Hi all,
I noticed that interleaved IR dump causes some tests to fail, which
would otherwise pass. I can name a few tests, but there are more:
- alias_alloc.lua
- jit_flush.lua
- unordered_jit.lua
- ffi_nosink.lua
Regards,
Stefan
Hi,
I just check:
tnew and tdup pass after changing variable name ("table").
hload_num hstore_pri hstore_str hstore_num still fail.
On 28 August 2016 at 21:31, Dikshit, Abhishek <Abhishek.Dikshit(a)cavium.com>
wrote:
> After changing the variable name I am not seeing any failure. Can you
> check and confirm if you are still seeing the failure and what is the
> failure ?
>
>
>
> On Sun, Aug 28, 2016 at 2:28 AM -0700, "Zhongwei Yao" <
> zhongwei.yao(a)linaro.org> wrote:
>
> Thanks!
>
> On 27 August 2016 at 03:25, Dikshit, Abhishek <Abhishek.Dikshit(a)cavium.com
> > wrote:
>
>> I will start with tnew
>>
>> -----Original Message-----
>> From: luajit [mailto:luajit-bounces@lists.linaro.org] On Behalf Of
>> Stefan Pejic
>> Sent: Friday, August 26, 2016 7:06 AM
>> To: luajit(a)lists.linaro.org
>> Subject: Re: Recent merged patches break test cases
>>
>> Hey,
>>
>> That pull request wasn't merged. Are you maybe talking about #69?
>> https://github.com/cbaylis/luajit-aarch64/pull/69
>>
>> Stefan
>>
>> On 08/26/2016 08:53 AM, Zhongwei Yao wrote:
>> > Hi, all,
>> > It seems this PL: https://github.com/cbaylis/luajit-aarch64/pull/65
>> > introduces following breaks:
>> > tnew hload_num hstore_pri hstore_str tdup hstore_num
>> >
>> > Is anyone working on fixing it?
>> >
>> > --
>> > Best regards,
>> > Zhongwei
>> >
>> >
>> > _______________________________________________
>> > luajit mailing list
>> > luajit(a)lists.linaro.org
>> > https://lists.linaro.org/mailman/listinfo/luajit
>> >
>> _______________________________________________
>> luajit mailing list
>> luajit(a)lists.linaro.org
>> https://lists.linaro.org/mailman/listinfo/luajit
>> _______________________________________________
>> luajit mailing list
>> luajit(a)lists.linaro.org
>> https://lists.linaro.org/mailman/listinfo/luajit
>>
>
>
>
> --
> Best regards,
> Zhongwei
>
--
Best regards,
Zhongwei
Hi All
I am getting following error when I try to dump the machine code for test/tnew.lua
$ ./luajit -jon -jdump=m ../test/tnew.lua
---- TRACE 1 start tnew.lua:1
---- TRACE 1 mcode 316
VM handler failed: ./jit/dis_arm.lua:439: attempt to call upvalue 'concat' (a nil value)
$ ./luajit -jon -jdump=m ../test/tnew.lua
---- TRACE 1 start tnew.lua:1
---- TRACE 1 mcode 384
VM handler failed: ./jit/dis_arm64.lua:755: attempt to call upvalue 'concat' (a nil value)
The error is seen on ARM64 and ARM32 but not on x64.
I am having tough time to debug the asm_tbar function for failure of test_tnew. Any suggestion is welcome.
Abhishek Dikshit
Hi, all,
It seems this PL: https://github.com/cbaylis/luajit-aarch64/pull/65
introduces following breaks:
tnew hload_num hstore_pri hstore_str tdup hstore_num
Is anyone working on fixing it?
--
Best regards,
Zhongwei
Hi All
I am new to both luajit and ARM. I have made the changes for the functions asm_tbar, asm_cnew and emit_isk13 for arm64 based on my limited understanding. I have attached the diffs. Please review the diffs and share any suggestions you have. The changes are mostly based on arm32 code for corresponding functions.
The diffs are corresponding to test case in the attached email.
Thanks
Abhishek Dikshit
Hi all,
There are more and more tests added to the project. And we are fixing
issues in the common IRs and even common codes. I think it is time to set
up CI for the project to help people to identify issues early.
We have in house CI set up for the project which will do test for arm32,
arm64, x86 and x86_64, but unfortunately it can not be accessed from public
domain. I will try to get the approval from the managers to copy the
scripts to Linaro CI if possible. I will get back to you once I got the
approval or be rejected.
Hi Alan & Fathi,
I saw there are some aarch64 and x86 machines available on ci.linaro.org.
Can we share use some of them for LuaJIT CI? If I am able to copy the
scripts to Linaro CI, the basic requirement for the infrastructure would be
:
- 1 aarch64 machine with ubuntu installed
- 1 x86_64 machine with ubuntu installed
- 1 gerrit project to host the CI scripts
- It is a GitHub project, so we might need some related plugins
installed in the Jenkins system.
- Might need help from your team to run a script to setup the host
environment or setup the host manually. (docker, makefile and some other
dependencies)
- We do not have many tests so far, so the job won't be too large. It
should be able to be finished within 15 minutes.
Can these requirement be satisfied?
Regards,
Zheng
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