[Bug c/42483] AVR fails all whopr/lto tests
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2010-01-11 23:21 --- Due to target defining linker name. Fixed 4.5 HEAD Author: hutchinsonandy Date: Mon Jan 11 23:12:14 2010 New Revision: 155819 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155819 Log: 2010-01-11 Andy Hutchinson * config/avr/avr.h (LINKER_NAME): Remove. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr.h -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42483
[Bug target/41885] AVR Rotate patterns do not correctly consider overlap.
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2010-01-08 23:02 --- Subject: Bug 41885 Author: hutchinsonandy Date: Fri Jan 8 23:01:45 2010 New Revision: 155746 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155746 Log: 2010-01-08 Andy Hutchinson PR target/41885 * avr.md (rotlqi3): Add CONST_INT_P check. (rotlhi3): Delete. (rotlhi3_8): Delete. (rotlsi3): Delete. (rotlsi3_8): Delete. (rotlsi3_16): Delete. (rotlsi3_24): Delete. (rotl3): New. (*rotw3): New. (*rotb3): New. *avr.c (avr_rotate_bytes): New function. *avr-proto.h (avr_rotate_bytes): New function. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr-protos.h trunk/gcc/config/avr/avr.c trunk/gcc/config/avr/avr.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41885
[Bug target/29560] [avr] Poor optimization for byte shifts
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-12-31 03:16 --- The same occurs in 4.5 I think the bug is invalid The expression 1<< pin will be promoted. This produces a defined result if pin>7 and <15 The expression can not then be lower to 8 bit shift - since a shift by >7 is undefined. If you use pin &7, the shift is indeed 8 bit shift. Though it still loads a int (HIMODE) value of 1 that should have been reduced to QIMODE. setpinx: /* prologue: function */ /* frame size = 0 */ /* stack size = 0 */ .L__stack_usage = 0 mov r25,r24 andi r25,lo8(7) ldi r18,lo8(1) ldi r19,hi8(1) mov r24,r18 rjmp 2f 1: lsl r24 2: dec r25 brpl 1b cpi r22,lo8(1) brne .L7 in r25,53-0x20 or r25,r24 out 53-0x20,r25 ret .L7: in r25,53-0x20 com r24 and r24,r25 out 53-0x20,r24 ret -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29560
[Bug middle-end/19154] miss-optimization of (x & pow2C) avr conditionals returning bool equivalent values
--- Comment #9 from hutchinsonandy at gcc dot gnu dot org 2009-12-30 16:14 --- Fixed on 4.5 - or at least significantly better enough to warrant closure. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19154
[Bug target/42457] [4.5 Regression] AVR fails to build with other than C family languages
--- Comment #6 from hutchinsonandy at gcc dot gnu dot org 2009-12-24 20:33 --- Fixed 4.5 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
[Bug target/42457] [4.5 Regression] AVR fails to build with other than C family languages
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2009-12-24 20:32 --- Subject: Bug 42457 Author: hutchinsonandy Date: Thu Dec 24 20:32:38 2009 New Revision: 155460 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155460 Log: 2009-12-24 Andy Hutchinson PR target/42457 * config/avr/avr.c (avr_extra_arch_macro): Remove static. (avr_cpu_cpp_builtins): Remove. * config/avr/avr.h (avr_extra_arch_macro): Add prototype. * config/avr/avr-c.c: New File. (avr_cpu_cpp_builtins) : Add. * config/avr/t-avr: Add make information for avr-c. * config.gcc (avr-*-*): Include avr-c.o as c and cpp object. Added: trunk/gcc/config/avr/avr-c.c Modified: trunk/gcc/ChangeLog trunk/gcc/config.gcc trunk/gcc/config/avr/avr.c trunk/gcc/config/avr/avr.h trunk/gcc/config/avr/t-avr -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
[Bug target/27192] call through function pointer goes to wrong address
--- Comment #7 from hutchinsonandy at gcc dot gnu dot org 2009-12-24 20:01 --- Fixed 4.5 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27192
[Bug target/35013] Incomplete check in RTL for "pm()" annotation
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-12-24 19:58 --- Fixed 4.5 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35013
[Bug target/35013] Incomplete check in RTL for "pm()" annotation
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-12-24 19:54 --- Subject: Bug 35013 Author: hutchinsonandy Date: Thu Dec 24 19:53:57 2009 New Revision: 155459 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155459 Log: 2009-12-24 Andy Hutchinson PR target/35013, 27192 * config/avr/avr.c (print_operand_address): Print correct program memory address. Add warning for large device offset addresses. (avr_assemble_integer): Ditto. (print_operand): Add warnings for incorrect addressing. (out_movqi_r_mr): Tag assembler with new address codes. (out_movhi_r_mr): Ditto. (out_movsi_r_mr): Ditto. (out_movqi_mr_r): Ditto. (out_movhi_mr_r): Ditto. (out_movsi_mr_r): Ditto. * config/avr/predicates.md (text_segment_operand): New predicate. * config/avr/avr.md (jump): Tag assembler with new address codes. (call_insn): Ditto. (call_value_insn): Ditto. (*tablejump_lib): Ditto. (*cbi): Ditto. (*sbi): Ditto. (indirect_jump): New define_expand. (jcindirect_jump): New pattern for constant expression jump. (njcindirect_jump): Renamed old indirect_jump. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr.c trunk/gcc/config/avr/avr.md trunk/gcc/config/avr/predicates.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35013
[Bug testsuite/38036] [4.4/4.5 Regression][AVR] FAIL: gcc.c-torture/execute/pr37573.c execution
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-12-24 14:46 --- Fixed already Rev 142978 (29th December 2008) -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38036
[Bug c/42483] New: AVR fails all whopr/lto tests
When LTO is enabled AVR fails all whopr/lto tests: Testsuite fails all lto/whopr tests: Example: Executing on host: /media/verbatim/gcchead/obj-dir/gcc/xgcc -B/media/verbatim/gcchead/obj-dir/gcc/ /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1-lib.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/lib/main.c -w -O2 -flto -fno-builtin-abs -DSTACK_SIZE=2048 -DNO_TRAMPOLINES -DSIGNAL_SUPPRESS -mmcu=atmega128 /home/andy/winavrfiles/avrtest/dejagnuboards/exit.c -Wl,-u,vfprintf -lprintf_flt -Wl,-Tbss=0x802000,--defsym=__heap_end=0x80 -lm -o /media/verbatim/gcchead/obj-dir/gcc/testsuite/gcc/abs-1.x6(timeout = 300) spawn /media/verbatim/gcchead/obj-dir/gcc/xgcc -B/media/verbatim/gcchead/obj-dir/gcc/ /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1-lib.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/lib/main.c -w -O2 -flto -fno-builtin-abs -DSTACK_SIZE=2048 -DNO_TRAMPOLINES -DSIGNAL_SUPPRESS -mmcu=atmega128 /home/andy/winavrfiles/avrtest/dejagnuboards/exit.c -Wl,-u,vfprintf -lprintf_flt -Wl,-Tbss=0x802000,--defsym=__heap_end=0x80 -lm -o /media/verbatim/gcchead/obj-dir/gcc/testsuite/gcc/abs-1.x6 /home/andy/local/avr/lib/gcc/avr/4.5.0/../../../../avr/bin/ld: -f may not be used without -shared compiler exited with status 1 output is: /home/andy/local/avr/lib/gcc/avr/4.5.0/../../../../avr/bin/ld: -f may not be used without -shared FAIL: gcc.c-torture/execute/builtins/abs-1.c compilation, -O2 -flto Note LTO was built using patch from http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457 -- Summary: AVR fails all whopr/lto tests Product: gcc Version: lto Status: UNCONFIRMED Severity: normal Priority: P3 Component: c AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: hutchinsonandy at gcc dot gnu dot org GCC host triplet: i686-pc-linux-gnu GCC target triplet: avr-*-* http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42483
[Bug target/42457] [4.5 Regression] AVR fails to build with other than C family languages
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 21:20 --- Testsuite fails all lto/whopr tests: Example: Executing on host: /media/verbatim/gcchead/obj-dir/gcc/xgcc -B/media/verbatim/gcchead/obj-dir/gcc/ /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1-lib.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/lib/main.c -w -O2 -flto -fno-builtin-abs -DSTACK_SIZE=2048 -DNO_TRAMPOLINES -DSIGNAL_SUPPRESS -mmcu=atmega128 /home/andy/winavrfiles/avrtest/dejagnuboards/exit.c -Wl,-u,vfprintf -lprintf_flt -Wl,-Tbss=0x802000,--defsym=__heap_end=0x80 -lm -o /media/verbatim/gcchead/obj-dir/gcc/testsuite/gcc/abs-1.x6(timeout = 300) spawn /media/verbatim/gcchead/obj-dir/gcc/xgcc -B/media/verbatim/gcchead/obj-dir/gcc/ /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/abs-1-lib.c /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.c-torture/execute/builtins/lib/main.c -w -O2 -flto -fno-builtin-abs -DSTACK_SIZE=2048 -DNO_TRAMPOLINES -DSIGNAL_SUPPRESS -mmcu=atmega128 /home/andy/winavrfiles/avrtest/dejagnuboards/exit.c -Wl,-u,vfprintf -lprintf_flt -Wl,-Tbss=0x802000,--defsym=__heap_end=0x80 -lm -o /media/verbatim/gcchead/obj-dir/gcc/testsuite/gcc/abs-1.x6 /home/andy/local/avr/lib/gcc/avr/4.5.0/../../../../avr/bin/ld: -f may not be used without -shared compiler exited with status 1 output is: /home/andy/local/avr/lib/gcc/avr/4.5.0/../../../../avr/bin/ld: -f may not be used without -shared FAIL: gcc.c-torture/execute/builtins/abs-1.c compilation, -O2 -flto -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
[Bug target/42457] [4.5 Regression] AVR fails to build with other than C family languages
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 21:18 --- Created an attachment (id=19375) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19375&action=view) PATCH to add avr-c.c Here is patch that I believe moves out the C related stuff to avr-c. builtin_define_std and cpp_define are undeclared - since they have no prototypes - which is pretty ugly. Compiler now builds without disabling LTO - BUT I now have LTO/WHPR problems with LD that is possibly a different bug. I would like opinion on this before concluding the problem is fixed or indeed solution is getting closer. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
[Bug c/42457] AVR fails to build unless LTO is disabled
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 03:20 --- Created an attachment (id=19366) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=19366&action=view) Make log file -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
[Bug c/42457] New: AVR fails to build unless LTO is disabled
AVR will fail to build with (default) LTO enabled.
LTO becomes enabled when a working libelf implementation is found.
Mike Stein brought this to my attention and I have been able to replicate it
now I have loaded libelf.
The undefined symbol `builtin_define_std' is defined in c-cppbuiltin.c. It
would appear this is not linked for some of lto compilations - yet target avr.c
is.
The root cause seems to be that avr target uses `builtin_define_std' inside a
staic function `avr_cpu_cpp_builtins' (avr.c) - whereas is used to be used
inside a macro (avr.h). On most (all?) other targets it is used inside a macro
When the macro is not used, there is no definition needed and no link error.
A relevant change on avr target was made 149519
2009-07-11 Anatoly Sokolov
* config/avr/avr.h (TARGET_CPU_CPP_BUILTINS): Redefine.
(avr_extra_arch_macro) Remove declatation.
* config/avr/avr.c (avr_cpu_cpp_builtins): New function.
(avr_extra_arch_macro) Declare as static.
* config/avr/avr-protos.h (avr_cpu_cpp_builtins): Dclare.
Using --disable-lto switch for configure allows build to complete.
--
Summary: AVR fails to build unless LTO is disabled
Product: gcc
Version: lto
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42457
[Bug testsuite/36903] Generic vectorizarion test failures
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 00:16 --- Fixed 4.5. AVR wont vectorize with word size of 1 byte! -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36903
[Bug testsuite/36903] Generic vectorizarion test failures
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-12-22 00:01 --- Subject: Bug 36903 Author: hutchinsonandy Date: Tue Dec 22 00:00:50 2009 New Revision: 155382 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155382 Log: 2009-12-21 Andy Hutchinson PR testsuite/36903 * gcc.dg/tree-ssa/gen-vect-11.c : Disable for avr target. It will not vectorize. * gcc.dg/tree-ssa/gen-vect-11a.c: Ditto. * gcc.dg/tree-ssa/gen-vect-2.c: Ditto. * gcc.dg/tree-ssa/gen-vect-25.c: Ditto. * gcc.dg/tree-ssa/gen-vect-26.c: Ditto. * gcc.dg/tree-ssa/gen-vect-28.c: Ditto. * gcc.dg/tree-ssa/gen-vect-32.c: Ditto. * gcc.dg/tree-ssa/pr23455.c: Test for 4 eliminations on avr target. * gcc.dg/tree-ssa/ssa-fre-26.c: XFAIL test for avr. * gcc.dg/tree-ssa/vrp47.c: Skip test for avr target due to low branch cost. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-11.c trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-11a.c trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-2.c trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-25.c trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-26.c trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-28.c trunk/gcc/testsuite/gcc.dg/tree-ssa/gen-vect-32.c trunk/gcc/testsuite/gcc.dg/tree-ssa/pr23455.c trunk/gcc/testsuite/gcc.dg/tree-ssa/ssa-fre-26.c trunk/gcc/testsuite/gcc.dg/tree-ssa/vrp47.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36903
[Bug rtl-optimization/23726] Missed optimizations for divmod
--- Comment #10 from hutchinsonandy at gcc dot gnu dot org 2009-12-13 21:05 --- Fixed 4.5 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23726
[Bug rtl-optimization/23726] Missed optimizations for divmod
--- Comment #9 from hutchinsonandy at gcc dot gnu dot org 2009-12-13 21:03 --- Subject: Bug 23726 Author: hutchinsonandy Date: Sun Dec 13 21:03:41 2009 New Revision: 155195 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=155195 Log: PR target/23726 * config/avr/predicates.md (pseudo_register_operand): New predicate for pseudos. * config/avr/avr.md (divmodqi4): Replace with define_insn_and_split to allow div/mod optimization. (udivmodqi4): Ditto. (divmodhi4): Ditto. (udivmodhi4): Ditto. (divmodsi4): Ditto. (udivmodsi4): Ditto. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr.md trunk/gcc/config/avr/predicates.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23726
[Bug middle-end/42303] New: Shift left as wide as mode is expanded into RTL as shift
Shift LEFT by the full length of integer is carried through to RTL as shift -
rather than being ZERO.
This hidden zero prevents further optimization. For example x = a + (B<<16)
should be reduced to x = a. Instead we get x = 0 | a;
This can also be seen in rotate expressions.
int16_t shiftzero(uint16_t x,uint16_t y)
{
return x + (y<<16);
}
shiftzero (uint16_tD. xD., uint16_tD. yD.)
{
uint16_tD. D.;
uint16_tD. D.;
int16_tD. D.;
# BLOCK 2 freq:1
# PRED: ENTRY [100.0%] (fallthru,exec)
D._2 = yD._1(D) << 16;
D._4 = D._2 + xD._3(D);
D._5 = (int16_tD.) D._4;
return D._5;
# SUCC: EXIT [100.0%]
}
/* NOTE Same on rotate - since expander is uses shift */
int16_t rotatezero(uint16_t x,uint16_t y)
{
return ROTATE (x,16);
}
rotatezero (uint16_tD. xD., uint16_tD. yD.)
{
uint16_tD. D.;
uint16_tD. D.;
int16_tD. D.;
# BLOCK 2 freq:1
# PRED: ENTRY [100.0%] (fallthru,exec)
D._2 = xD._1(D) << 16;
D._3 = D._2 | xD._1(D);
D._4 = (int16_tD.) D._3;
return D._4;
# SUCC: EXIT [100.0%]
}
--
Summary: Shift left as wide as mode is expanded into RTL as shift
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42303
[Bug target/42210] avr: optimizing assignment to a bit field
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-11-29 14:35
---
Same on 4.5 Head.
The backend patterns match against "MEM AND singlebit".
Combine never considers this.
Incoming RTL and Combine pass dump file extract:
;; Pred edge ENTRY [100.0%] (fallthru)
(note 4 0 2 2 [bb 2] NOTE_INSN_BASIC_BLOCK)
(insn 2 4 3 2 rot.c:25 (set (reg/v:QI 41 [ flag ])
(reg:QI 24 r24 [ flag ])) 4 {*movqi} (expr_list:REG_DEAD (reg:QI 24 r24
[ flag ])
(nil)))
(note 3 2 7 2 NOTE_INSN_FUNCTION_BEG)
(insn 7 3 8 2 rot.c:26 (set (reg:QI 43)
(and:QI (reg/v:QI 41 [ flag ])
(const_int 1 [0x1]))) 43 {andqi3} (expr_list:REG_DEAD (reg/v:QI 41
[ flag ])
(nil)))
(insn 8 7 9 2 rot.c:26 (set (reg:QI 44)
(ashift:QI (reg:QI 43)
(const_int 6 [0x6]))) 59 {*ashlqi3} (expr_list:REG_DEAD (reg:QI 43)
(nil)))
(insn 9 8 10 2 rot.c:26 (set (reg:QI 45)
(mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])) 4 {*movqi} (nil))
(insn 10 9 11 2 rot.c:26 (set (reg:QI 46)
(and:QI (reg:QI 45)
(const_int -65 [0xffbf]))) 43 {andqi3} (expr_list:REG_DEAD
(reg:QI 45)
(nil)))
(insn 11 10 12 2 rot.c:26 (set (reg:QI 47)
(ior:QI (reg:QI 46)
(reg:QI 44))) 46 {iorqi3} (expr_list:REG_DEAD (reg:QI 46)
(expr_list:REG_DEAD (reg:QI 44)
(nil
(insn 12 11 0 2 rot.c:26 (set (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(reg:QI 47)) 4 {*movqi} (expr_list:REG_DEAD (reg:QI 47)
(nil)))
;; End of basic block 2 -> ( 1)
Trying 2 -> 7:
Successfully matched this instruction:
(set (reg:QI 43)
(and:QI (reg:QI 24 r24 [ flag ])
(const_int 1 [0x1])))
deferring deletion of insn with uid = 2.
modifying insn i3 7 r43:QI=r24:QI&0x1
REG_DEAD: r24:QI
deferring rescan insn with uid = 7.
Trying 7 -> 8:
Failed to match this instruction:
(set (reg:QI 44)
(and:QI (ashift:QI (reg:QI 24 r24 [ flag ])
(const_int 6 [0x6]))
(const_int 64 [0x40])))
Trying 9 -> 10:
Failed to match this instruction:
(set (reg:QI 46)
(and:QI (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(const_int -65 [0xffbf])))
Trying 8 -> 11:
Failed to match this instruction:
(set (reg:QI 47)
(ior:QI (ashift:QI (reg:QI 43)
(const_int 6 [0x6]))
(reg:QI 46)))
Trying 10 -> 11:
Failed to match this instruction:
(set (reg:QI 47)
(ior:QI (and:QI (reg:QI 45)
(const_int -65 [0xffbf]))
(reg:QI 44)))
Trying 7, 8 -> 11:
Failed to match this instruction:
(set (reg:QI 47)
(ior:QI (and:QI (ashift:QI (reg:QI 24 r24 [ flag ])
(const_int 6 [0x6]))
(const_int 64 [0x40]))
(reg:QI 46)))
Failed to match this instruction:
(set (reg:QI 44)
(and:QI (ashift:QI (reg:QI 24 r24 [ flag ])
(const_int 6 [0x6]))
(const_int 64 [0x40])))
Trying 9, 10 -> 11:
Failed to match this instruction:
(set (reg:QI 47)
(ior:QI (and:QI (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(const_int -65 [0xffbf]))
(reg:QI 44)))
Failed to match this instruction:
(set (reg:QI 46)
(and:QI (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(const_int -65 [0xffbf])))
Trying 10, 8 -> 11:
Failed to match this instruction:
(set (reg:QI 47)
(ior:QI (and:QI (reg:QI 45)
(const_int -65 [0xffbf]))
(ashift:QI (reg:QI 43)
(const_int 6 [0x6]
Successfully matched this instruction:
(set (reg:QI 46)
(ashift:QI (reg:QI 43)
(const_int 6 [0x6])))
Failed to match this instruction:
(set (reg:QI 47)
(ior:QI (and:QI (reg:QI 45)
(const_int -65 [0xffbf]))
(reg:QI 46)))
Trying 11 -> 12:
Failed to match this instruction:
(set (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(ior:QI (reg:QI 46)
(reg:QI 44)))
Trying 8, 11 -> 12:
Failed to match this instruction:
(set (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(ior:QI (ashift:QI (reg:QI 43)
(const_int 6 [0x6]))
(reg:QI 46)))
Successfully matched this instruction:
(set (reg:QI 47)
(ashift:QI (reg:QI 43)
(const_int 6 [0x6])))
Failed to match this instruction:
(set (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(ior:QI (reg:QI 47)
(reg:QI 46)))
Trying 10, 11 -> 12:
Failed to match this instruction:
(set (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(ior:QI (and:QI (reg:QI 45)
(const_int -65 [0xffbf]))
(reg:QI 44)))
Successfully matched this instruction:
(set (reg:QI 47)
(and:QI (reg:QI 45)
(const_int -65 [0xffbf])))
Failed to match this instruction:
(set (mem/s/j:QI (const_int 35 [0x23]) [0 S1 A8])
(ior:QI (reg:QI 47)
(reg:QI 44)))
__SREG__ = 0x3f
__SP_H__ = 0x3e
__SP_L__ = 0x3d
__tmp_reg__ = 0
__zero_reg__ = 1
.global __do_copy_data
.global __do_clear_bss
.text
.global set_flag_good
[Bug c/42204] New: update_eliminables should be called in reload after something changes
This bug was previously mentioned in email from Denis. I have confirmed this on 4.5 (after long but educational tour of reload) http://gcc.gnu.org/ml/gcc/2006-03/msg00783.html When stacks slots increase from zero in reload() (in my case due to setup_save_areas() ) the register elimination conditions can change. For example the AVR port has: get_frame_size () > 0 as part of frame_pointer_required_p thus FP->SP elimination can change. New slots do restart the spill and reload loop : /* If we allocated another stack slot, redo elimination bookkeeping. */ if (starting_frame_size != get_frame_size ()) continue; BUT update_eliminables is never called before calculate_needs_all_insns(), so the wrong reloads are calculated (in my case marking the FP to be spilled in the just added caller save insn !) As a consequence, a spill and reload failure occurs. The solution would seem to be to add update_eliminables before calculate_needs_all_insns (global); -- Summary: update_eliminables should be called in reload after something changes Product: gcc Version: 4.5.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: c AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: hutchinsonandy at gcc dot gnu dot org GCC host triplet: i686-pc-linux-gnu GCC target triplet: avr-unknown-none http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42204
[Bug testsuite/38202] [avr] FAIL: gcc.dg/torture/pr37868.c
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-11-26 17:29 --- Fixed 4.5 Author: hutchinsonandy Date: Mon Nov 23 22:21:21 2009 New Revision: 154474 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154474 Log: * gcc.dg/torture/pr37868.c: Use smaller bitfield for 16bit int targets. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/torture/pr37868.c -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38202
[Bug target/37081] [avr] FAIL: gcc.dg/pr36998.c
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-11-26 17:25 --- Author: hutchinsonandy Date: Mon Nov 23 21:38:32 2009 New Revision: 154467 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154467 Log: * gcc.dg/pr36998.c: Frame pointer required for avr target. * gcc.dg/debug/pr41264-1.c: Use long if int size 16bit or smaller. * gcc.dg/pr39226.c: Use smaller bitfield if int size is 16bit. * gcc.dg/pr41340.c: Test requires long pointer target. * gcc.dg/profile-generate-3.c: Test requires profiling support. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/debug/pr41264-1.c trunk/gcc/testsuite/gcc.dg/pr36998.c trunk/gcc/testsuite/gcc.dg/pr39226.c trunk/gcc/testsuite/gcc.dg/pr41340.c trunk/gcc/testsuite/gcc.dg/profile-generate-3.c -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37081
[Bug target/21080] Excecution test failure for avr for pr17377 test case.
--- Comment #6 from hutchinsonandy at gcc dot gnu dot org 2009-11-26 17:20 --- Fixed on 4.5 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21078 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21080
[Bug target/41077] [avr] excessive prologue/epilogue for interrupt handler
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-11-26 17:12 --- R1 and R0 are used at assembler level outside of compiler scope. R1 is always expected to contain as a zero value - to avoid using a register to load this common value. R1 maybe used in assembler for other purposes - but must be restored to zero. R0 is used as a temporary - to get around awkward situation created by compiler like a = *a; Thus any ISR will always save these values - and clear r1. Gcc cannot "see" either r1 or r0 or otherwise track usage. While we may debate the merits of this approach. It has been this way from day one and will not change. :-( -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||WONTFIX http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41077
[Bug c/36470] sizeof UTF-32 is 2 on AVR
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2009-11-23 22:10 --- Subject: Bug 36470 Author: hutchinsonandy Date: Mon Nov 23 22:10:18 2009 New Revision: 154471 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154471 Log: PR testsuite/36470 * gcc.dg/utf-cvt.c: Skip int test for 16bit int targets. Enable short test for avr target. * gcc.dg/utf32-1.c: Enable test for avr and m32 targets. * gcc.dg/utf32-2.c: Ditto. * gcc.dg/utf32-3.c: Ditto. * gcc.dg/utf32-4.c: Enable test for non-32bit targets. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/utf-cvt.c trunk/gcc/testsuite/gcc.dg/utf32-1.c trunk/gcc/testsuite/gcc.dg/utf32-2.c trunk/gcc/testsuite/gcc.dg/utf32-3.c trunk/gcc/testsuite/gcc.dg/utf32-4.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36470
[Bug middle-end/41738] optabs expands rotate using wrong mode
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-11-22 19:04 --- Fixed 4.5 Back port needed. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41738
[Bug c/42114] c99-stdint test fails for ptrdiff test
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2009-11-22 19:02 --- Fixed 4.5 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42114
[Bug middle-end/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #11 from hutchinsonandy at gcc dot gnu dot org 2009-11-22 18:59 --- Fixed. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug target/36055] Missed optimsation of QI register loads
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-11-22 18:57 --- Fixed on 4.5 Revision 152958 2009-10-17 Andy Hutchinson * config/avr.md (*movqi): Add zero as equally preferable constraint as general register. (*movhi): Ditto. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36055
[Bug c/42149] New: ICE:verify_gimple failed for avr target
Test trunk/gcc/testsuite/gcc.dg/matrix/matrix-1.c (and friends) fails with ICE on AVR target. Note there is existing open bug PR37165 - but the failure noted was not ICE. Executing on host: /media/verbatim/gcchead/obj-dir/gcc/xgcc -B/media/verbatim/gcchead/obj-dir/gcc/ /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/matrix/matrix-1.c -O3 -fipa-matrix-reorg -fdump-ipa-matrix-reorg -fwhole-program -combine -DSTACK_SIZE=2048 -DNO_TRAMPOLINES -DSIGNAL_SUPPRESS -mmcu=atmega128 /home/andy/winavrfiles/avrtest/dejagnuboards/exit.c -Wl,-u,vfprintf -lprintf_flt -Wl,-Tbss=0x802000,--defsym=__heap_end=0x80 -lm -o ./matrix-1.exe(timeout = 300) /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/matrix/matrix-1.c: In function 'mem_init': /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/matrix/matrix-1.c:52:1: error: type mismatch in array reference struct __file * struct __file * D.3778 = __iob[2]; /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/matrix/matrix-1.c:52:1: error: type mismatch in array reference struct __file * struct __file * D.3778 = __iob[2]; /media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/matrix/matrix-1.c:52:1: internal compiler error: verify_gimple failed Please submit a full bug report, -- Summary: ICE:verify_gimple failed for avr target Product: gcc Version: 4.5.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: c AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: hutchinsonandy at gcc dot gnu dot org GCC host triplet: i686-pc-linux-gnu GCC target triplet: avr-unknown-none http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42149
[Bug c/42147] New: ICE: verify_stmts failed
Test trunk/gcc/testsuite/gcc.dg/pr34668-1.c fails with ICE for AVR target.
PR34668-1.c:
/* PR c/34668 */
/* { dg-do compile } */
/* { dg-options "--combine -O2" } */
/* { dg-additional-sources "pr34668-2.c" } */
struct optab { unsigned code; };
extern struct optab optab_table[1];
void
init_optab (struct optab *op)
{
op->code = 0xdead;
}
void
set_conv_libfunc (void)
{
init_optab (&optab_table[0]);
}
-
PR34668-2.c:
/* PR c/34668 */
/* { dg-do compile } */
struct optab { unsigned code; };
extern struct optab optab_table[1];
---
Executing on host: /media/verbatim/gcchead/obj-dir/gcc/xgcc
-B/media/verbatim/gcchead/obj-dir/gcc/
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/pr34668-1.c --combine -O2
-DSTACK_SIZE=2048 -DNO_TRAMPOLINES
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/pr34668-2.c -S
-DSIGNAL_SUPPRESS -mmcu=atmega128 -o pr34668-1.s(timeout = 300)
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/pr34668-2.c: In function
'set_conv_libfunc':
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/pr34668-2.c:5:15: error:
type mismatch in array reference
struct optab
struct optab
# .MEM_3 = VDEF <.MEM_1(D)>
optab_table[0].code = 57005;
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/pr34668-2.c:5:15: internal
compiler error: verify_stmts failed
Please submit a full bug report,
--
Summary: ICE: verify_stmts failed
Product: gcc
Version: 4.5.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42147
[Bug c/42146] New: ICE in tree_low_cst
The test trunk/gcc/testsuite/gcc.dg/large-size-array-2.c causes ICE on AVR
target.
/* PR c/25309 */
/* { dg-do compile } */
/* { dg-options "-O2" } */
static char * name[] = {
[0x8000] = "bar"
};
/* { dg-error "too large" "" { target { { ! lp64 } && { ! llp64 } } } 6 } */
llp645503.c:2:6: error: size of array 'dummy' is negative
Executing on host: /media/verbatim/gcchead/obj-dir/gcc/xgcc
-B/media/verbatim/gcchead/obj-dir/gcc/
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/large-size-array-2.c -O2
-DSTACK_SIZE=2048 -DNO_TRAMPOLINES -S -DSIGNAL_SUPPRESS -mmcu=atmega128 -o
large-size-array-2.s(timeout = 300)
/media/verbatim/gcchead/trunk/gcc/testsuite/gcc.dg/large-size-array-2.c:6:3:
internal compiler error: in tree_low_cst, at tree.c:6199
--
Summary: ICE in tree_low_cst
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42146
[Bug c/36470] sizeof UTF-32 is 2 on AVR
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-11-21 15:06 --- Test and now passes for avr and m32 targets. gcc.dg/utf32-1.c gcc.dg/utf32-2.c gcc.dg/utf32-3.c I will remove XFAIL in cleanup patch. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36470
[Bug c/42114] c99-stdint test fails for ptrdiff test
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-11-21 01:35 --- Subject: Bug 42114 Author: hutchinsonandy Date: Sat Nov 21 01:34:51 2009 New Revision: 154392 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154392 Log: PR Testsuite/42114 * gcc-dg/c99-stdint-1.c: Condition test for target without signal.h. XFAIL ptrdiff range test for avr. * gcc-dg/c99-stdint-2.c: XFAIL for avr target. * gcc-dg/c99-stdint-5.c: Condition test for target without signal.h. * gcc-dg/c99-stdint-6.c: Ditto. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/c99-stdint-1.c trunk/gcc/testsuite/gcc.dg/c99-stdint-2.c trunk/gcc/testsuite/gcc.dg/c99-stdint-5.c trunk/gcc/testsuite/gcc.dg/c99-stdint-6.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42114
[Bug c/42114] c99-stdint test fails for ptrdiff test
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-11-20 02:07 --- I found c99 limit now which explains it. I was tempted to make PTRDIFF_TYPE signed 32 bits to solve c99 compliance - however that completely useless as we cannot declare array exceeding > 32767 bytes anyway (http://gcc.gnu.org/ml/gcc-patches/2009-11/msg00023.html), despite it being physically possible and with SIZE_TYPE unsigned 16 bits. Skipping the paricular test would seem appropriate - keeping the others which are useful. So should I skip ptrdiff limit test based on pointers being <32 bits? Perhaps ints < 32 bits Or do I use stdint definition of INTPTR_MAX <65535 ? -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42114
[Bug c/42114] New: c99-stdint test fails for ptrdiff test
Test gcc.dg/c99-stdint-1.c checks for range of__PTRDIFF_TYPE__ and fails for
avr target.
test_misc_limits (void)
{
CHECK_SIGNED_LIMITS_2(__PTRDIFF_TYPE__, PTRDIFF_MIN, PTRDIFF_MAX, -65535L,
65535L);
On AVR pointers and integers are 16bit.
INTPTR_MIN =-32768
INTPTR_MAX = 32767
UINTPTR_MAX = 65535U
__PTRDIFF_MAX__ is set by gcc at 32767 and so stdint.h gives PTRDIFF_MAX 32767
and PTRDIFF_MIN -32768
I am not sure why minimum range limits are applied in this test or if
test/gcc/target is wrong.
--
Summary: c99-stdint test fails for ptrdiff test
Product: gcc
Version: 4.5.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42114
[Bug target/21078] Testsuite reports excecution failure for gcc.c-torture/execute/20010122.c for some optimization levels
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-11-15 04:10 --- Subject: Bug 21078 Author: hutchinsonandy Date: Sun Nov 15 04:10:20 2009 New Revision: 154188 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154188 Log: PR target/21078, 21080 * config/avr/avr.c (avr_return_addr_rtx): New function for builtin_return_address. (expand_prologue): Calculate stack usage. (avr_asm_function_end_prologue): Output stack size and offset label. * config/avr/avr.h (RETURN_ADDR_RTX): Replace. (machine_function): Add stack_usage. * config/avr/avr-protos.h (avr_return_addr_rtx): New function. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr-protos.h trunk/gcc/config/avr/avr.c trunk/gcc/config/avr/avr.h -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21078
[Bug middle-end/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #10 from hutchinsonandy at gcc dot gnu dot org 2009-11-12 02:36 --- Subject: Bug 41440 Author: hutchinsonandy Date: Thu Nov 12 02:35:49 2009 New Revision: 154112 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154112 Log: PR middle-end/41440 * cfgexpand.c (expand_gimple_basic_block): Append NOP to a fallthru, single successor block, ending with jump created by RTL expander. Modified: trunk/gcc/ChangeLog trunk/gcc/cfgexpand.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug target/41894] wrong code with -fno-split-wide-types
--- Comment #8 from hutchinsonandy at gcc dot gnu dot org 2009-11-02 00:54 --- The problem seems related to use of R28+r29 - which is also frame pointer. avr_hard_regno_mode_ok allows R28 in HIMODE but not any other mode. (This hack was made to avoid reload problem where r29 was used as well as R28 frame pointer) It looks like the "not ok" QI subreg 28/29 of "ok" HImode r28 is not handled well. The mode restriction is ignored for input reload. (Perhaps because reload decides to use available frame pointer) The output reload needed is 16bit HImode but must be accessible as two QImode subregs. Somehow the mode of output reload gets mangled so we get movw r28,r18 Originally ~2005 the rejection of other modes by avr_hard_regno_mode_ok for r28 was only applied when reload was in progress and frame_pointer_required. So I reapplied this condition and the bug was solved. This function does not need frame pointer so r28 and r29 are indeed freely available. /* Otherwise disallow all regno/mode combinations that span r28:r29. */ if (reload_in_progress && frame_pointer_needed && regno <= (REG_Y + 1) && (regno + GET_MODE_SIZE (mode)) >= (REG_Y + 1)) return 0; This also produces better code! One concern is that the modes deemed ok or not are now not constant - and perhaps is the reason why condition was removed. In Gcc 4.5 the spill is exactly the same but it uses R16/r17 with the correct output reload subreg - so seems to figure out that r28 is "not ok" - suggesting problem is fixed. This may take a while to resolve, with some further research and testing. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41894
[Bug target/41894] wrong code with -fno-split-wide-types
--- Comment #6 from hutchinsonandy at gcc dot gnu dot org 2009-11-01 17:27 --- Created an attachment (id=18946) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18946&action=view) dump file -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41894
[Bug target/41894] wrong code with -fno-split-wide-types
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2009-11-01 17:27 --- Created an attachment (id=18945) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18945&action=view) dump file -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41894
[Bug target/41894] wrong code with -fno-split-wide-types
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-11-01 17:24
---
The problem is still present on 4.3.5 head
I cannot reproduce on 4.5
It looks like reload issue with SUBREG.
Instruction 18 gets reloaded. The output reload is HImode. I will post dump
files but here is extract that appears to highlight problem.
>From lreg dump file:
;; Pred edge 3 [50.0%] (fallthru)
(note 16 15 17 4 [bb 4] NOTE_INSN_BASIC_BLOCK)
(note 17 16 18 4 NOTE_INSN_DELETED)
(insn 18 17 54 4 pr41894.c:29 (set (subreg:QI (reg/v:HI 43 [ tmp ]) 0)
(plus:QI (reg:QI 42 [ D.1188 ])
(const_int -1 [0x]))) 15 {addqi3} (expr_list:REG_DEAD
(reg:QI 42 [ D.1188 ])
(nil)))
(jump_insn 54 18 55 4 (set (pc)
(label_ref 26)) 101 {jump} (nil))
;; End of basic block 4 -> ( 6)
;; lr out 28 [r28] 32 [__SP_L__] 34 [argL] 43
;; live out 28 [r28] 32 [__SP_L__] 34 [argL] 43
from greg dump file:
Spilling for insn 18.
Using reg 18 for reload 0
Spilling for insn 18.
Using reg 18 for reload 0
Reloads for insn # 18
Reload 0: reload_in (QI) = (reg:QI 24 r24 [orig:42 D.1188 ] [42])
reload_out (HI) = (reg/v:HI 28 r28 [orig:43 tmp ] [43])
LD_REGS, RELOAD_OTHER (opnum = 0)
reload_in_reg: (reg:QI 24 r24 [orig:42 D.1188 ] [42])
reload_out_reg: (reg/v:HI 28 r28 [orig:43 tmp ] [43])
reload_reg_rtx: (reg:HI 18 r18)
deleting insn with uid = 2.
;; Register dispositions:
42 in 24 43 in 28 44 in 24
;; Hard regs used: 18 19 24 25 28 29 32
Giving
(insn 57 17 18 4 pr41894.c:29 (set (reg:QI 18 r18)
(reg:QI 24 r24 [orig:42 D.1188 ] [42])) 4 {*movqi} (nil))
(insn 18 57 58 4 pr41894.c:29 (set (reg:QI 18 r18)
(plus:QI (reg:QI 18 r18)
(const_int -1 [0x]))) 15 {addqi3} (nil))
(insn 58 18 54 4 pr41894.c:29 (set (reg/v:HI 28 r28 [orig:43 tmp ] [43])
(reg:HI 18 r18)) 8 {*movhi} (nil))
(jump_insn 54 58 55 4 (set (pc)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41894
[Bug target/41894] wrong code with -fno-split-wide-types
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-10-31 23:02 --- Please post entire assembler code. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added CC||hutchinsonandy at gcc dot ||gnu dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41894
[Bug target/21080] Excecution test failure for avr for pr17377 test case.
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2009-10-31 17:02 --- Anatoly, can we implement this patch to correct __builtin_return_address and thus remove these old bug reports and associated testsuite failures? It also provides a means to document stack usage which would seem useful. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added CC||aesok at pautinka dot net http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21080
[Bug target/21078] Testsuite reports excecution failure for gcc.c-torture/execute/20010122.c for some optimization levels
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-10-31 15:27 --- *** This bug has been marked as a duplicate of 21080 *** -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||DUPLICATE http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21078
[Bug target/21080] Excecution test failure for avr for pr17377 test case.
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-10-31 15:27 --- *** Bug 21078 has been marked as a duplicate of this bug. *** -- Bug 21080 depends on bug 21078, which changed state. Bug 21078 Summary: Testsuite reports excecution failure for gcc.c-torture/execute/20010122.c for some optimization levels http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21078 What|Old Value |New Value Status|UNCONFIRMED |RESOLVED Resolution||DUPLICATE http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21080
[Bug target/41885] AVR Rotate patterns do not correctly consider overlap.
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-10-31 00:38 --- Subject: Bug 41885 Author: hutchinsonandy Date: Sat Oct 31 00:38:10 2009 New Revision: 153773 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=153773 Log: PR target/41885 * gcc.target/avr/torture/pr41885.c: New test. Added: trunk/gcc/testsuite/gcc.target/avr/torture/pr41885.c Modified: trunk/gcc/testsuite/ChangeLog -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41885
[Bug target/41885] New: AVR Rotate patterns do not correctly consider overlap.
Rotate patterns that split byte sized rotates into moves do not correctly consider overlap of operands. This was noted on a similar but different shift bug http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39635 and detailed here http://lists.gnu.org/archive/html/avr-gcc-list/2009-03/msg00158.html Bug raised to reference separate testcase and fix. -- Summary: AVR Rotate patterns do not correctly consider overlap. Product: gcc Version: 4.5.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: target AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: hutchinsonandy at gcc dot gnu dot org GCC host triplet: i486-linux-gnu GCC target triplet: avr-*-* http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41885
[Bug middle-end/19154] miss-optimization of (x & pow2C) avr conditionals returning bool equivalent values
--- Comment #8 from hutchinsonandy at gcc dot gnu dot org 2009-10-24 15:39 --- Subject: Bug 19154 Author: hutchinsonandy Date: Sat Oct 24 15:39:23 2009 New Revision: 153531 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=153531 Log: PR middle-end/19154 * avr.md (QIDI): Add new mode iterator. (sbrx_branch): Create new zero extract bit, test and jump patterns for all QI-DI modes combinations. (sbrx_and_branch): Create new and based bit test and jump patterns for QI-SI modes. avr.c (avr_out_sbxx_branch): Use only bit number. Modified: trunk/gcc/ChangeLog -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19154
[Bug middle-end/19154] miss-optimization of (x & pow2C) avr conditionals returning bool equivalent values
--- Comment #7 from hutchinsonandy at gcc dot gnu dot org 2009-10-24 15:36 --- Subject: Bug 19154 Author: hutchinsonandy Date: Sat Oct 24 15:36:40 2009 New Revision: 153530 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=153530 Log: PR middle-end/19154 * avr.md (QIDI): Add new mode iterator. (sbrx_branch): Create new zero extract bit, test and jump patterns for all QI-DI modes combinations. (sbrx_and_branch): Create new and based bit test and jump patterns for QI-SI modes. avr.c (avr_out_sbxx_branch): Use only bit number. Modified: trunk/gcc/config/avr/avr.c trunk/gcc/config/avr/avr.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19154
[Bug middle-end/41738] optabs expands rotate using wrong mode
-- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Target Milestone|--- |4.4.3 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41738
[Bug middle-end/41738] optabs expands rotate using wrong mode
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-10-17 22:42 --- Subject: Bug 41738 Author: hutchinsonandy Date: Sat Oct 17 22:42:17 2009 New Revision: 152961 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=152961 Log: 2009-10-17 Andy Hutchinson PR middle-end/41738 * optabs.c (expand_binop): Make mode of shift count expression mode of shift count not target. Remove indent nit. Modified: trunk/gcc/ChangeLog trunk/gcc/optabs.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41738
[Bug middle-end/41738] New: optabs expands rotate using wrong mode
I have been adding rotate capability to AVR port and have come this bug in
optabs.c: expand_binop()
This occurs during a rotate expansion. For example
target = op0 rotated by op1
In the particular situation (code extract below) it tries a reverse rotate of
(bits - op1). Where this expression is then expanded into an integer,
a negation or subtraction depending on type of op1 and target.
The expansion of the subtraction is using the mode of the target - it should be
using the mode of op1.
The mode of the rotation amount need not be the same as the target.
target:DI = Op0:DI rotate op1:HI
In my testcase it is not and I get asserts latter in simplfy_rtx with the
mismatched inner/outer modes.
The negation mode looks equally wrong.
I found this on testcase 20010226-1.c with gcc 4.42 (line 55) - however, it
would appear to be on 4.5 and perhaps earlier versions.
Revision 126370 added this particular code to optabs.c in July 2007
/* If we were trying to rotate, and that didn't work, try rotating
the other direction before falling back to shifts and bitwise-or. */
if (((binoptab == rotl_optab
&& optab_handler (rotr_optab, mode)->insn_code != CODE_FOR_nothing)
|| (binoptab == rotr_optab
&& optab_handler (rotl_optab, mode)->insn_code != CODE_FOR_nothing))
&& mclass == MODE_INT)
{
optab otheroptab = (binoptab == rotl_optab ? rotr_optab : rotl_optab);
rtx newop1;
unsigned int bits = GET_MODE_BITSIZE (mode);
if (CONST_INT_P (op1))
newop1 = GEN_INT (bits - INTVAL (op1));
else if (targetm.shift_truncation_mask (mode) == bits - 1)
newop1 = negate_rtx (mode, op1);
else
newop1 = expand_binop (mode, sub_optab,
GEN_INT (bits), op1,
NULL_RTX, unsignedp, OPTAB_DIRECT);
--
Summary: optabs expands rotate using wrong mode
Product: gcc
Version: 4.4.2
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41738
[Bug middle-end/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #9 from hutchinsonandy at gcc dot gnu dot org 2009-10-02 00:30 --- Checked earlier version. They all insert on edge before RTL is expanded. Making this 4.5 regression. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Component|c |middle-end Known to fail||4.5.0 Known to work||4.4.2 4.3.2 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #8 from hutchinsonandy at gcc dot gnu dot org 2009-10-01 01:49 --- Created an attachment (id=18681) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18681&action=view) Fix After lengthy and productive correspondence with Michael Matz, we have arrived at a conclusion. The patch above is ok but it is preferred to fix the problem at the point of RTL expansion. A block that ends with a jump created by RTL expander on an otherwise fallthru edge, is padded with mov(r,r) NOP. Thus commit_one_edge_insertion() will no longer find and interpret this as a jump to successor causing the insertion point to be wrong. The problem breaks cc0 targets during CSE. Non cc0 targets can equally get erroneous insertion made inside RTL expander loop pattern. These may or may not break. Regression tested on Rev15226 (20090927) Native configuration i686-pc-linux-gnu with no regression on C tests (G++ still running) Also Target is avr-unknown-none Host is i686-pc-linux-gnu Producing 28 less failure on C testsuite, no regressions. 2009-09-30 Andy Hutchinson PR middle-end/41440 * cfgexpand.c (expand_gimple_basic_block): Append NOP to a fallthru, single successor block, ending with jump created by RTL expander. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug debug/41474] [4.5 Regression] 951116-1.c ICEs in mem_loc_descriptor, at dwarf2out.c:11279
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-09-27 19:15 --- Same for AVR target. Numerous example (though not same testcase) /home/andy/gccsvn/gcc/testsuite/gcc.c-torture/compile/pr38123.c:13:1: internal compiler error: in mem_loc_descriptor, at dwarf2out.c:11292 Was ok on September 12th and has been failing for a last week or so. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added CC||hutchinsonandy at gcc dot ||gnu dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41474
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #6 from hutchinsonandy at gcc dot gnu dot org 2009-09-27 17:30
---
Created an attachment (id=18663)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18663&action=view)
Patch to cfgrtl.c commit_one_edge_insertion()
The problem is in cfgrtl.c function commit_one_edge_insertion()
It does copy value for phi nodes to previous edge. If it find jump on edge it
places copy before it. It assumes that edge is fall thru edge.
For a fall thru edge, the last instruction can be anything - specifically a
jump just created by RTL expander.
The bug has been present since Rev 146817
I have attached patch that correct the problem.
Index: cfgrtl.c
===
--- cfgrtl.c(revision 152217)
+++ cfgrtl.c(working copy)
@@ -1482,8 +1482,9 @@
happens on the fr30 for example.
We know this block has a single successor, so we can just emit
-the queued insns before the jump. */
- if (JUMP_P (BB_END (bb)))
+the queued insns before the jump. Fallthru edges can have
+jumps created by RTL expander, so ignore these. */
+ if (!(e->flags & EDGE_FALLTHRU) && JUMP_P (BB_END (bb)))
before = BB_END (bb);
else
{
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2009-09-24 03:30
---
Forget previous post, I misunderstood PHI node. However, I was close
The PHI in block 5 has incoming ZERO constant. With "detailed" expansion dump
file , it can be seen that this constant value copy is inserted on BB edge 4-5
but the RTL shows it is actually added one instruction before end of block 4.
The edge 4->5 falls through and insertion should be after jump that is last
insn of BB 4 (and the end of memcpy do loop expansion)
Hopefully, I right this time.
;; Function main (main)
Pass statistics:
CFG Statistics for main
-
Number ofMemory
instances used
-
Basic blocks 9576b
Edges 8352b
-
Total memory used by CFG data 928b
-
Coalesced label blocks: 0 (Max so far: 0)
main ()
{
unsigned intD.3 src.29D.2127;
unsigned intD.3 D.2128;
unsigned charD.10 D.2124;
unsigned charD.10 D.2126;
unsigned intD.3 ivtmp.25D.2120;
unsigned intD.3 ivtmp.20D.2115;
unsigned charD.10 dstD.1321[100];
unsigned charD.10 srcD.1320[100];
unsigned intD.3 iD.1319;
unsigned charD.10 D.2091;
unsigned charD.10 D.2090;
# BLOCK 2 freq:100
# PRED: ENTRY [100.0%] (fallthru,exec)
ivtmp.20D.2115_23 = (unsigned intD.3) &srcD.1320[0];
ivtmp.25D.2120_26 = (unsigned intD.3) &dstD.1321[0];
src.29D.2127_31 = (unsigned intD.3) &srcD.1320;
D.2128_32 = src.29D.2127_31 + 100;
# SUCC: 3 [100.0%] (fallthru,exec)
# BLOCK 3 freq:9900
# PRED: 3 [99.0%] (true,exec) 2 [100.0%] (fallthru,exec)
# ivtmp.20D.2115_2 = PHI
# ivtmp.25D.2120_24 = PHI
D.2124_27 = (unsigned charD.10) ivtmp.20D.2115_2;
D.2126_29 = (unsigned charD.10) ivtmp.20D.2115_23;
D.2090_30 = D.2124_27 - D.2126_29;
# .MEMD.2094_12 = VDEF <.MEMD.2094_20>
MEM[index: ivtmp.20D.2115_2] = D.2090_30;
ivtmp.20D.2115_16 = ivtmp.20D.2115_2 + 1;
# .MEMD.2094_13 = VDEF <.MEMD.2094_12>
MEM[index: ivtmp.25D.2120_24] = 0;
ivtmp.25D.2120_25 = ivtmp.25D.2120_24 + 1;
if (ivtmp.20D.2115_16 != D.2128_32)
goto ;
else
goto ;
# SUCC: 3 [99.0%] (true,exec) 4 [1.0%] (false,exec)
# BLOCK 4 freq:100
# PRED: 3 [1.0%] (false,exec)
# .MEMD.2094_14 = VDEF <.MEMD.2094_13>
dstD.1321 = srcD.1320;
# SUCC: 5 [100.0%] (fallthru,exec)
# BLOCK 5 freq:1815
# PRED: 7 [98.9%] (true,exec) 4 [100.0%] (fallthru,exec)
# iD.1319_19 = PHI
# ivtmp.25D.2120_17 = PHI
# VUSE <.MEMD.2094_14>
D.2091_7 = MEM[index: ivtmp.25D.2120_17];
ivtmp.25D.2120_10 = ivtmp.25D.2120_17 + 1;
D.2090_22 = (unsigned charD.10) iD.1319_19;
if (D.2091_7 != D.2090_22)
goto ;
else
goto ;
# SUCC: 6 [4.5%] (true,exec) 7 [95.5%] (false,exec)
# BLOCK 6 freq:82
# PRED: 5 [4.5%] (true,exec)
# VUSE <.MEMD.2094_14>
abortD.823 ();
# SUCC:
# BLOCK 7 freq:1733
# PRED: 5 [95.5%] (false,exec)
iD.1319_9 = iD.1319_19 + 1;
if (iD.1319_9 != 100)
goto ;
else
goto ;
# SUCC: 5 [98.9%] (true,exec) 8 [1.1%] (false,exec)
# BLOCK 8 freq:18
# PRED: 7 [1.1%] (false,exec)
# VUSE <.MEMD.2094_14>
exitD.869 (0);
# SUCC:
}
Partition map
Partition 2 (ivtmp.20_2 - 2 )
Partition 7 (D.2091_7 - 7 )
Partition 9 (i_9 - 9 )
Partition 10 (ivtmp.25_10 - 10 )
Partition 11 (.MEM_11(D) - 11 )
Partition 12 (.MEM_12 - 12 )
Partition 13 (.MEM_13 - 13 )
Partition 14 (.MEM_14 - 14 )
Partition 16 (ivtmp.20_16 - 16 )
Partition 17 (ivtmp.25_17 - 17 )
Partition 19 (i_19 - 19 )
Partition 22 (D.2090_22 - 22 )
Partition 23 (ivtmp.20_23 - 23 )
Partition 24 (ivtmp.25_24 - 24 )
Partition 25 (ivtmp.25_25 - 25 )
Partition 26 (ivtmp.25_26 - 26 )
Partition 27 (D.2124_27 - 27 )
Partition 29 (D.2126_29 - 29 )
Partition 30 (D.2090_30 - 30 )
Partition 31 (src.29_31 - 31 )
Partition 32 (D.2128_32 - 32 )
Partition map
Partition 0 (ivtmp.20_2 - 2 )
Partition 1 (i_9 - 9 )
Partition 2 (ivtmp.25_10 - 10 )
Partition 3 (ivtmp.20_16 - 16 )
Partition 4 (ivtmp.25_17 - 17 )
Partition 5 (i_19 - 19 )
Partition 6 (ivtmp.20_23 - 23 )
Partition 7 (ivtmp.25_24 - 24 )
Partition 8 (ivtmp.25_25 - 25 )
Partition 9 (ivtmp.25_26 - 26 )
Live on entry to BB2 :
Live on entry to BB3 : ivtmp.20_23 ivtmp.25_26
Live on entry to BB4 : ivtmp.25_26
Live on entry to BB5 :
Live on entry to BB6 :
Live on entry to BB7 : ivtmp.25_10 i_19
Live on entry to BB8 :
Conflict graph:
0: 6
3: 6
6: 0, 3
7: 9
8: 9
9: 7, 8
After sorting:
Sorted Coalesce list:
(2) ivtmp.20_2 <-> ivtmp.20_16
(2) i_9 <-> i_19
(2) ivtmp.25_10 <-> ivtmp.25_17
(2) ivtmp.25_24 <-> ivtmp.25_25
(1)
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2009-09-24 02:21 --- Problem is located in ssa pass. It seems to name initialization of loop index to something other than the loop index increment and tested in the last for loop. This "dead"initialization is deleted in latter passes but the damage is done and the data flow is obviously meesed up. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Keywords||ice-on-valid-code http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2009-09-24 02:15 --- Created an attachment (id=18642) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18642&action=view) ssa dump -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2009-09-24 02:15 --- Created an attachment (id=18641) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18641&action=view) cleanup_cfg dump -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug c/41440] SEG FAULT in CSE.C caused by bad RTL expansion
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2009-09-23 02:49 --- Created an attachment (id=18632) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18632&action=view) Expanded RTL Expanded RTL dump -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug c/41440] New: SEG FAULT in CSE.C caused by bad RTL expansion
ned int) &dst[0];
src.33_31 = (unsigned int) &src;
D.2132_32 = src.33_31 + 100;
:
# ivtmp.24_1 = PHI
# ivtmp.29_24 = PHI
D.2128_27 = (unsigned char) ivtmp.24_1;
D.2130_29 = (unsigned char) ivtmp.24_23;
D.2090_30 = D.2128_27 - D.2130_29;
MEM[index: ivtmp.24_1] = D.2090_30;
ivtmp.24_16 = ivtmp.24_1 + 1;
MEM[index: ivtmp.29_24] = 0;
ivtmp.29_25 = ivtmp.29_24 + 1;
if (ivtmp.24_16 != D.2132_32)
goto ;
else
goto ;
:
dst = src;
:
# i_19 = PHI
# ivtmp.29_17 = PHI
D.2091_7 = MEM[index: ivtmp.29_17];
ivtmp.29_10 = ivtmp.29_17 + 1;
D.2090_22 = (unsigned char) i_19;
if (D.2091_7 != D.2090_22)
goto ;
else
goto ;
:
abort ();
:
i_9 = i_19 + 1;
if (i_9 != 100)
goto ;
else
goto ;
:
exit (0);
}
=
RTL EXPANDER FROM AVR.MD
(define_expand "movmemhi"
[(parallel [(set (match_operand:BLK 0 "memory_operand" "")
(match_operand:BLK 1 "memory_operand" ""))
(use (match_operand:HI 2 "const_int_operand" ""))
(use (match_operand:HI 3 "const_int_operand" ""))])]
""
"{
int prob;
HOST_WIDE_INT count;
enum machine_mode mode;
rtx label = gen_label_rtx ();
rtx loop_reg;
rtx jump;
/* Copy pointers into new psuedos - they will be changed. */
rtx addr0 = copy_to_mode_reg (Pmode, XEXP (operands[0], 0));
rtx addr1 = copy_to_mode_reg (Pmode, XEXP (operands[1], 0));
/* Create rtx for tmp register - we use this as scratch. */
rtx tmp_reg_rtx = gen_rtx_REG (QImode, TMP_REGNO);
if (GET_CODE (operands[2]) != CONST_INT)
FAIL;
count = INTVAL (operands[2]);
if (count <= 0)
FAIL;
/* Work out branch probability for latter use. */
prob = REG_BR_PROB_BASE - REG_BR_PROB_BASE / count;
/* See if constant fit 8 bits. */
mode = (count < 0x100) ? QImode : HImode;
/* Create loop counter register. */
loop_reg = copy_to_mode_reg (mode, gen_int_mode (count, mode));
/* Now create RTL code for move loop. */
/* Label at top of loop. */
emit_label (label);
/* Move one byte into scratch and inc pointer. */
emit_move_insn (tmp_reg_rtx, gen_rtx_MEM (QImode, addr1));
emit_move_insn (addr1, gen_rtx_PLUS (Pmode, addr1, const1_rtx));
/* Move to mem and inc pointer. */
emit_move_insn (gen_rtx_MEM (QImode, addr0), tmp_reg_rtx);
emit_move_insn (addr0, gen_rtx_PLUS (Pmode, addr0, const1_rtx));
/* Decrement count. */
emit_move_insn (loop_reg, gen_rtx_PLUS (mode, loop_reg, constm1_rtx));
/* Compare with zero and jump if not equal. */
emit_cmp_and_jump_insns (loop_reg, const0_rtx, NE, NULL_RTX, mode, 1,
label);
/* Set jump probability based on loop count. */
jump = get_last_insn ();
add_reg_note (jump, REG_BR_PROB, GEN_INT (prob));
DONE;
}")
--
Summary: SEG FAULT in CSE.C caused by bad RTL expansion
Product: gcc
Version: 4.5.0
Status: UNCONFIRMED
Severity: major
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41440
[Bug target/36609] AVR wrong code using incorrect RTL for test reversal pattern
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2008-09-08 22:45 --- 4.4 fixed. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36609
[Bug target/36609] AVR wrong code using incorrect RTL for test reversal pattern
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-09-08 22:26 --- Subject: Bug 36609 Author: hutchinsonandy Date: Mon Sep 8 22:25:42 2008 New Revision: 140124 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=140124 Log: PR target/36609 * config/avr/avr.c (avr_reorg): Create RTL for reversed compare with zero. * config/avr/avr.md (QISI) : Define mode iterator. (negated_tst) : Redefine as split using mode macro. (reversed_tstqi): Define insn as reversed compare with zero. (reversed_tsthi): Ditto. (reversed_tstsi): Ditto. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr.c trunk/gcc/config/avr/avr.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36609
[Bug target/36609] AVR wrong code using incorrect RTL for test reversal pattern
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2008-07-23 10:43 --- Created an attachment (id=15945) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15945&action=view) Patch to target files http://gcc.gnu.org/ml/gcc-patches/2008-07/msg00391.html -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36609
[Bug testsuite/36903] New: Generic vectorizarion test failures
For AVR the following test fail. Probably since the day they were created.
For AVR there are no "vectypes" so they all fail message test
I don't know if these should be skipped as invalid or there is some significant
issue that warrant XFAIL.
FAIL: gcc.dg/tree-ssa/gen-vect-11.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/tree-ssa/gen-vect-11a.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/tree-ssa/gen-vect-2.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/tree-ssa/gen-vect-25.c scan-tree-dump-times vect "vectorized 2
loops" 1
FAIL: gcc.dg/tree-ssa/gen-vect-26.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/tree-ssa/gen-vect-26.c scan-tree-dump-times vect "Alignment of
access forced using peeling" 1
FAIL: gcc.dg/tree-ssa/gen-vect-28.c scan-tree-dump-times vect "vectorized 1
loops" 1
FAIL: gcc.dg/tree-ssa/gen-vect-28.c scan-tree-dump-times vect "Alignment of
access forced using peeling" 1
FAIL: gcc.dg/tree-ssa/gen-vect-32.c scan-tree-dump-times vect "vectorized 1
loops" 1
--
Summary: Generic vectorizarion test failures
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: testsuite
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36903
[Bug middle-end/36884] ifcvt poor optimization
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-07-22 02:37
---
4.2.2 has same issue.
Can someone benchmark -fno-if-conversion for AVR?
Transforms are made independent of target. BRANCH_COST is already 0 for avr.
Here is a related example. This time result is a power of two so it needs a
more complicated shift.
int z;
int foo2(void)
{
if ( ++z >= 0)
return 16;
else
return 0;
}
produces:
24.LM2:
25/*DEBUG: 0x00 12 */
26 8091 lds r24,z
27 0004 9091 lds r25,(z)+1
28/*DEBUG: 0x44 8 */
29 0008 0196 adiw r24,1
30/*DEBUG: 0x51 12 */
31 000a 9093 sts (z)+1,r25
32 000e 8093 sts z,r24
33/*DEBUG: 0x94 16 */
34 0012 292F mov r18,r25
35 0014 220F lsl r18
36 0016 330B sbc r19,r19
37 0018 2795 ror r18
38/*DEBUG: 0xd4 12 */
39 001a 2095 com r18
40 001c 3095 com r19
41/*DEBUG: 0xf2 28 */
42 001e 2227 clr r18
43 0020 330F lsl r19
44 0022 221F rol r18
45 0024 3327 clr r19
46/*DEBUG: 0x13 4 44 */
47 0026 2295 swap r18
48 0028 3295 swap r19
49 002a 307F andi r19,0xf0
50 002c 3227 eor r19,r18
51 002e 207F andi r18,0xf0
52 0030 3227 eor r19,r18
53.LM3:
54/*DEBUG: 0x19 6 4 */
55 0032 C901 movw r24,r18
56/* epilogue start */
57/*DEBUG: 0x1a 1 4 */
58 0034 0895 ret
If I played around with mode of operands I'm sure I could do worse.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36884
[Bug middle-end/36884] New: ifcvt poor optimization
If conversion is causing extraordinary bad code for AVR.
This occurs on 4.3 4.4 is no better.
Testcase:
int z;
int foo2(void)
{
return ( ++z >= 0);
}
Conversion replaces if using store flag. However, there appears to be no
account of relative costs and rather bad interactions with mode narrowing and
widening
After conversion we have:
z++
Extract upper byte of 'z'
Sign extend
One complement
Shift right 15
RTL prior to ce1 pass:
;; Pred edge ENTRY [100.0%] (fallthru)
(note 3 0 2 2 [bb 2] NOTE_INSN_BASIC_BLOCK)
(note 2 3 5 2 NOTE_INSN_FUNCTION_BEG)
(insn 5 2 6 2 test1.c:15 (set (reg:HI 43 [ z ])
(mem/c/i:HI (symbol_ref:HI ("z") ) [3 z+0 S2
A8])) 10 {*movhi} (nil))
(insn 6 5 7 2 test1.c:15 (set (reg:HI 41 [ z.5 ])
(plus:HI (reg:HI 43 [ z ])
(const_int 1 [0x1]))) 25 {addhi3} (expr_list:REG_DEAD (reg:HI 43 [
z ])
(nil)))
(insn 7 6 8 2 test1.c:15 (set (mem/c/i:HI (symbol_ref:HI ("z") ) [3 z+0 S2 A8])
(reg:HI 41 [ z.5 ])) 10 {*movhi} (nil))
(insn 8 7 9 2 test1.c:15 (set (reg:HI 44)
(const_int 0 [0x0])) 10 {*movhi} (nil))
(insn 9 8 10 2 test1.c:15 (set (cc0)
(reg:HI 41 [ z.5 ])) 95 {tsthi} (expr_list:REG_DEAD (reg:HI 41 [ z.5 ])
(nil)))
(jump_insn 10 9 25 2 test1.c:15 (set (pc)
(if_then_else (lt (cc0)
(const_int 0 [0x0]))
(label_ref 12)
(pc))) 109 {branch} (expr_list:REG_BR_PROB (const_int 2100 [0x834])
(nil)))
;; End of basic block 2 -> ( 4 3)
;; lr out 28 [r28] 32 [__SP_L__] 34 [argL] 44
;; live out 28 [r28] 32 [__SP_L__] 34 [argL] 44
RTL after ce1:
(note 2 3 5 2 NOTE_INSN_FUNCTION_BEG)
(insn 5 2 6 2 test1.c:15 (set (reg:HI 43 [ z ])
(mem/c/i:HI (symbol_ref:HI ("z") ) [3 z+0 S2
A8])) 10 {*movhi} (nil))
(insn 6 5 7 2 test1.c:15 (set (reg:HI 41 [ z.5 ])
(plus:HI (reg:HI 43 [ z ])
(const_int 1 [0x1]))) 25 {addhi3} (expr_list:REG_DEAD (reg:HI 43 [
z ])
(nil)))
(insn 7 6 8 2 test1.c:15 (set (mem/c/i:HI (symbol_ref:HI ("z") ) [3 z+0 S2 A8])
(reg:HI 41 [ z.5 ])) 10 {*movhi} (nil))
(insn 8 7 9 2 test1.c:15 (set (reg:HI 44)
(const_int 0 [0x0])) 10 {*movhi} (nil))
(insn 9 8 35 2 test1.c:15 (set (cc0)
(reg:HI 41 [ z.5 ])) 95 {tsthi} (expr_list:REG_DEAD (reg:HI 41 [ z.5 ])
(nil)))
(insn 35 9 36 2 test1.c:15 (set (reg:QI 49)
(subreg:QI (reg:HI 41 [ z.5 ]) 1)) 4 {*movqi} (nil))
(insn 36 35 37 2 test1.c:15 (set (reg:HI 48)
(sign_extend:HI (reg:QI 49))) 84 {extendqihi2} (nil))
(insn 37 36 38 2 test1.c:15 (set (reg:HI 50)
(not:HI (reg:HI 48))) 82 {one_cmplhi2} (nil))
(insn 38 37 32 2 test1.c:15 (set (reg:HI 44)
(lshiftrt:HI (reg:HI 50)
(const_int 15 [0xf]))) 71 {lshrhi3} (nil))
(insn 32 38 33 2 test1.c:16 (set (reg:QI 24 r24)
(subreg:QI (reg:HI 44) 0)) 4 {*movqi} (nil))
(insn 33 32 23 2 test1.c:16 (set (reg:QI 25 r25 [+1 ])
(subreg:QI (reg:HI 44) 1)) 4 {*movqi} (expr_list:REG_DEAD (reg:HI 44)
(nil)))
Final (annotated) code:
22/* prologue: frame size=0 */
23/* prologue end (size=0) */
24.LM2:
25 8091 lds r24,z
26 0004 9091 lds r25,(z)+1
27 0008 0196 adiw r24,1
28 000a 9093 sts (z)+1,r25
29 000e 8093 sts z,r24
30 0012 892F mov r24,r25
31 0014 9927 clr r25 ;SIGN EXTEND
32 0016 87FD sbrc r24,7 ;ditto
33 0018 9095 com r25 ; ditto
34 001a 8095 com r24 ; ONE'SCOMPLEMENT
35 001c 9095 com r25 ; ditto
36.LM3:
37 001e 8827 clr r24 ;LSHIFT 15
38 0020 990F lsl r25 ; ditto
39 0022 881F rol r24 ; ditto
40 0024 9927 clr r25 ; ditto
41/* epilogue: frame size=0 */
42 0026 0895 ret
--
Summary: ifcvt poor optimization
Product: gcc
Version: 4.3.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36884
[Bug tree-optimization/36792] [4.4 Regression] Revision 137631 causes many failures
--- Comment #6 from hutchinsonandy at gcc dot gnu dot org 2008-07-20 13:57 --- Similarly Platform: avr-unknown-none 4.4.0 20080718 (experimental) [trunk revision 137967] FAIL: gcc.dg/tree-ssa/ssa-fre-13.c scan-tree-dump fre "Inserted .* &a" FAIL: gcc.dg/tree-ssa/ssa-fre-13.c scan-tree-dump fre "Replaced tmp1_.\\(D\\)->data" FAIL: gcc.dg/tree-ssa/ssa-fre-14.c scan-tree-dump fre "Inserted .* &a" FAIL: gcc.dg/tree-ssa/ssa-fre-14.c scan-tree-dump fre "Replaced tmp1.data" FAIL: gcc.dg/tree-ssa/ssa-fre-17.c scan-tree-dump fre "Replaced f.doms\\[0\\].dom with i_" FAIL: gcc.dg/tree-ssa/ssa-fre-7.c scan-tree-dump-times fre "Inserted pretmp" 1 FAIL: gcc.dg/tree-ssa/ssa-fre-7.c scan-tree-dump-times fre "Replaced a.u.f with pretmp" 3 FAIL: gcc.dg/tree-ssa/ssa-fre-7.c scan-tree-dump-times fre "Replaced a.u.k with j" 1 FAIL: gcc.dg/tree-ssa/ssa-fre-7.c scan-tree-dump fre "= VIEW_CONVERT_EXPR\\(j_" FAIL: gcc.dg/tree-ssa/ssa-fre-7.c scan-tree-dump optimized "return j" FAIL: gcc.dg/tree-ssa/ssa-fre-8.c scan-tree-dump-times fre "Replaced u.f with pretmp" 2 FAIL: gcc.dg/tree-ssa/ssa-fre-8.c scan-tree-dump-times fre "Inserted pretmp" 2 FAIL: gcc.dg/tree-ssa/ssa-fre-9.c scan-tree-dump-times fre "Eliminated: 1" 2 FAIL: gcc.dg/tree-ssa/ssa-fre-9.c scan-tree-dump-times fre "Insertions: 1" 2 FAIL: gcc.dg/tree-ssa/ssa-pre-15.c scan-tree-dump optimized "= 0;" -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added ------------ CC||hutchinsonandy at gcc dot ||gnu dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36792
[Bug testsuite/36285] gcc.dg/compat/struct-by-value-xxx improper test for AVR target
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2008-07-02 22:32 --- Subject: Bug 36285 Author: hutchinsonandy Date: Wed Jul 2 22:31:11 2008 New Revision: 137396 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=137396 Log: PR testsuite/36285 * gcc.dg/compat/struct-by-value-16_x.c: Move some tests to new file to reduce code size. * gcc.dg/compat/struct-by-value-16_y.c: Ditto. * gcc.dg/compat/struct-by-value-16a_main.c: New file. * gcc.dg/compat/struct-by-value-16a_x.c: Ditto. * gcc.dg/compat/struct-by-value-16a_y.c: Ditto. * gcc.dg/compat/struct-by-value-17_x.c: Move some tests to new file to reduce code size. * gcc.dg/compat/struct-by-value-17_y.c: Ditto. * gcc.dg/compat/struct-by-value-17a_main.c: New file. * gcc.dg/compat/struct-by-value-17a_x.c: Ditto. * gcc.dg/compat/struct-by-value-17a_y.c: Ditto. * gcc.dg/compat/struct-by-value-18_x.c: Move some tests to new file to reduce code size. * gcc.dg/compat/struct-by-value-18_y.c: Ditto. * gcc.dg/compat/struct-by-value-18a_main.c: New file. * gcc.dg/compat/struct-by-value-18a_x.c: Ditto. * gcc.dg/compat/struct-by-value-18a_y.c: Ditto. Added: trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-16a_main.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-16a_x.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-16a_y.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-17a_main.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-17a_x.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-17a_y.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-18a_main.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-18a_x.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-18a_y.c Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-16_x.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-16_y.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-17_x.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-17_y.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-18_x.c trunk/gcc/testsuite/gcc.dg/compat/struct-by-value-18_y.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36285
[Bug middle-end/36598] Failed optimisation of return of struct argment in memcpy-1.c
--- Comment #5 from hutchinsonandy at gcc dot gnu dot org 2008-06-30 22:39 --- Subject: Bug 36598 Author: hutchinsonandy Date: Mon Jun 30 22:38:34 2008 New Revision: 137298 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=137298 Log: PR target/36598 * gcc.dg/memcpy-1.c: Mark test XFAIL for avr target. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/memcpy-1.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36598
[Bug target/36609] New: AVR wrong code using incorrect RTL for test reversal pattern
I found problem with AVR test instructions such used by c/c++ code (x>0)
and demonstrated by testsuite test gcc.c-torture/execute/pr22493-1.c
#include
extern void abort ();
extern void exit (int);
void f(int i)
{
if (i>0)
abort();
i = -i;
if (i<0)
return;
abort ();
}
int main(int argc, char *argv[])
{
f(INT_MIN);
exit (0);
}
Compiling this with:
avr-gcc -c -mmcu=atmega128 -I. -w -Os -fwrapv -mmcu=atmega128 -Wall
-Wstrict-prototypes -Wa,-ahlms=test.s -std=gnu99 test.c -o test.o
produces wrong code.
An integer test x>0 is translated into RTL:
CC0=Rx
from which we create AVR assembler code
TST Rx
which is equivalent to CMP Rx,0
AVR target also have negated_tst RTL patterns to reverse tests in avr_reorg so
that we can match the available branches on AVR. For example AVR has BRLT and
BRGE but not BRLE and BRGT.
So
CC0 = Rx
Jump LE
becomes reorganized as
CC0 = NEG:xx(Rx)
Jump GE
which emits the following code via define_insn "*negated_tst.
CMP 0,Rx
BRGE
For such test reversals the final code is correct. However, the intermediate
RTL used by translation is WRONG
CC0=Rx
Jump LE
is NOT equivalent to
CC0 = NEG:xx(Rx)
Jump GE
For test reversal the problem is hidden - 2 wrongs making a right in this case.
BUT the bug appears when we have a normal code sequence such as
NEG:xx(Rx)
CC0 = Rx
which the combine phase will match against the RTL of *negated_tst and result
in assembler code of:
CMP 0,Rx
which is incorrect. Consider the case where Rx= -32768
NEG:HI(-32768) => -32768
CC0 = -32768 => Negative correct!
but the replacment:
CMP 0,-32768 =>Positive - WRONG
The fix is to redefine the "*negated_tst" RTL patterns to use correct COMPARES
and have avr_reorg to emit the corrected RTL.
(define_insn "*reversed_tsthi"
[(set (cc0)
(compare (const_int 0)
(match_operand:HI 0 "register_operand" "r")))
Test reversals will be unaffected. But now combine won't find any bad RTL.
I have not studied to see if the above RTL is ever generated. If it is, it is
perfectly safe.
--
Summary: AVR wrong code using incorrect RTL for test reversal
pattern
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC build triplet: i686-pc-cygwin
GCC host triplet: i686-pc-cygwin
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36609
[Bug c/36598] Failed optimisation of return of struct argment in memcpy-1.c
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2008-06-22 16:11 --- Quite possibly due to cost metrics. They are far from ideal. Will mark test XFAIL until we can investigate and fix. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added CC||eric dot weddington at atmel ||dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36598
[Bug c/36598] Failed optimisation of return of struct argment in memcpy-1.c
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-06-22 14:33 --- Thanks for information --param sra-max-structure-size=32 does indeed remove test failure and produces optimal code. But changing the testcase does not remove the optimization problem - unless sra-max-structure-size was always used. So there is problem somewhere else to fix. See also: http://gcc.gnu.org/ml/gcc-patches/2007-12/msg01144.html Andy -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36598
[Bug c/36598] New: Failed optimisation of return of struct argment in memcpy-1.c
Testcase gcc.dg/memcpy-1.c fails on AVR target.
Failure occurs because the return value is not simplified to avoid memcpy. This
test works on i686 and I can't see why same optimization should not apply to
AVR
Test is:
/* { dg-do compile } */
/* { dg-options "-O2 -fdump-tree-optimized" } */
/* { dg-final { scan-tree-dump-times "nasty_local" 0 "optimized" } } */
/* { dg-final { cleanup-tree-dump "optimized" } } */
struct a {int a,b,c;} a;
int test(struct a a)
{
struct a nasty_local;
__builtin_memcpy (&nasty_local,&a, sizeof(a));
return nasty_local.a;
}
On i686 we get:
;; Function test (test)
Analyzing Edge Insertions.
test (struct a a)
{
:
return a.a;
}
BUT on AVR we get:
;; Function test (test)
Analyzing Edge Insertions.
test (struct a a)
{
struct a nasty_local;
:
nasty_local = a;
return nasty_local.a;
}
I have confirmed the final AVR code is suboptimal.
--
Summary: Failed optimisation of return of struct argment in
memcpy-1.c
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC build triplet: i686-pc-linux-gnu
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36598
[Bug testsuite/36284] gcc.dg-struct-layout fails AVR target - multiple reasons
--- Comment #3 from hutchinsonandy at gcc dot gnu dot org 2008-06-19 02:16 --- Testcase fixed for 4.4 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36284
[Bug middle-end/36571] Default untyped return for AVR is byte register.
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2008-06-18 23:00 --- Subject: Bug 36571 Author: hutchinsonandy Date: Wed Jun 18 22:59:41 2008 New Revision: 136926 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136926 Log: PR c/36571 * gcc.dg/builtin-return-1.c: Mark as xfail on AVR. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/builtin-return-1.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36571
[Bug c/36571] New: Default untyped return for AVR is byte register.
Test testsuite/gcc.dg/builtin-return-1.c will fail for avr target. Default implementation of builtin-return-1 is untyped return of a single register. Which is R25 and only 1 byte. This is not int as test expects. So high byte may and does get overwritten before outer return. The test may be invalid, but this is also reminder that AVR target does not have a complete definition for untyped return. The default is pretty useless. Test with be marked XFAIL. -- Summary: Default untyped return for AVR is byte register. Product: gcc Version: 4.4.0 Status: UNCONFIRMED Severity: minor Priority: P3 Component: c AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: hutchinsonandy at gcc dot gnu dot org GCC build triplet: i686-pc-linux-gnu GCC host triplet: i686-pc-linux-gnu GCC target triplet: avr-unknown-nne http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36571
[Bug testsuite/36284] gcc.dg-struct-layout fails AVR target - multiple reasons
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-06-18 21:54 --- Subject: Bug 36284 Author: hutchinsonandy Date: Wed Jun 18 21:53:20 2008 New Revision: 136920 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136920 Log: PR testsuite/36284 * gcc.dg/compat/struct-layout-1_generate.c (switchfiles): Require int32plus for generated main files. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/compat/struct-layout-1_generate.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36284
[Bug c++/36460] No space between >'s not always handled in C++0x
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-06-18 21:49 --- Subject: Bug 36460 Author: hutchinsonandy Date: Wed Jun 18 21:48:50 2008 New Revision: 136919 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136919 Log: PR c/36460 * gcc.dg/utf32-1.c: Mark xfail for AVR target. * gcc.dg/utf32-2.c: Ditto. * gcc.dg/utf32-3.c: Ditto. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/utf32-1.c trunk/gcc/testsuite/gcc.dg/utf32-2.c trunk/gcc/testsuite/gcc.dg/utf32-3.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36460
[Bug tree-optimization/36561] New: store using long array index not hoisted out of loop
Testsuite test gcc.dg/tree-ssa/loop35.c fails for avr target for test3().
This test uses long array index. Tests with int or char index get hoisted as
expected.
void test3(unsigned long b)
{
unsigned i;
/* And here. */
for (i = 0; i < 100; i++)
{
arr[b+8].X += i;
arr[b+9].X += i;
}
}
Richard Guenther indicates:
> This is because the alias-oracle for store-motion doesn't handle conversions
> to sizetype well in the offset disambiguation.
--
Summary: store using long array index not hoisted out of loop
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Severity: minor
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC target triplet: avr-unkown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36561
[Bug target/36336] ICE push_reload - psuedo reg_equiv_constant
--- Comment #8 from hutchinsonandy at gcc dot gnu dot org 2008-06-15 18:36 --- Fixed 4.4 and 4.3 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36336
[Bug target/36336] ICE push_reload - psuedo reg_equiv_constant
--- Comment #7 from hutchinsonandy at gcc dot gnu dot org 2008-06-15 18:35 --- Subject: Bug 36336 Author: hutchinsonandy Date: Sun Jun 15 18:34:45 2008 New Revision: 136819 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136819 Log: PR target/36336 * config/avr/avr.h (LEGITIMIZE_RELOAD_ADDRESS): Add check for reg_equiv_constant. Modified: branches/gcc-4_3-branch/gcc/ChangeLog branches/gcc-4_3-branch/gcc/config/avr/avr.h -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36336
[Bug target/36336] ICE push_reload - psuedo reg_equiv_constant
--- Comment #6 from hutchinsonandy at gcc dot gnu dot org 2008-06-15 18:25 --- Subject: Bug 36336 Author: hutchinsonandy Date: Sun Jun 15 18:24:27 2008 New Revision: 136818 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136818 Log: PR target/36336 * config/avr/avr.h (LEGITIMIZE_RELOAD_ADDRESS): Add check for reg_equiv_constant. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr.h -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36336
[Bug target/36494] Char arrays gets corrupted in avr programs.
--- Comment #1 from hutchinsonandy at gcc dot gnu dot org 2008-06-11 02:35 --- Please post command line used to compile and assembler produced. I cannot reproduce this on 4.4 or 4.2.2. At high optimisation (-Os) it compiles to load PORTC with 3 directly. (as one would expect): 13/* prologue: frame size=0 */ 14/* prologue end (size=0) */ 15 8FE0 ldi r24,lo8(15) 16 0002 87B9 out 39-0x20,r24 17 0004 83E0 ldi r24,lo8(3) 18 0006 90E0 ldi r25,hi8(3) 19 0008 88B9 out 40-0x20,r24 20 000a 80E0 ldi r24,lo8(0) 21 000c 90E0 ldi r25,hi8(0) 22/* epilogue: frame size=0 */ 23 000e 0895 ret -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added CC||eric dot weddington at atmel ||dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36494
[Bug c/36470] sizeof UTF-32 is 2 on AVR
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-06-10 03:21 --- Fixing the testcase is easy, That is how i found problem. Two wrongs make a right! It is not a priority but it seems like the suggestions is right way to fix. I'll mark as XFAIL sometime soon. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36470
[Bug testsuite/36285] gcc.dg/compat/struct-by-value-xxx improper test for AVR target
--- Comment #2 from hutchinsonandy at gcc dot gnu dot org 2008-06-09 22:49 --- Subject: Bug 36285 Author: hutchinsonandy Date: Mon Jun 9 22:48:37 2008 New Revision: 136604 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136604 Log: PR testsuite/36285 * gcc.dg/compat/struct-return-10_y.c: Remove test_va. * gcc.dg/compat/vector-defs.h: Add check for availability of 8 byte double. * gcc.dg/compat/struct-layout-1_test.h: Change bitfield to match integer size. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.dg/compat/struct-layout-1_test.h trunk/gcc/testsuite/gcc.dg/compat/struct-return-10_y.c trunk/gcc/testsuite/gcc.dg/compat/vector-defs.h -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36285
[Bug middle-end/36447] [4.4 Regression] simplify_subreg ICE with right shift more than length type AVR
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2008-06-09 22:39 --- Subject: Bug 36447 Author: hutchinsonandy Date: Mon Jun 9 22:38:34 2008 New Revision: 136602 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136602 Log: PR middle-end/36447 * simplify-rtx.c (simplify_subreg): Add check for shift count greater than size. Modified: trunk/gcc/ChangeLog trunk/gcc/simplify-rtx.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36447
[Bug c/36470] New: sizeof UTF-32 is 2 on AVR
AVR fails gcc.dg/utf32-1.c execution test.
As AVR has 16 bit int, I tried fixing this by using unsigned long for char_32.
This failed, I was surprised to find this is apparently due to incorrect size
of UTF-32 characters so that:
sizeof (U'\0')
sizeof (U'\u2029')
sizeof (U'\U00064321')
all give value of 2.
So it would seem size has been set as sizeof(int) perhaps?
I can assign 32 bit values ok - just that sizeof says they are only 16 bits.
The modified testcase using 32 bit char_32t compiles to abort(), so I figure
its not a target issue.
/** Change this to unsigned long for AVR/
typedef unsigned int char32_t;
extern void abort (void);
char32_tc0 = U'a';
char32_tc1 = U'\0';
char32_tc2 = U'\u0024';
char32_tc3 = U'\u2029';
char32_tc4 = U'\U00064321';
#define A 0x0061
#define D 0x0024
#define X 0x2029
#define Y 0x00064321
int main ()
{
if (sizeof (U'a') != sizeof (char32_t))
abort ();
if (sizeof (U'\0') != sizeof (char32_t))
abort ();
if (sizeof (U'\u0024') != sizeof (char32_t))
abort ();
if (sizeof (U'\u2029') != sizeof (char32_t))
abort ();
if (sizeof (U'\U00064321') != sizeof (char32_t))
abort ();
if (c0 != A)
abort ();
if (c1 != 0x)
abort ();
if (c2 != D)
abort ();
if (c3 != X)
abort ();
if (c4 != Y)
abort ();
}
--
Summary: sizeof UTF-32 is 2 on AVR
Product: gcc
Version: 4.4.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hutchinsonandy at gcc dot gnu dot org
GCC host triplet: i686-pc-cygwin
GCC target triplet: avr-unknown-none
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36470
[Bug target/36467] [avr] Missed optimization with pointer arithmetic and mul*
--- Comment #4 from hutchinsonandy at gcc dot gnu dot org 2008-06-08 18:20 --- It makes sense in one respect We don't have fast shift by 4 bits and code defaults to loop for Os. Seems we should be selective as MUL is indeed shorter. Though I think gcc may be confused by our poor cost data and perhaps was alsp mislead into using shift instead of MUL. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |NEW Ever Confirmed|0 |1 Last reconfirmed|-00-00 00:00:00 |2008-06-08 18:20:52 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36467
[Bug target/30243] [4.1/4.2/4.3/4.4 Regression][avr] signbit() causes an internal compiler error
--- Comment #12 from hutchinsonandy at gcc dot gnu dot org 2008-06-07 15:49 --- Subject: Bug 30243 Author: hutchinsonandy Date: Sat Jun 7 15:48:25 2008 New Revision: 136531 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136531 Log: Backports from 4.4 PR target/27386 * config/avr/avr.h: (PUSH_ROUNDING): Remove. PR target/30243 * builtins.c (expand_builtin_signbit): Don't take lowpart when register is already smaller or equal to required mode. PR target/34932 * config/avr/avr.md (*addhi3_zero_extend2): Remove. * config/avr/avr.h (MAX_OFILE_ALIGNMENT): Define. Modified: branches/gcc-4_3-branch/gcc/ChangeLog branches/gcc-4_3-branch/gcc/builtins.c branches/gcc-4_3-branch/gcc/config/avr/avr.h branches/gcc-4_3-branch/gcc/config/avr/avr.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30243
[Bug target/34932] [avr] ICE in reload
--- Comment #7 from hutchinsonandy at gcc dot gnu dot org 2008-06-07 15:49 --- Subject: Bug 34932 Author: hutchinsonandy Date: Sat Jun 7 15:48:25 2008 New Revision: 136531 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136531 Log: Backports from 4.4 PR target/27386 * config/avr/avr.h: (PUSH_ROUNDING): Remove. PR target/30243 * builtins.c (expand_builtin_signbit): Don't take lowpart when register is already smaller or equal to required mode. PR target/34932 * config/avr/avr.md (*addhi3_zero_extend2): Remove. * config/avr/avr.h (MAX_OFILE_ALIGNMENT): Define. Modified: branches/gcc-4_3-branch/gcc/ChangeLog branches/gcc-4_3-branch/gcc/builtins.c branches/gcc-4_3-branch/gcc/config/avr/avr.h branches/gcc-4_3-branch/gcc/config/avr/avr.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34932
[Bug target/27386] AVR: wrong code generated when passing three uint64_t arguments to function
--- Comment #17 from hutchinsonandy at gcc dot gnu dot org 2008-06-07 15:49 --- Subject: Bug 27386 Author: hutchinsonandy Date: Sat Jun 7 15:48:25 2008 New Revision: 136531 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136531 Log: Backports from 4.4 PR target/27386 * config/avr/avr.h: (PUSH_ROUNDING): Remove. PR target/30243 * builtins.c (expand_builtin_signbit): Don't take lowpart when register is already smaller or equal to required mode. PR target/34932 * config/avr/avr.md (*addhi3_zero_extend2): Remove. * config/avr/avr.h (MAX_OFILE_ALIGNMENT): Define. Modified: branches/gcc-4_3-branch/gcc/ChangeLog branches/gcc-4_3-branch/gcc/builtins.c branches/gcc-4_3-branch/gcc/config/avr/avr.h branches/gcc-4_3-branch/gcc/config/avr/avr.md -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27386
[Bug target/27386] AVR: wrong code generated when passing three uint64_t arguments to function
--- Comment #16 from hutchinsonandy at gcc dot gnu dot org 2008-06-04 22:06 --- Fixed 4.4 -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED Target Milestone|--- |4.3.2 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27386
[Bug target/30243] [4.1/4.2/4.3/4.4 Regression][avr] signbit() causes an internal compiler error
--- Comment #11 from hutchinsonandy at gcc dot gnu dot org 2008-06-04 22:05 --- Fixed 4.4 Back port to 4.3.2 when its open. -- hutchinsonandy at gcc dot gnu dot org changed: What|Removed |Added Status|NEW |RESOLVED Resolution||FIXED Target Milestone|4.2.5 |4.3.2 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30243
[Bug target/27386] AVR: wrong code generated when passing three uint64_t arguments to function
--- Comment #15 from hutchinsonandy at gcc dot gnu dot org 2008-06-04 22:03 --- Subject: Bug 27386 Author: hutchinsonandy Date: Wed Jun 4 22:02:57 2008 New Revision: 136377 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136377 Log: PR target/27386 * config/avr/avr.h: (PUSH_ROUNDING): Remove. Modified: trunk/gcc/ChangeLog trunk/gcc/config/avr/avr.h -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27386
[Bug target/30243] [4.1/4.2/4.3/4.4 Regression][avr] signbit() causes an internal compiler error
--- Comment #10 from hutchinsonandy at gcc dot gnu dot org 2008-06-04 22:00 --- Subject: Bug 30243 Author: hutchinsonandy Date: Wed Jun 4 21:59:54 2008 New Revision: 136376 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=136376 Log: PR target/30243 * builtins.c (expand_builtin_signbit): Don't take lowpart when register is already smaller or equal to required mode. Modified: trunk/gcc/ChangeLog trunk/gcc/builtins.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30243
