 pierre
|
2916235cfe
Fix make_simple_ref for pic address with big offsets
|
12 years ago |
 sergei
|
d82387ff72
* MIPS: fixed parameter management to properly align records smaller than 32 bits on big-endian targets.
|
12 years ago |
|
sergei
|
ef38193043
* MIPS: Once again fixed g_external_wrapper. My initial assumption about .cpload was wrong: it is optimized into position-dependent sequence only if "-KPIC -mno-shared" is passed to GAS command line, without -KPIC it is ignored altogether. Therefore r23566 breaks some non-PIC tests.
|
12 years ago |
|
sergei
|
9bc758d62b
* Forgot to set base register to NR_GP
|
12 years ago |
|
sergei
|
a0c1e1b07f
* MIPS: fixed a_call_name and a_call_reg methods to workaround a bug in GAS <2.21, see comments in source for details.
|
12 years ago |
|
sergei
|
8bfbee747a
* MIPS: Rewrote g_intf_wrapper method to support instance offsets greater than 32767 and PIC (the latter part untested yet).
|
12 years ago |
|
pierre
|
7e79042d58
Reorganize for better pic support (not yet working completly ...)
|
12 years ago |
|
sergei
|
162896e7d7
* MIPS code generator fixes and improvements:
|
12 years ago |
 Jonas Maebe
|
69c29a415f
* pass the procdef to getintparaloc instead of only the proccalloption, so
|
12 years ago |
|
sergei
|
db72b8d7fd
* TCGMips.a_loadfpu_reg_cgpara: temps of type Double need 8-byte alignment, according to description of sdcX/ldcX instructions. Using TCGSize2Size to specify alignment is somewhat weird, but it is being used in other CPU back-ends and looks working.
|
12 years ago |
|
pierre
|
ff78c930c5
Fix problem in 64bit substraction
|
13 years ago |
|
pierre
|
15f4e1279b
Add RS_GP to list of saved registers if pi_needs_got flag is set
|
13 years ago |
|
pierre
|
c72f27e53e
Remove more TABs in source
|
13 years ago |
|
pierre
|
80b0891c7c
Remove TABs in source
|
13 years ago |
|
pierre
|
9ad6285fe2
* Fix 64-bit OP_ADD and OP_SUB operations (no overflow support yet)
|
13 years ago |
|
pierre
|
05a326a83e
Make sure f2X+1 is also considered used if f2X is used
|
13 years ago |
|
pierre
|
ffd01794fb
Reserve R25 () register for PIC function calling
|
13 years ago |
|
pierre
|
bc0c94c204
fpcdefs.inc: Set fpc_compiler_has_fixup_jmps for powerpcXX and mips CPUs.
|
13 years ago |
|
pierre
|
6f12197d3b
Fix set of frame poointer value for big local stack size
|
13 years ago |
|
pierre
|
a771b32097
Use R9 instead of AT for big local stack size to avoid error due to AT being changed by pseudo-instructions
|
13 years ago |
|
pierre
|
2e4c704f06
Use A_J instead of A_BA for a_jmp_XXX unless we generate PIC code to avoid branch out of range error (like for packages/fpmake compilation)
|
13 years ago |
|
pierre
|
7982b34416
Start of PIC code support
|
13 years ago |
|
pierre
|
7492f196f8
* Handle OS_F32 <-> OS_F64 float conversions
|
13 years ago |
|
Jonas Maebe
|
7a0ae38700
+ also specify the parameter def when allocating a parameter via
|
13 years ago |
|
pierre
|
74eb121d1c
Fix big endian issues
|
13 years ago |
|
pierre
|
0462dc3ca7
* Use A_NEGU (without overflow check) for OP_NEG op_reg_reg
|
13 years ago |
|
pierre
|
c87ee7f100
Merge of 21562 by Foxsen
|
13 years ago |
|
pierre
|
e8c7635d49
Merge of rev21561 by Foxsen
|
13 years ago |
|
pierre
|
d6828965fd
Merge of revision 21552 by Foxsen.
|
13 years ago |
|
pierre
|
93e0dd9c2f
* Patch from Fuxin Zhang: other mips and mipsel CPUs changes
|
13 years ago |
