fdurastante/amg4psblas
1
0
Fork 0
mirror of https://github.com/sfilippone/amg4psblas.git synced 2026-10-20 12:52:16 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update documentation for release

Browse Source
This commit is contained in:
sfilippone
2025-12-23 11:21:53 +01:00
parent 43a75e3171
commit 06f8f83114
18 changed files with 970 additions and 917 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/html/userhtmlse4.html
+73 -59
View File
@@ -351,7 +351,7 @@ class="content">Preconditioner types, corresponding strings and default choices.
</div><hr class="endfloat" />
</div>
<!--l. 98--><p class="indent" > <span
<!--l. 97--><p class="indent" > <span
class="cmr-12">Note that the module </span><code class="lstinline"><span style="color:#000000">amg_prec_mod</span></code><span
class="cmr-12">, containing the definition of the preconditioner</span>
<span
@@ -370,7 +370,7 @@ class="cmr-12">4.1</span><!--tex4ht:ref: sec:examples --></a><span
class="cmr-12">).</span>
<br
class="newline" />
<!--l. 105--><p class="indent" > <span
<!--l. 104--><p class="indent" > <span
class="cmbx-12">Remark 1. </span><span
class="cmr-12">Coarsest-level solvers based on the LU factorization, such as those</span>
<span
@@ -385,11 +385,24 @@ class="cmr-12">problems. However, this does not necessarily correspond to the sh
<span
class="cmr-12">on parallel</span><span
class="cmr-12">&#x00A0;computers.</span>
<!--l. 112--><p class="indent" > <span
class="cmbx-12">Remark 2. </span><span
class="cmr-12">Memory allocation on GPUs is a costly operation implying a</span>
<span
class="cmr-12">synchronization; therefore, it is convenient to preallocate internal preconditioner</span>
<span
class="cmr-12">workspace with the method </span><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">prec%allocate_wrk(info)</span></span></span> <span
class="cmr-12">before invoking an iterative</span>
<span
class="cmr-12">method, and release it upon exit with </span><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">prec%deallocate_wrk(info)</span></span></span><span
class="cmr-12">.</span>
<h4 class="subsectionHead"><span class="titlemark"><span
class="cmr-12">4.1 </span></span> <a
id="x7-140004.1"></a><span
class="cmr-12">Examples</span></h4>
<!--l. 116--><p class="noindent" ><span
<!--l. 121--><p class="noindent" ><span
class="cmr-12">The code reported in Figure</span><span
class="cmr-12">&#x00A0;</span><a
href="#x7-14001r1"><span
@@ -418,7 +431,7 @@ class="cmr-12">and </span><code class="lstinline"><span style="color:#000000">ps
class="cmr-12">must be used by the example</span>
<span
class="cmr-12">program.</span>
<!--l. 126--><p class="indent" > <span
<!--l. 131--><p class="indent" > <span
class="cmr-12">The part of the code dealing with reading and assembling the sparse matrix and the</span>
<span
class="cmr-12">right-hand side vector and the deallocation of the relevant data structures, performed</span>
@@ -451,7 +464,7 @@ href="userhtmlli3.html#XPSBLASGUIDE"><span
class="cmr-12">21</span></a><span
class="cmr-12">]</span></span><span
class="cmr-12">.</span>
<!--l. 138--><p class="indent" > <span
<!--l. 143--><p class="indent" > <span
class="cmr-12">The setup and application of the default multilevel preconditioner for the real single</span>
<span
class="cmr-12">precision and the complex, single and double precision, versions are obtained</span>
@@ -461,6 +474,9 @@ class="cmr-12">&#x00A0;</span><a
href="userhtmlse5.html#x8-160005"><span
class="cmr-12">5</span><!--tex4ht:ref: sec:userinterface --></a> <span
class="cmr-12">for</span>
<span
class="cmr-12">details). If these versions are installed, the corresponding codes are available in</span>
<span class="obeylines-h"><span class="verb"><span
@@ -470,7 +486,7 @@ class="cmr-12">.</span>
<!--l. 144--><p class="indent" > <a
<!--l. 148--><p class="indent" > <a
id="x7-14001r1"></a><hr class="float"><div class="float"
>
@@ -478,7 +494,7 @@ class="cmr-12">.</span>
<div class="center"
>
<!--l. 145--><p class="noindent" >
<!--l. 149--><p class="noindent" >
@@ -535,7 +551,7 @@ class="cmr-12">.</span>
&#x00A0;&#x00A0;call&#x00A0;psb_exit(ctxt)
&#x00A0;&#x00A0;stop
</pre>
<!--l. 255--><p class="nopar" > </div>
<!--l. 259--><p class="nopar" > </div>
@@ -548,7 +564,7 @@ class="content">setup and application of the default multilevel preconditioner (
</div><hr class="endfloat" />
<!--l. 264--><p class="indent" > <span
<!--l. 267--><p class="indent" > <span
class="cmr-12">Different versions of the multilevel preconditioner can be obtained by changing the</span>
<span
class="cmr-12">default values of the preconditioner parameters. The code reported in Figure</span><span
@@ -557,42 +573,40 @@ href="#x7-14002r2"><span
class="cmr-12">2</span><!--tex4ht:ref: fig:ex2 --></a> <span
class="cmr-12">shows</span>
<span
class="cmr-12">how to set a V-cycle preconditioner which applies 1 block-Jacobi sweep as pre-</span>
class="cmr-12">how to set a V-cycle preconditioner which applies 1 block-Jacobi sweep as pre- and</span>
<span
class="cmr-12">and post-smoother, and solves the coarsest-level system with 8 block-Jacobi</span>
class="cmr-12">post-smoother, and solves the coarsest-level system with 8 block-Jacobi sweeps. Note</span>
<span
class="cmr-12">sweeps. Note that the ILU(0) factorization (plus triangular solve) is used as</span>
class="cmr-12">that the ILU(0) factorization (plus triangular solve) is used as local solver for the</span>
<span
class="cmr-12">local solver for the block-Jacobi sweeps, since this is the default associated</span>
class="cmr-12">block-Jacobi sweeps, since this is the default associated with block-Jacobi and set</span>
<span
class="cmr-12">with block-Jacobi and set by</span><span
class="cmr-12">by</span><span
class="cmr-12">&#x00A0;</span><code class="lstinline"><span style="color:#000000">P</span><span style="color:#000000">%</span><span style="color:#000000">init</span></code><span
class="cmr-12">. Furthermore, specifying block-Jacobi as</span>
class="cmr-12">. Furthermore, specifying block-Jacobi as coarsest-level solver implies that</span>
<span
class="cmr-12">coarsest-level solver implies that the coarsest-level matrix is distributed among</span>
<span
class="cmr-12">the processes. Figure</span><span
class="cmr-12">the coarsest-level matrix is distributed among the processes. Figure</span><span
class="cmr-12">&#x00A0;</span><a
href="#x7-14003r3"><span
class="cmr-12">3</span><!--tex4ht:ref: fig:ex3 --></a> <span
class="cmr-12">shows how to set a W-cycle preconditioner using the</span>
class="cmr-12">shows how</span>
<span
class="cmr-12">Coarsening based on Compatible Weighted Matching, aggregates of size at</span>
class="cmr-12">to set a W-cycle preconditioner using the Coarsening based on Compatible</span>
<span
class="cmr-12">most 8 and smoothed prolongators. It applies 2 hybrid Gauss-Seidel sweeps as</span>
class="cmr-12">Weighted Matching, aggregates of size at most 8 and smoothed prolongators. It</span>
<span
class="cmr-12">pre- and post-smoother, and solves the coarsest-level system with the parallel</span>
class="cmr-12">applies 2 hybrid Gauss-Seidel sweeps as pre- and post-smoother, and solves the</span>
<span
class="cmr-12">flexible Conjugate Gradient method (KRM) coupled with the block-Jacobi</span>
class="cmr-12">coarsest-level system with the parallel flexible Conjugate Gradient method (KRM)</span>
<span
class="cmr-12">preconditioner having ILU(0) on the blocks. Default parameters are used for stopping</span>
class="cmr-12">coupled with the block-Jacobi preconditioner having ILU(0) on the blocks, with</span>
<span
class="cmr-12">criterion of the coarsest solver. Note that, also in this case, specifying KRM as</span>
class="cmr-12">default parameters used for the coarsest solver. Note that specifying KRM as</span>
<span
class="cmr-12">coarsest-level solver implies that the coarsest-level matrix is distributed among the</span>
<span
class="cmr-12">processes.</span>
<!--l. 291--><p class="indent" > <span
<!--l. 299--><p class="indent" > <span
class="cmr-12">The code fragments shown in Figures</span><span
class="cmr-12">&#x00A0;</span><a
href="#x7-14002r2"><span
@@ -605,7 +619,7 @@ class="cmr-12">are included in the example program</span>
class="cmr-12">file </span><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">amg_dexample_ml.f90</span></span></span> <span
class="cmr-12">too.</span>
<!--l. 294--><p class="indent" > <span
<!--l. 302--><p class="indent" > <span
class="cmr-12">Finally, Figure</span><span
class="cmr-12">&#x00A0;</span><a
href="#x7-14004r4"><span
@@ -620,7 +634,7 @@ class="cmr-12">nonsymmetric. The corresponding example program is available in t
<span class="obeylines-h"><span class="verb"><span
class="cmtt-12">amg_dexample_1lev.f90</span></span></span><span
class="cmr-12">.</span>
<!--l. 301--><p class="indent" > <span
<!--l. 309--><p class="indent" > <span
class="cmr-12">For all the previous preconditioners, example programs where the sparse matrix</span>
<span
class="cmr-12">and the right-hand side are generated by discretizing a PDE with Dirichlet</span>
@@ -631,7 +645,7 @@ class="cmr-12">.</span>
<!--l. 304--><p class="indent" > <a
<!--l. 312--><p class="indent" > <a
id="x7-14002r2"></a><hr class="float"><div class="float"
>
@@ -639,7 +653,7 @@ class="cmr-12">.</span>
<div class="center"
>
<!--l. 318--><p class="noindent" >
<!--l. 326--><p class="noindent" >
<div class="minipage"><pre class="verbatim" id="verbatim-8">
...&#x00A0;...
!&#x00A0;build&#x00A0;a&#x00A0;V-cycle&#x00A0;preconditioner&#x00A0;with&#x00A0;1&#x00A0;block-Jacobi&#x00A0;sweep&#x00A0;(with
@@ -653,7 +667,7 @@ class="cmr-12">.</span>
&#x00A0;&#x00A0;call&#x00A0;P%smoothers_build(A,desc_A,info)
...&#x00A0;...
</pre>
<!--l. 333--><p class="nopar" > </div></div>
<!--l. 341--><p class="nopar" > </div></div>
<br /><div class="caption"
><span class="id">Listing 2: </span><span
class="content">setup of a multilevel preconditioner based on the default decoupled coarsening</span></div><!--tex4ht:label?: x7-14002r2 -->
@@ -664,7 +678,7 @@ class="content">setup of a multilevel preconditioner based on the default decoup
<!--l. 340--><p class="indent" > <a
<!--l. 348--><p class="indent" > <a
id="x7-14003r3"></a><hr class="float"><div class="float"
>
@@ -672,7 +686,7 @@ class="content">setup of a multilevel preconditioner based on the default decoup
<div class="center"
>
<!--l. 362--><p class="noindent" >
<!--l. 370--><p class="noindent" >
<div class="minipage"><pre class="verbatim" id="verbatim-9">
...&#x00A0;...
!&#x00A0;build&#x00A0;a&#x00A0;W-cycle&#x00A0;preconditioner&#x00A0;with&#x00A0;2&#x00A0;hybrid&#x00A0;Gauss-Seidel&#x00A0;sweeps
@@ -692,7 +706,7 @@ class="content">setup of a multilevel preconditioner based on the default decoup
&#x00A0;&#x00A0;call&#x00A0;P%smoothers_build(A,desc_A,info)
...&#x00A0;...
</pre>
<!--l. 383--><p class="nopar" > </div></div>
<!--l. 391--><p class="nopar" > </div></div>
<br /> <div class="caption"
><span class="id">Listing 3: </span><span
class="content">setup of a multilevel preconditioner based on the coupled coarsening using
@@ -704,7 +718,7 @@ weighted matching</span></div><!--tex4ht:label?: x7-14003r3 -->
<!--l. 390--><p class="indent" > <a
<!--l. 398--><p class="indent" > <a
id="x7-14004r4"></a><hr class="float"><div class="float"
>
@@ -712,7 +726,7 @@ weighted matching</span></div><!--tex4ht:label?: x7-14003r3 -->
<div class="center"
>
<!--l. 402--><p class="noindent" >
<!--l. 410--><p class="noindent" >
<div class="minipage"><pre class="verbatim" id="verbatim-10">
...&#x00A0;...
!&#x00A0;set&#x00A0;RAS&#x00A0;with&#x00A0;overlap&#x00A0;2&#x00A0;and&#x00A0;ILU(0)&#x00A0;on&#x00A0;the&#x00A0;local&#x00A0;blocks
@@ -723,7 +737,7 @@ weighted matching</span></div><!--tex4ht:label?: x7-14003r3 -->
!&#x00A0;solve&#x00A0;Ax=b&#x00A0;with&#x00A0;preconditioned&#x00A0;BiCGSTAB
&#x00A0;&#x00A0;call&#x00A0;psb_krylov(&#8217;BICGSTAB&#8217;,A,P,b,x,tol,desc_A,info)
</pre>
<!--l. 414--><p class="nopar" > </div></div>
<!--l. 422--><p class="nopar" > </div></div>
<br /> <div class="caption"
><span class="id">Listing 4: </span><span
class="content">setup of a one-level Schwarz preconditioner.</span></div><!--tex4ht:label?: x7-14004r4 -->
@@ -735,7 +749,7 @@ class="content">setup of a one-level Schwarz preconditioner.</span></div><!--tex
class="cmr-12">4.2 </span></span> <a
id="x7-150004.2"></a><span
class="cmr-12">GPU example</span></h4>
<!--l. 426--><p class="noindent" ><span
<!--l. 434--><p class="noindent" ><span
class="cmr-12">The code discussed here shows how to set up a program exploiting the combined GPU</span>
<span
class="cmr-12">capabilities of PSBLAS and AMG4PSBLAS. The code example is available in the</span>
@@ -743,14 +757,14 @@ class="cmr-12">capabilities of PSBLAS and AMG4PSBLAS. The code example is availa
class="cmr-12">source distribution directory </span><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">amg4psblas/examples/gpu</span></span></span><span
class="cmr-12">.</span>
<!--l. 431--><p class="indent" > <span
<!--l. 439--><p class="indent" > <span
class="cmr-12">First of all, we need to include the appropriate modules and declare some auxiliary</span>
<span
class="cmr-12">variables:</span>
<!--l. 433--><p class="indent" > <a
<!--l. 441--><p class="indent" > <a
id="x7-15001r5"></a><hr class="float"><div class="float"
>
@@ -758,7 +772,7 @@ class="cmr-12">variables:</span>
<div class="center"
>
<!--l. 452--><p class="noindent" >
<!--l. 460--><p class="noindent" >
<div class="minipage"><pre class="verbatim" id="verbatim-11">
program&#x00A0;amg_dexample_gpu
&#x00A0;&#x00A0;use&#x00A0;psb_base_mod
@@ -777,7 +791,7 @@ program&#x00A0;amg_dexample_gpu
&#x00A0;
</pre>
<!--l. 471--><p class="nopar" > </div></div>
<!--l. 479--><p class="nopar" > </div></div>
<br /> <div class="caption"
><span class="id">Listing 5: </span><span
class="content">setup of a GPU-enabled test program part one.</span></div><!--tex4ht:label?: x7-15001r5 -->
@@ -785,7 +799,7 @@ class="content">setup of a GPU-enabled test program part one.</span></div><!--te
</div><hr class="endfloat" />
<!--l. 478--><p class="indent" > <span
<!--l. 486--><p class="indent" > <span
class="cmr-12">In this particular example we are choosing to employ a </span><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">HLG</span></span></span> <span
class="cmr-12">data structure for</span>
@@ -793,14 +807,14 @@ class="cmr-12">data structure for</span>
class="cmr-12">sparse matrices on GPUs; for more information please refer to the PSBLAS users&#8217;</span>
<span
class="cmr-12">guide.</span>
<!--l. 482--><p class="indent" > <span
<!--l. 490--><p class="indent" > <span
class="cmr-12">We then have to initialize the GPU environment, and pass the appropriate MOLD</span>
<span
class="cmr-12">variables to the build methods (see also the PSBLAS users&#8217; guide).</span>
<!--l. 485--><p class="indent" > <a
<!--l. 493--><p class="indent" > <a
id="x7-15002r6"></a><hr class="float"><div class="float"
>
@@ -808,7 +822,7 @@ class="cmr-12">variables to the build methods (see also the PSBLAS users&#8217;
<div class="center"
>
<!--l. 501--><p class="noindent" >
<!--l. 509--><p class="noindent" >
<div class="minipage"><pre class="verbatim" id="verbatim-12">
&#x00A0;&#x00A0;call&#x00A0;psb_init(ctxt)
&#x00A0;&#x00A0;call&#x00A0;psb_info(ctxt,iam,np)
@@ -823,7 +837,7 @@ class="cmr-12">variables to the build methods (see also the PSBLAS users&#8217;
&#x00A0;
</pre>
<!--l. 516--><p class="nopar" > </div></div>
<!--l. 524--><p class="nopar" > </div></div>
<br /> <div class="caption"
><span class="id">Listing 6: </span><span
class="content">setup of a GPU-enabled test program part two.</span></div><!--tex4ht:label?: x7-15002r6 -->
@@ -831,7 +845,7 @@ class="content">setup of a GPU-enabled test program part two.</span></div><!--te
</div><hr class="endfloat" />
<!--l. 523--><p class="indent" > <span
<!--l. 531--><p class="indent" > <span
class="cmr-12">Finally, we convert the input matrix, the descriptor and the vectors to use a</span>
<span
class="cmr-12">GPU-enabled internal storage format. We then preallocate the preconditioner</span>
@@ -842,7 +856,7 @@ class="cmr-12">GPU environment</span>
<!--l. 527--><p class="indent" > <a
<!--l. 535--><p class="indent" > <a
id="x7-15003r7"></a><hr class="float"><div class="float"
>
@@ -850,7 +864,7 @@ class="cmr-12">GPU environment</span>
<div class="center"
>
<!--l. 557--><p class="noindent" >
<!--l. 565--><p class="noindent" >
<div class="minipage"><pre class="verbatim" id="verbatim-13">
&#x00A0;&#x00A0;call&#x00A0;desc_a%cnv(mold=igmold)
&#x00A0;&#x00A0;call&#x00A0;a%cscnv(info,mold=agmold)
@@ -877,7 +891,7 @@ class="cmr-12">GPU environment</span>
&#x00A0;
</pre>
<!--l. 584--><p class="nopar" > </div></div>
<!--l. 592--><p class="nopar" > </div></div>
<br /> <div class="caption"
><span class="id">Listing 7: </span><span
class="content">setup of a GPU-enabled test program part three.</span></div><!--tex4ht:label?: x7-15003r7 -->
@@ -885,7 +899,7 @@ class="content">setup of a GPU-enabled test program part three.</span></div><!--
</div><hr class="endfloat" />
<!--l. 592--><p class="indent" > <span
<!--l. 600--><p class="indent" > <span
class="cmr-12">It is very important to employ smoothers and coarsest solvers that are suited to the</span>
<span
class="cmr-12">GPU, i.e. methods that do NOT employ triangular system solve kernels. Methods that</span>
@@ -893,30 +907,30 @@ class="cmr-12">GPU, i.e. methods that do NOT employ triangular system solve kern
class="cmr-12">satisfy this constraint include:</span>
<ul class="itemize1">
<li class="itemize">
<!--l. 596--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
<!--l. 604--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">JACOBI</span></span></span>
</li>
<li class="itemize">
<!--l. 597--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
<!--l. 605--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">BJAC</span></span></span> <span
class="cmr-12">with the following methods on the local blocks:</span>
<ul class="itemize2">
<li class="itemize">
<!--l. 599--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
<!--l. 607--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">INVK</span></span></span>
</li>
<li class="itemize">
<!--l. 600--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
<!--l. 608--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">INVT</span></span></span>
</li>
<li class="itemize">
<!--l. 601--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
<!--l. 609--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">AINV</span></span></span></li></ul>
</li>
<li class="itemize">
<!--l. 603--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
<!--l. 611--><p class="noindent" ><span class="obeylines-h"><span class="verb"><span
class="cmtt-12">POLY</span></span></span></li></ul>
<!--l. 605--><p class="noindent" ><span
<!--l. 613--><p class="noindent" ><span
class="cmr-12">and their </span><span
class="cmmi-12">&#x2113;</span><sub><span
class="cmr-8">1</span></sub> <span