Commit Graph

485 Commits

Author SHA1 Message Date
Jonas Maebe
8bcb6e689d * only check validity of assembler instructions in pass_2, after all
local operands have been resolved (mantis #8950)

git-svn-id: trunk@7516 -
2007-05-29 16:53:28 +00:00
Jonas Maebe
a0b57eddb5 * new internal set format for big endian systems. Advantages:
* 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 -
2007-05-19 17:15:15 +00:00
daniel
182fca72f2 * Change spill_* routines to return Taicpu instead of Tai to increase
strong typing.
  * Fix PowerPC R0 register allocation

git-svn-id: trunk@7317 -
2007-05-12 15:43:16 +00:00
daniel
3ad8257ecb * Fix replace spilling of shld/shrd.
git-svn-id: trunk@7252 -
2007-05-03 20:56:09 +00:00
daniel
f3660976bc + Allow replace spilling for "opcode register,const" and "opcode const,register"
git-svn-id: trunk@7183 -
2007-04-28 19:11:17 +00:00
daniel
c43855a8f7 * Improve comment for [0..31]-x optimization
git-svn-id: trunk@7177 -
2007-04-27 13:35:25 +00:00
daniel
ea79e69176 + Add optimization to optimize [0..31]-x set expression to use one less
register. This construction is used in our Shootout meteor contest
    implementation.

git-svn-id: trunk@7176 -
2007-04-26 21:48:18 +00:00
peter
29ffdefacb * secrel32 relocation support required by dwarf2 under windows
git-svn-id: trunk@7172 -
2007-04-25 09:06:36 +00:00
florian
f87e96dfb0 * properly release open array value parameters on x86-64 or if they contain automated types, resolves #8664
git-svn-id: trunk@7100 -
2007-04-13 19:20:56 +00:00
Jonas Maebe
5f169f97ff * fixed SSE2 substraction when both operands are on the 80x87 fpu stack
git-svn-id: trunk@6959 -
2007-03-23 00:21:46 +00:00
Tomas Hajny
30b78920cd * OS/2 compilation fixes (not completely finished yet)
git-svn-id: trunk@6937 -
2007-03-20 00:49:20 +00:00
florian
d55dbf92d3 * fixed assembling of cvt(t)ss2si
git-svn-id: trunk@6918 -
2007-03-18 17:35:22 +00:00
florian
4fc85cb084 * new test
* fix compilation on i386

git-svn-id: trunk@6861 -
2007-03-14 21:09:05 +00:00
florian
bae83ee777 * set expectloc correctly on x86-64
* decide properly if we inline trunc or not on x86-64

git-svn-id: trunk@6860 -
2007-03-14 21:05:15 +00:00
florian
f4840620f0 * location.size of trunc/round must be OS_S64
* fixed assembling of cvt*2si
+ optimized trunc/round for x86_64/sse
* only floats need no type cast for trunc/round

git-svn-id: trunk@6859 -
2007-03-14 20:56:14 +00:00
pierre
d1d6413443 * use S_IQ opsize for A_FISTP instruction
git-svn-id: trunk@6848 -
2007-03-14 17:09:42 +00:00
florian
bcb1cf0d28 * make no use of int registers in trunc to reduce register pressure
git-svn-id: trunk@6844 -
2007-03-14 10:23:52 +00:00
pierre
f07349a774 * force S_IQ size for movq instruction, needed for nasm
git-svn-id: trunk@6835 -
2007-03-13 23:17:09 +00:00
florian
c4f65df9f9 * fixed expectloc setting
git-svn-id: trunk@6833 -
2007-03-13 22:45:41 +00:00
florian
8aa259401c * fixed x86-64
git-svn-id: trunk@6828 -
2007-03-13 21:12:40 +00:00
florian
1cd41b68f7 * compiler handled round/trunc for x86
git-svn-id: trunk@6827 -
2007-03-13 20:23:24 +00:00
florian
e6e1a3a09b * fixed assembling of sse instructions on x86-64 with regs > xmm7
git-svn-id: trunk@6714 -
2007-03-04 19:16:20 +00:00
florian
9a0f769b2f * cleaned up set conversion
* fixed conversion of var sets

git-svn-id: trunk@6644 -
2007-02-25 16:29:39 +00:00
florian
05e0a804cc * fixes cardinal in reg -> real on x86-64, resolves #8225
git-svn-id: trunk@6588 -
2007-02-21 16:53:57 +00:00
daniel
9256c180de * Fix register allocation bug that happens when a coalesced
move is spilled in its entirely.

git-svn-id: trunk@6408 -
2007-02-10 22:47:13 +00:00
florian
08f48f5fc9 * completed jrcxz support
git-svn-id: trunk@6401 -
2007-02-10 21:14:05 +00:00
florian
a19ed91cc3 * fix for jcxz, jecxz and jrcxz on 64 bit platforms
git-svn-id: trunk@6400 -
2007-02-10 21:05:27 +00:00
Jonas Maebe
9813a8708c * fixed encoding of jcxz
git-svn-id: trunk@6398 -
2007-02-10 20:38:38 +00:00
Jonas Maebe
1863fb4c55 * make more use of location_force_register in in-code, should fix
cycle on x86-64

git-svn-id: trunk@6376 -
2007-02-08 14:14:32 +00:00
Jonas Maebe
21cef2cdb1 * r6372 for x86(-64)
git-svn-id: trunk@6373 -
2007-02-07 23:10:54 +00:00
Jonas Maebe
f38d29cfab * optimized previous patch for "byte in set" (Thorsten Engler)
* fixed storing/loading of setdefs to/from ppu + test (Thorsten Engler)
  * fixed ppudump for new setdef format in ppus

git-svn-id: trunk@6363 -
2007-02-07 20:22:29 +00:00
Jonas Maebe
fc3852e59a * in non-TP modes, 'ord_larger_than_256 in set' now always returns
false instead of triggering a range error (mantis #8258,
    based on patch provided by Thorsten Engler)

git-svn-id: trunk@6357 -
2007-02-07 18:57:19 +00:00
Jonas Maebe
1a32e4fdf5 * fixed pushing of words in intel assembler (indirectly related to
wrong bug report 7808)

git-svn-id: trunk@6274 -
2007-01-30 21:52:11 +00:00
Jonas Maebe
cc67ed17b5 * fixed uninitialised variable (mantis 8198, already tested by
webtbs/tw3931a.pp)

git-svn-id: trunk@6183 -
2007-01-25 11:43:15 +00:00
Jonas Maebe
cdb67cfc8b * fixed bug with set comparisons after r6172
git-svn-id: trunk@6181 -
2007-01-24 22:28:25 +00:00
florian
2579cd139f + support for sets with size 1 and 2
git-svn-id: trunk@6172 -
2007-01-24 20:06:56 +00:00
Jonas Maebe
c866400823 + support for VMTOFFSET in assembler readers to get VMT offset of virtual
methods (mantis #8153)

git-svn-id: trunk@6168 -
2007-01-24 18:26:23 +00:00
Jonas Maebe
c1df4454fe * cmpxchg8b doesn't have a size suffix
git-svn-id: trunk@6164 -
2007-01-24 15:15:04 +00:00
Jonas Maebe
11528c9da6 * better fix for tw8171 (now also works for with-statements)
git-svn-id: trunk@6158 -
2007-01-24 09:55:45 +00:00
Jonas Maebe
ed13c7f60e + support for [recordfield.field] in assembler blocks in methods in
intel assembler (mantis 8171)

git-svn-id: trunk@6150 -
2007-01-23 17:34:11 +00:00
Jonas Maebe
302a2a3ec2 * fixed calling qualified methods in assembler + tests for ppc/ppc64/
i386/x86_64

git-svn-id: trunk@6112 -
2007-01-21 22:16:42 +00:00
florian
2da51bce72 + cmpxchg16b
git-svn-id: trunk@6095 -
2007-01-21 11:04:19 +00:00
florian
5e2042cff8 * correct assembling of test rm8,reg8
git-svn-id: trunk@5972 -
2007-01-14 18:08:37 +00:00
Jonas Maebe
abef29c1d5 * fixed x86 vector mmloads and use scalar loads in ncgcnv
git-svn-id: trunk@5967 -
2007-01-14 13:36:26 +00:00
Jonas Maebe
fc99b0ad48 * fixed floating point precision conversions for loc_mmregister and
between loc_fpuregister and loc_mmregister

git-svn-id: trunk@5942 -
2007-01-13 15:00:00 +00:00
Jonas Maebe
e815b923d5 * a_loadfpu_* gets two size parameters: fromsize and tosize
* fixed downsizing the precision of floating point values
  * floating point constants are now treated using only the minimal
    precision required (e.g. 2.0 is now a single, 1.1 extended etc)
    (Delphi compatible)

git-svn-id: trunk@5927 -
2007-01-12 18:33:51 +00:00
florian
36a500c38c * don't coalescate with frame/stack pointer
git-svn-id: trunk@5880 -
2007-01-11 10:47:07 +00:00
florian
c238bb15d1 * .init section support
git-svn-id: trunk@5835 -
2007-01-07 10:03:52 +00:00
florian
db3f8575ae * fixed library creating on x86_64-linux
git-svn-id: trunk@5828 -
2007-01-06 20:27:59 +00:00
florian
c759b114bf * allow second operand being am imm in intel assembler, fixed #7719
git-svn-id: trunk@5823 -
2007-01-05 23:15:25 +00:00