* reordered fields of tsetting to avoid unaligned fields
* fixed compilation of arm after disableddircache was added
* increased ppu version
git-svn-id: trunk@13327 -
vmt entries of virtual methods that can never be called with references
to FPC_ABSTRACTERROR. Some virtual methods are always considered to be
reachable: published methods, and methods used as getter/setter for a
published property.
git-svn-id: trunk@13238 -
alignment for each memory reference (mantis #12137, and
test/packages/fcl-registry/tregistry1.pp on sparc). This also
enables better code generation for packed records in many cases.
o several changes were made to the compiler to minimise the chances
of accidentally forgetting to set the alignment of memory references
in the future:
- reference_reset*() now has an extra alignment parameter
- location_reset() can now only be used for non LOC_(C)REFERENCE,
use location_reset_ref() for those (split the tloc enum so the
compiler can catch errors using range checking)
git-svn-id: trunk@12719 -
svn+ssh://jonas@svn.freepascal.org/FPC/svn/fpc/branches/wpo
........
r11878 | jonas | 2008-10-11 02:25:18 +0200 (Sat, 11 Oct 2008) | 19 lines
+ initial implementation of whole-program optimisation framework
+ implementation of whole-program devirtualisation
o use:
a) generate whole-program optimisation information (no need
to completely compile the program and all of its units
with -OW/-FW, only the main program is sufficient)
fpc -OWdevirtcalls -FWmyprog.wpo myprog
b) use it to optimise the program
fpc -B -Owdevirtcalls -Fwmyprog.wpo myprog
(the -B is not required, but only sources recompiled during
the second pass will actually be optimised -- if you want,
you can even rebuild the rtl devirtualised for a particular
program; and these options can obviously also be used
together with regular optimisation switches)
o warning:
- there are no checks yet to ensure that you do not use
units optimised for a particular program with another
program (or with a changed version of the same program)
........
r11881 | jonas | 2008-10-11 19:35:52 +0200 (Sat, 11 Oct 2008) | 13 lines
* extracted code to detect constructed class/object types from
tcallnode.gen_vmt_tree into its own method to avoid clutter
* detect x.classtype.create constructs (with classtype = the
system.tobject.classtype method), and treat them as if a
"class of x" has been instantiated rather than a
"class of tobject". this required storing the instantiated
classrefs in their own array though, because at such a
point we don't have a "class of x" tdef available (so
now "x", and all other defs instantiated via a classref,
are now stored as tobjectdefs in a separate array)
+ support for devirtualising class methods (including
constructors)
........
r11882 | jonas | 2008-10-11 20:44:02 +0200 (Sat, 11 Oct 2008) | 7 lines
+ -Owoptvmts whole program optimisation which replaces vmt entries
with method names of child classes in case the current class'
method can never be called (e.g., because this class is never
instantiated). As a result, such methods can then be removed
by dead code removal/smart linking (not much effect for either
the compiler, lazarus or a trivial lazarus app though).
........
r11889 | jonas | 2008-10-12 14:29:54 +0200 (Sun, 12 Oct 2008) | 2 lines
* some comment fixes
........
r11891 | jonas | 2008-10-12 18:49:13 +0200 (Sun, 12 Oct 2008) | 4 lines
* fixed twpofilereader.getnextnoncommentline() when reusing a previously
read line
* fixed skipping of unnecessary wpo feedback file sections
........
r11892 | jonas | 2008-10-12 23:42:43 +0200 (Sun, 12 Oct 2008) | 31 lines
+ symbol liveness wpo information extracted from smartlinked programs
(-OW/-Owsymbolliveness)
+ use symbol liveness information to improve devirtualisation (don't
consider classes created in code that has been dead code stripped).
This requires at least two passes of using wpo (first uses dead code
info to locate classes that are constructed only in dead code,
second pass uses this info to potentially further devirtualise).
I.e.:
1) generate initial liveness and devirtualisation feedback
fpc -FWtt.wpo -OWall tt.pp -Xs- -CX -XX
2) use previously generated feedback, and regenerate new feedback
based on this (i.e., disregard classes created in dead code)
fpc -FWtt-1.wpo -OWall -Fwtt.wo -Owall tt.pp -Xs- -CX -XX
3) use the newly generated feedback (in theory, it is possible
that even more opportunities pop up afterwards; you can
continue until the program does not get smaller anymore)
fpc -Fwtt-1.wpo -Owall tt.pp -CX -XX
* changed all message() to cgmessage() calls so the set codegenerror
* changed static fsectionhandlers field to a regular field called
fwpocomponents
* changed registration of wpocomponents: no longer happens in the
initialization section of their unit, but in the InitWpo routine
(which has been moved from the woinfo to the wpo unit). This way
you can register different classes based on the target/parameters.
+ added static method to twpocomponentbase for checking whether
the command line parameters don't conflict with the requested
optimisations (e.g. generating liveness info requires that
smartlinking is turned on)
+ added static method to twpocomponentbase to request the
section name
........
r11893 | jonas | 2008-10-12 23:53:57 +0200 (Sun, 12 Oct 2008) | 3 lines
* fixed comment error (twpodeadcodeinfo keeps a list of live,
not dead symbols)
........
r11895 | jonas | 2008-10-13 00:13:59 +0200 (Mon, 13 Oct 2008) | 2 lines
+ documented -OW<x>, -Ow<x>, -FW<x> and -Fw<x> wpo parameters
........
r11899 | jonas | 2008-10-14 22:14:56 +0200 (Tue, 14 Oct 2008) | 2 lines
* replaced hardcoded string with objdumpsearchstr constant
........
r11900 | jonas | 2008-10-14 22:15:25 +0200 (Tue, 14 Oct 2008) | 2 lines
* reset wpofeedbackinput and wpofeedbackoutput in wpodone
........
r11901 | jonas | 2008-10-14 22:16:07 +0200 (Tue, 14 Oct 2008) | 2 lines
* various additional comments and comment fixes
........
r11902 | jonas | 2008-10-15 18:09:42 +0200 (Wed, 15 Oct 2008) | 5 lines
* store vmt procdefs in the ppu files so we don't have to use a hack to
regenerate them for whole-program optimisation
* fixed crash when performing devirtualisation optimisation on programs
that do not construct any classes/objects with optimisable vmts
........
r11935 | jonas | 2008-10-19 12:24:26 +0200 (Sun, 19 Oct 2008) | 4 lines
* set the vmt entries of non-class virtual methods of not instantiated
objects/classes to FPC_ABSTRACTERROR so the code they refer to can
be thrown away if it is not referred to in any other way either
........
r11938 | jonas | 2008-10-19 20:55:02 +0200 (Sun, 19 Oct 2008) | 7 lines
* record all classrefdefs/objdefs for which a loadvmtaddrnode is generated,
and instead of marking all classes that derive from instantiated
classrefdefs as instantiated, only mark those classes from the above
collection that derive from instantiated classrefdefs as
instantiated (since to instantiate a class, you have to load its vmt
somehow -- this may be broken by using assembler code though)
........
r12212 | jonas | 2008-11-23 12:26:34 +0100 (Sun, 23 Nov 2008) | 3 lines
* fixed to work with the new vmtentries that are always available and
removed previously added code to save/load vmtentries to ppu files
........
r12304 | jonas | 2008-12-05 22:23:30 +0100 (Fri, 05 Dec 2008) | 4 lines
* check whether the correct wpo feedback file is used in the current
compilation when using units that were compiled using wpo information
during a previous compilation run
........
r12308 | jonas | 2008-12-06 18:03:39 +0100 (Sat, 06 Dec 2008) | 2 lines
* abort compilation if an error occurred during wpo initialisation
........
r12309 | jonas | 2008-12-06 18:04:28 +0100 (Sat, 06 Dec 2008) | 3 lines
* give an error message instead of crashing with an io exception if the
compiler is unable to create the wpo feedback file specified using -FW
........
r12310 | jonas | 2008-12-06 18:12:43 +0100 (Sat, 06 Dec 2008) | 3 lines
* don't let the used wpo feedback file influence the interface crc (there's
a separate check for such changes)
........
r12316 | jonas | 2008-12-08 19:08:25 +0100 (Mon, 08 Dec 2008) | 3 lines
* document the format of the sections of the wpo feedback file inside the
feedback file itself
........
r12330 | jonas | 2008-12-10 22:26:47 +0100 (Wed, 10 Dec 2008) | 2 lines
* use sysutils instead of dos to avoid command line length limits
........
r12331 | jonas | 2008-12-10 22:31:11 +0100 (Wed, 10 Dec 2008) | 3 lines
+ support for testing whole program optimisation tests (multiple
compilations using successively generated feedback files)
........
r12332 | jonas | 2008-12-10 22:31:40 +0100 (Wed, 10 Dec 2008) | 2 lines
+ whole program optimisation tests
........
r12334 | jonas | 2008-12-10 22:38:07 +0100 (Wed, 10 Dec 2008) | 2 lines
- removed unused local variable
........
r12339 | jonas | 2008-12-11 18:06:36 +0100 (Thu, 11 Dec 2008) | 2 lines
+ comments for newly added fields to tobjectdef for devirtualisation
........
r12340 | jonas | 2008-12-11 18:10:01 +0100 (Thu, 11 Dec 2008) | 2 lines
* increase ppu version (was no longer different from trunk due to merging)
........
git-svn-id: trunk@12341 -
* give a note if the visibility of a vmt entry is lower than
the previous (parent) entry
* refactor vmt method collection using the new always available
vmt entries
git-svn-id: trunk@12159 -
* store a node's verbosity in the node so that e.g. disabling warnings
also disables warnings for this node in pass_1
(the above together fix mantis #12076)
* save/restore verbosity with {$push}/{$pop} (mantis #12075)
* if warnings/notes/hints are turned off, also do not count encountered
ones for the totals (otherwise -Sew cannot be used properly in
combination with {$warnings off}, because disabled warnings will still
trigger a compiler error) -- this required adding -vw/-vn/-vh to all
tests using -Sew/-Sen/-Seh
- removed some superfluous state saving/restoring from firstpass()
git-svn-id: trunk@12025 -
* record symtables are now handled properly as well if a record is copied, resolves#9144
* fixed a memory leak when copying classes
* increased PPU version to 91
git-svn-id: trunk@11273 -
from architectures with a larger maximum floating point precision
for an architecture with a smaller maximum fp precision can also be
used by native compilers on the target architectures (final fix for
mantis #9165)
git-svn-id: trunk@11076 -
the code using {$pascalmainname x} + storing it in the ppu file
(and give a warning if it's overridden multiple times + test)
git-svn-id: trunk@10406 -
a) cpu64bitaddr, which means that we are generating a compiler which
will generate code for targets with a 64 bit address space/abi
b) cpu64bitalu, which means that we are generating a compiler which
will generate code for a cpu with support for 64 bit integer
operations (possibly running in a 32 bit address space, depending
on the cpu64bitaddr define)
All cpus which had cpu64bit set now have both the above defines set,
and none of the 32 bit cpus have cpu64bitalu set (and none will
compile with it currently)
+ pint and puint types, similar to aint/aword (not pword because that
that conflicts with pword=^word)
* several changes from aint/aword to pint/pword
* some changes of tcgsize2size[OS_INT] to sizeof(pint)
git-svn-id: trunk@10320 -
for qwordbool) + test:
o assigning true to such a variable now sets them to $ff/$ffff/$ffffffff
o these types are now all signed
o converting an integer type to a byte/word/long/qwordbool using an
explicit type cast keeps the integer's original value stored in the
bool, instead of forcing it to ord(true)/ord(false)
(mantis #10233 and #10613, implemented for all architectures, testsuite
tested for ppc32, sparc and x86)
* fixed some places where the rtl depended on longbool(true) having the
value 1
* extended several boolean tests (and adapted some to no longer assume
that byte/word/long/qwordbool(true)=1)
+ support for converting to qwordbool in second_int_to_bool for x86, ppc
and sparc
git-svn-id: trunk@9898 -
* -Xg now produces a .dbg file with debuginfo
that can be used by gdb. The main executable gets
a debuglink section that references the .dbg file.
git-svn-id: trunk@9778 -
perform the transformation of x:=f(hiddencomplexresult, ..)
-> f(x, ...) (the compiler now performs some very
conservative escape analysis for such types)
git-svn-id: trunk@8361 -
finally blocks
* fixed the handling of function results of inlined functions with
exit statements
+ test for the above two issues and for bug fixed in r8091
* copy may_be_in_reg field inside ttempcreatenode.dogetcopy (allows
some more temps which were needlessly forced into memory to be in
registers)
git-svn-id: trunk@8108 -
* fixed some (harmless) ptrint warnings
- removed some commented code in agppcmpw
* added one missing field in a typed constant in cp1251
git-svn-id: trunk@8081 -
It is done in the following way:
When unit is compiled, .rc file are compiled to .res and list of unit's resource files is stored in .ppu
Before final linking all program's .res files are collected into global .res file (.res files are easily concatenated).
Then this global .res files is compiled to single .or file, which is linked to executable.
As a result global resource index is created and the problem is fixed.
Old resource processing behavior still supported when tresinfo.rcbin is not set for target.
New resource processing is activated for windows and linux. Cross compiled windres can be used to compile .rc files on linux.
git-svn-id: trunk@7515 -
* varsets ({$packset x}) are now supported on big endian targets
* gdb now displays sets properly on big endian systems
* cleanup of generic set code (in, include/exclude, helpers), all
based on "bitpacked array[] of 0..1" now
* there are no helpers available yet to convert sets from the old to
the new format, because the set format will change again slightly
in the near future (so that e.g. a set of 24..31 will be stored in
1 byte), and creating two classes of set conversion helpers would
confuse things (i.e., it's not recommended to use trunk currently for
programs which load sets stored to disk by big endian programs compiled
by previous FPC versions)
* cross-endian compiling has been tested and still works, but one case
is not supported: compiling a compiler for a different endianess
using a starting compiler from before the current revision (so first
cycle natively, and then use the newly created compiler to create a
cross-compiler)
git-svn-id: trunk@7395 -
* symtables based on TFPHashObjectList and TFPObjectList
* rename torddef.typ to torddef.ordtype
* rename tfloatdef.typ to tfloatdef.floattype
* rename tdef.deftype to tdef.typ
* remove obsolete browser code, browcol is kept so the ide
can still be compiled
git-svn-id: trunk@5192 -
Not sure how it ever worked, nor how exactly symtable ppu entries
work -- but it's now stored with the recorddef (which also means
that bitpacking is disabled currently for objects and classes,
since they are based on tabstractrecorddef rather than trecorddef)
git-svn-id: trunk@4679 -
+ use {$bitpacking on/+} to change the meaning of "packed"
into "bitpacked" for arrays. This is the default for MacPas.
You can also define individual arrays as "bitpacked", but
this is not encouraged since this keyword is not known by
other compilers and therefore makes your code unportable.
+ pack(unpackedarray,index,packedarray) to pack
length(packedarray) elements starting at
unpackedarray[index] into packedarray.
+ unpack(packedarray,unpackedarray,index) to unpack
packedarray into unpackedarray, with the first
element being stored at unpackedarray[index]
* todo:
* "open packed arrays" and rtti for packed arrays are not
yet supported
* gdb does not properly support bitpacked arrays
git-svn-id: trunk@4449 -
the external linker
* create import libraries for dll imports, this uses
the new objdata framework to generate the binary
object files directly without needing an assembler pass
* store import_dll and import_name in ppu
* external linker uses import libraries
* internal linker uses import info from symtables,
no dlls are needed anymore
git-svn-id: trunk@3255 -
* change booleans in arraydef to set
* set option in arraydef to indicate a constant string so
a nicer type can be shown instead of array[0..x] of char
git-svn-id: trunk@3051 -
tarrayconstructornode.inset_typeconvs() and fixed them:
* integers < 32 bit are converted to 32 bit (this was previously
done in the code generator for some targets, and not for others)
* currency is also converted to double for targets where currency = int64
* single is converted to double, except for x86_64 (is at least
necessary on darwin/ppc, darwin/i386 and linux/i386)
* enums are converted to 32 bit ints
* procvars are converted to pointers
* proper errors are given for various unsupported types
NOTE: in C, floating point constants are by default double, while in
FPC they are of type extended. On platforms where extended <> double,
such constants when passed to C varargs are automatically converted
to double by default (gives warning). If you want to pass them as
single or extended or get rid of the warning, use an explicit typecast
* increased ppu version because of introduction of new node flag
(nf_cvarargs for tarrayconstructornode)
* fixed tests/test/cg/tprintf
* changed tests/test/cg/cdecl/taoc5 to use explicit typecasts for
floating point constants passed to C varargs functions.
git-svn-id: trunk@2949 -
* support for case aware filesystems (Windows), they do now only
one lookup if a file exists
* add -WI option to generate import section for DLL imports or
let the linker handle it. Default is still import section until
the Makefiles are fixed, then the generation can be left to the
linker
git-svn-id: trunk@2274 -
assembler and debuginfo and use these vars to override
the defaults for the target after the parameters are read
* remove not-maintained and tested gdb code
git-svn-id: trunk@1201 -
cdecl/cppdecl procedures on Mac OS/Mac OS X to fix compatibility with
GPC (slightly more efficient than Metrowerks behaviour below, but
less efficient in most cases than our previous scheme)
+ "mwpascal" procedure directive to support the const record parameter
behaviour of Metrowerks Pascal, which passes all const records by
reference
+ support for Macintosh PowerPC alignment (if the first field of a record
or union has an alignment > 4, then the record or union size must be
padded to a multiple of this size)
limited circumstances for now (only procedures, and only if they have
no or only vs_out/vs_var parameters).
* fixed ppudump for inline procedures
* fixed ppudump for ppc
+ refaddr to treference added
+ refsymbol to treference added
* top_local stuff moved to an extra record to save memory
+ aint introduced
* tppufile.get/putint64/aint implemented
and aliases for result and function name are added using absolutesym
* vs_hidden parameter for funcret passed in parameter
* vs_hidden fixes
* writenode changed to printnode and released from extdebug
* -vp option added to generate a tree.log with the nodetree
* nicer printnode for statements, callnode
- -V option for generating bug report tracing
- more tracing for option parsing
- errors for cdecl and high()
- win32 import stabs
- win32 records<=8 are returned in eax:edx (turned off by default)
- heaptrc update
- more info for temp management in .s file with EXTDEBUG
* asmsymbollist global is removed and moved into a new class
tasmlibrarydata that will hold the info of a .a file which
corresponds with a single module. Added librarydata to tmodule
to keep the library info stored for the module. In the future the
objectfiles will also be stored to the tasmlibrarydata class
* all getlabel/newasmsymbol and friends are moved to the new class
* register allocation is now handled by a class and is mostly processor
independent (+rgobj.pas and i386/rgcpu.pas)
* temp allocation is now handled by a class (+tgobj.pas, -i386\tgcpu.pas)
* some small improvements and fixes to the optimizer
* some register allocation fixes
* some fpuvaroffset fixes in the unary minus node
* push/popusedregisters is now called rg.save/restoreusedregisters and
(for i386) uses temps instead of push/pop's when using -Op3 (that code is
also better optimizable)
* fixed and optimized register saving/restoring for new/dispose nodes
* LOC_FPU locations now also require their "register" field to be set to
R_ST, not R_ST0 (the latter is used for LOC_CFPUREGISTER locations only)
- list field removed of the tnode class because it's not used currently
and can cause hard-to-find bugs
parsed as unsigned constants (otherwise, $80000000 got sign extended
and became $ffffffff80000000), all constants in the longint range
become longints, all constants >$7fffffff and <=cardinal($ffffffff)
are cardinals and the rest are int64's.
* added lots of longint typecast to prevent range check errors in the
compiler and rtl
* type casts of symbolic ordinal constants are now preserved
* fixed bug where the original resulttype wasn't restored correctly
after doing a 64bit rangecheck
+ cs_local_browser ($L+) is disabled if cs_browser ($Y+)
is not enabled when quitting global section
* local vars and procedures are not yet stored into PPU
(first steps to have objects without vmt if no virtual !!)
+ added fpu_used field for tabstractprocdef :
sets this level to 2 if the functions return with value in FPU
(is then set to correct value at parsing of implementation)
THIS MIGHT refuse some code with FPU expression too complex
that were accepted before and even in some cases
that don't overflow in fact
( like if f : float; is a forward that finally in implementation
only uses one fpu register !!)
Nevertheless I think that it will improve security on
FPU operations !!
* most other changes only for UseBrowser code
(added symtable references for record and objects)
local switch for refs to args and local of each function
(static symtable still missing)
UseBrowser still not stable and probably broken by
the definition hash array !!
(accepted are popstack cdecl and pascal)
+ added C vars with the following syntax
var C calias 'true_c_name';(can be followed by external)
reason is that you must add the Cprefix
which is target dependent
