From: Wolfgang Bangerth Date: Thu, 3 Sep 2009 18:43:29 +0000 (+0000) Subject: A few more edits. X-Git-Tag: v8.0.0~7192 X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=e89cd7232fd1e226b4260df2309ccc61f382694d;p=dealii.git A few more edits. git-svn-id: https://svn.dealii.org/trunk@19379 0785d39b-7218-0410-832d-ea1e28bc413d --- diff --git a/deal.II/examples/step-37/doc/intro.dox b/deal.II/examples/step-37/doc/intro.dox index ce94b64b38..d1660590fd 100644 --- a/deal.II/examples/step-37/doc/intro.dox +++ b/deal.II/examples/step-37/doc/intro.dox @@ -261,7 +261,7 @@ considerably. In fact, all the remainder is just to make a slightly more clever use of data in order to gain some extra speed. It does not change the code structure, though. -If one would implement the above code, one would soon realize that the +If one would implement the code above, one would soon realize that the operations done by the call fe_values.reinit(cell) takes about as much time as the other steps together (at least if the mesh is unstructured; deal.II can recognize that the gradients are often unchanged @@ -274,7 +274,7 @@ point, i.e., as many times as there are elements in one of the loops above. The Jacobian does not depend on the basis function, but it is different on different quadrature points in general. The trick is now to first apply the operation that leads us to temp_vector only -with the gradient of the reference cell. That transforms the vector of +with the gradient on the reference cell. That transforms the vector of values on the local dofs to a vector of gradients on the quadrature points. There, we apply first the Jacobian that we factored out from the gradient, then we apply the weights of the quadrature, and we apply with the @@ -383,27 +383,31 @@ there that make the implementation even more efficient, namely: be implemented more efficiently. Obviously, we need some adapted data structures for that, but it isn't too hard to realize that. What is nice is that matrix-matrix products are close to the processor's peak - performance (and these are the most expensive part in the code). This - makes that the matrix-free matrix-vector product is slower for linear + performance if the matrices are not too large (and these are the + most expensive part in the code). This + implies that the matrix-free matrix-vector product is slower for linear and quadratic elements, but on par with third order elements and faster for even higher order elements. -And one more gain with this implementation is that we do not have to build +An additional gain with this implementation is that we do not have to build the sparse matrix itself, which can be quite expensive depending on the underlying differential equation. +

Parallelization issues

-We mentioned in the philosophical section above that parallelization is an -issue where sparse matrix-vector products are not that suited. There is a +We mentioned in the philosophical section above that parallelization +with multiple cores in a shared memory machines is an +issue where sparse matrix-vector products are not particularly well +suited because processors are memory bandwidth limited. There is a lot of data traffic involved, and the access patterns in the vector are not -very regular. Different rows might have different numbers of nonzero +very regular. Different rows might have different %numbers of nonzero elements. The matrix-free implementation, however, is better in this respect. It does not need to save all the elements (only the product of transposed Jacobian, weights, and Jacobian, for all quadrature points on all cells, which is about 4 times the size of the solution vector in 2D and 9 -times the size of the solution vector in 3D), whereas number of nonzeros +times the size of the solution vector in 3D), whereas the number of nonzeros grow with the element order. Moreover, most of the work is done on a very regular pattern: Perform matrix-vector products with the same matrix, perform (equally many) transformations on the vector related quadrature @@ -419,23 +423,27 @@ matrix-vector product run efficiently on a GPU. For our program, we choose to follow a simple strategy: We let several processors work together by splitting the cells into several chunks. The -threading building blocks implemenation of a parallel for loop -implements this concept by the command apply_to_subranges. +threading building blocks implemenation of a %parallel for loop +implements this concept using the parallel::apply_to_subranges() function. +

Combination with multigrid

-Now we made big efforts to implement a matrix-vector product that does not -need to have access to the matrix elements. In many user codes, however, +Above, we have gone to significant efforts to implement a +matrix-vector product that does actually store the matrix elements. In many user codes, however, there is more than just performing un uncertain number of matrix-vector products — one wants to do as little of these operations as possible when solving linear equation systems. In theory, we could use the CG method -without preconditioning, however, that would not by very efficient. Often -one uses preconditioners for improving speed. One such option is SSOR, which -is not possible here because there one needs to do substitutions based on -the matrix entries. - -Multigrid methods as shown in @ref step_16 "step-16" are very fast, and -actually even suitable for our purpose since they can be designed based +without preconditioning, however, that would not by very efficient. Rather, +one uses preconditioners for improving speed. On the other hand, most +of the more frequently used preconditioners such as Jacobi, SSOR, ILU +or algebraic multigrid (AMG) can now no longer be used here because +their implementation requires knowledge of the elements of the system +matrix. + +One solution is to use multigrid methods as shown in @ref step_16 +"step-16". They are known to be very fast, and +they are suitable for our purpose here since they can be designed based purely on matrix-vector products. All one needs to do is to find a smoother that works with matrix-vector products only. One such candidate would be a damped Jacobi iteration, but that one is often not sufficiently good in