[gem5-users] Re: HPCG on RISCV

Wed, 28 Sep 2022 06:12:49 -0700 

Dear Jason,
I am trying to use --debug-start but in FS mode it is very difficult to find the tick on which the application is started! 

However, I am writing the following very simple c++ program:

#include <cmath>
#include <stdio.h>

int main(){

    int dim = 4096;

    double result;

    for (int iter = 0; iter < 2; iter++){
        result = 0;
        for (int i = 0; i < dim; i++){
            for (int j = 0; j < dim; j++){
                result += sqrt(i) * sqrt(j);
            }
        }
        printf("Result: %lf\n", result); //Result: 30530733453.127449
    }
}


I cross-compile it using: riscv64-linux-gnu-g++ -static -O3 -o  
test_riscv test_riscv.cpp




While in X86 (without cross-compilation of course), QEMU-RISCV,  
GEM5-SE the result is the same (30530733453.127449), in GEM5-FS the  
result is different! In addition, the result is also different between  
the 2 iterations.


Please reproduce the error if you want in order to verify my result.
Ηow can the issue be resolved?

Thank you in advance!

Best regards,
Nikos


Quoting Jason Lowe-Power <ja...@lowepower.com>:


Hi Nikos,

You can use --debug-start to start the debugging after some number of
ticks. Also, I would expect that the difference should come up quickly, so
no need to run the program to the end.

For the FS mode one, you will want to just start the trace as the
application starts. This could be a bit of a pain.

I'm not really sure what fundamentally could be different. FS and SE mode
use the exact same code for executing instructions, so I don't think that's
the problem. Have you tried running for smaller inputs or just one
iteration?

Jason



On Wed, Sep 21, 2022 at 9:04 AM Νικόλαος Ταμπουρατζής <
ntampourat...@ece.auth.gr> wrote:


Dear Bobby,

Iam trying to add --debug-flags=Exec (building the gem5 for gem5.opt
not for gem5.fast which I had) but the debug traces exceed the 20GB
(and it is not finished yet) for less than 1 simulated second. How can
I reduce the size of the debug-flags (or set something more specific)?

In contrast I build the HPCG benchmark with DHPCG_DEBUG flag. If you
want, you can compare these two output files
(hpcg20010909T014640_SE_Mode & HPCG-Benchmark_3.1_FS_Mode). As you can
see, something goes wrong with the accuracy of calculations in FS mode
(benchmark uses double precission). You can find the files here:
http://kition.mhl.tuc.gr:8000/d/68d82f3533/

Best regards,
Nikos

Quoting Jason Lowe-Power <ja...@lowepower.com>:

> That's quite odd that it works in SE mode but not FS mode!
>
> I would suggest running with --debug-flags=Exec for both and then
perform a
> diff to see how they differ.
>
> Cheers,
> Jason
>
> On Tue, Sep 20, 2022 at 2:45 PM Νικόλαος Ταμπουρατζής <
> ntampourat...@ece.auth.gr> wrote:
>
>> Dear Bobby,
>>
>> In QEMU I get the same (correct) results that I get in SE mode
>> simulation. I get invalid results in FS simulation (in both
>> riscv-fs.py and riscv-ubuntu-run.py). I cannot access real RISCV
>> hardware at this moment, however, if you want you may execute my xhpcg
>> binary (http://kition.mhl.tuc.gr:8000/f/4ca25fdd3c/) with the
>> following configuration:
>>
>> ./xhpcg --nx=16 --ny=16 --nz=16 --npx=1 --npy=1 --npz=1 --rt=0.1
>>
>> Please let me know if you have any updates!
>>
>> Best regards,
>> Nikos
>>
>>
>> Quoting Jason Lowe-Power <ja...@lowepower.com>:
>>
>> > Hi Nikos,
>> >
>> > I notice you said the following in your original email:
>> >
>> > In addition, I used the RISCV Ubuntu image
>> >> (https://github.com/gem5/gem5-resources/tree/stable/src/riscv-ubuntu
),
>> >> I installed the gcc compiler, compile it (through qemu) and I get
>> >> wrong results too.
>> >
>> >
>> > Is this saying you get the wrong results is QEMU? If so, the bug is in
>> GCC
>> > or the HPCG workload, not in gem5. If not, I would test in QEMU to
make
>> > sure the binary works there. Another way you could test to see if the
>> > problem is your binary or gem5 would be to run it on real hardware. We
>> have
>> > access to some RISC-V hardware here at UC Davis, if you don't have
access
>> > to it.
>> >
>> > Cheers,
>> > Jason
>> >
>> > On Tue, Sep 20, 2022 at 12:58 AM Νικόλαος Ταμπουρατζής <
>> > ntampourat...@ece.auth.gr> wrote:
>> >
>> >> Dear Bobby,
>> >>
>> >> 1) I use the original riscv-fs.py which is provided in the latest
gem5
>> >> release.
>> >> I run the gem5 once (./build/RISCV/gem5.fast -d ./HPCG_FS_results
>> >> ./configs/example/gem5_library/riscv-fs.py) in order to download the
>> >> riscv-bootloader-vmlinux-5.10 and riscv-disk-img.
>> >> After this I mount the riscv-disk-img (sudo mount -o loop
>> >> riscv-disk-img /mnt), put the xhpcg executable and I do the following
>> >> changes in riscv-fs.py to boot the riscv-disk-img with executable:
>> >>
>> >> image = CustomDiskImageResource(
>> >>      local_path = "/home/cossim/.cache/gem5/riscv-disk-img",
>> >> )
>> >>
>> >> # Set the Full System workload.
>> >> board.set_kernel_disk_workload(
>> >>                     kernel=Resource("riscv-bootloader-vmlinux-5.10"),
>> >>                     disk_image=image,
>> >> )
>> >>
>> >> Finally, in the gem5/src/python/gem5/components/boards/riscv_board.py
>> >> I change the last line to "return ["console=ttyS0",
>> >> "root={root_value}", "rw"]" in order to allow the write permissions
in
>> >> the image.
>> >>
>> >>
>> >> 2) The HPCG benchmark after some iterations calculates if the results
>> >> are valid or not valid. In the case of FS it gives invalid results.
As
>> >> I see from the results, one (at least) problem is that produces
>> >> different results in each HPCG execution (with the same
configuration).
>> >>
>> >> Here is the HPCG output and riscv-fs.py
>> >> (http://kition.mhl.tuc.gr:8000/d/68d82f3533/). You may reproduce the
>> >> results in the video if you use the xhpcg executable
>> >> (http://kition.mhl.tuc.gr:8000/f/4ca25fdd3c/)
>> >>
>> >> Please help me in order to solve it!
>> >>
>> >> Finally, I get invalid results in the HPL benchmark in FS mode too.
>> >>
>> >> Best regards,
>> >> Nikos
>> >>
>> >>
>> >> Quoting Bobby Bruce <bbr...@ucdavis.edu>:
>> >>
>> >> > I'm going to need a bit more information to help:
>> >> >
>> >> > 1. In what way have you modified
>> >> > ./configs/example/gem5_library/riscv-fs.py? Can you attach the
script
>> >> here?
>> >> > 2. What error are you getting or in what way are the results
invalid?
>> >> >
>> >> > -
>> >> > Dr. Bobby R. Bruce
>> >> > Room 3050,
>> >> > Kemper Hall, UC Davis
>> >> > Davis,
>> >> > CA, 95616
>> >> >
>> >> > web: https://www.bobbybruce.net
>> >> >
>> >> >
>> >> > On Mon, Sep 19, 2022 at 1:43 PM Νικόλαος Ταμπουρατζής <
>> >> > ntampourat...@ece.auth.gr> wrote:
>> >> >
>> >> >>
>> >> >> Dear gem5 community,
>> >> >>
>> >> >> I have successfully cross-compile the HPCG benchmark for RISCV
>> (Serial
>> >> >> version, without MPI and OpenMP). While it working properly in
gem5
>> SE
>> >> >> mode (./build/RISCV/gem5.fast -d ./HPCG_SE_results
>> >> >> ./configs/example/se.py -c xhpcg --options '--nx=16 --ny=16
--nz=16
>> >> >> --npx=1 --npy=1 --npz=1 --rt=0.1'), I get invalid results in FS
>> >> >> simulation using "./build/RISCV/gem5.fast -d ./HPCG_FS_results
>> >> >> ./configs/example/gem5_library/riscv-fs.py" (I mount the riscv
image
>> >> >> and put it).
>> >> >>
>> >> >> Can you help me please?
>> >> >>
>> >> >> In addition, I used the RISCV Ubuntu image
>> >> >> (
https://github.com/gem5/gem5-resources/tree/stable/src/riscv-ubuntu
>> ),
>> >> >> I installed the gcc compiler, compile it (through qemu) and I get
>> >> >> wrong results too.
>> >> >>
>> >> >> Here is the Makefile which I use, the hpcg executable for RISCV
>> >> >> (xhpcg), and a video that shows the results
>> >> >> (http://kition.mhl.tuc.gr:8000/f/4ca25fdd3c/).
>> >> >>
>> >> >> P.S. I use the latest gem5 version.
>> >> >>
>> >> >> Thank you in advance! :)
>> >> >>
>> >> >> Best regards,
>> >> >> Nikos
>> >> >> _______________________________________________
>> >> >> gem5-users mailing list -- gem5-users@gem5.org
>> >> >> To unsubscribe send an email to gem5-users-le...@gem5.org
>> >> >>
>> >>
>> >>
>> >> _______________________________________________
>> >> gem5-users mailing list -- gem5-users@gem5.org
>> >> To unsubscribe send an email to gem5-users-le...@gem5.org
>> >>
>>
>>
>> _______________________________________________
>> gem5-users mailing list -- gem5-users@gem5.org
>> To unsubscribe send an email to gem5-users-le...@gem5.org
>>


_______________________________________________
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org




_______________________________________________
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org














    

  • [gem5-users] HPCG... Νικόλαος Ταμπουρατζής
    • 

    • 

    • [gem5-users]... Bobby Bruce
      • 

      • 

      • [gem5-us... Νικόλαος Ταμπουρατζής
        • 

        • 

        • [gem... Jason Lowe-Power
          • 

          • 

          • ... Νικόλαος Ταμπουρατζής
            • 

            • 

            • ... Jason Lowe-Power
              • 

              • 

              • ... Νικόλαος Ταμπουρατζής
                • 

                • 

                • ... Jason Lowe-Power
                  • 

                • ... Νικόλαος Ταμπουρατζής
                  • 

                • ... Νικόλαος Ταμπουρατζής
                  • 

                • ... Νικόλαος Ταμπουρατζής
                  • 

                • ... Bobby Bruce
                  • 

                • ... Νικόλαος Ταμπουρατζής
                  • 

                • ... Jason Lowe-Power
                  • 

                • ... Νικόλαος Ταμπουρατζής
                  • 

                • ... Bobby Bruce
                  • 

                • ... Νικόλαος Ταμπουρατζής via gem5-users
                  • 

                • ... Bobby Bruce via gem5-users
                  • 

                • ... Νικόλαος Ταμπουρατζής via gem5-users
                  • 

                

              

            

          

        

      

    










					Reply via email to