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 -
+ 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
parameters are released before a call
* fix storeback of registers in case of different sizes (e.g., first
a "movl %eax,%edx" and later a "movb %dl,%al")
* fixed several bugs which were in there already for a long time, but
which only popped up now :) -O2/-O3 will now optimise less than in
the past (and correctly so), but -O2u/-O3u will optimise a bit more
* some more small improvements for -O3 are still possible
* foreach gets extra argument for passing local data to the
iterator function
* -CR checks also class typecasts at runtime by changing them
into as
* fixed compiler to cycle with the -CR option
* fixed stabs with elf writer, finally the global variables can
be watched
* removed a lot of routines from cga unit and replaced them by
calls to cgobj
* u32bit-s32bit updates for and,or,xor nodes. When one element is
u32bit then the other is typecasted also to u32bit without giving
a rangecheck warning/error.
* fixed pascal calling method with reversing also the high tree in
the parast, detected by tcalcst3 test
+ first parameter offset in stack now portable
* rename some constants
+ move some cpu stuff to other units
- remove unused constents
* fix stacksize for some targets
* fix generic size problems which depend now on EXTEND_SIZE constant
* removing frame pointer in routines is only available for : i386,m68k and vis targets
error was found
* generic changeregsize(reg,size) for i386 register resizing
* removed some more routines from cga unit
* fixed returnvalue handling
* fixed default stacksize of linux and go32v2, 8kb was a bit small :-)
* LOC_CONSTANT added for better constant handling
* secondadd splitted in multiple routines
* location_force_reg added for loading a location to a register
of a specified size
* secondassignment parses now first the right and then the left node
(this is compatible with Kylix). This saves a lot of push/pop especially
with string operations
* adapted some routines to use the new cg methods
* 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
optimizer bugs with int64 calculations (because of the carry flag usage)
* fixed another bug which caused wrong optimizations with complex
array expressions
* changed command line options meanings for optimizer:
O2 now means peepholopts, CSE and register renaming in 1 pass
O3 is the same, but repeated until no further optimizations are
possible or until 5 passes have been done (to avoid endless loops)
* changed aopt386 so it does this looping
* added some procedures from csopt386 to the interface because they're
used by rropt386 as well
* some changes to csopt386 and daopt386 so that newly added instructions
by the CSE get optimizer info (they were simply skipped previously),
this fixes some bugs
