891
--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Hmm, was this solve by r0-123608-g7894bc6be052 ?
^ permalink raw reply [flat|nested] 6+ messages in thread* [Bug rtl-optimization/50891] move2add_note_store fails to properly track register content
2011-10-27 22:59 [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content hjl.tools at gmail dot com
` (3 preceding siblings ...)
2021-09-13 5:56 ` pinskia at gcc dot gnu.org
@ 2021-09-13 11:40 ` hjl.tools at gmail dot com
4 siblings, 0 replies; 6+ messages in thread
From: hjl.tools at gmail dot com @ 2021-09-13 11:40 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50891
H.J. Lu <hjl.tools at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |DUPLICATE
Status|UNCONFIRMED |RESOLVED
--- Comment #5 from H.J. Lu <hjl.tools at gmail dot com> ---
Dup.
*** This bug has been marked as a duplicate of bug 56833 ***
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-09-13 11:40 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-10-27 22:59 [Bug rtl-optimization/50891] New: move2add_note_store fails to properly track register content hjl.tools at gmail dot com
2011-10-27 23:11 ` [Bug rtl-optimization/50891] " hjl.tools at gmail dot com
2011-10-27 23:21 ` hjl.tools at gmail dot com
2011-10-28 21:20 ` hjl.tools at gmail dot com
2021-09-13 5:56 ` pinskia at gcc dot gnu.org
2021-09-13 11:40 ` hjl.tools at gmail dot com
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).