Michael Beck [Mon, 19 Sep 2005 15:22:34 +0000 (15:22 +0000)]
removed buggy firm_copy_node() and restorted old copy_node()
[r6644]
Michael Beck [Sat, 17 Sep 2005 14:21:09 +0000 (14:21 +0000)]
tarval_not() han now handle tarval_b_true and tarval_b_false
[r6638]
Michael Beck [Fri, 16 Sep 2005 14:05:10 +0000 (14:05 +0000)]
updated
[r6632]
Michael Beck [Thu, 15 Sep 2005 12:14:54 +0000 (12:14 +0000)]
added function for convertion jump prediction into human readable names
[r6626]
Michael Beck [Thu, 15 Sep 2005 12:14:23 +0000 (12:14 +0000)]
dump now jump prediction, Free place (stack/heap)
[r6625]
Michael Beck [Thu, 15 Sep 2005 11:03:15 +0000 (11:03 +0000)]
added a generic function pointer to an opcode
[r6624]
Michael Beck [Thu, 15 Sep 2005 10:01:28 +0000 (10:01 +0000)]
cond_jmp_predicate to Cond nodes added
[r6623]
Michael Beck [Thu, 15 Sep 2005 10:01:14 +0000 (10:01 +0000)]
cond_jmp_predicate type and access function for Cond nodes added
[r6622]
Michael Beck [Wed, 14 Sep 2005 14:43:26 +0000 (14:43 +0000)]
irg_malloc_function added
[r6618]
Michael Beck [Wed, 14 Sep 2005 11:11:34 +0000 (11:11 +0000)]
renamed is_forking_op() to is_irn_forking()
cleaned up the comments
fixed typos
[r6617]
Michael Beck [Wed, 14 Sep 2005 11:08:14 +0000 (11:08 +0000)]
is_irn_constlike() added
renamed is_forking_op() to is_irn_forking()
[r6616]
Michael Beck [Wed, 14 Sep 2005 10:17:24 +0000 (10:17 +0000)]
irop_flag_constlike flag and is_op_constlike() function added
[r6615]
Michael Beck [Mon, 12 Sep 2005 14:07:54 +0000 (14:07 +0000)]
Added extended block statistic
[r6610]
Michael Beck [Mon, 12 Sep 2005 14:06:30 +0000 (14:06 +0000)]
uses new get_extbb_leader() function
[r6609]
Michael Beck [Mon, 12 Sep 2005 14:05:47 +0000 (14:05 +0000)]
ensure that the leader is the first one in the extended block block list
[r6608]
Michael Beck [Thu, 8 Sep 2005 14:45:21 +0000 (14:45 +0000)]
BugFix: Unreachable Blocks are always leader blocks
[r6604]
Michael Beck [Thu, 8 Sep 2005 14:19:28 +0000 (14:19 +0000)]
get_extbb_leader() and get_extbb_node_nr() added
[r6603]
Michael Beck [Thu, 8 Sep 2005 12:06:32 +0000 (12:06 +0000)]
added additional graph properties, these can be set in the entity
for external allocated entities.
fixed some typos
added more comments
[r6602]
Michael Beck [Thu, 8 Sep 2005 12:04:44 +0000 (12:04 +0000)]
when doing CSE of Proj-Nodes, places Proj's into the right block, this
prevents SSA-violations in irg_verify()
tracks now dataflow and control flow changes to correctly invalidate
analysis info
[r6601]
Michael Beck [Thu, 8 Sep 2005 12:01:55 +0000 (12:01 +0000)]
additional graph properties added
[r6600]
Michael Beck [Thu, 8 Sep 2005 12:00:43 +0000 (12:00 +0000)]
dump additional graph properties for method entities
[r6599]
Michael Beck [Thu, 8 Sep 2005 09:59:37 +0000 (09:59 +0000)]
C99 construct removed
[r6598]
FIRM Projekt Account [Thu, 8 Sep 2005 08:08:43 +0000 (08:08 +0000)]
Make a typedef to generate JNI constants
[r6596]
FIRM Projekt Account [Thu, 8 Sep 2005 08:07:35 +0000 (08:07 +0000)]
renamed old IrgVrfy() into new IrgVerify()
[r6595]
Michael Beck [Wed, 7 Sep 2005 16:16:33 +0000 (16:16 +0000)]
Handle dead blocks hopefully right ...
[r6593]
Michael Beck [Wed, 7 Sep 2005 08:31:30 +0000 (08:31 +0000)]
moved get_irn_generic_attr() to public i/f
[r6587]
Michael Beck [Tue, 6 Sep 2005 11:56:26 +0000 (11:56 +0000)]
fixed SH's dominance checker, can now handle dead blocks
ir_vrfy() renamed to ir_verify() with new flag argument:
dominance checking can now be enforced
[r6579]
Michael Beck [Tue, 6 Sep 2005 08:04:25 +0000 (08:04 +0000)]
BugFix:
We cannot replace old Phi nodes with current one BEFORE we
have handled all Phi's in the current block, else we might
generate wrong references between Phi nodes of a block
clean up is_despensible: uses goto's now, but more readable
[r6578]
Michael Beck [Mon, 5 Sep 2005 09:41:24 +0000 (09:41 +0000)]
backported from irargs.c
[r6577]
Michael Beck [Fri, 2 Sep 2005 15:44:09 +0000 (15:44 +0000)]
removed SLOOOWWW call to is_in_irgs_obstack()
should only be needed for debugging
[r6576]
Michael Beck [Fri, 2 Sep 2005 13:56:10 +0000 (13:56 +0000)]
creates libfirm_xmalloc.a if DISABLE_LIBIBERTY is enabled
[r6573]
Michael Beck [Fri, 2 Sep 2005 12:31:25 +0000 (12:31 +0000)]
irtools.h removed from public header
[r6569]
Daniel Grund [Thu, 1 Sep 2005 15:35:34 +0000 (15:35 +0000)]
checked for dead blocks before getting dom depht.
[r6567]
Daniel Grund [Thu, 1 Sep 2005 15:13:55 +0000 (15:13 +0000)]
*** empty log message ***
[r6566]
Michael Beck [Thu, 1 Sep 2005 15:13:47 +0000 (15:13 +0000)]
get_irn_generic_attr() added, needed for backends that use it's own
opcodes
[r6565]
Michael Beck [Thu, 1 Sep 2005 09:36:27 +0000 (09:36 +0000)]
used PTR_TO_INT() and INT_TO_PTR() macros for 64bit safety
[r6561]
Michael Beck [Thu, 1 Sep 2005 09:35:10 +0000 (09:35 +0000)]
removed wrong const
some comments added
[r6560]
Michael Beck [Thu, 1 Sep 2005 09:32:38 +0000 (09:32 +0000)]
ed PTR_TO_INT() and INT_TO_PTR() macros for 64bit safety
[r6559]
Michael Beck [Thu, 1 Sep 2005 09:32:06 +0000 (09:32 +0000)]
used HASH_PTR() now
[r6558]
Michael Beck [Thu, 1 Sep 2005 09:31:07 +0000 (09:31 +0000)]
renamed function (clashed with buildin)
[r6557]
Michael Beck [Thu, 1 Sep 2005 09:27:22 +0000 (09:27 +0000)]
bitfields of size 1 should be unsigned if 1 is stored there
[r6556]
Michael Beck [Thu, 1 Sep 2005 09:17:25 +0000 (09:17 +0000)]
used HASHPTR() now
[r6555]
Sebastian Hack [Wed, 31 Aug 2005 17:39:47 +0000 (17:39 +0000)]
Removed dominance computation if irg_vrfy
[r6554]
Michael Beck [Wed, 31 Aug 2005 17:26:07 +0000 (17:26 +0000)]
newer version that can handle x86_64
[r6553]
Michael Beck [Wed, 31 Aug 2005 17:24:18 +0000 (17:24 +0000)]
irtools.h is not installed anymore
[r6552]
Michael Beck [Wed, 31 Aug 2005 17:24:03 +0000 (17:24 +0000)]
INT_TO_PTR() and PTR_TO_INT() macros added
[r6551]
Sebastian Hack [Wed, 31 Aug 2005 13:47:57 +0000 (13:47 +0000)]
Added private data facility for edges
[r6546]
Michael Beck [Wed, 31 Aug 2005 09:38:21 +0000 (09:38 +0000)]
removed warnings: single bit type should be unsigned if 0 and 1 should be stored there
[r6543]
Michael Beck [Wed, 31 Aug 2005 09:36:58 +0000 (09:36 +0000)]
Fixes for GCC 4.0: void * and pointer to function are not compatible anymore
[r6542]
Michael Beck [Wed, 31 Aug 2005 09:36:08 +0000 (09:36 +0000)]
fixes for gcc 4.0: void * and pointer to function are not compatible anymore
[r6541]
Michael Beck [Wed, 31 Aug 2005 09:34:57 +0000 (09:34 +0000)]
Fixes for gcc4.0: void * and pointer to functions are NOT compatible anymore
[r6540]
Michael Beck [Wed, 31 Aug 2005 09:25:48 +0000 (09:25 +0000)]
C99 feature removed
[r6539]
Sebastian Hack [Mon, 29 Aug 2005 16:41:03 +0000 (16:41 +0000)]
Fixed a bug
[r6536]
Sebastian Hack [Mon, 29 Aug 2005 16:37:05 +0000 (16:37 +0000)]
Added little more debugging output
[r6535]
Sebastian Hack [Mon, 29 Aug 2005 16:32:37 +0000 (16:32 +0000)]
A little better debug output
[r6534]
Daniel Grund [Mon, 29 Aug 2005 16:28:48 +0000 (16:28 +0000)]
*** empty log message ***
[r6533]
Sebastian Hack [Mon, 29 Aug 2005 14:53:33 +0000 (14:53 +0000)]
Fixed a bug again
[r6531]
Sebastian Hack [Mon, 29 Aug 2005 14:37:23 +0000 (14:37 +0000)]
Fixed a bug
[r6527]
Sebastian Hack [Mon, 29 Aug 2005 14:30:34 +0000 (14:30 +0000)]
Added dominance check
[r6526]
Daniel Grund [Mon, 29 Aug 2005 11:18:33 +0000 (11:18 +0000)]
Added mod and mux
[r6524]
Sebastian Hack [Mon, 29 Aug 2005 10:00:42 +0000 (10:00 +0000)]
Added get_irn_edge()
[r6523]
Michael Beck [Fri, 26 Aug 2005 15:01:15 +0000 (15:01 +0000)]
removed wrong INLINE spec
[r6510]
Michael Beck [Fri, 26 Aug 2005 12:17:55 +0000 (12:17 +0000)]
BugFix:
new_rd_*() functions that call optimize_node() did not
work if current_ir_graph was not set!
So, the current_ir_graph is now set and restored.
Moreover, to speed up the default new_*() and new_d_*()
cases, new_bd_*() functions have been introduced which use
current_ir_graph, but not current_block and replace
the use of _rd_ functions in the comfortable i/f
[r6509]
Michael Beck [Fri, 26 Aug 2005 12:02:46 +0000 (12:02 +0000)]
BugFix: in place_late, do NOT count keep-alive edges as consuming ones,
or nodes might be placed wrong!
[r6508]
Michael Beck [Thu, 25 Aug 2005 12:42:12 +0000 (12:42 +0000)]
BugFix: predecessors of End nodes might be Blocks
[r6495]
Michael Beck [Wed, 24 Aug 2005 13:59:29 +0000 (13:59 +0000)]
BugFix:
When entities are freed, type information is used, so we cannot
free one type before all entities are freed
[r6485]
Daniel Grund [Tue, 23 Aug 2005 20:27:13 +0000 (20:27 +0000)]
del_pset in now called correctly
[r6482]
Michael Beck [Tue, 23 Aug 2005 18:14:02 +0000 (18:14 +0000)]
Set array border to 0, needed for fixed add_compound_ent_value(), one node more
[r6481]
Michael Beck [Tue, 23 Aug 2005 18:13:51 +0000 (18:13 +0000)]
Set array border to 0, needed for fixed add_compound_ent_value()
[r6480]
Daniel Grund [Tue, 23 Aug 2005 18:03:21 +0000 (18:03 +0000)]
Bugfixes
[r6479]
Michael Beck [Tue, 23 Aug 2005 16:32:01 +0000 (16:32 +0000)]
BugFix: add_compound_ent_value() did not generate right indexes for compound pathes
[r6478]
Michael Beck [Tue, 23 Aug 2005 13:22:36 +0000 (13:22 +0000)]
allow character values to be converted into long
some cleanup
[r6476]
Michael Beck [Tue, 23 Aug 2005 13:20:35 +0000 (13:20 +0000)]
moke mode sort of character modes irms_twos_complement
[r6475]
Michael Beck [Mon, 22 Aug 2005 16:43:00 +0000 (16:43 +0000)]
added %P for compound-graph-paths
[r6470]
Daniel Grund [Thu, 18 Aug 2005 16:58:50 +0000 (16:58 +0000)]
Added reflection for mulh
[r6463]
Michael Beck [Thu, 18 Aug 2005 12:57:45 +0000 (12:57 +0000)]
node function copy_irn extracted from firm_copy_node()
[r6461]
Michael Beck [Thu, 18 Aug 2005 11:49:28 +0000 (11:49 +0000)]
removed warning
[r6458]
Michael Beck [Thu, 18 Aug 2005 11:47:58 +0000 (11:47 +0000)]
enabled type_visited() function again
[r6457]
Michael Beck [Thu, 18 Aug 2005 11:44:25 +0000 (11:44 +0000)]
removed definition of old type_visited flag
restored type_visited() function declaration
[r6456]
Michael Beck [Wed, 17 Aug 2005 14:13:45 +0000 (14:13 +0000)]
resreuctures compound path to make allocation/deallocation simplier
[r6455]
Michael Beck [Tue, 16 Aug 2005 13:00:55 +0000 (13:00 +0000)]
fixed copy'n'paste error
nd fixed
[r6439]
Michael Beck [Tue, 16 Aug 2005 12:34:01 +0000 (12:34 +0000)]
changed option %g to %G (floating point uses %g)
[r6438]
Sebastian Hack [Tue, 16 Aug 2005 12:23:34 +0000 (12:23 +0000)]
Changed %g to %G to enable float dumping
[r6436]
Sebastian Hack [Tue, 16 Aug 2005 12:12:38 +0000 (12:12 +0000)]
Added Minus for Ints
[r6435]
Michael Beck [Tue, 16 Aug 2005 10:18:35 +0000 (10:18 +0000)]
create_abstraction() now returns an error code if the file could not
be opened.
[r6434]
Götz Lindenmaier [Sun, 14 Aug 2005 11:22:51 +0000 (11:22 +0000)]
bugfix
[r6419]
FIRM Projekt Account [Fri, 12 Aug 2005 16:31:20 +0000 (16:31 +0000)]
new_rd_Unknown sets the current_ir_graph right.
This fixes crashes in jack, but a general solution is needed: all
rd functions calling optimize_node must do that!
Restructure IR construction!
[r6418]
Michael Beck [Fri, 12 Aug 2005 13:11:02 +0000 (13:11 +0000)]
install scalar_replace.h
[r6415]
Michael Beck [Fri, 12 Aug 2005 10:28:51 +0000 (10:28 +0000)]
changed new_type_pointer() to new semantic
[r6413]
Michael Beck [Fri, 12 Aug 2005 09:42:23 +0000 (09:42 +0000)]
Unknown nodes are optimized (CSE) now.
This was not done earlier, why?
[r6412]
Michael Beck [Fri, 12 Aug 2005 09:41:48 +0000 (09:41 +0000)]
changed back to old visit order
removed all calles form for () headers
[r6411]
Michael Beck [Fri, 12 Aug 2005 09:40:28 +0000 (09:40 +0000)]
current_ir_graph must be set EVEN iff the raw interface is used!!!
[r6410]
Michael Beck [Fri, 12 Aug 2005 09:39:20 +0000 (09:39 +0000)]
new_d_type_array now allocates only ONE Uknown node for the array
bounds, and Unknown nodes are optimized now, so lesser nodes
[r6409]
Michael Beck [Thu, 11 Aug 2005 16:40:38 +0000 (16:40 +0000)]
typo fixed
[r6408]
Michael Beck [Thu, 11 Aug 2005 16:39:57 +0000 (16:39 +0000)]
used now always the visited functions, no more access to type_visited flag directly
[r6407]
Michael Beck [Thu, 11 Aug 2005 16:38:32 +0000 (16:38 +0000)]
renamed type_visited to firm_type_visited
removed unused fields from union attributes
renamed new_type_pointer_mode() to new_type_pointer(), now always a
mode argument is needed
[r6406]
Michael Beck [Thu, 11 Aug 2005 16:36:59 +0000 (16:36 +0000)]
Bugfix: unset debug info is now initialized with NULL
renamed new_type_pointer_mode() to new_type_pointer(), it's now
always necessary to give a mode when construction a pointer type
renamed type_visited to firm_type_visited and removed from public i/f
removed #if 0 code
[r6405]
Michael Beck [Thu, 11 Aug 2005 16:33:49 +0000 (16:33 +0000)]
renamed global flag type_visited into firm_type_visited
[r6404]
Michael Beck [Thu, 11 Aug 2005 16:32:41 +0000 (16:32 +0000)]
renamed new_type_pointer_mode() into new_type_pointer()
[r6403]
Michael Beck [Thu, 11 Aug 2005 16:31:15 +0000 (16:31 +0000)]
used Block_not_block_visited() instead of explicit code
[r6402]