libfirm
15 years ago- more simplification in skip_Tuple
Michael Beck [Thu, 18 Dec 2008 17:38:10 +0000 (17:38 +0000)]
- more simplification in skip_Tuple

[r24791]

15 years ago- always skip tuple, its useless to prevent that
Michael Beck [Thu, 18 Dec 2008 17:36:54 +0000 (17:36 +0000)]
- always skip tuple, its useless to prevent that

[r24790]

15 years ago- implemented Builtin node to represent builtins ...
Michael Beck [Thu, 18 Dec 2008 15:31:11 +0000 (15:31 +0000)]
- implemented Builtin node to represent builtins ...

[r24785]

15 years ago- removed useless end -> node dependency
Michael Beck [Wed, 17 Dec 2008 15:08:46 +0000 (15:08 +0000)]
- removed useless end -> node dependency

[r24754]

15 years agoremove duplicated macros
Matthias Braun [Mon, 15 Dec 2008 13:02:28 +0000 (13:02 +0000)]
remove duplicated macros

[r24687]

15 years ago- renamed cons_flags to ir_cons_flags
Michael Beck [Mon, 15 Dec 2008 12:20:32 +0000 (12:20 +0000)]
- renamed cons_flags to ir_cons_flags

[r24681]

15 years ago- fixed warning
Michael Beck [Mon, 15 Dec 2008 12:18:50 +0000 (12:18 +0000)]
- fixed warning

[r24680]

15 years agoAdded cons_flags parameter to new_Load and new_Store functions for volatility, alignm...
Moritz Kroll [Sun, 14 Dec 2008 10:52:07 +0000 (10:52 +0000)]
Added cons_flags parameter to new_Load and new_Store functions for volatility, alignment and pin state

[r24655]

15 years agoFix generating bit representation of INF modes with explicit one, i.e. x87 long double.
Christoph Mallon [Sat, 13 Dec 2008 23:15:05 +0000 (23:15 +0000)]
Fix generating bit representation of INF modes with explicit one, i.e. x87 long double.

[r24637]

15 years ago- BugFix: the mode of a Load must be its load_mode, else a Byte load would be congrue...
Michael Beck [Sat, 13 Dec 2008 22:46:17 +0000 (22:46 +0000)]
- BugFix: the mode of a Load must be its load_mode, else a Byte load would be congruent a float load (which is true to some extend but not we want to handle here)
- BugFix: two Blocks are never congruent: this cannot be detected by combo either and fixed the IJmp predecessor problem of combo_load_mode.c

[r24633]

15 years agoFix tarval_mod() for negative dividend.
Christoph Mallon [Sat, 13 Dec 2008 22:20:39 +0000 (22:20 +0000)]
Fix tarval_mod() for negative dividend.

[r24632]

15 years agooops set_entity_align did not what it appeared to do
Matthias Braun [Sat, 13 Dec 2008 22:04:02 +0000 (22:04 +0000)]
oops set_entity_align did not what it appeared to do

[r24630]

15 years ago- C99 feature removed
Michael Beck [Sat, 13 Dec 2008 21:38:14 +0000 (21:38 +0000)]
- C99 feature removed
- fixed an FIXME

[r24629]

15 years agoIgnore End keeps.
Sebastian Buchwald [Sat, 13 Dec 2008 20:45:12 +0000 (20:45 +0000)]
Ignore End keeps.

[r24628]

15 years agoalign mode_E konstants at 16 bytes
Matthias Braun [Sat, 13 Dec 2008 19:06:54 +0000 (19:06 +0000)]
align mode_E konstants at 16 bytes

[r24619]

15 years agoredefine mode_E to 96 bits
Matthias Braun [Sat, 13 Dec 2008 18:59:51 +0000 (18:59 +0000)]
redefine mode_E to 96 bits

[r24618]

15 years agosome adjustment to float convs (skip some strictconvs when no_fold_const is not set...
Matthias Braun [Sat, 13 Dec 2008 18:23:51 +0000 (18:23 +0000)]
some adjustment to float convs (skip some strictconvs when no_fold_const is not set, but always strict conv in int->float converstion)

[r24617]

15 years agofix new_Block users
Matthias Braun [Sat, 13 Dec 2008 18:08:33 +0000 (18:08 +0000)]
fix new_Block users

[r24615]

15 years agonew_Block should not set curr_block too
Matthias Braun [Sat, 13 Dec 2008 17:58:30 +0000 (17:58 +0000)]
new_Block should not set curr_block too

[r24610]

15 years agointroduce fp_no_float_fold flag for frontends which do not query the backends float...
Matthias Braun [Sat, 13 Dec 2008 17:56:13 +0000 (17:56 +0000)]
introduce fp_no_float_fold flag for frontends which do not query the backends float arithmetic mode yet

[r24609]

15 years agobackend can report back float point arithmetic mode
Matthias Braun [Sat, 13 Dec 2008 17:11:45 +0000 (17:11 +0000)]
backend can report back float point arithmetic mode

[r24603]

15 years agoNext attempt in fixing tarval float->int conversion.
Christoph Mallon [Sat, 13 Dec 2008 16:58:29 +0000 (16:58 +0000)]
Next attempt in fixing tarval float->int conversion.

[r24602]

15 years agoIgnore End keeps.
Christoph Mallon [Sat, 13 Dec 2008 16:03:42 +0000 (16:03 +0000)]
Ignore End keeps.

[r24600]

15 years agofix fc_can_lossless_conv_to (I hope this is correct)
Matthias Braun [Sat, 13 Dec 2008 15:48:54 +0000 (15:48 +0000)]
fix fc_can_lossless_conv_to (I hope this is correct)

[r24599]

15 years agono point in first setting the wrong mode and then correcting it, set the right mode...
Matthias Braun [Sat, 13 Dec 2008 14:52:12 +0000 (14:52 +0000)]
no point in first setting the wrong mode and then correcting it, set the right mode directly

[r24598]

15 years agoClean up new_const_code_irg() a bit.
Christoph Mallon [Sat, 13 Dec 2008 07:07:57 +0000 (07:07 +0000)]
Clean up new_const_code_irg() a bit.

[r24585]

15 years ago- skip_Id nodes when returning the entity initializer: this should fix ldst_opt_id.c
Michael Beck [Fri, 12 Dec 2008 20:31:19 +0000 (20:31 +0000)]
- skip_Id nodes when returning the entity initializer: this should fix ldst_opt_id.c

[r24582]

15 years agoput bison and flex sources into the package
Matthias Braun [Fri, 12 Dec 2008 19:15:29 +0000 (19:15 +0000)]
put bison and flex sources into the package

[r24581]

15 years ago- do not remove keeps for now, only Bads, NoMem and doublets
Michael Beck [Fri, 12 Dec 2008 18:29:26 +0000 (18:29 +0000)]
- do not remove keeps for now, only Bads, NoMem and doublets

[r24579]

15 years ago- remove NoMem from keep list
Michael Beck [Fri, 12 Dec 2008 18:28:32 +0000 (18:28 +0000)]
- remove NoMem from keep list

[r24578]

15 years ago- Changed wrong (and slow) implementation of fix_atsrt_block(),
Michael Beck [Fri, 12 Dec 2008 18:27:12 +0000 (18:27 +0000)]
- Changed wrong (and slow) implementation of fix_atsrt_block(),
  the old version fails because nodes from the startblock can be kept now ...

[r24577]

15 years ago- Do not remove keep-alives during optimize_graph_df(), kill only Bads and doublets
Michael Beck [Fri, 12 Dec 2008 15:20:40 +0000 (15:20 +0000)]
- Do not remove keep-alives during optimize_graph_df(), kill only Bads and doublets

[r24571]

15 years ago- implemented remove_End_Bads_and_doublets()
Michael Beck [Fri, 12 Dec 2008 15:19:11 +0000 (15:19 +0000)]
- implemented remove_End_Bads_and_doublets()

[r24570]

15 years agodon't have new_immBlock set the current block
Matthias Braun [Fri, 12 Dec 2008 14:46:36 +0000 (14:46 +0000)]
don't have new_immBlock set the current block

[r24566]

15 years agoUpdate NEWS.
Christoph Mallon [Thu, 11 Dec 2008 15:00:58 +0000 (15:00 +0000)]
Update NEWS.

[r24533]

15 years agokeep the API consistent with other enable functions
Matthias Braun [Thu, 11 Dec 2008 11:57:55 +0000 (11:57 +0000)]
keep the API consistent with other enable functions

[r24520]

15 years agoless includes
Matthias Braun [Wed, 10 Dec 2008 21:23:04 +0000 (21:23 +0000)]
less includes

[r24506]

15 years agoadd missing includes
Matthias Braun [Wed, 10 Dec 2008 20:41:30 +0000 (20:41 +0000)]
add missing includes

[r24505]

15 years agoavoid some unnecessary includes in public headers
Matthias Braun [Wed, 10 Dec 2008 20:39:24 +0000 (20:39 +0000)]
avoid some unnecessary includes in public headers

[r24504]

15 years ago- changed handling of block walks: we might reach a block through a kept node only ...
Michael Beck [Wed, 10 Dec 2008 19:59:57 +0000 (19:59 +0000)]
- changed handling of block walks: we might reach a block through a kept node only ...
- this should fix the gigo_return problem

[r24502]

15 years agomove some typedefs to firm_types, move functions from firm.h into firm_common.h
Matthias Braun [Wed, 10 Dec 2008 19:05:14 +0000 (19:05 +0000)]
move some typedefs to firm_types, move functions from firm.h into firm_common.h

[r24498]

15 years agoFix comments.
Christoph Mallon [Wed, 10 Dec 2008 16:28:02 +0000 (16:28 +0000)]
Fix comments.

[r24491]

15 years ago- replaced pset by pset_new in adjust call
Michael Beck [Wed, 10 Dec 2008 15:27:39 +0000 (15:27 +0000)]
- replaced pset by pset_new in adjust call
- removed useless set
- add support for returns_twice
- some cosmetic changes

[r24488]

15 years ago- add a currently non-functioning mtp_property_returns_twice property
Michael Beck [Wed, 10 Dec 2008 13:06:02 +0000 (13:06 +0000)]
- add a currently non-functioning mtp_property_returns_twice property
- let dumper dump all properties

[r24476]

15 years agoit's called mux now, not Psi
Matthias Braun [Wed, 10 Dec 2008 10:25:24 +0000 (10:25 +0000)]
it's called mux now, not Psi

[r24472]

15 years ago- fixed assertion
Michael Beck [Tue, 9 Dec 2008 12:28:02 +0000 (12:28 +0000)]
- fixed assertion
- maybe we should initially mark labeled blocks as non-removable

[r24451]

15 years ago- do not move the return out of labeled blocks, this must be still reachable
Michael Beck [Tue, 9 Dec 2008 12:26:37 +0000 (12:26 +0000)]
- do not move the return out of labeled blocks, this must be still reachable

[r24450]

15 years ago- do not combine blocks if both have a label
Michael Beck [Tue, 9 Dec 2008 11:55:35 +0000 (11:55 +0000)]
- do not combine blocks if both have a label

[r24449]

15 years agoinclude be.h in firm.h, assert that get_value/set_value have pos>=0
Matthias Braun [Tue, 9 Dec 2008 10:08:25 +0000 (10:08 +0000)]
include be.h in firm.h, assert that get_value/set_value have pos>=0

[r24442]

15 years agochange functions to use the usual ir_ prefix
Matthias Braun [Tue, 9 Dec 2008 08:04:17 +0000 (08:04 +0000)]
change functions to use the usual ir_ prefix

[r24432]

15 years agoavoid structs in public API where easily possible (this helps people generating bindi...
Matthias Braun [Tue, 9 Dec 2008 07:54:48 +0000 (07:54 +0000)]
avoid structs in public API where easily possible (this helps people generating bindings and makes the API more stable)

[r24428]

15 years ago- if an empty labeled block is killed, move the label to the successor block
Michael Beck [Tue, 9 Dec 2008 00:40:19 +0000 (00:40 +0000)]
- if an empty labeled block is killed, move the label to the successor block

[r24423]

15 years ago- do not optimize away labeled blocks
Michael Beck [Tue, 9 Dec 2008 00:27:14 +0000 (00:27 +0000)]
- do not optimize away labeled blocks

[r24422]

15 years ago- removed unused immediate fp mode
Michael Beck [Mon, 8 Dec 2008 23:18:08 +0000 (23:18 +0000)]
- removed unused immediate fp mode
- add architecture for trampoline builder
- implemented trampoline builder for x86

[r24418]

15 years agoRepairing one case in the tarval module broke another, of course. Repair a bit more...
Christoph Mallon [Mon, 8 Dec 2008 21:56:14 +0000 (21:56 +0000)]
Repairing one case in the tarval module broke another, of course.  Repair a bit more and hopefully break a bit less.

[r24413]

15 years agoCygwin has no strtold but long double works
Moritz Kroll [Mon, 8 Dec 2008 20:51:48 +0000 (20:51 +0000)]
Cygwin has no strtold but long double works

[r24410]

15 years agoRepair converting long double to integer and hopefully do not break something else.
Christoph Mallon [Mon, 8 Dec 2008 19:03:01 +0000 (19:03 +0000)]
Repair converting long double to integer and hopefully do not break something else.

[r24406]

15 years agoCopy the sign bit when truncating a float tarval.
Christoph Mallon [Mon, 8 Dec 2008 17:18:11 +0000 (17:18 +0000)]
Copy the sign bit when truncating a float tarval.

[r24403]

15 years agoBefore deconving a node, make sure it is a downconv.
Christoph Mallon [Mon, 8 Dec 2008 10:46:22 +0000 (10:46 +0000)]
Before deconving a node, make sure it is a downconv.

[r24387]

15 years agoa note for IPD users (which will not be included in the release packages
Matthias Braun [Mon, 8 Dec 2008 09:30:33 +0000 (09:30 +0000)]
a note for IPD users (which will not be included in the release packages

[r24385]

15 years agoHelper script to generate configure stuff
Moritz Kroll [Sat, 6 Dec 2008 22:58:36 +0000 (22:58 +0000)]
Helper script to generate configure stuff

[r24367]

15 years ago- BugFix: NEVER skip BAD nodes when creating Phi's, or data cycles loops
Michael Beck [Fri, 5 Dec 2008 16:09:22 +0000 (16:09 +0000)]
- BugFix: NEVER skip BAD nodes when creating Phi's, or data cycles loops
  can arise in dead loops. We already fixed this bug in Phi-optimization, but
  it still exists in SSA-construction, ARG!
  This fixes compilerhang5.c
- clean up if condition by using DeMorgan

[r24332]

15 years ago- BugFix: wrong assertion fixed
Michael Beck [Fri, 5 Dec 2008 14:45:20 +0000 (14:45 +0000)]
- BugFix: wrong assertion fixed

[r24331]

15 years ago- BugFix: mixing of long and int lead to overflow on 64bit platform
Michael Beck [Fri, 5 Dec 2008 14:35:44 +0000 (14:35 +0000)]
- BugFix: mixing of long and int lead to overflow on 64bit platform

[r24329]

15 years ago- add missing else
Michael Beck [Fri, 5 Dec 2008 12:08:38 +0000 (12:08 +0000)]
- add missing else

[r24320]

15 years ago- fixed handling of long double constants
Michael Beck [Thu, 4 Dec 2008 15:10:26 +0000 (15:10 +0000)]
- fixed handling of long double constants

[r24298]

15 years ago- simplified code
Michael Beck [Thu, 4 Dec 2008 14:23:28 +0000 (14:23 +0000)]
- simplified code
- add support for Psi(Cmp(!mode_float), Const(mode_float), Const(mode_float))

[r24295]

15 years ago- simplied code by removing a goto
Michael Beck [Thu, 4 Dec 2008 14:21:39 +0000 (14:21 +0000)]
- simplied code by removing a goto

[r24294]

15 years ago- make ia32_am_scale unsigned
Michael Beck [Thu, 4 Dec 2008 14:20:26 +0000 (14:20 +0000)]
- make ia32_am_scale unsigned

[r24293]

15 years ago- renamed unique_id() to ia32_unique_id() and make it public
Michael Beck [Thu, 4 Dec 2008 14:19:43 +0000 (14:19 +0000)]
- renamed unique_id() to ia32_unique_id() and make it public

[r24292]

15 years ago- fixed warning
Michael Beck [Wed, 3 Dec 2008 14:22:21 +0000 (14:22 +0000)]
- fixed warning

[r24260]

15 years ago- BufFix: add missing initializtion
Michael Beck [Wed, 3 Dec 2008 14:20:22 +0000 (14:20 +0000)]
- BufFix: add missing initializtion

[r24259]

15 years agoAdded first version of partial specification for IR nodes
Moritz Kroll [Wed, 3 Dec 2008 11:07:26 +0000 (11:07 +0000)]
Added first version of partial specification for IR nodes

[r24252]

15 years agoincrement version for next release
Matthias Braun [Mon, 1 Dec 2008 17:23:20 +0000 (17:23 +0000)]
increment version for next release

[r24196]

15 years agoMention improved peephole optimisations for x86.
Christoph Mallon [Mon, 1 Dec 2008 14:06:59 +0000 (14:06 +0000)]
Mention improved peephole optimisations for x86.

[r24190]

15 years agoincrease version number
Matthias Braun [Mon, 1 Dec 2008 14:06:31 +0000 (14:06 +0000)]
increase version number

[r24189]

15 years agoarchop is no more
Matthias Braun [Mon, 1 Dec 2008 14:06:18 +0000 (14:06 +0000)]
archop is no more

[r24188]

15 years agoRELEASE notes update
Matthias Braun [Mon, 1 Dec 2008 13:54:56 +0000 (13:54 +0000)]
RELEASE notes update

[r24187]

15 years agobad blocks have cf_probability 0
Matthias Braun [Mon, 1 Dec 2008 13:54:47 +0000 (13:54 +0000)]
bad blocks have cf_probability 0

[r24186]

15 years agoPerform dark rituals and resurrect ProduceVal. We are not ready to dispose it, yet...
Christoph Mallon [Mon, 1 Dec 2008 12:43:11 +0000 (12:43 +0000)]
Perform dark rituals and resurrect ProduceVal.  We are not ready to dispose it, yet, because the Spiller is too clever for the Keep-trick.

[r24182]

15 years agoKeeps should float too
Matthias Braun [Mon, 1 Dec 2008 12:06:09 +0000 (12:06 +0000)]
Keeps should float too

[r24179]

15 years agowe have a bugtracker
Matthias Braun [Mon, 1 Dec 2008 11:59:57 +0000 (11:59 +0000)]
we have a bugtracker

[r24178]

15 years agoSmall cleanup.
Christoph Mallon [Mon, 1 Dec 2008 11:25:41 +0000 (11:25 +0000)]
Small cleanup.

[r24176]

15 years agowe can't guarantee that there won't be dead blocks, so remove them
Matthias Braun [Mon, 1 Dec 2008 10:32:12 +0000 (10:32 +0000)]
we can't guarantee that there won't be dead blocks, so remove them

[r24175]

15 years agoRemove ProduceVal. Replace it by Xor0 (to produce 0), Sbb0 (to produce 0/-1) and...
Christoph Mallon [Fri, 28 Nov 2008 18:10:03 +0000 (18:10 +0000)]
Remove ProduceVal. Replace it by Xor0 (to produce 0), Sbb0 (to produce 0/-1) and by Keep to make Cltd register pressure faithful.

[r24128]

15 years agoUnconditionally include stdio.h
Christoph Mallon [Fri, 28 Nov 2008 16:00:39 +0000 (16:00 +0000)]
Unconditionally include stdio.h

[r24125]

15 years agoUnconditionally include string.h
Christoph Mallon [Fri, 28 Nov 2008 15:59:08 +0000 (15:59 +0000)]
Unconditionally include string.h

[r24124]

15 years agoUnconditionally include stdlib.h.
Christoph Mallon [Fri, 28 Nov 2008 15:08:27 +0000 (15:08 +0000)]
Unconditionally include stdlib.h.

[r24123]

15 years agoInclude alloca.h only, if HAVE_ALLOCA_H is set.
Christoph Mallon [Fri, 28 Nov 2008 14:54:23 +0000 (14:54 +0000)]
Include alloca.h only, if HAVE_ALLOCA_H is set.

[r24121]

15 years agoRemove bashism.
Christoph Mallon [Fri, 28 Nov 2008 14:37:27 +0000 (14:37 +0000)]
Remove bashism.

[r24117]

15 years agoremove archop stuff which wasn't really used anyway
Matthias Braun [Fri, 28 Nov 2008 12:45:17 +0000 (12:45 +0000)]
remove archop stuff which wasn't really used anyway

[r24112]

15 years agofix warning
Matthias Braun [Fri, 28 Nov 2008 12:44:56 +0000 (12:44 +0000)]
fix warning

[r24111]

15 years agofix wrong printf usage
Matthias Braun [Fri, 28 Nov 2008 12:44:15 +0000 (12:44 +0000)]
fix wrong printf usage

[r24109]

15 years agoRemoved mode parameter from Const and Const_type constructors (now derived from tarval)
Moritz Kroll [Thu, 27 Nov 2008 20:49:12 +0000 (20:49 +0000)]
Removed mode parameter from Const and Const_type constructors (now derived from tarval)

[r24088]

15 years agoEmit a block header for the start block of functions.
Christoph Mallon [Thu, 27 Nov 2008 16:22:10 +0000 (16:22 +0000)]
Emit a block header for the start block of functions.

[r24085]

15 years ago- BugFix: drop wrong dependency of mode_P
Michael Beck [Thu, 27 Nov 2008 15:09:38 +0000 (15:09 +0000)]
- BugFix: drop wrong dependency of mode_P

[r24071]

15 years ago- BugFix: drop wrong dependency of mode_P
Michael Beck [Thu, 27 Nov 2008 15:04:02 +0000 (15:04 +0000)]
- BugFix: drop wrong dependency of mode_P

[r24069]

15 years ago- drop dependency of predefined modes
Michael Beck [Thu, 27 Nov 2008 15:00:26 +0000 (15:00 +0000)]
- drop dependency of predefined modes

[r24068]

15 years ago- BugFix: use mode_is_reference() instead of != mode_P
Michael Beck [Thu, 27 Nov 2008 14:55:09 +0000 (14:55 +0000)]
- BugFix: use mode_is_reference() instead of != mode_P

[r24067]

15 years agoSame typo as r24062, different file.
Christoph Mallon [Thu, 27 Nov 2008 14:51:14 +0000 (14:51 +0000)]
Same typo as r24062, different file.

[r24066]