Hi Ben,
Could you please give an example of some of the things you
typically cache in libMesh ?
thanks,
df
On Thu, 24 Jun 2010, Kirk, Benjamin (JSC-EG311) wrote:
> In my case I cache what I can, but still the behavior is as follows:
>
> 1) calculate residual
> 2) calculate jacobian. solve fo
On Thu, 24 Jun 2010 19:28:10 -0500, "Kirk, Benjamin (JSC-EG311)"
wrote:
> In my case I cache what I can, but still the behavior is as follows:
>
> 1) calculate residual
> 2) calculate jacobian. solve for update using (1) as rhs.
> 3) compute residual again and check against (1)
>
> In my case i
In my case I cache what I can, but still the behavior is as follows:
1) calculate residual
2) calculate jacobian. solve for update using (1) as rhs.
3) compute residual again and check against (1)
In my case it is common to do this once at each time step - that is,
solve the nonlinear problem v
thanks jed. I can't seem to find a stored profile. I'd have to recreate
one. But i'm thinking roughly twice as many functions evaluations as
jacobian evaluations.
On Thu, 24 Jun 2010, Jed Brown wrote:
> On Thu, 24 Jun 2010 14:28:03 -0500, David Fuentes wrote:
>> On 6/24/10, Jed Brown wr
On Thu, 24 Jun 2010 14:28:03 -0500, David Fuentes wrote:
> On 6/24/10, Jed Brown wrote:
> > On Thu, 24 Jun 2010 13:59:21 -0500 (CDT), David Fuentes
> > wrote:
> >>
> >> I typically use Petsc Nonlinear Solvers in 3D and my bottle neck is
> >> typically in the assembly with Petsc SNESSolve taking
On 6/24/10, Jed Brown wrote:
> On Thu, 24 Jun 2010 13:59:21 -0500 (CDT), David Fuentes
> wrote:
>>
>> I typically use Petsc Nonlinear Solvers in 3D and my bottle neck is
>> typically in the assembly with Petsc SNESSolve taking ~10% of the time
>> about ~50% in the jacobian, ~30% in the residual,
On Thu, 24 Jun 2010 13:59:21 -0500 (CDT), David Fuentes
wrote:
>
> I typically use Petsc Nonlinear Solvers in 3D and my bottle neck is
> typically in the assembly with Petsc SNESSolve taking ~10% of the time
> about ~50% in the jacobian, ~30% in the residual, and the
> rest is distributed.
So
I typically use Petsc Nonlinear Solvers in 3D and my bottle neck is
typically in the assembly with Petsc SNESSolve taking ~10% of the time
about ~50% in the jacobian, ~30% in the residual, and the
rest is distributed.
On Thu, 24 Jun 2010, Jed Brown wrote:
> On Thu, 24 Jun 2010 13:23:37 -05
cool thanks. I'll keep you posted.
I'll move my current assembly routines and
overwrite the FEMSystem::assembly in a derived class to compare.
df
On Thu, 24 Jun 2010, Roy Stogner wrote:
>
> On Thu, 24 Jun 2010, David Fuentes wrote:
>
>> I typically use 1-2, p=1, variables per system.
>
On Thu, 24 Jun 2010 13:23:37 -0500 (CDT), David Fuentes
wrote:
> Thank you Roy,
>
>
> I typically use 1-2, p=1, variables per system.
In 2D or 3D? Virtual calls compile to "mov, mov, jmp", the indirect
call typically costs 5 or 6 cycles (because data dependence interferes
with OoO), compared
On Thu, 24 Jun 2010, David Fuentes wrote:
> I typically use 1-2, p=1, variables per system.
That is just about the worst-case scenario. ;-)
> I'll try it out.
Thanks! Let me know what the profiling says. There are a couple ugly
template tricks we could experiment with if the virtual functio
Thank you Roy,
I typically use 1-2, p=1, variables per system.
I'll try it out.
df
On Thu, 24 Jun 2010, Roy Stogner wrote:
>
> On Thu, 24 Jun 2010, David Fuentes wrote:
>
>> My initial work with libMesh was based on the ex13 approach to solving
>> my nonlinear system of equations. I re
On Thu, 24 Jun 2010, David Fuentes wrote:
> My initial work with libMesh was based on the ex13 approach to solving
> my nonlinear system of equations. I really like the FEMSystem design and
> am thinking of switching to the FEMSystem approach seen in ex18 with
> petsc's nonlinear solvers. Have yo
Hello,
My initial work with libMesh was based on the ex13 approach to solving
my nonlinear system of equations. I really like the FEMSystem design and
am thinking of switching to the FEMSystem approach seen in ex18 with
petsc's nonlinear solvers. Have you seen any significant performance
differe
14 matches
Mail list logo
