Commit Graph

42113 Commits

Author SHA1 Message Date
sergei
9c1f917e3a * a_call_ref functionality cannot be implemented efficiently at code generator level, because references need specific preparations at earlier points. Moved this support to tcgcallnode and its x86 descendants, and got rid of all ifdef's around.
+ x86 targets now directly call procedure variables located in references.
- a_call_ref method removed from tcg and thlcgobj.

git-svn-id: trunk@26666 -
2014-02-03 13:28:56 +00:00
sergei
0d3f36eebf - Remove references to global variable 'cg' from methods of tcg and some of its descendants.
git-svn-id: trunk@26665 -
2014-02-03 12:27:48 +00:00
sergei
7104157a57 * Changed tabstractprocdef.address_size to return the def, which can be used to get other properties like size and alignment. Removes one of i8086 ifdef's.
git-svn-id: trunk@26664 -
2014-02-03 10:47:54 +00:00
sergei
53556f529d * gen_alloc_symtable: don't use current_procinfo.procdef, the necessary procdef is passed in 'pd' parameter.
git-svn-id: trunk@26663 -
2014-02-03 10:21:56 +00:00
marco
b7d94992eb * moved unxsockh for solaris, mantis #25657
git-svn-id: trunk@26662 -
2014-02-03 10:12:32 +00:00
michael
7afd5f6afe * Fixed bug ID #25091
git-svn-id: trunk@26661 -
2014-02-03 10:00:33 +00:00
sergei
949907f1f3 * Don't emit "Temps allocated between..." comment in assembler listing if procedure has no temps.
git-svn-id: trunk@26660 -
2014-02-03 09:55:46 +00:00
sergei
1626667374 * Object helper routines review/improvement for alignment-sensitive targets:
- Replaced duplicate local declarations of pvmt/tvmt with pobjectvmt/tobjectvmt
    (name change needed because tvmt is already used for class-style VMTs)
  - Removed 'packed' attribute from tobjectvmt, since it is always aligned.
  * Use appropriate typecasts to generate aligned memory accesses where possible.

git-svn-id: trunk@26659 -
2014-02-03 01:18:19 +00:00
nickysn
f4f35ad47b * changed ExceptTryLevel from longint to nativeint (because i8086)
git-svn-id: trunk@26658 -
2014-02-03 00:48:11 +00:00
sergei
e0082262cc * Don't generate VMT checking code twice (gen_load_vmt_register generates it already).
git-svn-id: trunk@26657 -
2014-02-03 00:27:05 +00:00
sergei
a5d5ce8eb2 * TP-style object destructors do not need actual VMT in hidden parameter, a constant flag indicating need to free memory is sufficient.
* Destructor called from exception block in constructor must have this flag set in order to free memory.
+ Test.

git-svn-id: trunk@26656 -
2014-02-03 00:20:42 +00:00
florian
be17962748 * delete test file immediately after they were used, external tools might get into trouble trying deleting them due to the uncommon names
git-svn-id: trunk@26655 -
2014-02-02 19:10:42 +00:00
marco
e7639e582c * try fix #25657
git-svn-id: trunk@26654 -
2014-02-02 15:42:21 +00:00
Jonas Maebe
868b4f50e9 * use spilling_operation_type() instead of a big table to determine whether
instructions overwrite operand 0 (fixes wrongly handled rlwimi and some
    others, and also adds support for loads/stores)

git-svn-id: trunk@26653 -
2014-02-02 15:32:54 +00:00
Jonas Maebe
de0ced1522 * A_INSLWI, A_INSLWI_, A_INSRWI, A_INSRWI_ also modify operand 0
git-svn-id: trunk@26652 -
2014-02-02 15:32:00 +00:00
Jonas Maebe
6a4afd3a33 * changed uses of "exit" in constructors to if/then constructs, to
avoid the constructor result/SSA bug in 2.6.x

git-svn-id: trunk@26651 -
2014-02-02 09:09:51 +00:00
Jonas Maebe
4e9c54278e * the function result of constructors always is self, always is valid and is
not stored in procdef.funcretsym -> fix check so that SSA is not performed
    on it in constructors when exit is used

git-svn-id: trunk@26650 -
2014-02-02 09:09:44 +00:00
florian
e6b68d19c8 * reordered decision how arguments for inlining are handled
* improved comments for inline argument handling
* replace arguments more aggressively with a copy of their original node tree when inlining

git-svn-id: trunk@26649 -
2014-02-01 22:26:23 +00:00
Jeppe Johansen
07b2982e77 Don't do ARM FoldShiftLdrStr peephole optimization if there's an offset in the reference.
Use UXTH+UXTB instructions instead of two shifts on processors that supports that.
Eliminate internalerror when constant pointers are typecast as arrays.

git-svn-id: trunk@26647 -
2014-02-01 13:29:35 +00:00
florian
2a7dcf6c7b * formatting
git-svn-id: trunk@26646 -
2014-02-01 13:08:30 +00:00
florian
e3f9982ee1 * make_not_regable is a matter of code generation so do it in pass_1
git-svn-id: trunk@26645 -
2014-02-01 13:08:28 +00:00
florian
fbcfb63826 + implement tcasenode.pass_typecheck
git-svn-id: trunk@26644 -
2014-02-01 13:08:26 +00:00
florian
d71bfe4c6f * var parameters can be put in registers if the callee is inlined
git-svn-id: trunk@26643 -
2014-02-01 13:08:24 +00:00
florian
cd6f4d4469 * fix tsubscriptnode to handle left.location being an fpu or mm register as well.
git-svn-id: trunk@26642 -
2014-02-01 13:08:21 +00:00
Tomas Hajny
77967432ef * always use forward slashes in path to fpmake similarly to paths to other tools in Makefiles
git-svn-id: trunk@26641 -
2014-02-01 00:11:09 +00:00
masta
81517fdf70 Rerun peephole optimizers on the whole block
Up until now if a peephole optimizer matched we've only restarted from
the current instruction. This patch restarts optimizations on the whole
block if the previous run had at least one match.

As this can take considerable time this will only be enabled if -O3 or
higher is specified.

git-svn-id: trunk@26640 -
2014-01-31 23:21:52 +00:00
marco
4fa8998541 * add SMsgDlgClose, Patch by stocki Mantis #25616
git-svn-id: trunk@26638 -
2014-01-31 19:56:50 +00:00
masta
3f8549365a Don't schedule LDRD on ARM
The load scheduler does not handle LDRD correctly right now, but it does
not prevent A_LDR with PF_D set from beeing scheduled.

git-svn-id: trunk@26637 -
2014-01-30 21:53:03 +00:00
florian
28a8dc42c6 * fix tcgx86.a_op_reg_reg on x86-64 and make use of it
git-svn-id: trunk@26636 -
2014-01-30 20:24:41 +00:00
florian
7d7bf1d877 + make use of SHLX/SHRX on CPUs supporting BMI2
git-svn-id: trunk@26635 -
2014-01-30 20:24:39 +00:00
florian
8fba876904 * apply cse to static symbols on x86-64 as well, if their address can be stored
git-svn-id: trunk@26634 -
2014-01-30 20:24:36 +00:00
florian
d8c0f11ff9 + cs_userbp optimizer switch, so on x86-64 the compiler can make use of rbp if it is not needed as frame pointer
git-svn-id: trunk@26633 -
2014-01-30 19:44:17 +00:00
florian
a3f58e84be * rbp can be used for normal purpose under certain conditions so it shouldn't interfere with all other registers
git-svn-id: trunk@26632 -
2014-01-30 19:44:14 +00:00
florian
ecdfe75841 * tests fail if dead values are optimized so disable dead value optimization for the tests
git-svn-id: trunk@26631 -
2014-01-30 19:44:12 +00:00
florian
546f4a5d46 * force records with LOC_CONSTANT into memory when they are subscripted
git-svn-id: trunk@26630 -
2014-01-30 19:44:09 +00:00
florian
1d73f49330 * do not force the index location to a register for vec. nodes if the index is already in a register
git-svn-id: trunk@26629 -
2014-01-30 17:34:35 +00:00
florian
cd235d8409 + simplify shl/shr x, if (x mod bitsizeof(left))=0
git-svn-id: trunk@26628 -
2014-01-30 17:34:33 +00:00
florian
b1f54b2853 * indentation fixed
git-svn-id: trunk@26627 -
2014-01-30 17:34:31 +00:00
florian
8989a40b4f * more flags to mark changes fixed
git-svn-id: trunk@26626 -
2014-01-30 17:34:28 +00:00
florian
809549a31f * addr nodes can be also considered in cse
git-svn-id: trunk@26625 -
2014-01-30 17:19:06 +00:00
florian
3a11dd1055 * create the addr node for cse with create_internal to avoid errors when taking the address of a temp.
git-svn-id: trunk@26624 -
2014-01-30 17:19:04 +00:00
florian
fc7dc9e54c * do cse after the init/final code has been inserted, this might enable new cse opportunities
git-svn-id: trunk@26623 -
2014-01-30 17:19:01 +00:00
lacak
c4a981fa8e fcl-db: bufdataset: TBufDataset autoincrement fields (ftAutoInc) are not restarted when dataset is closed and again reopened. See bug #25628
git-svn-id: trunk@26622 -
2014-01-30 09:03:40 +00:00
lacak
03e76b5f8b fcl-db: bufdataset: Check if blob field is not ReadOnly when blob stream is created (in bmWrite mode). Bug #25648
git-svn-id: trunk@26621 -
2014-01-30 08:19:52 +00:00
Jeppe Johansen
257b1affaa Fixed previous fix of LsrAnd2Lsr optimization.
Added an extra condition for <ARMv6 processors in MLA/MLS optimization.

git-svn-id: trunk@26620 -
2014-01-29 22:22:58 +00:00
Jeppe Johansen
f773334374 Fixed LsrAnd2Lsr peephole optimization for ARM.
git-svn-id: trunk@26619 -
2014-01-29 21:35:28 +00:00
Jonas Maebe
859676d7d3 * fixed r26519 for darwin/x86-64, see comments (mantis #25644)
git-svn-id: trunk@26618 -
2014-01-29 21:26:45 +00:00
Jonas Maebe
c05da62b0e * detect static data use by record methods and methods of nested classes/
records, so we avoid inlining them (mantis #25598)

git-svn-id: trunk@26617 -
2014-01-29 21:26:38 +00:00
Jonas Maebe
b611882337 * some minor formatting fixes
git-svn-id: trunk@26616 -
2014-01-29 19:26:17 +00:00
Jonas Maebe
a6d28b5630 * pass on the "eval" state (whether or not subexpressions should be evaluated)
when encountering a "(" in a preprocessor expression (broken after r25465,
    mantis #25296)

git-svn-id: trunk@26615 -
2014-01-29 19:26:12 +00:00