Hi all,
I submitted another small pull request that fixes IR_ADD/SUB. It allows
them to work with 64-bit values. The issue can be reproduces with
vload.lua test.
The next thing that is a blocker for quite a lot of tests is
asm_head_side_base. Is anyone working on that? If not, I will look into
it.
Thanks,
Stefan
Hi,
I submitted pull request which contains implementation of asm_tvptr
function. Note that it isn't fully tested, since I had difficulties
creating simple examples that would cover all parts of the function.
Now, I'm going to work on asm_fstore function.
Regards,
Djordje
Hi All
I was trying one of the test cases in the "test" directory and it seems to fail while trying to dump the machine code both on arm32 and arm64.
./luajit -jdump=m ../test/tnew.lua
---- TRACE 1 start tnew.lua:1
---- TRACE 1 mcode 316
VM handler failed: ./jit/dis_arm.lua:441: attempt to call upvalue 'concat' (a nil value)
Am I doing something wrong ? or is this a known issue ?
Test case:tnew.lua
---------------------------
for i=1, 100 do
table = {}
end
assert(type(table) == "table", "failed to create table")
---------------------------
Abhishek Dikshit
Hi, all,
I run into Segment Fault in:
======
x = 0
for i = 1, 100 do
x = tostring(i)
end
y = "101"
======
And it turns out the JITed code overwrites native stack position at trace
head. It can be walked around by set SPS_FIXED to 0.
The JITed code is doing register spill and assuming there is "2 fixed spill
slots" (defined by SPS_FIXED). But there seems no "fixed spill slots" in
arm64 interpreter implementation. Is it true? Could anyone provide some
info of aarch64 frame layout?
Any input will help! Thanks.
--
Best regards,
Zhongwei
Hi everyone,
I fixed asm_fload function and implemented asm_strref function (although
asm_strref is not fully tested, yet). The next thing on my list is
working on the asm_tvptr function. If someone is already working on that
part, let me know.
Djordje
--
Đorđe Kovačević
Software engineer
Phone +381-(0)21-480-1207
djordje.lj.kovacevic(a)rt-rk.com
RT-RK Institute for Computer Based Systems
Narodnog Fronta 23a
21000 Novi Sad
Serbia
www.rt-rk.com
Hi All
The immr and imms values are decoded to get an immediate value. Is there a way to encode from immediate value to immr and imms field for ARMv8 ?
Instruction
[cid:image005.png@01D1E75A.8DE84F20]
Routine.
[cid:image006.png@01D1E75A.8DE84F20]
The DecodeBitMasks routine is not very clear. Any help is appreciated.
Abhishek Dikshit
Hi All,
I am looking forward to contribute in development of LuaJIT for ARM64.
Please feel free to assign work to me even if it is a minor bug fix.
Thanks
Abhishek Dikshit
Hello all,
I finished asm_ahustore and submitted a pull request. Apart from
asm_ahustore, it contains some additional changes to the interpreter
code, that caused me major problems. It enabled a few tests in the
current test suite (array, hash and a few more), but it didn't solve
everything for upvale tests, which fail because of asm_fuseahuref.
I thought to take a look at unimplemented paths of asm_fuseahuref, if
that's alright with everybody.
Stefan
Hi All
I am trying to make this small loop work on ARM64.
./luajit -jon -e 'x=0; for i=1,400000 do x=x+i end;'
The code flow reaches the point
Unimplemented: asm_conv
When I looked into it, it seems that luajit is trying to convert from INT to FLOAT.
Any ideas why it might be doing that ? Or is it a bug somewhere I need to look ?
Thanks
Abhishek Dikshit