mirror of
https://github.com/sfilippone/amg4psblas.git
synced 2026-10-06 14:44:55 +00:00
New GPU comments and sample program in docs.
This commit is contained in:
Salvatore Filippone
Binary file not shown.
+15
-15
@@ -112,73 +112,73 @@ href="userhtmlsu7.html#x16-150004.2" id="QQ2-16-21"><span
|
||||
class="cmr-12">GPU example</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">5 </span><a
|
||||
href="userhtmlse5.html#x17-160005" id="QQ2-17-23"><span
|
||||
href="userhtmlse5.html#x17-160005" id="QQ2-17-25"><span
|
||||
class="cmr-12">User Interface</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.1 </span><a
|
||||
href="userhtmlsu8.html#x18-170005.1" id="QQ2-18-24"><span
|
||||
href="userhtmlsu8.html#x18-170005.1" id="QQ2-18-26"><span
|
||||
class="cmr-12">Method init</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.2 </span><a
|
||||
href="userhtmlsu9.html#x19-180005.2" id="QQ2-19-25"><span
|
||||
href="userhtmlsu9.html#x19-180005.2" id="QQ2-19-27"><span
|
||||
class="cmr-12">Method set</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.3 </span><a
|
||||
href="userhtmlsu10.html#x20-190005.3" id="QQ2-20-33"><span
|
||||
href="userhtmlsu10.html#x20-190005.3" id="QQ2-20-35"><span
|
||||
class="cmr-12">Method hierarchy</span><span
|
||||
class="cmr-12">_build</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.4 </span><a
|
||||
href="userhtmlsu11.html#x21-200005.4" id="QQ2-21-34"><span
|
||||
href="userhtmlsu11.html#x21-200005.4" id="QQ2-21-36"><span
|
||||
class="cmr-12">Method smoothers</span><span
|
||||
class="cmr-12">_build</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.5 </span><a
|
||||
href="userhtmlsu12.html#x22-210005.5" id="QQ2-22-35"><span
|
||||
href="userhtmlsu12.html#x22-210005.5" id="QQ2-22-37"><span
|
||||
class="cmr-12">Method build</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.6 </span><a
|
||||
href="userhtmlsu13.html#x23-220005.6" id="QQ2-23-36"><span
|
||||
href="userhtmlsu13.html#x23-220005.6" id="QQ2-23-38"><span
|
||||
class="cmr-12">Method apply</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.7 </span><a
|
||||
href="userhtmlsu14.html#x24-230005.7" id="QQ2-24-37"><span
|
||||
href="userhtmlsu14.html#x24-230005.7" id="QQ2-24-39"><span
|
||||
class="cmr-12">Method free</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.8 </span><a
|
||||
href="userhtmlsu15.html#x25-240005.8" id="QQ2-25-38"><span
|
||||
href="userhtmlsu15.html#x25-240005.8" id="QQ2-25-40"><span
|
||||
class="cmr-12">Method descr</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.9 </span><a
|
||||
href="userhtmlsu16.html#x26-250005.9" id="QQ2-26-39"><span
|
||||
href="userhtmlsu16.html#x26-250005.9" id="QQ2-26-41"><span
|
||||
class="cmr-12">Auxiliary Methods</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">6 </span><a
|
||||
href="userhtmlse6.html#x27-310006" id="QQ2-27-45"><span
|
||||
href="userhtmlse6.html#x27-310006" id="QQ2-27-47"><span
|
||||
class="cmr-12">Adding new smoother and solver objects to AMG4PSBLAS</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">7 </span><a
|
||||
href="userhtmlse7.html#x28-320007" id="QQ2-28-46"><span
|
||||
href="userhtmlse7.html#x28-320007" id="QQ2-28-48"><span
|
||||
class="cmr-12">Error Handling</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">A </span><a
|
||||
href="userhtmlse8.html#x29-33000A" id="QQ2-29-47"><span
|
||||
href="userhtmlse8.html#x29-33000A" id="QQ2-29-49"><span
|
||||
class="cmr-12">License</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">B </span><a
|
||||
href="userhtmlse9.html#x30-34000B" id="QQ2-30-48"><span
|
||||
href="userhtmlse9.html#x30-34000B" id="QQ2-30-50"><span
|
||||
class="cmr-12">Contributor Covenant Code of Conduct</span></a></span>
|
||||
<br /> <span class="likesectionToc" ><a
|
||||
href="userhtmlli5.html#x31-39000B" id="QQ2-31-53"><span
|
||||
href="userhtmlli5.html#x31-39000B" id="QQ2-31-55"><span
|
||||
class="cmr-12">References</span></a></span>
|
||||
</div>
|
||||
|
||||
|
||||
+15
-15
@@ -112,73 +112,73 @@ href="userhtmlsu7.html#x16-150004.2" id="QQ2-16-21"><span
|
||||
class="cmr-12">GPU example</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">5 </span><a
|
||||
href="userhtmlse5.html#x17-160005" id="QQ2-17-23"><span
|
||||
href="userhtmlse5.html#x17-160005" id="QQ2-17-25"><span
|
||||
class="cmr-12">User Interface</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.1 </span><a
|
||||
href="userhtmlsu8.html#x18-170005.1" id="QQ2-18-24"><span
|
||||
href="userhtmlsu8.html#x18-170005.1" id="QQ2-18-26"><span
|
||||
class="cmr-12">Method init</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.2 </span><a
|
||||
href="userhtmlsu9.html#x19-180005.2" id="QQ2-19-25"><span
|
||||
href="userhtmlsu9.html#x19-180005.2" id="QQ2-19-27"><span
|
||||
class="cmr-12">Method set</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.3 </span><a
|
||||
href="userhtmlsu10.html#x20-190005.3" id="QQ2-20-33"><span
|
||||
href="userhtmlsu10.html#x20-190005.3" id="QQ2-20-35"><span
|
||||
class="cmr-12">Method hierarchy</span><span
|
||||
class="cmr-12">_build</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.4 </span><a
|
||||
href="userhtmlsu11.html#x21-200005.4" id="QQ2-21-34"><span
|
||||
href="userhtmlsu11.html#x21-200005.4" id="QQ2-21-36"><span
|
||||
class="cmr-12">Method smoothers</span><span
|
||||
class="cmr-12">_build</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.5 </span><a
|
||||
href="userhtmlsu12.html#x22-210005.5" id="QQ2-22-35"><span
|
||||
href="userhtmlsu12.html#x22-210005.5" id="QQ2-22-37"><span
|
||||
class="cmr-12">Method build</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.6 </span><a
|
||||
href="userhtmlsu13.html#x23-220005.6" id="QQ2-23-36"><span
|
||||
href="userhtmlsu13.html#x23-220005.6" id="QQ2-23-38"><span
|
||||
class="cmr-12">Method apply</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.7 </span><a
|
||||
href="userhtmlsu14.html#x24-230005.7" id="QQ2-24-37"><span
|
||||
href="userhtmlsu14.html#x24-230005.7" id="QQ2-24-39"><span
|
||||
class="cmr-12">Method free</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.8 </span><a
|
||||
href="userhtmlsu15.html#x25-240005.8" id="QQ2-25-38"><span
|
||||
href="userhtmlsu15.html#x25-240005.8" id="QQ2-25-40"><span
|
||||
class="cmr-12">Method descr</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span class="subsectionToc" ><span
|
||||
class="cmr-12">5.9 </span><a
|
||||
href="userhtmlsu16.html#x26-250005.9" id="QQ2-26-39"><span
|
||||
href="userhtmlsu16.html#x26-250005.9" id="QQ2-26-41"><span
|
||||
class="cmr-12">Auxiliary Methods</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">6 </span><a
|
||||
href="userhtmlse6.html#x27-310006" id="QQ2-27-45"><span
|
||||
href="userhtmlse6.html#x27-310006" id="QQ2-27-47"><span
|
||||
class="cmr-12">Adding new smoother and solver objects to AMG4PSBLAS</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">7 </span><a
|
||||
href="userhtmlse7.html#x28-320007" id="QQ2-28-46"><span
|
||||
href="userhtmlse7.html#x28-320007" id="QQ2-28-48"><span
|
||||
class="cmr-12">Error Handling</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">A </span><a
|
||||
href="userhtmlse8.html#x29-33000A" id="QQ2-29-47"><span
|
||||
href="userhtmlse8.html#x29-33000A" id="QQ2-29-49"><span
|
||||
class="cmr-12">License</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
class="cmr-12">B </span><a
|
||||
href="userhtmlse9.html#x30-34000B" id="QQ2-30-48"><span
|
||||
href="userhtmlse9.html#x30-34000B" id="QQ2-30-50"><span
|
||||
class="cmr-12">Contributor Covenant Code of Conduct</span></a></span>
|
||||
<br /> <span class="likesectionToc" ><a
|
||||
href="userhtmlli5.html#x31-39000B" id="QQ2-31-53"><span
|
||||
href="userhtmlli5.html#x31-39000B" id="QQ2-31-55"><span
|
||||
class="cmr-12">References</span></a></span>
|
||||
</div>
|
||||
|
||||
|
||||
@@ -137,32 +137,32 @@ class="cmr-12">Auxiliary Methods</span></a></span>
|
||||
class="cmr-12"> </span><span
|
||||
class="cmr-12"> </span><span class="subsubsectionToc" ><span
|
||||
class="cmr-12">5.9.1 </span><a
|
||||
href="userhtmlsu16.html#x26-260005.9.1" id="QQ2-26-40"><span
|
||||
href="userhtmlsu16.html#x26-260005.9.1" id="QQ2-26-42"><span
|
||||
class="cmr-12">Method: dump</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span
|
||||
class="cmr-12"> </span><span class="subsubsectionToc" ><span
|
||||
class="cmr-12">5.9.2 </span><a
|
||||
href="userhtmlsu16.html#x26-270005.9.2" id="QQ2-26-41"><span
|
||||
href="userhtmlsu16.html#x26-270005.9.2" id="QQ2-26-43"><span
|
||||
class="cmr-12">Method: clone</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span
|
||||
class="cmr-12"> </span><span class="subsubsectionToc" ><span
|
||||
class="cmr-12">5.9.3 </span><a
|
||||
href="userhtmlsu16.html#x26-280005.9.3" id="QQ2-26-42"><span
|
||||
href="userhtmlsu16.html#x26-280005.9.3" id="QQ2-26-44"><span
|
||||
class="cmr-12">Method: sizeof</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span
|
||||
class="cmr-12"> </span><span class="subsubsectionToc" ><span
|
||||
class="cmr-12">5.9.4 </span><a
|
||||
href="userhtmlsu16.html#x26-290005.9.4" id="QQ2-26-43"><span
|
||||
href="userhtmlsu16.html#x26-290005.9.4" id="QQ2-26-45"><span
|
||||
class="cmr-12">Method: allocate</span><span
|
||||
class="cmr-12">_wrk</span></a></span>
|
||||
<br /> <span
|
||||
class="cmr-12"> </span><span
|
||||
class="cmr-12"> </span><span class="subsubsectionToc" ><span
|
||||
class="cmr-12">5.9.5 </span><a
|
||||
href="userhtmlsu16.html#x26-300005.9.5" id="QQ2-26-44"><span
|
||||
href="userhtmlsu16.html#x26-300005.9.5" id="QQ2-26-46"><span
|
||||
class="cmr-12">Method: free</span><span
|
||||
class="cmr-12">_wrk</span></a></span>
|
||||
<br /> <span class="sectionToc" ><span
|
||||
|
||||
@@ -215,7 +215,7 @@ class="cmr-12">pass it as follows:</span>
|
||||
|
||||
|
||||
|
||||
<pre class="verbatim" id="verbatim-11">
|
||||
<pre class="verbatim" id="verbatim-13">
|
||||
  ! sparse matrix and preconditioner
|
||||
  type(psb_dspmat_type) :: a
|
||||
  type(amg_dprec_type)  :: prec
|
||||
|
||||
@@ -36,7 +36,7 @@ class="cmr-12">AMG4PSBLAS is freely distributable under the following copyright
|
||||
|
||||
|
||||
|
||||
<pre class="verbatim" id="verbatim-12">
|
||||
<pre class="verbatim" id="verbatim-14">
|
||||
                           AMG4PSBLAS  version 1.0
|
||||
              Algebraic MultiGrid Preconditioners Package
|
||||
             based on PSBLAS (Parallel Sparse BLAS version 3.7)
|
||||
@@ -78,7 +78,7 @@ class="cmr-12">abide by its terms:</span>
|
||||
|
||||
|
||||
|
||||
<pre class="verbatim" id="verbatim-13">
|
||||
<pre class="verbatim" id="verbatim-15">
|
||||
                           MLD2P4  version 2.2
|
||||
  MultiLevel Domain Decomposition Parallel Preconditioners Package
|
||||
             based on PSBLAS (Parallel Sparse BLAS version 3.5)
|
||||
@@ -127,7 +127,7 @@ class="cmr-12">here.</span>
|
||||
|
||||
|
||||
|
||||
<pre class="verbatim" id="verbatim-14">
|
||||
<pre class="verbatim" id="verbatim-16">
|
||||
// ***********************************************************************
|
||||
//
|
||||
//        MatchboxP: A C++ library for approximate weighted matching
|
||||
|
||||
+140
-17
@@ -29,17 +29,21 @@ class="cmr-12">4.2 </span></span> <a
|
||||
id="x16-150004.2"></a><span
|
||||
class="cmr-12">GPU example</span></h4>
|
||||
<!--l. 422--><p class="noindent" ><span
|
||||
class="cmr-12">The code reported in Figure</span><span
|
||||
class="cmr-12"> </span><a
|
||||
href="#x16-15001r5"><span
|
||||
class="cmr-12">5</span><!--tex4ht:ref: fig:gpu-ex1 --></a> <span
|
||||
class="cmr-12">shows how to set up a program exploiting the combined</span>
|
||||
class="cmr-12">The code discussed here shows how to set up a program exploiting the combined GPU</span>
|
||||
<span
|
||||
class="cmr-12">GPU capabilities of PSBLAS and AMG4PSBLAS.</span>
|
||||
class="cmr-12">capabilities of PSBLAS and AMG4PSBLAS. The code example is availabile in the</span>
|
||||
<span
|
||||
class="cmr-12">source distribution directory </span><span class="obeylines-h"><span class="verb"><span
|
||||
class="cmtt-12">amg4psblas/tests/gpu</span></span></span><span
|
||||
class="cmr-12">.</span>
|
||||
<!--l. 427--><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. 425--><p class="indent" > <a
|
||||
<!--l. 429--><p class="indent" > <a
|
||||
id="x16-15001r5"></a><hr class="float"><div class="float"
|
||||
>
|
||||
|
||||
@@ -47,24 +51,143 @@ class="cmr-12">GPU capabilities of PSBLAS and AMG4PSBLAS.</span>
|
||||
|
||||
<div class="center"
|
||||
>
|
||||
<!--l. 437--><p class="noindent" >
|
||||
<!--l. 451--><p class="noindent" >
|
||||
<div class="minipage"><pre class="verbatim" id="verbatim-10">
|
||||
! build a one-level RAS with overlap 2 and ILU(0) on the local blocks.
|
||||
call P%init(’AS’,info)
|
||||
call P%set(’SUB_OVR’,2,info)
|
||||
call P%build(A,desc_A,info)
|
||||
... ...
|
||||
! solve Ax=b with preconditioned BiCGSTAB
|
||||
  call psb_krylov(’BICGSTAB’,A,P,b,x,tol,desc_A,info)
|
||||
program amg_d_pde3d
|
||||
  use psb_base_mod
|
||||
  use amg_prec_mod
|
||||
  use psb_krylov_mod
|
||||
  use psb_util_mod
|
||||
  use psb_gpu_mod
|
||||
  use data_input
|
||||
  use amg_d_pde3d_base_mod
|
||||
  use amg_d_pde3d_exp_mod
|
||||
  use amg_d_pde3d_gauss_mod
|
||||
  use amg_d_genpde_mod
|
||||
  implicit none
|
||||
  .......
|
||||
  ! GPU variables
|
||||
  type(psb_d_hlg_sparse_mat) :: agmold
|
||||
  type(psb_d_vect_gpu)       :: vgmold
|
||||
  type(psb_i_vect_gpu)       :: igmold
|
||||
 
|
||||
</pre>
|
||||
<!--l. 448--><p class="nopar" ></div></div>
|
||||
<!--l. 473--><p class="nopar" ></div></div>
|
||||
<br /> <div class="caption"
|
||||
><span class="id">Listing 5: </span><span
|
||||
class="content">setup of a one-level Schwarz preconditioner.</span></div><!--tex4ht:label?: x16-15001r5 -->
|
||||
class="content">setup of a GPU-enabled test program part one.</span></div><!--tex4ht:label?: x16-15001r5 -->
|
||||
|
||||
|
||||
|
||||
</div><hr class="endfloat" />
|
||||
<!--l. 481--><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</span>
|
||||
|
||||
|
||||
|
||||
<!--l. 483--><p class="indent" > <a
|
||||
id="x16-15002r6"></a><hr class="float"><div class="float"
|
||||
>
|
||||
|
||||
|
||||
|
||||
<div class="center"
|
||||
>
|
||||
<!--l. 499--><p class="noindent" >
|
||||
<div class="minipage"><pre class="verbatim" id="verbatim-11">
|
||||
  call psb_init(ctxt)
|
||||
  call psb_info(ctxt,iam,np)
|
||||
  !
|
||||
  ! BEWARE: if you have NGPUS  per node, the default is to
|
||||
  ! attach to mod(IAM,NGPUS)
|
||||
  !
|
||||
  call psb_gpu_init(ictxt)
|
||||
  ......
|
||||
  t1 = psb_wtime()
|
||||
  call prec%smoothers_build(a,desc_a,info, amold=agmold, vmold=vgmold, imold=igmold)
|
||||
 
|
||||
</pre>
|
||||
<!--l. 514--><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?: x16-15002r6 -->
|
||||
|
||||
|
||||
|
||||
</div><hr class="endfloat" />
|
||||
<!--l. 521--><p class="indent" > <span
|
||||
class="cmr-12">Finally, we convert the input matrix, the descriptor and the vectors, then</span>
|
||||
<span
|
||||
class="cmr-12">preallocate the preconditioner workspace before entering the Krylov method. At the</span>
|
||||
<span
|
||||
class="cmr-12">end of the code, we close the GPU environment</span>
|
||||
|
||||
|
||||
|
||||
<!--l. 524--><p class="indent" > <a
|
||||
id="x16-15003r7"></a><hr class="float"><div class="float"
|
||||
>
|
||||
|
||||
|
||||
|
||||
<div class="center"
|
||||
>
|
||||
<!--l. 553--><p class="noindent" >
|
||||
<div class="minipage"><pre class="verbatim" id="verbatim-12">
|
||||
  call desc_a%cnv(mold=igmold)
|
||||
  call a%cscnv(info,mold=agmold)
|
||||
  call psb_geasb(x,desc_a,info,mold=vgmold)
|
||||
  call psb_geasb(b,desc_a,info,mold=vgmold)
|
||||
  !
|
||||
  ! iterative method parameters
|
||||
  !
|
||||
  call psb_barrier(ctxt)
|
||||
  call prec%allocate_wrk(info)
|
||||
  t1 = psb_wtime()
|
||||
  call psb_krylov(s_choice%kmethd,a,prec,b,x,s_choice%eps,&
|
||||
       & desc_a,info,itmax=s_choice%itmax,iter=iter,err=err,itrace=s_choice%itrace,&
|
||||
       & istop=s_choice%istopc,irst=s_choice%irst)
|
||||
  call prec%deallocate_wrk(info)
|
||||
  call psb_barrier(ctxt)
|
||||
  tslv = psb_wtime() - t1
|
||||
  ......
|
||||
  call psb_gpu_exit()
|
||||
  call psb_exit(ctxt)
|
||||
  stop
|
||||
 
|
||||
</pre>
|
||||
<!--l. 580--><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?: x16-15003r7 -->
|
||||
|
||||
|
||||
|
||||
</div><hr class="endfloat" />
|
||||
<!--l. 588--><p class="indent" > <span
|
||||
class="cmr-12">It is very important to employ solvers that are suited to the GPU, i.e. solvers that</span>
|
||||
<span
|
||||
class="cmr-12">do NOT employ triangular system solve kernels. Solvers that satisfy this constraint</span>
|
||||
<span
|
||||
class="cmr-12">include:</span>
|
||||
<ul class="itemize1">
|
||||
<li class="itemize"><span class="obeylines-h"><span class="verb"><span
|
||||
class="cmtt-12">JACOBI</span></span></span>
|
||||
</li>
|
||||
<li class="itemize"><span class="obeylines-h"><span class="verb"><span
|
||||
class="cmtt-12">INVK</span></span></span>
|
||||
</li>
|
||||
<li class="itemize"><span class="obeylines-h"><span class="verb"><span
|
||||
class="cmtt-12">INVT</span></span></span>
|
||||
</li>
|
||||
<li class="itemize"><span class="obeylines-h"><span class="verb"><span
|
||||
class="cmtt-12">AINV</span></span></span></li></ul>
|
||||
<!--l. 597--><p class="noindent" ><span
|
||||
class="cmr-12">and their </span><span class="obeylines-h"><span class="verb"><span
|
||||
class="cmtt-12">L1</span></span></span> <span
|
||||
class="cmr-12">variants.</span>
|
||||
|
||||
|
||||
|
||||
|
||||
+159
-18
@@ -419,41 +419,182 @@ call P%build(A,desc_A,info)
|
||||
|
||||
\subsection{GPU example\label{sec:gpu-example}}
|
||||
|
||||
The code reported in Figure~\ref{fig:gpu-ex1} shows how to set up a
|
||||
The code discussed here shows how to set up a
|
||||
program exploiting the combined GPU capabilities of PSBLAS and
|
||||
AMG4PSBLAS.
|
||||
AMG4PSBLAS. The code example is availabile in the source distribution
|
||||
directory \verb|amg4psblas/tests/gpu|.
|
||||
|
||||
First of all, we need to include the appropriate modules and
|
||||
declare some auxiliary variables:
|
||||
\begin{listing}[h!]
|
||||
\ifpdf
|
||||
\begin{minted}[breaklines=true,bgcolor=bg,fontsize=\small]{fortran}
|
||||
! build a one-level RAS with overlap 2 and ILU(0) on the local blocks.
|
||||
call P%init('AS',info)
|
||||
call P%set('SUB_OVR',2,info)
|
||||
call P%build(A,desc_A,info)
|
||||
... ...
|
||||
! solve Ax=b with preconditioned BiCGSTAB
|
||||
call psb_krylov('BICGSTAB',A,P,b,x,tol,desc_A,info)
|
||||
program amg_d_pde3d
|
||||
use psb_base_mod
|
||||
use amg_prec_mod
|
||||
use psb_krylov_mod
|
||||
use psb_util_mod
|
||||
use psb_gpu_mod
|
||||
use data_input
|
||||
use amg_d_pde3d_base_mod
|
||||
use amg_d_pde3d_exp_mod
|
||||
use amg_d_pde3d_gauss_mod
|
||||
use amg_d_genpde_mod
|
||||
implicit none
|
||||
.......
|
||||
! GPU variables
|
||||
type(psb_d_hlg_sparse_mat) :: agmold
|
||||
type(psb_d_vect_gpu) :: vgmold
|
||||
type(psb_i_vect_gpu) :: igmold
|
||||
\end{minted}
|
||||
\else
|
||||
\begin{center}
|
||||
\begin{minipage}{.90\textwidth}
|
||||
{\small
|
||||
\begin{verbatim}
|
||||
! build a one-level RAS with overlap 2 and ILU(0) on the local blocks.
|
||||
call P%init('AS',info)
|
||||
call P%set('SUB_OVR',2,info)
|
||||
call P%build(A,desc_A,info)
|
||||
... ...
|
||||
! solve Ax=b with preconditioned BiCGSTAB
|
||||
call psb_krylov('BICGSTAB',A,P,b,x,tol,desc_A,info)
|
||||
\end{verbatim}
|
||||
program amg_d_pde3d
|
||||
use psb_base_mod
|
||||
use amg_prec_mod
|
||||
use psb_krylov_mod
|
||||
use psb_util_mod
|
||||
use psb_gpu_mod
|
||||
use data_input
|
||||
use amg_d_pde3d_base_mod
|
||||
use amg_d_pde3d_exp_mod
|
||||
use amg_d_pde3d_gauss_mod
|
||||
use amg_d_genpde_mod
|
||||
implicit none
|
||||
.......
|
||||
! GPU variables
|
||||
type(psb_d_hlg_sparse_mat) :: agmold
|
||||
type(psb_d_vect_gpu) :: vgmold
|
||||
type(psb_i_vect_gpu) :: igmold
|
||||
|
||||
\end{verbatim}
|
||||
}
|
||||
\end{minipage}
|
||||
\end{center}
|
||||
\fi
|
||||
\caption{setup of a one-level Schwarz preconditioner.\label{fig:gpu-ex1}}
|
||||
\caption{setup of a GPU-enabled test program part one.\label{fig:gpu-ex1}}
|
||||
\end{listing}
|
||||
|
||||
We then have to initialize the GPU environment, and pass the
|
||||
appropriate MOLD variables to the build methods
|
||||
\begin{listing}[h!]
|
||||
\ifpdf
|
||||
\begin{minted}[breaklines=true,bgcolor=bg,fontsize=\small]{fortran}
|
||||
call psb_init(ctxt)
|
||||
call psb_info(ctxt,iam,np)
|
||||
!
|
||||
! BEWARE: if you have NGPUS per node, the default is to
|
||||
! attach to mod(IAM,NGPUS)
|
||||
!
|
||||
call psb_gpu_init(ictxt)
|
||||
......
|
||||
t1 = psb_wtime()
|
||||
call prec%smoothers_build(a,desc_a,info, amold=agmold, vmold=vgmold, imold=igmold)
|
||||
|
||||
\end{minted}
|
||||
\else
|
||||
\begin{center}
|
||||
\begin{minipage}{.90\textwidth}
|
||||
{\small
|
||||
\begin{verbatim}
|
||||
call psb_init(ctxt)
|
||||
call psb_info(ctxt,iam,np)
|
||||
!
|
||||
! BEWARE: if you have NGPUS per node, the default is to
|
||||
! attach to mod(IAM,NGPUS)
|
||||
!
|
||||
call psb_gpu_init(ictxt)
|
||||
......
|
||||
t1 = psb_wtime()
|
||||
call prec%smoothers_build(a,desc_a,info, amold=agmold, vmold=vgmold, imold=igmold)
|
||||
|
||||
\end{verbatim}
|
||||
}
|
||||
\end{minipage}
|
||||
\end{center}
|
||||
\fi
|
||||
\caption{setup of a GPU-enabled test program part two.\label{fig:gpu-ex2}}
|
||||
\end{listing}
|
||||
Finally, we convert the input matrix, the descriptor and the vectors,
|
||||
then preallocate the preconditioner workspace before entering the
|
||||
Krylov method. At the end of the code, we close the GPU environment
|
||||
\begin{listing}[h!]
|
||||
\ifpdf
|
||||
\begin{minted}[breaklines=true,bgcolor=bg,fontsize=\small]{fortran}
|
||||
call desc_a%cnv(mold=igmold)
|
||||
call a%cscnv(info,mold=agmold)
|
||||
call psb_geasb(x,desc_a,info,mold=vgmold)
|
||||
call psb_geasb(b,desc_a,info,mold=vgmold)
|
||||
|
||||
!
|
||||
! iterative method parameters
|
||||
!
|
||||
call psb_barrier(ctxt)
|
||||
call prec%allocate_wrk(info)
|
||||
t1 = psb_wtime()
|
||||
call psb_krylov(s_choice%kmethd,a,prec,b,x,s_choice%eps,&
|
||||
& desc_a,info,itmax=s_choice%itmax,iter=iter,err=err,itrace=s_choice%itrace,&
|
||||
& istop=s_choice%istopc,irst=s_choice%irst)
|
||||
call prec%deallocate_wrk(info)
|
||||
call psb_barrier(ctxt)
|
||||
tslv = psb_wtime() - t1
|
||||
|
||||
......
|
||||
call psb_gpu_exit()
|
||||
call psb_exit(ctxt)
|
||||
stop
|
||||
|
||||
|
||||
\end{minted}
|
||||
\else
|
||||
\begin{center}
|
||||
\begin{minipage}{.90\textwidth}
|
||||
{\small
|
||||
\begin{verbatim}
|
||||
call desc_a%cnv(mold=igmold)
|
||||
call a%cscnv(info,mold=agmold)
|
||||
call psb_geasb(x,desc_a,info,mold=vgmold)
|
||||
call psb_geasb(b,desc_a,info,mold=vgmold)
|
||||
|
||||
!
|
||||
! iterative method parameters
|
||||
!
|
||||
call psb_barrier(ctxt)
|
||||
call prec%allocate_wrk(info)
|
||||
t1 = psb_wtime()
|
||||
call psb_krylov(s_choice%kmethd,a,prec,b,x,s_choice%eps,&
|
||||
& desc_a,info,itmax=s_choice%itmax,iter=iter,err=err,itrace=s_choice%itrace,&
|
||||
& istop=s_choice%istopc,irst=s_choice%irst)
|
||||
call prec%deallocate_wrk(info)
|
||||
call psb_barrier(ctxt)
|
||||
tslv = psb_wtime() - t1
|
||||
|
||||
......
|
||||
call psb_gpu_exit()
|
||||
call psb_exit(ctxt)
|
||||
stop
|
||||
|
||||
\end{verbatim}
|
||||
}
|
||||
\end{minipage}
|
||||
\end{center}
|
||||
\fi
|
||||
\caption{setup of a GPU-enabled test program part three.\label{fig:gpu-ex3}}
|
||||
\end{listing}
|
||||
|
||||
It is very important to employ solvers that are suited
|
||||
to the GPU, i.e. solvers that do NOT employ triangular
|
||||
system solve kernels. Solvers that satisfy this constraint include:
|
||||
\begin{itemize}
|
||||
\item \verb|JACOBI|
|
||||
\item \verb|INVK|
|
||||
\item \verb|INVT|
|
||||
\item \verb|AINV|
|
||||
\end{itemize}
|
||||
and their \verb|L1| variants.
|
||||
|
||||
%%% Local Variables:
|
||||
%%% mode: latex
|
||||
|
||||
Reference in New Issue
Block a user