-
Notifications
You must be signed in to change notification settings - Fork 64
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[RISCV64]Fedora(on QEMU) run test,FatalError: stack reset position two #23
Comments
Hi, We have tested OpenJDK 8 (u222) (https://maxine-vm.readthedocs.io/en/latest/Status.html#platform-label) for RISC-V. Thanks |
Hello Prof.kotselidis @kotselidis !
|
I would suggest yes to change it, regarding cross compilation I am not sure we have tested it, @gigiblender do you remember? |
Hi @zdlgv5, you don't need to cross-compile the JDK. Fedora comes with 8u212 in its repositories and this should work with maxine. See http://fedora.riscv.rocks/koji/buildinfo?buildID=87633 https://github.com/beehive-lab/Maxine-VM-internal/blob/develop/docs/Status.rst looks wrong to me, it should say 212 for RISC-V not 222. |
Oh sorry, I think I miss-read your comment, it looks like you are cross-compiling the image not the JDK itself. This should work @kotselidis, see https://github.com/beehive-lab/Maxine-VM-internal/blob/develop/docs/Working-with-RISCV-on-QEMU.rst#building-maxine-vm-for-risc-v. |
@zdlgv5 I would use https://github.com/AdoptOpenJDK/openjdk8-binaries/releases/tag/jdk8u212-b04 instead of jdk8u222 to make sure the versions match. |
@zakkak thank you for reply! |
No unfortunately I don't. Maybe @gigiblender, @bakaliosdim, or @evankar can help you here. |
@zakkak I think xalan was running fine when I tested it. If @bakaliosdim or @evankar still have the different results on SpecJVM or DaCapo, I am happy to update the status page with those. If not, then I can run the benchmarks and get the results again |
I had tested all DaCapo RISCV64-test listed on https://github.com/beehive-lab/Maxine-VM/blob/develop/docs/Status.rst#dacapo-912-bach-mr1 and have the different results on Here is my image: https://github.com/zdlgv5/Maxine-VM_RISCV64_Dacapo_Test/blob/master/com.oracle.max.vm.native.zip, which followed the instructions in Build and Usage Instructions and Running Maxine VM on the Fedora RISC-V VM |
@zdlgv5 I just ran the tests again and for me
Can you run again the test so we can see if they behave the same? |
@bakaliosdim Oh sorry , I think that I forgot to point out that I add I just ran again with no param e.g. The results are below:
The result of sunflow:
Qemu start script:
Fedora info:
Host Ubuntu info:
|
I suggest adding |
I ran the benchmarks again (on a dual core setup in Qemu but in single threaded mode for the benchmarks) before and after #19 and obtained the results below. I will list only the failures: Before the trampolines patch #19 (0ab4793) :
SpecJVM2008:
After the trampolines, current develop (e213a84) :
SpecJVM2008:
As @zakkak mentioned, most of the regressed benchmarks fail in the deoptimization phase and others bailout due to the method body being larger than 1 MB |
Nice digging @gigiblender , so it looks like someone needs to fix deopt and possibly the 1MB bailout or we should beter revert #19 and keep in the backlog. |
I had followed the instructions in Build and Usage Instructions and Running Maxine VM on the Fedora RISC-V VM but it seems had some problems about stack, Is it the same question as #20 ? Thx a lot.
The text was updated successfully, but these errors were encountered: