[Bug target/20331] [3.4/4.0/4.1 Regression] Wrong code generation for the argument of the pure function in PIC

2005-03-09 Thread kkojima at gcc dot gnu dot org

--- Additional Comments From kkojima at gcc dot gnu dot org  2005-03-09 
23:58 ---
I'm taking a look at the uses of MEM_READONLY_P, though I'm not
sure which one has the similar issue.  It looks that modified_in_p
has the same problem with modified_between_p.
BTW, it's better to change the component of this PR, isn't it?
rtl-optimization?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20331


[Bug target/20331] [3.4/4.0/4.1 Regression] Wrong code generation for the argument of the pure function in PIC

2005-03-08 Thread amylaar at gcc dot gnu dot org

--- Additional Comments From amylaar at gcc dot gnu dot org  2005-03-08 
21:58 ---
(In reply to comment #1)
 I've traced what's going on:
   http://gcc.gnu.org/ml/gcc-patches/2005-03/msg00525.html
 It includes a patch for comment.
 

The sh64 port is partcularily exposed in emit_libcall_block because it uses
paradoxical subregs to set up the offset part of the PIC address, but I
think your patch makes sense.  Except that this is only the tip of the iceberg.
The problem was introduced in a maga-patch to remove RTX_UNCHANGING_P; there
are a number of other places that need to be fixed up similarily.

2004-08-18  Richard Henderson  [EMAIL PROTECTED]

* rtl.h (MEM_READONLY_P): Replace RTX_UNCHANGING_P.
* alias.c (true_dependence): Update to match new semantics.
(canon_true_dependence, write_dependence_p): Likewise.
(anti_dependence, output_dependence): Update write_dependence_p args.
(unchanging_anti_dependence): Remove.
* calls.c (purge_mem_unchanging_flag): Remove.
(fixup_tail_calls): Don't call it.
(expand_call): Don't add unchanging memory to function usage.
* expr.c (emit_block_move_via_libcall): Likewise.
(clear_storage_via_libcall): Don't clobber RTX_UNCHANGING_P mems.
(get_subtarget): Don't use RTX_UNCHANGING_P.
(expand_assignment, store_constructor, expand_expr_real_1): Likewise.
(do_tablejump): Set MEM_READONLY_P, not RTX_UNCHANGING_P.
* combine.c (get_last_value_validate): Use MEM_READONLY_P.
* cse.c (insert): Don't use RTX_UNCHANGING_P.
(cse_insn, canon_hash): Use MEM_READONLY_P.
* emit-rtl.c (set_mem_attributes_minus_bitpos): Use MEM_READONLY_P
instead of RTX_UNCHANGING_P.
* explow.c (maybe_set_unchanging): Remove.
* expr.h (maybe_set_unchanging): Remove.
* flow.c (insn_dead_p, mark_used_regs): Use anti_dependence.
* function.c (assign_stack_temp_for_type): Don't use RTX_UNCHANGING_P.
(assign_parm_setup_reg, expand_function_start): Likewise.
* integrate.c (copy_rtx_and_substitute): Likewise.
* ra-rewrite.c (emit_colors): Likewise.
* regmove.c (copy_src_to_dest, regmove_optimize): Likewise.
(fixup_match_1): Likewise.
* reload1.c (reload, alter_reg): Likewise.
* local-alloc.c (validate_equiv_mem): Check MEM_READONLY_P,
not RTX_UNCHANGING_P.
(equiv_init_varies_p): Likewise.
* loop-invariant.c (check_maybe_invariant): Likewise.
* resource.c (mark_referenced_resources, mark_set_resources): Likewise.
* loop.c (note_addr_stored): Likewise.
(prescan_loop): Likewise. Don't check function usage for clobbered
unchanging memory.
* rtlanal.c (rtx_unstable_p): Check MEM_READONLY_P,
not RTX_UNCHANGING_P.
(rtx_varies_p, modified_between_p, modified_in_p): Likewise.
* varasm.c (force_const_mem): Likewise.
* stmt.c (expand_decl): Don't set RTX_UNCHANGING_P.
* web.c (entry_register): Likewise.
* tree-gimple.h (get_base_address): Move decl ...
* tree.h: ... here.
* doc/rtl.texi (MEM_READONLY_P): Replace RTX_UNCHANGING_P.

* config/alpha/alpha.c (alpha_set_memflags_1): Rewrite to be
called via for_each_rtx.  Copy MEM_SCALAR_P, MEM_NOTRAP_P too.
(alpha_set_memflags): Update to match.

* config/darwin.c (machopic_indirect_data_reference): Set
MEM_READONLY_P instead of RTX_UNCHANGING_P.
(machopic_indirect_call_target): Likewise.
(machopic_legitimize_pic_address): Likewise.
* config/arm/arm.c (legitimize_pic_address, arm_gen_load_multiple,
arm_gen_store_multiple, arm_gen_movmemqi): Likewise.
* config/arm/arm.md (load_multiple, store_multiple): Likewise.
* config/frv/frv.md (symGOT2reg): Likewise.
* config/i386/i386.c (legitimize_pic_address,
legitimize_tls_address, ix86_split_to_parts): Likewise.
* config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
* config/ia64/ia64.md (load_fptr): Likewise.
* config/m32r/m32r.c (m32r_legitimize_pic_address): Likewise.
* config/m68k/m68k.c (legitimize_pic_address): Likewise.
* config/mcore/mcore.c (block_move_sequence): Likewise.
* config/mn10300/mn10300.md (symGOT2reg): Likewise.
* config/pa/pa.c (legitimize_pic_address): Likewise.
* config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
(rs6000_emit_move): Likewise.
* config/s390/s390.c (legitimize_pic_address): Likewise.
(legitimize_tls_address): Likewise.
* config/s390/s390.md (casesi): Likewise.
* config/sh/sh.c (prepare_move_operands, sh_reorg): Likewise.
* config/sh/sh.md (symGOT2reg): Likewise.
* config/sparc/sparc.c (legitimize_pic_address): Likewise.
* config/v850/v850.md (casesi): Likewise.

* config/ia64/ia64.c 

[Bug target/20331] [3.4/4.0/4.1 Regression] Wrong code generation for the argument of the pure function in PIC

2005-03-06 Thread kkojima at gcc dot gnu dot org

--- Additional Comments From kkojima at gcc dot gnu dot org  2005-03-07 
00:20 ---
I've traced what's going on:
  http://gcc.gnu.org/ml/gcc-patches/2005-03/msg00525.html
It includes a patch for comment.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20331


[Bug target/20331] [3.4/4.0/4.1 Regression] Wrong code generation for the argument of the pure function in PIC

2005-03-05 Thread giovannibajo at libero dot it


-- 
   What|Removed |Added

Summary|[3.4/4.0/41 Regression] |[3.4/4.0/4.1 Regression]
   |Wrong code generation for   |Wrong code generation for
   |the argument of the pure|the argument of the pure
   |function in PIC |function in PIC


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20331


[Bug target/20331] [3.4/4.0/4.1 Regression] Wrong code generation for the argument of the pure function in PIC

2005-03-05 Thread pinskia at gcc dot gnu dot org


-- 
   What|Removed |Added

   Target Milestone|--- |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20331