* Introduce MatchInstruction and MatchOperand
MatchInstruction allows to match an instruction by condition and
oppostfix. MatchOperand checks if an operand is a register and matches
another operand. In the future this could be overloaded with other
versions not only accepting TRegister.
* Optimize cmp,moveq,movne sequence on ARM
This patch implements an peephole optimizer for the following sequence:
cmp reg,const1
movne reg,const2
moveq reg,const1
* Small improvements to the ARM peephole optimizer
Most instructions in the ARM ISA have taicpu(p).oper[0]^.typ = top_reg
as the only option, so there is no need to check for it if we're
looking at those instructions.
* Remove redundant mov instructions on ARM
This is an addition to the ARM PeepHole Optimizer.
It folds code like this:
mov reg1, reg2
add reg1, reg1, (const|reg)
git-svn-id: trunk@21024 -
* Added missing call to paramanager.freecgpara.
* Handle safecall stuff after deallocating registers of the main call in order to keep allocations balanced.
git-svn-id: trunk@21021 -
left-aligned. Other targets only require this if the total size of the
parameter was > 8 bytes (this last part used to be ensured by
the ppc64-specific version of a_load_ref_cgpara() that was removed in
r20962)
git-svn-id: trunk@21008 -
record has only a single field (and if so, which one) when checking
whether we have to treat the record as if it was actually its first
field that's passed as a parameter (fixes problems with records
with methods, e.g. crash while compiling tests/test/trhlp35)
git-svn-id: trunk@20967 -
on AIX due to the extra inserted local symbols (although in case e.g
multiple procedures contain a typed constant with the same name, only
one can be accessible)
* removed '.' suffix of local symbols inserted for debug info of typed
constants for AIX (gdb couldn't find those symbols; it seems the
address calculated by gdb for typed constants is still not correct
though)
git-svn-id: trunk@20966 -
be generated for a single object file on AIX, automatically switch to
an indirection scheme that uses much less TOC entries, but is slower to
access global variables
* manually merged infrastructure for target-specific code generation
switches from JVM branch (-CTxxx switches)
* -CTsmalltoc switch for AIX that forces the indirection scheme for TOC
entries from the start in case the automatic one does not conserve
a sufficient amount (or in case you bump up against the global TOC
entries limit during linking)
git-svn-id: trunk@20963 -
if the value has to be shifted first (it could be a regvar)
* also take the cgpara's shiftval into account when loading a constant
git-svn-id: trunk@20961 -
itself (e.g. a 3-byte record as field insde a 4 byte record), make
the parent record non-regable either. The reason is that the parent
record will be forced to memory in that case when subscripting, and
the assignment will happen to this temporary memory location.
Fixes tests/test/tw15357.pp on ppc64 when compiled with -O2 and
without -Cg (not sure why enabling PIC disables global regvars)
git-svn-id: trunk@20959 -
assume that all virtual methods that can be called via interfaces that
appear in the program are in fact called (fixes WPO when applied to
MSEIDE)
git-svn-id: trunk@20952 -
a) we don't care about them
b) for 32 bit binaries on 64 bit platforms, they are indented differently
than defined symbols
(fixes -OWsymbolliveness for 32 bit binaries on 64 bit Linux platforms)
git-svn-id: trunk@20950 -
before the change order was (from better to worse): shortstring, ansistring, widestring, unicodestring
now the order is (from better to worse): widestring, unicodestring, ansistring, shortstring
git-svn-id: trunk@20907 -
* created separate version of fpc_AnsiStr_Concat_multi() after all
because it contains a punicodechar(unicodestring) typecast, which
can't be supported on the JVM target (and splitting it out in the
generic code seems like a bit overkill), and restored original
generic version of fpc_AnsiStr_Concat_multi() (slightly faster
than version partially adapted for jvm)
git-svn-id: branches/jvmbackend@20903 -
based on the actual target platform rather than always on the host
platform
* on Unix, use single rather than double quotes for quoting, so it also
properly handles $, ! and `, which keep their special meaning when
appearing in scripts inside double quotes
* since sysutils.executeprocess() can only deal with double-quoted
strings in parameters, re-quote parameters on Unix when they turn
out not to be for scripts but for direct execution (which is most
of the time, but unfortunately doing the reverse is not possible
because parameters used in scripts sometimes contain script-specific
code that must not be quoted, such as `cat link.res`)
-> always use cfileutl.RequotedExecuteProcess() rather than
sysutils.ExecuteProcess() in the compiler (added a bunch of dummy
ExecuteProcess string constants to common units to prevent accidental
usage)
git-svn-id: branches/jvmbackend@20901 -
regular mangledname has already been set (just like we don't internalerror
when changing the regular mangledname even if it has already been set);
this happens e.g. when specifying the external name of some interface
variables, such as in the Win32 system unit
git-svn-id: branches/jvmbackend@20893 -
o support for the new codepage-aware ansistrings in the jvm branch
o empty ansistrings are now always represented by a nil pointer rather than
by an empty string, because an empty string also has a code page which
can confuse code (although this will make ansistrings harder to use
in Java code)
o more string helpers code shared between the general and jvm rtl
o support for indexbyte/word in the jvm rtl (warning: first parameter
is an open array rather than an untyped parameter there, so
indexchar(pcharvar^,10,0) will be equivalent to
indexchar[pcharvar^],10,0) there, which is different from what is
intended; changing it to an untyped parameter wouldn't help though)
o default() support is not yet complete
o calling fpcres is currently broken due to limitations in
sysutils.executeprocess() regarding handling unix quoting and
the compiler using the same command lines for scripts and directly
calling external programs
o compiling the Java compiler currently requires adding ALLOW_WARNINGS=1
to the make command line
git-svn-id: branches/jvmbackend@20887 -
support the stabx as understood by gdb, which is a mixture of regular
stabs and stabx (e.g., class/object definitions are completely different
in real stabx). This means that gdb can be used to debug aix programs,
but the native dbx will complain about lots of debug information
constructs
git-svn-id: trunk@20842 -
on AIX. Adding ".align" statements only align relative to the section
start, but does not influence the alignment of the section itself
git-svn-id: trunk@20835 -
to low addresses in case the target cpu/OS does not support this (enabled
for reads on AIX/PPC because on AIX the first 16KB are readable)
git-svn-id: trunk@20832 -
* in case of a jump/call to an AT_FUNCTION symbols on
systems_dotted_function_names platforms, prepend a dot to the symbol
name in assembler code
git-svn-id: trunk@20829 -
o the last bytes of records passed by value whose size is not a multiple
of the register size must be passed in the upper (leftmost) bytes of
a register
git-svn-id: trunk@20804 -
o .short/.long/.llong are automatically aligned to 2/4/8 byte multiples
by the AIX assembler (and for compatibility reasons, also by the
GNU assembler when targeting AIX) -> change to .vbyte statements
o .ascii does not allow non-ASCII characters in the AIX assembler
-> change to .byte sequences like gcc on AIX
git-svn-id: trunk@20803 -
pass on whether the symbol refers to code or data (important for
AIX weak symbols; note that AIX does not support undefined weak
external symbols, and FPC does not yet support weakly defining
non-external symbols, so this functionality isn't really used yet)
git-svn-id: trunk@20802 -
of the AVR-specific ifdef'ed variant
o since the only special character we use in mangled names on all platforms
is $, added a new field to tasminfo called "dollarsign" that holds the
character $'s should be replaced with (if it doesn't have to be replaced,
leave it at $)
git-svn-id: trunk@20801 -
* renamed some ppc64 stackframe offset constants to the same names as their
32 bit counterparts so they can be used in shared ppc code
git-svn-id: trunk@20794 -
* fixed code dealing with symbol+offsets on 64 bit platforms in
case of offsets > 32 bit (and to use smallint() typecasts
instead of lo(), since the return value of lo() depends on
the type passed in)
git-svn-id: trunk@20793 -
hence contains a dummy "main" that refers to the system unit
FPC_SYSTEMMAIN, which in turn jumps to the main program's PASCALMAIN
* we call FPC_SYSTEMMAIN instead of jumping to it like on Darwin, so that
the TOC gets set correctly
git-svn-id: trunk@20792 -
o doubles are aligned to 8 bytes outside records, but to 4 bytes
inside them
o same as for Darwin/PPC: if the first field of a record has 8 byte
alignment (including double!), then the struct itself gets 8 byte
alignment
git-svn-id: trunk@20791 -
* omit several error messages when in_generic is true because the error can be checked only during specialisation, resolves the issue reported in #21592
git-svn-id: trunk@20699 -
in LOC_CMMREGISTER, not in LOC_CFPUREGISTER (and they also are there in
case the softfp calling convention is used), and this change broke SPARC
git-svn-id: trunk@20667 -
for inherited calls (that is only required when not using them for
dispatch purposes, and calling an inherited method is obviously
a dispatch situation)
git-svn-id: trunk@20664 -
always points to the previous r7 on the stack (with the saved return
address coming right after it) so that the debugger and crashreporter
can use it for backtraces as specified in the ABI
o changed NR_FRAME_POINTER_REG and RS_FRAME_POINTER_REG from a symbolic
into a typed constant, and added a new method to tprocinfo that can
be used to initialze it (so it can be inited to r7/r11 depending on
the target platform)
* allow using r9 on Darwin, it was only used by the system on iOS up to
2.x, which we no longer support
* prefer using r9 and r12 before r4..r11 on Darwin, because they are
volatile and hence do not have to be saved
git-svn-id: trunk@20661 -
o new eabihf (hard float) abi
o vfpv3_d16 variant of VFP (default variant used by EABI assemblers: VFPv3
with only 16 double registers instead of 32) and pass it to GNU as
o make the odd numbered single precision floating point VFP registers
available for explicit allocation for use by the calling convention
* fixed copy/paste error in stdname of S30 register
-> use -dFPC_ARMHF to create an ARM eabi hard float compiler
(mantis #21554)
git-svn-id: trunk@20660 -
in code and not in constants. In the case of primitive types constant nodes
are used while complex types like arrays, records and objects use a local
variable which is initialized to zero once at the entry of the method (the
variable is reused if Default() is used for the same type multiple times in
the same method). For this a new compilerproc was added which uses FillChar
to initialize the given memory area to zero.
This fixes Mantis #9420.
+ psystem.pas: Added Default symbol to system unit
+ htypechk.pas: Added function "is_valid_for_default" which checks recursively
whether the given type can be used with Default at all.
Forbidden types are files, helpers, ObjC and C++ types. This
check is used for records, arrays and objects only if the mode
is a non-Delphi one, as Delphi ignores these types on lower
levels.
+ msg/errore.msg: Added error message for unsupported types for Default()
+ symconst.pas: Added a new enum value vo_is_default_var which is used for the
local variables utilized by Default() so their initalization
and finalization can be avoided.
+ pexpr.pas: Add handling of Default() intrinsic to "statement_syssym"
+ ninl.pas: Extended tinlinenode by a method which returns the correct node for
a Default() and used that method in handle_typecheck.
* ncgutil.pas: Check for new flag "vo_is_default_var" when initializing and
finalizing local variables.
* ppu.pas: increase PPU version
+ psub.pas:
* Added a new routine which zeros defaultvars of a symtable.
* Use this routine inside "initializevars".
* Also use this routine to initialize the staticsymtable of the unit/program.
* Adjusted ppudump, because of the new enum value.
+ Added implementation of fpc_zeromem to system unit.
+ Added tests for Default()
git-svn-id: trunk@20629 -
self parameter for invoking inherited method calls (if you are in a
category method or a method of a child class, the extended or parent
class must have already been loaded by the run time, so we can
directly load the ISA pointer rather than detouring via the "class"
method)
git-svn-id: trunk@20624 -
0, just like gcc/clang. This results in more efficient code on
non-fragile ABI platforms and removes a dependency of the name
of the isa field in header translations
git-svn-id: trunk@20623 -
involved and the classes are not exactly equal (but one is a child of the
other): we have to look up the real class definition first in those cases
git-svn-id: trunk@20622 -
class before its VMT has been built, because that's the process that looks
for inherited abstract methods that are overridden -> instead use a new
synthetic method kind (tsk_callthrough_nonabstract) that will call
through to another method (in this case a constructor) if the owning
class does not contain any abstract methods, and generates an
empty routine otherwise
git-svn-id: branches/jvmbackend@20589 -
constructors (simply do nothing), because creating an abstract class
is not possible in the JVM (and while the JVM only triggers an
exception when you actually execute code that tries to create an
abstract class, the Android platform checks this using the bytecode
verifier at class verification time and therefore stumbled over the
old code)
git-svn-id: branches/jvmbackend@20547 -
on Dalvik, because those values are obtained via an "and 255" operation
which again turns them into longint values as var as the Dalvik type
system is concerned
git-svn-id: branches/jvmbackend@20527 -
to correctly handle non-pasbool types (can't just "not" their value),
and for Dalvik (a "not pasbool" became "xori 1", which turned the value
from a boolean into an integer as far as the Dalvik type checker was
concerned)
git-svn-id: branches/jvmbackend@20525 -
after an 8/16 bit operation on Dalvik (even if the value cannot have
become too big), because otherwise its type checker complains
git-svn-id: branches/jvmbackend@20520 -
Mac OS X/iOS version for the compiler code
o such a parameter is not passed, extract the information from the
environment variables MACOSX_DEPLOYMENT_TARGET/IPHONEOS_DEPLOYMENT_TARGET,
like gcc
o if neither the parameter nor the environment variable is used, use preset
default values
o pass on this version setting on to the Darwin linker
o use this setting to determine which version of the startup code (crt1.o
etc) to use, if any (based on information gathered from the GCC sources)
o define a symbol called MAC_OS_X_VERSION_MIN_REQUIRED/
IPHONE_OS_VERSION_MIN_REQUIRED based on this parameter
o for usage information, see
http://wiki.freepascal.org/FPC_New_Features_Trunk#Support_for_specifying_and_querying_the_deployment_version
git-svn-id: trunk@20503 -
the register parameters are not put on the stack, so don't count them
(only makes a practical difference for fixed stack targets, fixes
random crashes with webtbs/tw3870 on Darwin/i386)
git-svn-id: trunk@20499 -
in a compilation module parsed using a different code page, because this
changes the default ansistring type and {$codepage xxx} is also not
allowed in the middle of a unit (mantis #21445)
git-svn-id: trunk@20488 -
(mantis #21323)
* don't print a message about the ppc binary exiting with an error if no
source file is specified (if a no option or -h is used, the help message
will be printed; in other cases, a source file may not have been specified
but that's an error like any other in that case -- that message only
existed because the compiler binary also exits with a non-zero exit code
in case the help is printed)
- removed alpha and ia64 support for selecting ppc binary
+ added mips and mipsel support for selecting ppc binary
git-svn-id: trunk@20487 -
if expr1 has sideeffects. This can't be done safely even in case of
short boolean evaluation, because expr1 may return the inverse the
second time its called (and "0 or 1" is not the same as "0", and
neither is "1 and 0"), based on comment by Michael Karcher
* perform a full string compare when comparing stringconstnodes
before the string constant labels have been generated (patch by
Michael Karcher, mantis #21255)
git-svn-id: trunk@20485 -
* fixed printed source code positions in error message in case the external
name of a formal and of a real definition don't match
git-svn-id: trunk@20425 -
* remove unnecessary ldr after str to the same memoy location, however, to do this optimization safely, we should add support for volatile variables
git-svn-id: trunk@20399 -
