Commit Graph

1507 Commits

Author SHA1 Message Date
florian
78878f59b1 + generic TAOptObj.AllocRegBetween
- removed x86 specific AllocRegBetween

git-svn-id: trunk@38445 -
2018-03-07 22:17:35 +00:00
nickysn
6e79c8ba86 + support adding an OPR_LOCAL to a OPR_REFERENCE in
tx86intreader.AddReferences(); this allows things like mov ax,[cs:[local]] to
  work

git-svn-id: trunk@38444 -
2018-03-07 17:31:25 +00:00
nickysn
d3cb61b651 * refactored tx86intreader.AddReferences to use a pair of case statements,
depending on the combination of operand types; this is done, so that adding
  OPR_LOCAL with OPR_REFERENCE operands can be supported later.

git-svn-id: trunk@38443 -
2018-03-07 16:43:39 +00:00
marco
f0042a4719 * vcmppd hardcoded primitives like vcmpeqpd.
* required increasing maxinfolen to 9 

git-svn-id: trunk@38404 -
2018-03-03 23:32:54 +00:00
marco
f21a141144 * mantis #32001, add 32 vcmpps variants.
git-svn-id: trunk@38403 -
2018-03-03 23:10:03 +00:00
nickysn
2cee948b72 + support segment overrides in inline asm references to local variables or parameters on x86
git-svn-id: trunk@38392 -
2018-03-01 17:59:57 +00:00
nickysn
df6a870873 * print an "invalid segment override expression" error message in case a segment
override uses an invalid (non-segment) register in the x86 inline asm reader
  (both for intel and att syntax)

git-svn-id: trunk@38391 -
2018-03-01 15:34:16 +00:00
nickysn
c93c15429d + introduce tx86intreader.SetSegmentOverride(); using it everywhere adds more
checks for duplicated segment overrides (which cause a warning in TP mode and
  an error in all the other compiler modes)

git-svn-id: trunk@38384 -
2018-02-28 17:16:50 +00:00
nickysn
6f8abde786 + support [const+[ref]] in the x86 intel syntax asm reader
git-svn-id: trunk@38383 -
2018-02-28 16:24:45 +00:00
nickysn
8ffde52974 + also support [expr1]+[expr2] for adding references in the x86 intel syntax asm reader
git-svn-id: trunk@38381 -
2018-02-28 15:48:48 +00:00
florian
2385c47c28 * compilation on x86-64 fixed
git-svn-id: trunk@38370 -
2018-02-27 21:54:12 +00:00
florian
8c5606b41d + support mmx shifting
git-svn-id: trunk@38367 -
2018-02-27 21:40:12 +00:00
nickysn
9c814e42c9 + support segment overrides inside references (e.g. [es:bx]) in the x86 intel
syntax inline asm reader

git-svn-id: trunk@38365 -
2018-02-27 17:29:46 +00:00
nickysn
1cf1ab8ab9 * consume the register in the reference before the check for invalid reference
syntax in the x86 intel syntax asm reader; this is preparation for support of
  segment overrides inside the reference expression (i.e. [es:bx] instead of
  es:[bx])

git-svn-id: trunk@38363 -
2018-02-27 16:35:55 +00:00
nickysn
f623038da6 + fixed a bug in the previous commit, when adding two references, the first
containing a base register, the second - an index register with a scalefactor.
  The scalefactor was ignored in this case.

git-svn-id: trunk@38354 -
2018-02-26 17:30:08 +00:00
nickysn
45fdd7655d + support concatenation of references in x86 intel syntax inline asm:
[expr1][expr2] = [expr1+expr2]
  [expr1[expr2]] = [expr1+expr2]
  This is compatible with TP7's inline asm, and perhaps also with tasm/masm/delphi.

git-svn-id: trunk@38352 -
2018-02-26 17:17:47 +00:00
florian
c5f8567ed7 * getsubreg => getsupreg
git-svn-id: trunk@38344 -
2018-02-25 15:34:14 +00:00
florian
fc6c0e8ef4 + AndShlToShl optimization
* moved topsize2memsize to cpubase

git-svn-id: trunk@38343 -
2018-02-25 15:34:12 +00:00
florian
6e811d057c * MovOpMov2Op cannot be applied for L, Q, Q as operand sizes
+ test

git-svn-id: trunk@38337 -
2018-02-25 10:19:52 +00:00
florian
98bd768521 * remove type cast of the parameter of int(...), if an instruction is available which can be used directly
git-svn-id: trunk@38335 -
2018-02-24 23:13:27 +00:00
florian
3b779278e2 + (slightly) patch by Emelyanov Roman to add support of SEH directive in FPC internal assembler with INTEL syntax, resolves #29894
git-svn-id: trunk@38331 -
2018-02-24 16:14:08 +00:00
nickysn
506d5fe30e + support bracketless references in the x86 intel syntax; ugly, but TP7 (and
perhaps also Delphi, TASM and MASM)-compatible

git-svn-id: trunk@38296 -
2018-02-20 17:18:09 +00:00
florian
e92422383a * compilation fixed
git-svn-id: trunk@38280 -
2018-02-18 09:29:36 +00:00
florian
91514da267 * factored out TX86AsmOptimizer.PostPeepholeOptCall
+ use TX86AsmOptimizer.PostPeepholeOptCall on x86-64

git-svn-id: trunk@38278 -
2018-02-17 23:25:01 +00:00
florian
1b3627add1 + peephole optimization MovMov2Mov 5
git-svn-id: trunk@38273 -
2018-02-17 21:10:51 +00:00
florian
5eb59196d5 * remove sequential moves to the same register
+ test

git-svn-id: trunk@38267 -
2018-02-17 12:45:17 +00:00
florian
99f1fe54af * check for registers removed which is not needed anymore
git-svn-id: trunk@38265 -
2018-02-17 12:45:13 +00:00
florian
f66a91499d * FPC uses meanwhile more mov instructions, so extended taicpu.is_same_reg_move to support them
git-svn-id: trunk@38264 -
2018-02-17 09:45:19 +00:00
pierre
55289c13e4 Use af_no_stabs for clang assembler which does not support stabs debug format
git-svn-id: trunk@38238 -
2018-02-14 15:09:35 +00:00
florian
381cf78ff1 * there is no vmovq for mmx registers
git-svn-id: trunk@38208 -
2018-02-11 17:50:41 +00:00
florian
31f78ea2b6 + implementation of the vectorcall calling convention by J. Gareth Moreton
+ tests

git-svn-id: trunk@38206 -
2018-02-11 17:50:37 +00:00
florian
092223f400 * comment links to test now
git-svn-id: trunk@38204 -
2018-02-11 16:38:33 +00:00
florian
e5ebc65cce * if si and di are allocated on i8086, using an index in references is not possible anymore
git-svn-id: trunk@38203 -
2018-02-11 15:54:37 +00:00
nickysn
ae6a4030c3 + support evaluation of recordtype*constant in the intel syntax inline asm.
The recordtype itself is evaluated to 0. This makes e.g.
  'test [di+recordtype*5], 1' work. This is TP7 compatible.

git-svn-id: trunk@38200 -
2018-02-11 03:42:45 +00:00
nickysn
f0765421eb + also set the operand size in constants like recordtype (without addressing a
record field). This makes e.g.
    test [di + recordtype], 1
  work and use the size of recordtype to determine the operand size; recordtype
  itself is evaluated to 0, so if recordtype's size is 2 bytes, the above
  instruction assembles as:
    test word ptr [di], 1
  Ugly, but TP7 compatible.

git-svn-id: trunk@38176 -
2018-02-09 17:43:31 +00:00
nickysn
7338437dcd * allow recordtype.recordfield constants to set the operand size; this makes things like
test [di + recordtype.recordfield], 1
  work, as long as the size of recordfield is a valid operand size for the target

git-svn-id: trunk@38175 -
2018-02-09 17:36:54 +00:00
nickysn
f829f70186 * use the 'size' return value of BuildConstSymbolExpression to set the operand size
git-svn-id: trunk@38174 -
2018-02-09 17:22:39 +00:00
nickysn
384715be8d + add an extra output parameter 'size' to tx86intreader.BuildConstSymbolExpression,
which allows const symbol expressions to also have a size sometimes. Why?
  Because TP7 (and perhaps Delphi) allows not specifying the size in e.g.
    test [di+recordtype.recordfield], 1
  in this case, the operand size (byte ptr, word ptr, dword ptr, qword ptr) is
  determined by the size of recordtype.recordfield; this already happens with
  variables, but in this case, this is a type.field, which is resolved to a
  constant.
  This commit only adds a dummy 'size' parameter, which is always initialized to
  0 and not used. The actual implementation of the above will follow in separate
  commits.

git-svn-id: trunk@38173 -
2018-02-09 16:52:12 +00:00
nickysn
4b339fd133 * convert the 'var' parameters of tx86intreader.BuildConstSymbolExpression to 'out'
git-svn-id: trunk@38172 -
2018-02-09 16:20:34 +00:00
nickysn
84611d716b * convert the 'var' parameters of tx86intreader.BuildRecordOffsetSize to 'out'
git-svn-id: trunk@38171 -
2018-02-09 16:15:40 +00:00
nickysn
391f85f828 + for TP7 compatibility, allow the '&', '$' and '?' characters in the x86 intel
syntax inline asm reader

git-svn-id: trunk@38167 -
2018-02-08 16:54:33 +00:00
nickysn
5f56f6ec11 * always pass a valid constsize to ConcatConstSymbol; this prevents a compiler
internal error on i8086, generated by 'DB xx' or 'DB OFFSET xx'

git-svn-id: trunk@38153 -
2018-02-07 16:17:05 +00:00
nickysn
a792a92cfb * prevent the spurious warning, generated by 'DD xx' and 'DD OFFSET xx' on i8086
git-svn-id: trunk@38152 -
2018-02-07 16:14:05 +00:00
nickysn
3ad3807b6f + accept the 'DB/DW/DD OFFSET xx' syntax
git-svn-id: trunk@38150 -
2018-02-07 16:07:59 +00:00
nickysn
3b1bd8de3f + actually set hasofs to true if 'offset' is encountered in BuildConstSymbolExpression
git-svn-id: trunk@38149 -
2018-02-07 16:03:40 +00:00
nickysn
d591e698ac + added parameter 'hasofs' to ConcatConstSymbol. Will be used to support 'dd xx'
vs 'dd offset xx' being different on i8086.

git-svn-id: trunk@38148 -
2018-02-07 15:23:48 +00:00
nickysn
70038f1c0b + added extra boolean out parameter 'hasofs' to
tx86intreader.BuildConstSymbolExpression; it returns whether the 'OFFSET'
  keyword has been used in the expression. This will be used for disambiguation
  between 'dd xx' and 'dd offset xx', because they should produce different
  results on i8086 (the first generates a far pointer, i.e. the same as
  'dw xx, SEG xx', the second - a 32-bit offset)

git-svn-id: trunk@38147 -
2018-02-07 15:09:54 +00:00
nickysn
629be9cee1 + add a 'constsize' parameter to ConcatConstSymbol, so it can be used to
distinguish between e.g. 16-bit and 32-bit offsets on i8086

git-svn-id: trunk@38141 -
2018-02-06 16:22:41 +00:00
florian
e6a6938787 * make fpu/mmx/xmm/ymm registers numbers instead of flags to have enough space for zmm/bnd/k registers
git-svn-id: trunk@38116 -
2018-02-04 20:29:41 +00:00
nickysn
1f0e311fdd + allow segment override prefixes to be used as a standalone opcode in the intel
syntax inline assembler; this is TP7 compatible and allows compiling ugly
  code, such as 'seges; db $67,$66; lodsw'

git-svn-id: trunk@38096 -
2018-02-01 17:31:59 +00:00