From b83706cd983b17ba27a569d45992daa3737fc573 Mon Sep 17 00:00:00 2001
From: Martin Bauer <martin.bauer@fau.de>
Date: Thu, 18 Apr 2019 11:14:11 +0200
Subject: [PATCH] Bugfixes in pystencils_walberla and lbmpy_walberla
- missing pragma once added
- static variable to member in overlap sweep, when called with changing
block sizes this lead to wrong results
---
.../templates/GpuPackInfo.tmpl.h | 1 +
.../templates/SweepInnerOuter.tmpl.cpp | 20 +++++++++----------
.../templates/SweepInnerOuter.tmpl.h | 2 ++
3 files changed, 12 insertions(+), 11 deletions(-)
diff --git a/pystencils_walberla/templates/GpuPackInfo.tmpl.h b/pystencils_walberla/templates/GpuPackInfo.tmpl.h
index 74bd12b..19d68f5 100644
--- a/pystencils_walberla/templates/GpuPackInfo.tmpl.h
+++ b/pystencils_walberla/templates/GpuPackInfo.tmpl.h
@@ -1,3 +1,4 @@
+#pragma once
#include "stencil/Directions.h"
#include "core/cell/CellInterval.h"
#include "cuda/GPUField.h"
diff --git a/pystencils_walberla/templates/SweepInnerOuter.tmpl.cpp b/pystencils_walberla/templates/SweepInnerOuter.tmpl.cpp
index 3cf5069..de1c5b5 100644
--- a/pystencils_walberla/templates/SweepInnerOuter.tmpl.cpp
+++ b/pystencils_walberla/templates/SweepInnerOuter.tmpl.cpp
@@ -85,38 +85,36 @@ void {{class_name}}::inner( IBlock * block{%if target is equalto 'gpu'%} , cudaS
void {{class_name}}::outer( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream {% endif %} )
{
- static std::vector<CellInterval> layers;
-
{{kernel|generate_block_data_to_field_extraction|indent(4)}}
- if( layers.size() == 0 )
+ if( layers_.size() == 0 )
{
CellInterval ci;
{{field}}->getSliceBeforeGhostLayer(stencil::T, ci, 1, false);
- layers.push_back(ci);
+ layers_.push_back(ci);
{{field}}->getSliceBeforeGhostLayer(stencil::B, ci, 1, false);
- layers.push_back(ci);
+ layers_.push_back(ci);
{{field}}->getSliceBeforeGhostLayer(stencil::N, ci, 1, false);
ci.expand(Cell(0, 0, -1));
- layers.push_back(ci);
+ layers_.push_back(ci);
{{field}}->getSliceBeforeGhostLayer(stencil::S, ci, 1, false);
ci.expand(Cell(0, 0, -1));
- layers.push_back(ci);
+ layers_.push_back(ci);
{{field}}->getSliceBeforeGhostLayer(stencil::E, ci, 1, false);
ci.expand(Cell(0, -1, -1));
- layers.push_back(ci);
+ layers_.push_back(ci);
{{field}}->getSliceBeforeGhostLayer(stencil::W, ci, 1, false);
ci.expand(Cell(0, -1, -1));
- layers.push_back(ci);
+ layers_.push_back(ci);
}
{%if target is equalto 'gpu'%}
{
auto parallelSection_ = parallelStreams_.parallelSection( stream );
- for( auto & ci: layers )
+ for( auto & ci: layers_ )
{
parallelSection_.run([&]( auto s ) {
{{kernel|generate_call(stream='s', cell_interval='ci')|indent(16)}}
@@ -124,7 +122,7 @@ void {{class_name}}::outer( IBlock * block{%if target is equalto 'gpu'%} , cudaS
}
}
{% else %}
- for( auto & ci: layers )
+ for( auto & ci: layers_ )
{
{{kernel|generate_call(cell_interval='ci')|indent(8)}}
}
diff --git a/pystencils_walberla/templates/SweepInnerOuter.tmpl.h b/pystencils_walberla/templates/SweepInnerOuter.tmpl.h
index 733083a..c2c1d4d 100644
--- a/pystencils_walberla/templates/SweepInnerOuter.tmpl.h
+++ b/pystencils_walberla/templates/SweepInnerOuter.tmpl.h
@@ -97,6 +97,8 @@ private:
{%if target is equalto 'gpu'%}
cuda::ParallelStreams parallelStreams_;
{% endif %}
+
+ std::vector<CellInterval> layers_;
};
--
GitLab