gh-mirror/mikepaul-LuaJIT
1
0
Fork 0
mirror of https://github.com/LuaJIT/LuaJIT.git synced 2025-02-08 15:34:09 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
2,301 Commits 3 Branches 28 Tags 16 MiB
77852cbe0b
Commit Graph

2301 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Michael Munday
77852cbe0b Don't bother saving BASE before a call unless it will be modified. 
BASE is callee-saved anyway, so we don't need to save it in RB.
 2017-01-17 17:23:53 -05:00
Michael Munday
e151edea53 Avoid saving/restoring floating point registers when entering the interpreter. 
We only need to worry about doing this if we actually use those
floating point registers.
 2017-01-17 14:05:31 -05:00
Michael Munday
695c59703c Swap register assignments for BASE and RB. 
Feels more natural this way round. Puts all parameters in the range
[r4,r7] and BASE is now the register used as the literal pool, which
seems appropriate.
 2017-01-17 09:46:02 -05:00
Michael Munday
a40e905aec Implement return hooks. 2017-01-16 16:28:21 -05:00
Michael Munday
828bf84999 Rename TMPR2 as TMPR0. 
TMPR2 was r0 and so cannot be used in address calculations (or
shift values). Renaming it TMPR0 makes this more obvious.
 2017-01-16 16:14:51 -05:00
Michael Munday
70b5f6b966 Make TMPR1 r1 and RA r4. 
This avoids using the link register as a temporary.
 2017-01-16 16:11:24 -05:00
Michael Munday
a9d61d0044 Implement debug.sethook(). 2017-01-16 15:52:23 -05:00
Michael Munday
9e9a1cf0a5 Replace lay with la where possible. 2017-01-11 16:55:50 -05:00
Michael Munday
a3bb1cee5d Add and use branch on count instructions where possible. 2017-01-11 16:38:35 -05:00
Michael Munday
c8db1b960c Clean up register allocations and comments slightly. 
Move RB from r12 to r13 so that it no longer overlaps with the
GOT pointer (to avoid potential problems with PIC compilation).
 2017-01-11 16:16:51 -05:00
Michael Munday
084ab26280 Delete LREG and replace it's uses with RB. 
This frees up a register and fixes a bug where RB was not loaded
correctly into LREG.
 2017-01-11 14:42:24 -05:00
Michael Munday
65906fbc49 Fix typo in lj_ccallback.h 
For future reference only, we aren't using this bit of code yet.
 2017-01-11 12:46:30 -05:00
Michael Munday
5c697cc772 Fix ffi calls returning structs. 2017-01-11 12:13:59 -05:00
Michael Munday
f660d36138 Fix ffi calls with complex parameters. 2017-01-11 12:13:52 -05:00
ketank-new
db99c31890 Added example for 'TM' instruction 2017-01-11 17:04:09 +05:30
Michael Munday
938f964b6a Fix single-precision floating point parameters passed on stack. 
The opposite way round to the registers for some reason.
 2017-01-10 16:56:43 -05:00
Michael Munday
86a24eac4a Various fixes for FFI calls. 
Fixes the following scenarios:
 * Returning floating point value.
 * More than 4 GPR arguments.
 2017-01-10 16:11:30 -05:00
Michael Munday
e933353feb Use execute rather than loop for mvc and avoid jumps in fast path. 
Not sure if this works, the tests don't exercise the stack code.
 2017-01-10 14:12:06 -05:00
Michael Munday
660ddd1db2 Replace some instruction sequences with their SI equivalents. 2017-01-10 12:52:45 -05:00
Michael Munday
21655cf90b Add remaining (useful) SI instructions to DynASM. 2017-01-10 12:52:02 -05:00
Michael Munday
55af819e97 Remove TODO. 2017-01-10 12:04:17 -05:00
Michael Munday
102d848577 Minor instruction changes. 2017-01-10 11:58:24 -05:00
Michael Munday
d44390a2d3 Use tm{,y} instructions where possible. 2017-01-10 11:44:27 -05:00
Michael Munday
a038a08189 Fix SI (tm) action parsing. 2017-01-10 11:37:25 -05:00
Michael Munday
a8562b7f34 Allow displacements to be used directly without register values. 
Allows sllg r1, r1, 3(r0,r0) to be written as sllg r1, r1, 3.
 2017-01-10 10:50:41 -05:00
Michael Munday
dae61f59d6 Delete some unused function stubs. 2017-01-09 15:57:37 -05:00
Michael Munday
4c738134df Improve ins_NEXT performance. 
Prioritise critical path and reduce number of instructions. About
10% improvement on md5 benchmark.
 2017-01-09 14:16:44 -05:00
Michael Munday
99b3668995 Add stub dis_s390x.lua file to allow make install to work. 2017-01-09 11:20:13 -05:00
Michael Munday
bb98985db3 Add partial FFI support. 
Interestingly, enough to pass all the FFI tests. So s390x now
passes all the tests in LuaJIT-test-cleanup.
 2017-01-06 16:19:56 -05:00
Michael Munday
5dec8c2211 Fix vm_tsetr (needed by table.remove). 
The A argument was being loaded as 2-bytes instead of 1.
 2017-01-06 13:54:28 -05:00
Michael Munday
ed2d43d73c Fix VARG. 2017-01-06 11:47:56 -05:00
Michael Munday
2e05e7ca67 Implement math.floor/math.ceil. 2017-01-06 11:16:33 -05:00
Michael Munday
443814b6b3 Add more convert to/from fixed instructions to DynASM. 2017-01-06 11:16:04 -05:00
Michael Munday
a3501b062d Various fixes for coroutines. 
Now passing the tests.
 2017-01-05 23:39:25 -05:00
Michael Munday
d90293f55e Implement coroutines. 
TODO: delete LREG, caused problems while implementing this (x64
doesn't have LREG).
 2017-01-05 23:39:25 -05:00
Michael Munday
60fb35cb68 Implement rawget. 2017-01-05 23:39:25 -05:00
ketank-new
c08fa1c119 Enabled math.pow ,math.min and math.max 2017-01-06 10:01:36 +05:30
Michael Munday
1b16e7ce3c Fix math.pow. 
The second floating point argument is placed into f2, not f1.
Use the macros FARG{1,2} instead of using the registers directly.
 2017-01-05 10:52:27 -05:00
ketank-new
3da26860b7 Updated file to display table contents 2017-01-05 14:50:23 +05:30
niravthakkar
206c650689 Updated the memory parsing 
The order matters here, so just moved displacement check to end
 2017-01-05 14:48:09 +05:30
ketank-new
dcb1dd74ed Added table Manipulation Functions 2017-01-05 14:40:19 +05:30
ketank-new
fe42519c16 Added String processing Functions 2017-01-05 13:53:18 +05:30
Michael Munday
dfb95646f8 Implement call_tail. 2017-01-04 16:05:55 -05:00
Michael Munday
faef0fb092 Fix for __newindex metamethod. 2017-01-04 15:54:21 -05:00
Michael Munday
80191dcab7 Implement cont_cat. 
Required to pass cat tests.
 2017-01-04 15:34:30 -05:00
Michael Munday
e598c067c3 Fixes for negative string.sub arguments and __index metamethod calls. 2017-01-04 14:50:19 -05:00
Michael Munday
705784a4f9 Implement math.min and math.max. 
Replicates the standard Lua behaviour in the presence of NaNs.
 2017-01-04 11:48:07 -05:00
niravthakkar
7aae451d93 Reverting the changes, as its breaking the build 
The above expression works on CLI, but its failing here, not sure whats going wrong , Please let me know your comments on it
 2017-01-04 18:15:57 +05:30
ketank-new
b92584b497 Added test for OS based functions 2017-01-04 17:45:19 +05:30
niravthakkar
1fe2176241 Updated memory parsing 
The values of base and index registers have been passed as 0, if only displacement is passed
the displacement is assumed to be alphanumeric (since label might be used)
 2017-01-04 17:43:53 +05:30