Run a bit bigger problem. The results shouldn't be that noisy; unless you are 
playing some game in another window :-).

   Barry

On Aug 23, 2013, at 11:52 AM, Mark F. Adams <[email protected]> wrote:

> This code looks funny in MatSOR_SeqAIJ:
> 
>          x[i] = (1-omega)*x[i] + sum*idiag[i];
> 
> shouldn't this be:
> 
>          x[i] = (1-omega)*x[i] + omega*sum*idiag[i];
> 
> and I've done the first optimization (for the non-blocked version) and get 
> this (running each many time because my Mac is a little noisy).  Flop rates 
> go down a bit and total flops got down a lot (perhaps I have a flop counting 
> bug?).  6 old runs then7 new ones:
> 
> 12:40 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.3511e-01 1.0 1.49e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 34  0  0  0  14 34  0  0  0  1105
> 12:40 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.3611e-01 1.0 1.49e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 34  0  0  0  14 34  0  0  0  1097
> 12:40 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.3464e-01 1.0 1.49e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 34  0  0  0  14 34  0  0  0  1109
> 12:40 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.3487e-01 1.0 1.49e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 34  0  0  0  14 34  0  0  0  1107
> 12:41 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.3629e-01 1.0 1.49e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 34  0  0  0  15 34  0  0  0  1095
> 12:41 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.3492e-01 1.0 1.49e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 34  0  0  0  14 34  0  0  0  1107
> 12:41 dummy ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f make-local 
> runex54 | grep MatSOR
> MatSOR                60 1.0 1.2776e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 11 29  0  0  0  14 29  0  0  0   910
> 12:41 madams/sor-opt ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f 
> make-local runex54 | grep MatSOR
> MatSOR                60 1.0 1.2809e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 29  0  0  0  14 29  0  0  0   908
> 12:41 madams/sor-opt ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f 
> make-local runex54 | grep MatSOR
> MatSOR                60 1.0 1.2765e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 29  0  0  0  14 29  0  0  0   911
> 12:41 madams/sor-opt ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f 
> make-local runex54 | grep MatSOR
> MatSOR                60 1.0 1.3131e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 29  0  0  0  14 29  0  0  0   886
> 12:41 madams/sor-opt ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f 
> make-local runex54 | grep MatSOR
> MatSOR                60 1.0 1.2792e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 29  0  0  0  14 29  0  0  0   909
> 12:41 madams/sor-opt ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f 
> make-local runex54 | grep MatSOR
> MatSOR                60 1.0 1.2913e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 29  0  0  0  14 29  0  0  0   901
> 12:41 madams/sor-opt ~/Codes/petsc/src/ksp/ksp/examples/tutorials$  make -f 
> make-local runex54 | grep MatSOR
> MatSOR                60 1.0 1.2889e-01 1.0 1.16e+08 1.0 0.0e+00 0.0e+00 
> 0.0e+00 12 29  0  0  0  14 29  0  0  0   902
> 
> 
> On Aug 20, 2013, at 6:40 PM, Barry Smith <[email protected]> wrote:
> 
>> 
>> On Aug 20, 2013, at 4:39 PM, Jed Brown <[email protected]> wrote:
>> 
>>> Hmm, less clear flops benefit. Half of these are nonzero initial guess, for 
>>> which triangular storage would be worse
>> 
>>  Is it clear that triangular storage would be particularly worse? I think 
>> only a small percentage worse for a multiply. The problem with CSR for 
>> triangular solves is for each new row one needs to jump some amount in the 
>> index and double array wasting cache lines and limiting streaming. With 
>> multiply with triangular storage there is no jumping, just streaming two 
>> different index and double arrays and current CPUs have no problem managing 
>> 4 streams. 
>> 
>>  In other words CSR good for multiply, bad for triangular solve; triangular 
>> storage good for triangular solve and pretty good for multiply.
>> 
>>  Barry
>> 
>> 
>>  Barry
>> 
>>> (unless you add an Eisenstat optimization).
>>> 
>>> On Aug 20, 2013 4:26 PM, "Mark F. Adams" <[email protected]> wrote:
>>> Barry,
>>> 
>>> These are tests using the well load balanced problems in KSP on my Mac:
>>> 
>>> 3D Vector (ex56)
>>> 
>>> 8 proc:
>>> 
>>> rich/ssor(1)
>>>> KSPSolve               1 1.0 3.0143e-01 1.0 7.98e+07 1.1 1.0e+04 7.2e+02 
>>>> 8.5e+01 19 26 26 16  8 100100100100100  2047
>>> 
>>> cheb/jac(2)
>>>> KSPSolve               1 1.0 2.5836e-01 1.0 6.87e+07 1.1 1.4e+04 7.3e+02 
>>>> 8.8e+01 17 25 27 19  8 100100100100100  2053
>>> 
>>> 1 proc:
>>> 
>>> rich/ssor(1)
>>>> KSPSolve               1 1.0 1.8541e-01 1.0 3.20e+08 1.0 0.0e+00 0.0e+00 
>>>> 0.0e+00 10 22  0  0  0 100100  0  0  0  1727
>>> 
>>> cheb/jac(2)
>>>> KSPSolve               1 1.0 2.4841e-01 1.0 4.65e+08 1.0 0.0e+00 0.0e+00 
>>>> 0.0e+00 12 24  0  0  0 100100  0  0  0  1870
>>> 
>>> 2D Scalar (ex54) 4 procs
>>> 
>>> cheb/jac(2)
>>>> KSPSolve               1 1.0 2.3614e-01 1.0 3.51e+07 1.0 2.6e+03 8.7e+02 
>>>> 7.8e+02 91100 98 93 95 100100100100100   592
>>> 
>>> rich/ssor(1)
>>>> KSPSolve               1 1.0 2.0144e-01 1.0 2.29e+07 1.0 1.8e+03 8.8e+02 
>>>> 7.1e+02 89100 98 90 95 100100100100100   453
>>> 
>>> I'm not sure if this matters but I wanted to get off of the funny test 
>>> problem that I was using.
>>> 
>>> Mark
>>> 
>>> On Aug 17, 2013, at 5:30 PM, Barry Smith <[email protected]> wrote:
>>> 
>>>> 
>>>> Mark,
>>>> 
>>>>  Why rich/eisenstat(2) ?   With Cheb/eisenstat you get the acceleration of 
>>>> Cheby on top of the SOR but rich/eisenstat(2) just seems like some strange 
>>>> implementation of sor?
>>>> 
>>>>  I am guessing you are getting your 15% potential improvement from
>>>> 
>>>>>>> 7.63/11.3
>>>> 0.6752212389380531
>>>>>>> .6752212389380531*5.9800e+00
>>>> 4.037823008849558
>>>>>>> .63/4.63
>>>> 0.13606911447084233
>>>> 
>>>> Anyways I would focus on rich/ssor()  smoothing.
>>>> 
>>>> Before thinking about implementing a new data structure I think it is 
>>>> relatively easy to reduce more the total number of flops done with ssor(1) 
>>>> smoother [and despite the current conventional wisdom that doing extra 
>>>> flops is in a good thing in HPC :-)).
>>>> 
>>>> Presmoothing: Note that since you are running ssor(1) and zero initial 
>>>> guess the current implementation is already good in terms of reducing 
>>>> total flops; it does one down solve (saving the intermediate values) and 
>>>> then one up solve (using those saved values).
>>>> 
>>>> Postsmoothing: Here it is running SSOR(1) with nonzero initial guess, the 
>>>> current implementation in MatSOR_SeqAIJ() is bad because it applies the 
>>>> entire matrix in both the down solve and the up solve.
>>>> 
>>>> while (its--) {
>>>>  if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
>>>>    for (i=0; i<m; i++) {
>>>>      n   = a->i[i+1] - a->i[i];
>>>>      idx = a->j + a->i[i];
>>>>      v   = a->a + a->i[i];
>>>>      sum = b[i];
>>>>      PetscSparseDenseMinusDot(sum,x,v,idx,n);
>>>>      x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
>>>>    }
>>>>    ierr = PetscLogFlops(2.0*a->nz);CHKERRQ(ierr);
>>>>  }
>>>>  if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
>>>>    for (i=m-1; i>=0; i--) {
>>>>      n   = a->i[i+1] - a->i[i];
>>>>      idx = a->j + a->i[i];
>>>>      v   = a->a + a->i[i];
>>>>      sum = b[i];
>>>>      PetscSparseDenseMinusDot(sum,x,v,idx,n);
>>>>      x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
>>>>    }
>>>>    ierr = PetscLogFlops(2.0*a->nz);CHKERRQ(ierr);
>>>>  }
>>>> 
>>>> (Note also that this partially explains why the flop rate in the table for 
>>>> MatSOR() s so good, two of the three ssor triangular solves are actually 
>>>> full matrix vector products). What it should do is apply the entire matrix 
>>>> in the down solve (but save the partial sums of the lower triangular part) 
>>>> and then it can do the upper solve applying only 1/2 the matrix and using 
>>>> the accumulated values.  So currently the combined pre and post smooth has 
>>>> 3 complete applications of the matrix (.5 and .5 from the pre smooth and 1 
>>>> and 1 from the up smooth. Adding the optimization will reduce it to 2.5 
>>>> applications of the matrix giving a reduction of 1/6  = 17 percent of the 
>>>> smooth flops or
>>>> 
>>>>>>> 17.*14/33
>>>> 7.212121212121212
>>>> 
>>>> 7 percent decrease in the total KSPSolve flops
>>>> 
>>>> Next we can eliminate 1/2 of the application of the matrix in the residual 
>>>> computation after the smoothing if we have saved the partial sums of the 
>>>> upper triangular solve in the smoother. We can estimate the flops of the 
>>>> matrix-multiple in the residual computation as one application of the 
>>>> matrix and hence equal to 1/3 of the flops in the ssor computations. 
>>>> Eliminating 1/2 of that means eliminating the equivalent of eliminating 
>>>> 1/6 of the flops in the MatSOR computation which is
>>>> 
>>>>>>> 14./(33*6)
>>>> 0.0707070707070707  (actually the same computation as above :-(
>>>> 
>>>> so another 7 percent of the total flops in the KSPSolve.  So in total we 
>>>> could save 14 percent of the flops (and lots of memory access) in the 
>>>> KSPSolve at the work of editing a couple of functions.
>>>> 
>>>> At that point we could compute the flop rate of all the SSOR triangular 
>>>> solves and the percent of the time of the KSPSolves() in the them to 
>>>> determine if a new data structure is warranted. Note that in reducing the 
>>>> total number of flops we are replacing two full matrix-vector products 
>>>> with 2 triangular solves hence the flop rate will be lower indicating that 
>>>> using a new data structure will actually help more.
>>>> 
>>>> Notes on the implementation:
>>>> 
>>>> * modifying the MatSolve_SeqAIJ() to better handle the nonzero initial 
>>>> guess is straightforward; edit one functions
>>>> 
>>>> * modifying the residual computation in the multigrid to reuse the upper 
>>>> triangular sums requires a tiny bit more thought. Basically 
>>>> MatSOR_SeqAIJ() would have an internal pointer to the accumulated solves 
>>>> for the upper triangular part and we would need to either (1) have a 
>>>> MatMult_SeqAIJ() implementation that would retrieve those values and use 
>>>> them appropriately (how would it know when the values are valid?) or (2) 
>>>> use the PCMGSetResidual() to provide a custom residual routine that again 
>>>> pulled out the accumulated sums and used them. Of course we will 
>>>> eventually figure out how to organize it cleanly for users.
>>>> 
>>>> 
>>>> Note also the even if a new custom data structure is introduced the work 
>>>> outlined above is not lost since the new data structure still needs those 
>>>> flop reduction optimizations. So step one, fix the MatSOR_SeqAIJ() for 
>>>> nonzero initial guess, get a roughly 6 percent improvement, step 2, 
>>>> optimize the residual computation, get another 6 percent improvement, step 
>>>> 3,  introduce a new data structure and get another roughly 5 percent 
>>>> improvement in KSPSolve() (the 6, 6  and 5 are my guesstimates).
>>>> 
>>>> I would be very interested in seeing new numbers.
>>>> 
>>>> Final note: the very unload balancing of this problem will limit how much 
>>>> these optimizations will help. For a perfectly load balanced problem I 
>>>> think these optimizations would a lot higher in percentage (maybe twice as 
>>>> much? more?).
>>>> 
>>>> 
>>>> Barry
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> On Aug 17, 2013, at 3:16 PM, "Mark F. Adams" <[email protected]> wrote:
>>>> 
>>>>> I would like to get an idea of how much benefit there would be with using 
>>>>> a special matrix type for SOR.  Here is an experiment on 128 cores of 
>>>>> Hopper (Cray XE6), 7 point stencil, with some embedded BCs that look like 
>>>>> higher order stencils at BCs.  32^3 subdomian on each core:
>>>>> 
>>>>> cheb/jacobi(2)
>>>>> KSPSolve              15 1.0 5.9800e+00 1.0 1.13e+09 3.2 6.2e+06 1.1e+03 
>>>>> 2.8e+02  7 29 67 46  7  26100100100 76 18765
>>>>> 
>>>>> 
>>>>> rich/eisenstat(2)
>>>>> KSPSolve              15 1.0 1.1563e+01 1.0 1.37e+09 3.4 5.4e+06 1.1e+03 
>>>>> 2.8e+02 12 32 66 44  7  38100100100 76 11659
>>>>> 
>>>>> 
>>>>> rich/sor
>>>>> KSPSolve              15 1.0 4.6355e+00 1.0 7.63e+08 4.5 3.2e+06 1.0e+03 
>>>>> 3.1e+02 10 21 57 31  8  33100100100 77 15708
>>>>> 
>>>>> 
>>>>> Complete log files attached.  The "projection" solve is the solver of 
>>>>> interest.
>>>>> 
>>>>> I have 2 Jacobi so that it has about the same amount of work a one 
>>>>> (s)sor.  There are voids in the domain which I believe accounts for the 
>>>>> large differences in the number of flops per core.  These were run with 
>>>>> the same processor group (i.e., all runs done in the same qsub script)
>>>>> 
>>>>> This shows only about 15% potential gain.  Should we conclude that there 
>>>>> is not much to gain from an optimized data structure?
>>>>> 
>>>>> Mark
>>>>> <log_eis><log_jac><log_sor>
>>>>> 
>>>>> 
>>>>> On Aug 16, 2013, at 7:53 PM, Jed Brown <[email protected]> wrote:
>>>>> 
>>>>>> "Mark F. Adams" <[email protected]> writes:
>>>>>>> Some hypre papers have shown that cheb/jacobi is faster for some
>>>>>>> problems but for me robustness trumps this for default solver
>>>>>>> parameters in PETSc.
>>>>>> 
>>>>>> Richardson/SOR is about the best thing out there in terms of reasonable
>>>>>> local work, low/no setup cost, and reliable convergence.  Cheby/Jacobi
>>>>>> just has trivial fine-grained parallelism, but it's not clear that buys
>>>>>> anything on a CPU.
>>>>>> 
>>>>>>> Jed's analysis suggests that Eisenstat's method saves almost 50% work
>>>>>>> but needs a specialized matrix to get good flop rates.  Something to
>>>>>>> think about doing …
>>>>>> 
>>>>>> Mine was too sloppy, Barry got it right.  Eisenstat is for Cheby/SOR,
>>>>>> however, and doesn't do anything for Richardson.
>>>>>> 
>>>>>> To speed Richardson/SSOR up with a new matrix format, I think we have to
>>>>>> cache the action of the lower triangular part in the forward sweep so
>>>>>> that the back-sweep can use it, and vice-versa.  With full caching and
>>>>>> triangular residual optimization, I think this brings 2 SSOR iterations
>>>>>> of the down-smoother plus a residual to 2.5 work units in the
>>>>>> down-smooth (zero initial guess) and 3 work units in the up-smooth
>>>>>> (nonzero initial guess).  (This is a strong smoother and frequently, one
>>>>>> SSOR would be enough.)
>>>>> 
>>>> 
>>> 
>> 
> 

Reply via email to