Dear Hoa, all,
Yes, the problem has been solved!! :)
I am able to execute HPCG benchmark without accuracy problems in RISCV
FS mode!
Thank you!!
Best regards,
Nikos
Quoting Hoa Nguyen via gem5-users :
Hi Nikos,
The problem you ran into with the FS mode seems to be the same problem
Hi Nikos,
The problem you ran into with the FS mode seems to be the same problem
described here [1] [2].
Can you try downloading the changes and let me know if it fixes the problem?
[1] https://gem5-review.googlesource.com/c/public/gem5/+/65272
[2]
Hi all,
I also ran into the same problem using another benchmark. I want to note
that this problem also appears when using the AtomicCPU.
Regards,
Hoa Nguyen
On Tue, Nov 1, 2022 at 3:02 AM Νικόλαος Ταμπουρατζής via gem5-users <
gem5-users@gem5.org> wrote:
> Dear Boddy,
>
> Thank you for the
Dear Boddy,
Thank you for the update! Please let me know when the accuracy issue
will be resolved because I cannot execute any benchmark in RISCV FS
mode (I am wondering if any other user faces the same problem).
Best regards,
Nikos
Quoting Bobby Bruce via gem5-users :
You mean this
You mean this bug? Unfortunately not, I've been very busy with the upcoming
gem5 release and haven't had time to investigate this further.
--
Dr. Bobby R. Bruce
Room 3050,
Kemper Hall, UC Davis
Davis,
CA, 95616
web: https://www.bobbybruce.net
On Mon, Oct 31, 2022 at 1:45 AM Νικόλαος
Dear Bobby, Jason, all,
Is there any update about the accuracy of RISC-V FS?
Best regards,
Nikos
Quoting Bobby Bruce :
Jason and I had a theory that this may be due to the "Rounding Mode" for
floating pointing being set incorrectly in FS mode. That's set via a macro
here:
Jason and I had a theory that this may be due to the "Rounding Mode" for
floating pointing being set incorrectly in FS mode. That's set via a macro
here:
https://gem5.googlesource.com/public/gem5/+/refs/tags/v22.0.0.2/src/arch/riscv/fp_inst.hh#36
I manually expanded the macro here:
Dear Hoa, all
I have ported successfully HPCG and many simple examples using gem5
ARM-FS and they are working properly. The problem is only in RISCV-FS
using double and float variables. Which option of objdump to use?
Best regards,
Nikos
Quoting Hoa Nguyen :
Hi,
It's quite odd that
Hi,
It's quite odd that both sqrt_i and result were zeroed out at the same
time. Does the problem appear in other ISA FS mode, e.g. x86 FS mode? Can
you show the objdump of the loop as well?
Regards,
Hoa Nguyen
On Thu, Oct 6, 2022, 04:06 Νικόλαος Ταμπουρατζής
wrote:
> Dear Jason, all,
>
> I
Dear Jason & Boddy,
Unfortunately, I have tried my simple example without the sqrt
function and the problem remains. Specifically, I have the following
simple code:
#include
#include
int main(){
int dim = 1024;
double result;
for (int iter = 0; iter < 2; iter++){
I have an idea...
Have you put a breakpoint in the implementation of the fsqrt_d function? I
would like to know if when running in SE mode and running in FS mode we are
using the same rounding mode. My hypothesis is that in FS mode the rounding
mode is set differently.
Cheers,
Jason
On Fri, Oct
Dear Boddy,
Thanks a lot for the effort! I looked in detail and I observe that the
problem is created only using float and double variables (in the case
of int it is working properly in FS mode). Specifically, in the case
of float the variables are set to "nan", while in the case of double
Hey Niko,
Thanks for this analysis. I jumped a little into this today but didn't get
as far as you did. I wanted to find a quick way to recreate the following:
https://gem5-review.googlesource.com/c/public/gem5/+/64211. Please feel
free to use this, if it helps any.
It's very strange to me that
In my previous results, I had used double (not float) for the
following variables: result, sq_i and sq_j. In the case of float
instead of double I get "nan" and not 0.00.
Quoting Νικόλαος Ταμπουρατζής :
Dear Jason, all,
I am trying to find the accuracy problem with RISCV-FS and I
Dear Jason, all,
I am trying to find the accuracy problem with RISCV-FS and I observe
that the problem is created (at least in my dummy example) because the
variables (double) are set to zero in random simulated time (for this
reason I get different results among executions of the same
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
#include
int main(){
int dim = 4096;
double result;
for (int iter =
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
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)?
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
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
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
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
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,
23 matches
Mail list logo