paweld/fpc
1
0
Fork 0
mirror of https://gitlab.com/freepascal.org/fpc/source.git synced 2025-04-30 15:53:40 +02:00
Code Issues Packages Projects Releases Wiki Activity
50,682 Commits 100 Branches 75 Tags 305 MiB
c028362f31
Commit Graph

416 Commits

Author SHA1 Message Date
nickysn
c028362f31 * added the missing pop directive after the push directive in handle_const_rox 
git-svn-id: trunk@36043 -
 2017-05-01 18:25:01 +00:00
nickysn
855cd616f0 + optimize rol(0, x) and ror(0, x) to 0; also optimize the case with all ones, 
e.g. rol32(ffffffff, x) = ffffffff, etc.

git-svn-id: trunk@36042 -
 2017-05-01 18:19:49 +00:00
nickysn
29429cca3c + mask the shift/rotate count value in the rol/ror/sar inline nodes, before 
checking for 0, so that things like sar(int32,32) can be optimized to int32,
  just like sar(int32,0)

git-svn-id: trunk@36031 -
 2017-05-01 11:45:23 +00:00
nickysn
b17d97050f + also simplify rol(x,0) and ror(x,0) to x 
git-svn-id: trunk@36024 -
 2017-04-30 20:45:51 +00:00
nickysn
8ac84ce43a + simplify sar(x,0) to x 
git-svn-id: trunk@36023 -
 2017-04-30 20:24:44 +00:00
nickysn
d8ec0eff8e + implemented inline code generation for 64-bit sar (the SarInt64 intrinsic) on 
less-than-64-bit CPUs that have a 64-bit OP_SHR/OP_SHL/OP_SAR implementation
  in their cg64 backend code generator. This is enabled only for i386 for now.

git-svn-id: trunk@36022 -
 2017-04-30 16:06:34 +00:00
nickysn
5393daa994 + added compile time const evaluation for Bsf/Bsr(const) as well 
git-svn-id: trunk@35941 -
 2017-04-24 20:35:08 +00:00
nickysn
32395bbcbb + added compile time const evaluation optimization for PopCnt(const) 
git-svn-id: trunk@35937 -
 2017-04-24 16:11:43 +00:00
nickysn
d7c8a081a1 + enable using the cg64 ops OP_SHR/OP_SHL/OP_SAR on i386 for implementing the 
64-bit in_sar/shl/shr_assign_x_y inline nodes

git-svn-id: trunk@35835 -
 2017-04-18 14:36:41 +00:00
nickysn
b14f277e8f + use rtl helpers for 64-bit shl/shr/sar/rol/ror modify in place operations, on 
platforms that don't have native 64-bit implementation of the corresponding
  64-bit shift/rotate operation

git-svn-id: trunk@35787 -
 2017-04-13 15:24:32 +00:00
nickysn
f514657e45 + implemented the sar/shl/shr/rol/ror inline modify-in-place nodes; the actual 
optimization of "i:=i shl/shr k", etc statements is not performed yet

git-svn-id: trunk@35780 -
 2017-04-12 00:41:15 +00:00
nickysn
8d885ffde6 + write the inlinenumber of inline nodes in the node tree 
git-svn-id: trunk@35764 -
 2017-04-10 14:54:50 +00:00
nickysn
256dc546ac + implemented the in_neg_assign_x and in_not_assign_x inline nodes, which will 
be used (TBD in a future commit) for optimizing x:=-x and x:=not x on CPUs
  that support performing these operations directly in memory (such as x86)

git-svn-id: trunk@35749 -
 2017-04-07 16:02:40 +00:00
nickysn
fc59649a98 + added inline nodes for handling and/or/xor in place (i.e. x:=x op y, where 
op=and/or/xor). They generate more optimal code on certain architectures
  (including x86). The new inline nodes aren't generated by the compiler yet,
  but will be used in the future, at certain optimization levels, whenever the
  pattern x:=x op y is detected by the compiler.

git-svn-id: trunk@35666 -
 2017-03-26 23:16:53 +00:00
florian
6d6a45c034 * add/sub nodes with nf_internal set should not do overflow checking, resolves #30889 
* do not convert succ/pred into add/sub nodes if nf_internal is set

git-svn-id: trunk@34896 -
 2016-11-13 21:18:39 +00:00
florian
a705525e9b * allow prec/succ on non-consecutive enums if pred/succ are generated internally 
git-svn-id: trunk@34813 -
 2016-11-06 15:22:39 +00:00
florian
0bc1e7e23b * "inherit" the nf_internal flag in tinlinenode.getaddsub_for_incdec 
git-svn-id: trunk@34802 -
 2016-11-05 23:06:06 +00:00
florian
64e2c2b914 + simplify succ/pred(<const>/x+/-x<const>) 
git-svn-id: trunk@34798 -
 2016-11-05 23:06:00 +00:00
florian
000cffa8e3 * make more use of nf_internal to avoid range check errors 
git-svn-id: trunk@34793 -
 2016-11-05 23:05:51 +00:00
florian
e4d60b2048 * factored out tinlinenode.getaddsub_for_incdec from first_IncDec so it can be used independently 
* get varstate right of inc/dec nodes which are converted into add/sub nodes

git-svn-id: trunk@34792 -
 2016-11-05 23:05:49 +00:00
florian
9c81e90e08 + extension to iso mode: reset/rewrite can take a file name as a second parameter 
git-svn-id: trunk@34726 -
 2016-10-16 07:55:08 +00:00
nickysn
d6007d650a * use tpointerdef.pointer_arithmetic_int/uint_type in tinlinenode.pass_typecheck 
for handling inc/dec(pointer, int)

git-svn-id: trunk@34655 -
 2016-10-08 12:06:57 +00:00
nickysn
e847971477 * use the actual ValSInt type from the system unit, instead of ptrsinttype when 
handling the code parameter of val()

git-svn-id: trunk@34608 -
 2016-10-03 21:45:08 +00:00
svenbarth
26a2ddd3d6 + extend Delete() intrinsics with Delphi compatible support for dynamic arrays. Also fixes Mantis #30306 
git-svn-id: trunk@34455 -
 2016-09-08 17:15:10 +00:00
svenbarth
17bb29ddc3 Check for tf_winlikewidestring in the target's flags instead of relying on the target being a Windows system when determining whether the system as the WideString type that's separate from UnicodeString 
git-svn-id: trunk@33900 -
 2016-06-04 08:45:22 +00:00
svenbarth
aaff6d0b97 Allow Delete() and Insert() to be used with generic types as arguments. 
+ added test

git-svn-id: trunk@33897 -
 2016-06-03 21:45:23 +00:00
svenbarth
a2c9c75e97 Convert Insert() and Delete() to intrinsics in preparation for dynamic array support for these two procedures. 
Since overloading compilerprocs does not work each procedure got its own unique name, but they are using the new compilerproc extension to map them to the Insert and Delete symbol so that error messages can be shown with the respective name for the procedure declarations instead of fpc_shortstr_delete for example.

git-svn-id: trunk@33895 -
 2016-06-03 21:25:49 +00:00
svenbarth
8c96fbed8f Fix for Mantis #28832 
ninl.pas, tinlinenode:
  * pass_typecheck: let code pass on to simplify() for undefineddefs in Low() and High()
  * simplify: create a constant 0 as fallback

+ added test

git-svn-id: trunk@33884 -
 2016-06-03 12:14:30 +00:00
Jonas Maebe
abe8d44f37 * changed allocating a temp with type sinttype and size sinttype*dims for 
with a ctemprefnode.create_offset() hack into a proper array, because
    the old way does not work at all with LLVM (and is ugly)

git-svn-id: trunk@33724 -
 2016-05-20 20:51:38 +00:00
svenbarth
8287773b16 Reverted revision 33036. This feature is too controverse to be left in. 
git-svn-id: trunk@33048 -
 2016-02-04 19:26:51 +00:00
svenbarth
ed94ca4b24 Add support for IfThen() instrinsic that works like the if-statement in that it evaluates only the expression that is indeed executed. 
The result type of the intrinsic is determined by the Then-expression to provide a bit of control. There might however be some situations in which this fails, for this exceptions need to be added (e.g. a constant string needs to be converted to a normal string).

compinnr.inc:
  + add new constant in_ifthen_x_y_z for the IfThen() intrinsic
psystem.pas: 
  + create_intern_symbols: add symbol for IfThen() intrinsic
pexpr.pas:
  * statement_syssym: parse parameters of IfThen() intrinsic and return corresponding inline node
ninl.pas, tinlinenode:
  + new method handle_ifthen() which converts the inline node to an if-node which assigns the expressions to a temp node that is returned
  * pass_typecheck: handle in_ifthen_x_y_z using handle_ifthen()
  * pass_1: in_ifthen_x_y_z does not need a first pass as it's already converted after the typecheck pass

+ added tests

git-svn-id: trunk@33036 -
 2016-01-31 14:29:12 +00:00
Jonas Maebe
8917633199 * implement sizeof/typeof completely at the node level, based on 
loadvmtaddr and vmt_def
   o give an error when trying to use sizeof on a class/object when
     targeting the JVM (can't get/load the data size there)

git-svn-id: trunk@32764 -
 2015-12-27 14:22:06 +00:00
svenbarth
74c79803f7 Extend tlocalvarsym with the ability to not register it if needed. 
git-svn-id: trunk@32375 -
 2015-11-20 11:29:21 +00:00
florian
70b8789262 o basic extended pascal support: 
+ automatically loaded helper unit
  * Makefiles adapted
  * mode switch
  * mode switches reorganized
+ TimeStamp support

git-svn-id: trunk@32330 -
 2015-11-15 15:13:36 +00:00
Jonas Maebe
2778bf55c7 * add "doregister" parameter to torddef.create, and don't register temporary 
defs created in the scanner for evaluating compile time expressions, or in
    the code generator for range checking

git-svn-id: trunk@32046 -
 2015-10-13 15:59:06 +00:00
florian
f00a55ae08 + tinlinenode.createintern 
* do not do range checking on constants created by pred/succ if these nodes are created internally

git-svn-id: trunk@31471 -
 2015-08-31 20:03:51 +00:00
Jonas Maebe
67b8aceaee * synchronized with privatetrunk till r30095 
git-svn-id: branches/hlcgllvm@30101 -
 2015-03-05 20:32:15 +00:00
Jonas Maebe
68303b8df1 * add explicit check that argument of sizeof(x) is a simple load node (this 
was already assumed later on)

git-svn-id: trunk@29816 -
 2015-02-23 22:47:35 +00:00
florian
0dfbba86f3 * properly firstpass nodes when inc(...) is converted into ...:=...+x, resolves #27471 
git-svn-id: trunk@29742 -
 2015-02-17 21:14:45 +00:00
florian
c9ba9665dd * set nf_internal for if nodes generated by assert, prevent warning for assert(true); resolves #26467 
git-svn-id: trunk@29293 -
 2014-12-14 19:41:22 +00:00
nickysn
a9e1419ee3 * extracted the pass 1 handling of inc/dec to a virtual method first_IncDec 
git-svn-id: trunk@29210 -
 2014-12-06 21:14:31 +00:00
pierre
a77735d7af Fix bug in in_const_swap_qword handling 
git-svn-id: trunk@29067 -
 2014-11-13 10:59:00 +00:00
Jonas Maebe
e8026ad8d3 * fixed the resultdef of length() intrinsic: it's ossinttype rather than 
sinttype (they're the same on most platforms, except on 8 bit ones where
    ossinttype is still 16 bit)
  * simplified and documented second_length(), and fixed some (harmless)
    errors in the used tcgsize for the resultdef (it had to be OS_SINT
    rather than OS_INT -- now it's just def_cgsize(resultdef) so it's
    always correct)

git-svn-id: branches/hlcgllvm@28387 -
 2014-08-12 18:59:13 +00:00
sergei
04d8e8a5dc * On 64-bit targets, handle abs(int64) internally, using the same code as for abs(longint), i.e. without branching. Both generic and x86-specific pass 2 code is already suitable for different operand sizes, only type checking needs removal of excessive conversions to 32 bits. 
git-svn-id: trunk@27989 -
 2014-06-17 18:45:11 +00:00
svenbarth
34394d6925 Fix for Mantis #26180. Accept undefineddef as first parameter type of an Assert if it is used inside a generic. 
ninl.pas, tinlinenode.pass_typecheck:
  * if the first parameter of an Assert is a undefineddef node then accept it as well if the node is part of a generic function/method

+ added test

git-svn-id: trunk@27875 -
 2014-06-06 14:39:27 +00:00
sergei
274c2fc82f * handle_ln_const: if/else branches were swapped, correct values are: ln(0)=-Inf and ln(-#)=NaN. 
* handle_ln_const and handle_sqrt_const handle all arguments, not just invalid ones, so calling code can be simplified.

git-svn-id: trunk@27837 -
 2014-06-01 01:20:06 +00:00
sergei
743326dc70 * abs(): when translated to generic fpc_abs_real() helper, typecast result back to original precision, similar to how it is done for sqr and sqrt. Fixes compilation on m68k after r27808 (m86k looks like the only target still needing this generic handling). 
git-svn-id: trunk@27818 -
 2014-05-25 11:55:55 +00:00
sergei
37625211f5 * Fixed sqrt procedure selection on arm-wince after r27809. 
git-svn-id: trunk@27815 -
 2014-05-23 21:58:23 +00:00
sergei
22e099d000 * sqrt(real): for targets with emulated floating point, invoke float64_sqrt or float32_sqrt from softfpu code. Testing on ARM CPU shows that float64_sqrt executes about twice faster and offers better accuracy than fpc_sqrt_real from genmath.inc. 
* softfpu.pp: changed float64_sqrt into a function, so it is consistent with other compiler-invoked routines.

git-svn-id: trunk@27809 -
 2014-05-21 15:07:13 +00:00
sergei
fc5f45f65c * sqr(real) and sqrt(real): remove typeconv node inserted by initial call processing (see explanation in comments), allowing these functions to be evaluated using precision of argument. In particular, sqrt(single) and sqrt(double) now emit 'sqrtss' and 'sqrtsd' instructions on x86 targets with -Cfsse3. Non-x86 targets already have the necessary support in code generators. 
* abs(real): handle the same way as sqrt and sqr, i.e. without casting to bestreal and back.

git-svn-id: trunk@27808 -
 2014-05-21 14:53:47 +00:00