> On 11 Aug 2023, at 1:14 AM, Mark Adams wrote:
>
> BTW, nice bug report ...
>>
>> So in the first step it coarsens from 150e6 to 5.4e6 DOFs instead of to
>> 2.6e6 DOFs.
>
> Yes, this is the critical place to see what is different and going wrong.
>
> My 3D tests were not that different
BTW, nice bug report ...
>
> So in the first step it coarsens from 150e6 to 5.4e6 DOFs instead of to
> 2.6e6 DOFs.
Yes, this is the critical place to see what is different and going wrong.
My 3D tests were not that different and I see you lowered the threshold.
Note, you can set the threshold
Hi, I'm trying to run a parallel matrix vector build and linear solution with
PETSc on 2 MPI processes + one V100 GPU. I tested that the matrix build and
solution is successful in CPUs only. I'm using cuda 11.5 and cuda enabled
openmpi and gcc 9.3. When I run the job with GPU enabled I get the
Hi Stephan,
Yes, MIS(A^T A) -> MIS(MIS(A)) change?
Yep, that is it.
This change was required because A^T A is super expensive. This change did
not do much to my tests but this is complex.
I am on travel now, but I can get to this in a few days. You provided me
with a lot of data and I can take
On Thu, Aug 10, 2023 at 2:30 AM maitri ksh wrote:
> I am unable to understand what possibly went wrong with my code, I could
> load a matrix (large sparse matrix) into petsc, write it out and read it
> back into Matlab but when I tried to use MatView to see the matrix-info, it
> produces error
Alright. Again, thank you very much for taking the time to answer my
beginner questions! Still a lot to learn..
Have a good day!
On 10.08.23 12:27, Stefano Zampini wrote:
Then just do the multiplications you need. My proposal was for the
example function you were showing.
On Thu, Aug 10,
Then just do the multiplications you need. My proposal was for the example
function you were showing.
On Thu, Aug 10, 2023, 12:25 Niclas Götting
wrote:
> You are absolutely right for this specific case (I get about 2400it/s
> instead of 2100it/s). However, the single square function will be
You are absolutely right for this specific case (I get about 2400it/s
instead of 2100it/s). However, the single square function will be
replaced by a series of gaussian pulses in the future, which will never
be zero. Maybe one could do an approximation and skip the second mult,
if the
If you do the mult of "pump" inside an if it should be faster
On Thu, Aug 10, 2023, 12:12 Niclas Götting
wrote:
> If I understood you right, this should be the resulting RHS:
>
> def rhsfunc5(ts, t, u, F):
> l.mult(u, F)
> pump.mult(u, tmp_vec)
> scale = 0.5 * (5 < t < 10)
>
If I understood you right, this should be the resulting RHS:
def rhsfunc5(ts, t, u, F):
l.mult(u, F)
pump.mult(u, tmp_vec)
scale = 0.5 * (5 < t < 10)
F.axpy(scale, tmp_vec)
It is a little bit slower than option 3, but with about 2100it/s
consistently ~10% faster than option 4.
I would use option 3. Keep a work vector and do a vector summation instead
of the multiple multiplication by scale and 1/scale.
I agree with you the docs are a little misleading here.
On Thu, Aug 10, 2023, 11:40 Niclas Götting
wrote:
> Thank you both for the very quick answer!
>
> So far, I
Thank you both for the very quick answer!
So far, I compiled PETSc with debugging turned on, but I think it should
still be faster than standard scipy in both cases. Actually, Stefano's
answer has got me very far already; now I only define the RHS of the ODE
and no Jacobian (I wonder, why the
I am unable to understand what possibly went wrong with my code, I could
load a matrix (large sparse matrix) into petsc, write it out and read it
back into Matlab but when I tried to use MatView to see the matrix-info, it
produces error of some 'corrupt argument, #valgrind'. Can anyone please
13 matches
Mail list logo