https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #31 from Jeffrey A. Law ---
It looks like you've already opened a new BZ. Thanks.
And yes, that's usually the preferred way, particularly if we're dealing with
multiple different testcases that aren't already confirmed to be the
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #30 from Arnd Bergmann ---
This seems to fix the majority of the failures I have run into, but I still see
the ICE with both the 0x2F25F020 and 0xB0981CD0 builds, see comments 18 through
21.
Should I open a new report for those?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
Jeffrey A. Law changed:
What|Removed |Added
Status|NEW |RESOLVED
CC|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #28 from Vladimir Makarov ---
(In reply to Vladimir Makarov from comment #27)
> (In reply to Vladimir Makarov from comment #26)
> > (In reply to Dominik Vogt from comment #24)
> > > While you're at it ... does it have the same or a
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #27 from Vladimir Makarov ---
(In reply to Vladimir Makarov from comment #26)
> (In reply to Dominik Vogt from comment #24)
> > While you're at it ... does it have the same or a similar cause as the Avr
> > bug?
> >
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #26 from Vladimir Makarov ---
(In reply to Dominik Vogt from comment #24)
> While you're at it ... does it have the same or a similar cause as the Avr
> bug?
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78883
>
> (A HImode
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #25 from Vladimir Makarov ---
Author: vmakarov
Date: Tue Jan 17 16:11:55 2017
New Revision: 244535
URL: https://gcc.gnu.org/viewcvs?rev=244535=gcc=rev
Log:
2017-01-17 Vladimir Makarov
PR target/79058
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #24 from Dominik Vogt ---
While you're at it ... does it have the same or a similar cause as the Avr bug?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78883
(A HImode quantity got allocated to r31+r32 (r31 is the last hardreg), in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
Vladimir Makarov changed:
What|Removed |Added
CC||vmakarov at gcc dot gnu.org
---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #22 from Dominik Vogt ---
That looks like a similar problem. I'm lacking some knowledge about how
register pairs are allocated for paradoxical subregs on bigendian systems
though. Deducing from the code quoted above and from what
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #21 from Arnd Bergmann ---
// reduced version of 0xB0981CD0 file
long a;
struct {
int su_flags;
} * b;
struct nilfs_segment_usage *e;
void fn1();
enum { NILFS_SEGMENT_USAGE_ACTIVE, NILFS_SEGMENT_USAGE_DIRTY } fn2() {
return
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #20 from Arnd Bergmann ---
// new reduced test case, build with "arm-linux-gnueabi-gcc-7.0.0 -c -Os
-mbig-endian"
struct nilfs_segment_usage {
int su_flags;
} a;
enum { NILFS_SEGMENT_USAGE_ACTIVE, NILFS_SEGMENT_USAGE_DIRTY } fn1();
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #19 from Arnd Bergmann ---
Created attachment 40516
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40516=edit
preprocessed linux-4.10/fs/nilfs2/sufile.c source, 0xB0981CD0 build
Another version of the source file, also still
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #18 from Arnd Bergmann ---
Created attachment 40515
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40515=edit
preprocessed linux-4.10/fs/nilfs2/sufile.c source, 0x2F25F020 build
The file is slightly different, because of
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #17 from Arnd Bergmann ---
(In reply to Dominik Vogt from comment #16)
> Or rather this one which avoids triggering an assertion failure in
> in_hard_reg_set_p ():
I tried this version, and while it fixes the test case I originally
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #16 from Dominik Vogt ---
Or rather this one which avoids triggering an assertion failure in
in_hard_reg_set_p ():
diff --git a/gcc/lra-constraints.c b/gcc/lra-constraints.c
--- a/gcc/lra-constraints.c
+++ b/gcc/lra-constraints.c
@@
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #15 from Dominik Vogt ---
There's some code to reload such paradoxical subregs in
lra-constraints.c:simplify_operand_subreg():
/* Force a reload for a paradoxical subreg. For paradoxical subreg,
IRA allocates hardreg to the
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #14 from Dominik Vogt ---
Isn't this more or less the same problem as the Avr issue?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78883
On Avr, the register allocator would allow r31:HI if the expression is a
paradoxical subreg of
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
Richard Biener changed:
What|Removed |Added
Priority|P3 |P1
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #13 from Andreas Krebbel ---
(In reply to Andreas Krebbel from comment #12)
> The splitter probably is the reason why the early clobber has been added.
> Handling the SImode part separately requires that the source reg does not
>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #12 from Andreas Krebbel ---
(In reply to Dominik Vogt from comment #11)
> gccint:
> > A operand which is read by the instruction can be tied to an earlyclobber
> > operand if its only use as an input occurs before the early result
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #11 from Dominik Vogt ---
gccint:
> A operand which is read by the instruction can be tied to an earlyclobber
> operand if its only use as an input occurs before the early result is written.
Mabe it's allowed here because of the
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
Andreas Krebbel changed:
What|Removed |Added
CC||krebbel at gcc dot gnu.org
---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #9 from Segher Boessenkool ---
With the code and flags in comment 2 i get a segmentation fault, instead
(with a powerpc64-linux host), somewhere during LRA.
insn 10 is
===
(insn 10 8 11 2 (set (reg:DI 120)
(and:DI
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #8 from Dominik Vogt ---
With the cross compiler and the reduced test case, reload generates a coredump.
Is that what you get for the minimized test?
Program received signal SIGSEGV, Segmentation fault.
0x802bb262 in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #7 from ktkachov at gcc dot gnu.org ---
(In reply to Dominik Vogt from comment #6)
> I'm trying to build an cross compiler but cannot figure out the --target
> configure option to use. Neither --target=arm nor --target=arm-linux nor
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #6 from Dominik Vogt ---
I'm trying to build an cross compiler but cannot figure out the --target
configure option to use. Neither --target=arm nor --target=arm-linux nor
--target=arm-gnu-linux work. gcc/configure spits out an
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #5 from Arnd Bergmann ---
Created attachment 40511
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40511=edit
rtl dump 256r.ed_dce and 257r.combine
Here are four dumps, I hope this is what you are asking for:
latest gcc
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #4 from Dominik Vogt ---
Can you please add the combine dump (and the dump before combine)?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
Arnd Bergmann changed:
What|Removed |Added
CC||vogt at linux dot vnet.ibm.com
---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
--- Comment #2 from ktkachov at gcc dot gnu.org ---
Reduced testcase:
enum { NILFS_SEGMENT_USAGE_ACTIVE, NILFS_SEGMENT_USAGE_DIRTY } a;
void fn2 (long long);
void fn1() {
int b = a & 1 << NILFS_SEGMENT_USAGE_DIRTY;
fn2 (b ? (long long) -1
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79058
ktkachov at gcc dot gnu.org changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Known to work|
32 matches
Mail list logo