--- Additional Comments From dje at gcc dot gnu dot org 2005-01-10 15:47
---
The testcases work correctly on AIX and powerpc64-linux, so this likely is due
to Darwin not implementing consistent aggregate padding rules in GCC for
backward compatibility.
--
http://gcc.gnu.org/bugzilla
--- Additional Comments From dje at gcc dot gnu dot org 2005-01-10 15:27
---
Patch applied to mainline and GCC 3.4 branch.
--
What|Removed |Added
Status
--- Additional Comments From dje at gcc dot gnu dot org 2005-01-07 18:22
---
Created an attachment (id=7893)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=7893&action=view)
swap library extension search order for -brtl
Attached is my proposed patch for this problem.
--- Additional Comments From dje at gcc dot gnu dot org 2005-01-07 15:40
---
patch applied
--
What|Removed |Added
Status|ASSIGNED|RESOLVED
--- Additional Comments From dje at gcc dot gnu dot org 2005-01-07 03:30
---
I finally figured out how to get legitmize_reload_address to force the sum into
a register: wrap the PLUS in another PLUS with constant offset zero. The outer
PLUS becomes the MEM and the inner PLUS becomes
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-31 23:55
---
XLC includes the option
strict_induction
Turns off induction variable optimizations
that have the potential to alter the semantics of a
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-31 19:38
---
Could we use/extend -ffinite-math-only option to cover this case and assert that
the loop will not be infinite?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19210
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-30 17:42
---
The patch slightly improves a few SPEC testcases.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19042
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-29 18:00
---
Increasing the value of MOVE_RATIO and CLEAR_RATIO decreases SPEC
performance. The uses of the macro may be overloaded and confusing the cost
model. Or it could be harmed by register allocator and spilling
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-28 00:10
---
libstdc++ is calling frexpl in libc.a, which expects an AIX format long double
value, not a double. Things work if one uses -mlong-double-128. At some point
we need to default to long-double-128 for AIX for
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-25 20:14
---
Is this compiled with -mlong-double-128 or using the default, which is
-mlong-double-64 on AIX? GCC currently does not build a long-double-128
multilib on AIX, which would be necessary for that mode to work
--
What|Removed |Added
CC||dje at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19115
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-23 19:55
---
Do you have the environment variable PSALLOC set on in the process building GCC?
I build on AIX systems with data, stack, and memory size unlimited, 2GB memory,
PSALLOC unset.
If running stage1 genattrtab
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-23 15:19
---
Fixing the problem at the RTL evel with loop header copying improves sixtrack
performance 12.5%.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19038
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-23 01:54
---
what release of gcc is used for bootstrap?
what are hard and soft limits, includig limits set in /etc/security?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19082
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-19 20:37
---
Other compilers for AIX can work around this problem, but this is asking a lot
from the compiler. I am recategorizing this as an enhancement request.
--
What|Removed |Added
--
What|Removed |Added
CC||dje at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19042
--
What|Removed |Added
CC||dje at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19046
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-16 22:58
---
I think it needs some experimentation to determine the best performance trade
off. I do not expect a value of 15 to be correct. I would guess at something
like
#define MOVE_RATIO (TARGET_STRING ? 2
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-16 21:39
---
The focus of the problem is the inner loop of functio thin6d at line 572. The
function consumes 97.5% of the cycles and the inner loop consumes 48.5%.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19038
Product: gcc
Version: 4.0.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: enhancement
Priority: P2
Component: rtl-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu
Version: 3.4.0
Status: UNCONFIRMED
Keywords: ABI
Severity: normal
Priority: P2
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot
: UNCONFIRMED
Keywords: wrong-debug
Severity: normal
Priority: P1
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet
--
What|Removed |Added
CC||dje at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18942
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-12 19:00
---
The new doloop generator is being more conservative for the signed int
argument and induction variable. Unsigned int produces better code. Maybe
the new algorithm could be more aggressive when one of the
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-12 17:38
---
The patch fixes the problem on big-endian systems.
--
What|Removed |Added
Status
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-11 17:55
---
Patch applied.
--
What|Removed |Added
Status|NEW |RESOLVED
--
What|Removed |Added
CC||geoffk at apple dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18916
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-09 16:32
---
AM_MAKEFLAGS removed from libgfortran/Makefile.am
--
What|Removed |Added
Status|NEW
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-09 00:12
---
Confirmed on AIX.
--
What|Removed |Added
Status|UNCONFIRMED |NEW
: build
Severity: critical
Priority: P1
Component: bootstrap
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet: powerpc-ibm-aix5.2.0.0
GCC host triplet
--- Additional Comments From dje at gcc dot gnu dot org 2004-12-05 06:09
---
For option (2), are you suggesting changing PREFERRED_RELOAD_CLASS from
((GET_CODE (X) == CONST_DOUBLE && GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT)
? NO_REGS : ...)
to
((GET_CODE (X) == CONS
--- Additional Comments From dje at gcc dot gnu dot org 2004-11-24 18:18
---
Allowing the middle-end to know that the L_R_A address is offsettable looks
like a better solution to me. The design is an issue for RTH. One
possibility is a target macro to decide if L_R_A addresses
--- Additional Comments From dje at gcc dot gnu dot org 2004-11-23 01:32
---
patches committed
--
What|Removed |Added
Status|ASSIGNED
--- Additional Comments From dje at gcc dot gnu dot org 2004-11-23 01:29
---
When AIX Bourne Shell is used for GCC bootstrap, bootstrap is extremely slow due
to the interaction between the configure script and the shell. To speed up the
process one has to use Bash. The AIX-specific
logical ops
Product: gcc
Version: 4.0.0
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc
--- Additional Comments From dje at gcc dot gnu dot org 2004-11-16 19:16
---
Fix included in PTF U488845.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11315
--
What|Removed |Added
AssignedTo|unassigned at gcc dot gnu |aldyh at gcc dot gnu dot org
|dot org |
Status|UNCONFIRMED
Priority: P2
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet: *-*-*
GCC host triplet: *-*-*
GCC target triplet: powerpc-*-*
http://gcc.gnu.org
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-28 15:40 ---
The GDB 5.3 error comes from the macro
#define ANOFFSET(secoff, whichone) \
((whichone == -1) \
? (internal_error (__FILE__, __LINE__, "Section index is uninitialized"), -1
) \
: secof
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-28 15:31 ---
gfortran -c -o applu.o -O3 -mcpu=power4 -ftree-loop-linear applu.f
applu.f: In function 'buts':
applu.f:633: internal compiler error: in build_classic_dist_vector, at tree-
data-ref.c:1848
--
--
What|Removed |Added
Severity|normal |critical
Status|UNCONFIRMED |NEW
Ever Confirmed||1
Component: debug
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet: powerpc-ibm-aix5.2.0.0
GCC host triplet: powerpc-ibm-aix5.2.0.0
GCC target triplet: powerpc-ibm-aix5.2.0.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18199
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-27 04:54 ---
Created an attachment (id=7414)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=7414&action=view)
rs6000.md patch
--
What|Removed
--
What|Removed |Added
Summary|SPEC CPU2000 173.applu tree-|SPEC CPU2000 173.applu tree-
|loop-linear failure |loop-linear ICE
http://gcc.gnu.o
Priority: P2
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot gnu dot org
GCC build triplet: *-*-*
GCC host triplet: *-*-*
GCC target triplet: *-*-*
http
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-26 21:25 ---
Also, do not enable when optimizing for size.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18154
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-26 20:06 ---
XLC chooses the straight-line code sequence versus compare and branch based on
a cost model. This should not be a uniform change in behavior for PowerPC.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id
0.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: enhancement
Priority: P2
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc dot gnu dot org
CC: gcc-bugs at gcc dot gnu dot org
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-25 18:33 ---
Correction. While the reload changes fix the crash, it looks like there still
is a bug because the resulting code does not access the correct SUBREG. It
does not access the LSW.
--
http://gcc.gnu.org
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-25 15:46 ---
GCC is using fix_truncdfdi2 because no unsfix_truncdfsi2 pattern is defined and
clock_t is an unsigned long. A short-term workaround is to implement an
unsfix_truncdfsi2 pattern for PPC64 that uses fctid and
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-24 22:52 ---
This problem appears to be a confluence of issues between fix_truncdfdi2 and
limitations of reload, exacerbated by Apple's LL64 mode.
The problematic code is
typedef unsigned long clock_t;
clock_t clock
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-24 02:59 ---
pr15286.c:72: error: unrecognizable insn:
(insn 496 167 381 0 (set (mem:SI (plus:SI (reg/f:SI 1 r1)
(const_int 204 [0xcc])) [15 clock_start+0 S4 A8])
(subreg:SI (reg:DI 32 f0) 4)) -1 (nil
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-22 18:39 ---
Daniel's DOM loop depth patch might help with this problem as well.
http://gcc.gnu.org/ml/gcc-patches/2004-10/msg00931.html
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18048
: mgrid loop performance regression with ivopts
Product: gcc
Version: 4.0.0
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: dje at gcc
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-15 20:16 ---
Proposed patch:
http://gcc.gnu.org/ml/gcc-patches/2004-10/msg01299.html
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18002
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-14 14:49 ---
reducing priority
--
What|Removed |Added
Severity|critical|normal
--- Additional Comments From dje at gcc dot gnu dot org 2004-10-11 15:03 ---
Failures fixed.
--
What|Removed |Added
Status|UNCONFIRMED |RESOLVED
--- Additional Comments From dje at gcc dot gnu dot org 2003-12-08 04:25 ---
http://gcc.gnu.org/ml/gcc-patches/2003-12/msg00731.html
--
What|Removed |Added
501 - 559 of 559 matches
Mail list logo