From fb607be28e4d67d94595a893ead6ae5003bc93e5 Mon Sep 17 00:00:00 2001
From: Dominik Thoennes <dominik.thoennes@fau.de>
Date: Thu, 19 Mar 2020 15:31:00 +0100
Subject: [PATCH] add open merge requests from pystencils_walberla
---
python/pystencils_walberla/jinja_filters.py | 31 +++++++++++--------
.../templates/Sweep.tmpl.h | 8 ++---
.../templates/SweepInnerOuter.tmpl.h | 14 ++++-----
3 files changed, 29 insertions(+), 24 deletions(-)
diff --git a/python/pystencils_walberla/jinja_filters.py b/python/pystencils_walberla/jinja_filters.py
index eb209237b..003315448 100644
--- a/python/pystencils_walberla/jinja_filters.py
+++ b/python/pystencils_walberla/jinja_filters.py
@@ -14,16 +14,18 @@ temporary_fieldMemberTemplate = """
private: std::set< {type} *, field::SwapableCompare< {type} * > > cache_{original_field_name}_;"""
temporary_fieldTemplate = """
-// Getting temporary field {tmp_field_name}
-auto it = cache_{original_field_name}_.find( {original_field_name} );
-if( it != cache_{original_field_name}_.end() )
{{
- {tmp_field_name} = *it;
-}}
-else
-{{
- {tmp_field_name} = {original_field_name}->cloneUninitialized();
- cache_{original_field_name}_.insert({tmp_field_name});
+ // Getting temporary field {tmp_field_name}
+ auto it = cache_{original_field_name}_.find( {original_field_name} );
+ if( it != cache_{original_field_name}_.end() )
+ {{
+ {tmp_field_name} = *it;
+ }}
+ else
+ {{
+ {tmp_field_name} = {original_field_name}->cloneUninitialized();
+ cache_{original_field_name}_.insert({tmp_field_name});
+ }}
}}
"""
@@ -42,7 +44,7 @@ def make_field_type(dtype, f_size, is_gpu):
if is_gpu:
return "cuda::GPUField<%s>" % (dtype,)
else:
- return "GhostLayerField<%s, %d>" % (dtype, f_size)
+ return "field::GhostLayerField<%s, %d>" % (dtype, f_size)
def get_field_fsize(field):
@@ -120,9 +122,9 @@ def field_extraction_code(field, is_temporary, declaration_only=False,
else:
prefix = "" if no_declaration else "auto "
if update_member:
- return "%s%s_ = block->uncheckedFastGetData< %s >(%sID);" % (prefix, field_name, field_type, field_name)
+ return "%s%s_ = block->getData< %s >(%sID);" % (prefix, field_name, field_type, field_name)
else:
- return "%s%s = block->uncheckedFastGetData< %s >(%sID);" % (prefix, field_name, field_type, field_name)
+ return "%s%s = block->getData< %s >(%sID);" % (prefix, field_name, field_type, field_name)
else:
assert field_name.endswith('_tmp')
original_field_name = field_name[:-len('_tmp')]
@@ -273,6 +275,9 @@ def generate_call(ctx, kernel_info, ghost_layers_to_include=0, cell_interval=Non
else:
spatial_shape_symbols = [TypedSymbol(s, SHAPE_DTYPE) for s in spatial_shape_symbols]
+ assert spatial_shape_symbols, "No shape parameters in kernel function arguments.\n"\
+ "Please be only use kernels for generic field sizes!"
+
indexing_dict = ast.indexing.call_parameters(spatial_shape_symbols)
sp_printer_c = CudaSympyPrinter()
kernel_call_lines += [
@@ -312,7 +317,7 @@ def generate_constructor_initializer_list(kernel_info, parameters_to_ignore=None
def generate_constructor_parameters(kernel_info, parameters_to_ignore=None):
if parameters_to_ignore is None:
parameters_to_ignore = []
-
+
varying_parameters = []
if hasattr(kernel_info, 'varying_parameters'):
varying_parameters = kernel_info.varying_parameters
diff --git a/python/pystencils_walberla/templates/Sweep.tmpl.h b/python/pystencils_walberla/templates/Sweep.tmpl.h
index 9de4b96c2..b4db0f35b 100644
--- a/python/pystencils_walberla/templates/Sweep.tmpl.h
+++ b/python/pystencils_walberla/templates/Sweep.tmpl.h
@@ -57,10 +57,10 @@ public:
{{ kernel| generate_destructor(class_name) |indent(4) }}
- void operator() ( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %} );
+ void operator() ( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %} );
void runOnCellInterval(const shared_ptr<StructuredBlockStorage> & blocks,
const CellInterval & globalCellInterval, cell_idx_t ghostLayers, IBlock * block
- {%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %});
+ {%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %});
@@ -74,8 +74,8 @@ public:
const CellInterval & globalCellInterval,
cell_idx_t ghostLayers=1 )
{
- return [kernel, blocks, globalCellInterval, ghostLayers] (IBlock * b{%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %}) {
- kernel->runOnCellInterval(blocks, globalCellInterval, ghostLayers, b{%if target is equalto 'gpu'%},stream {% endif %});
+ return [kernel, blocks, globalCellInterval, ghostLayers] (IBlock * b{%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %}) {
+ kernel->runOnCellInterval(blocks, globalCellInterval, ghostLayers, b{%if target is equalto 'gpu'%}, stream {% endif %});
};
}
diff --git a/python/pystencils_walberla/templates/SweepInnerOuter.tmpl.h b/python/pystencils_walberla/templates/SweepInnerOuter.tmpl.h
index 466f51071..6f6094d0e 100644
--- a/python/pystencils_walberla/templates/SweepInnerOuter.tmpl.h
+++ b/python/pystencils_walberla/templates/SweepInnerOuter.tmpl.h
@@ -54,17 +54,17 @@ class {{class_name}}
{
public:
{{class_name}}( {{kernel|generate_constructor_parameters}}, const Cell & outerWidth=Cell(1, 1, 1))
- : {{ kernel|generate_constructor_initializer_list }}, outerWidth_(outerWidth)
+ : {{ kernel|generate_constructor_initializer_list }}{% if kernel|generate_constructor_initializer_list|length %},{% endif %} outerWidth_(outerWidth)
{};
{{ kernel| generate_destructor(class_name) |indent(4) }}
- void operator() ( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %} );
+ void operator() ( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %} );
void runOnCellInterval(const shared_ptr<StructuredBlockStorage> & blocks,
const CellInterval & globalCellInterval, cell_idx_t ghostLayers, IBlock * block
- {%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %});
+ {%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %});
@@ -78,14 +78,14 @@ public:
const CellInterval & globalCellInterval,
cell_idx_t ghostLayers=1 )
{
- return [kernel, blocks, globalCellInterval, ghostLayers] (IBlock * b{%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %}) {
- kernel->runOnCellInterval(blocks, globalCellInterval, ghostLayers, b{%if target is equalto 'gpu'%},stream {% endif %});
+ return [kernel, blocks, globalCellInterval, ghostLayers] (IBlock * b{%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %}) {
+ kernel->runOnCellInterval(blocks, globalCellInterval, ghostLayers, b{%if target is equalto 'gpu'%}, stream {% endif %});
};
}
- void inner( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %} );
- void outer( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = 0{% endif %} );
+ void inner( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %} );
+ void outer( IBlock * block{%if target is equalto 'gpu'%} , cudaStream_t stream = nullptr{% endif %} );
void setOuterPriority(int priority ) {
{%if target is equalto 'gpu'%}
--
GitLab