From 7523a8898443fdf43e80bcca140f926aa5bf8200 Mon Sep 17 00:00:00 2001
From: Martin Bauer <martin.bauer@fau.de>
Date: Wed, 23 May 2018 10:55:57 +0200
Subject: [PATCH] Datahandling arrays are not filled by default any more
- updates in lbmpy benchmark
-> enables better numa placement
---
datahandling/serial_datahandling.py | 1 -
kerncraft_coupling/generate_benchmark.py | 12 ++++++------
timeloop.py | 1 +
3 files changed, 7 insertions(+), 7 deletions(-)
diff --git a/datahandling/serial_datahandling.py b/datahandling/serial_datahandling.py
index e91b37f1c..7f2ff29ea 100644
--- a/datahandling/serial_datahandling.py
+++ b/datahandling/serial_datahandling.py
@@ -104,7 +104,6 @@ class SerialDataHandling(DataHandling):
byte_offset = ghost_layers * np.dtype(dtype).itemsize
cpu_arr = create_numpy_array_with_layout(layout=layout_tuple, alignment=alignment,
byte_offset=byte_offset, **kwargs)
- cpu_arr.fill(np.inf)
if alignment and gpu:
raise NotImplementedError("Alignment for GPU fields not supported")
diff --git a/kerncraft_coupling/generate_benchmark.py b/kerncraft_coupling/generate_benchmark.py
index 7f80f743f..61b544563 100644
--- a/kerncraft_coupling/generate_benchmark.py
+++ b/kerncraft_coupling/generate_benchmark.py
@@ -19,7 +19,7 @@ void dummy(double *);
extern int var_false;
-{{kernelCode}}
+{{kernel_code}}
int main(int argc, char **argv)
@@ -32,8 +32,8 @@ int main(int argc, char **argv)
{%- for field_name, dataType, size in fields %}
// Initialization {{field_name}}
- double * {{field_name}} = aligned_malloc(sizeof({{dataType}}) * {{size}}, 32);
- for (int i = 0; i < {{size}}; ++i)
+ double * {{field_name}} = (double *) aligned_malloc(sizeof({{dataType}}) * {{size}}, 32);
+ for (unsigned long long i = 0; i < {{size}}; ++i)
{{field_name}}[i] = 0.23;
if(var_false)
@@ -60,7 +60,7 @@ int main(int argc, char **argv)
for (; repeat > 0; --repeat)
{
- {{kernelName}}({{callArgumentList}});
+ {{kernelName}}({{call_argument_list}});
// Dummy calls
{%- for field_name, dataType, size in fields %}
@@ -100,10 +100,10 @@ def generate_benchmark(ast, likwid=False):
args = {
'likwid': likwid,
- 'kernelCode': generate_c(ast),
+ 'kernel_code': generate_c(ast),
'kernelName': ast.function_name,
'fields': fields,
'constants': constants,
- 'callArgumentList': ",".join(call_parameters),
+ 'call_argument_list': ",".join(call_parameters),
}
return benchmark_template.render(**args)
diff --git a/timeloop.py b/timeloop.py
index db830c3cd..9a4a3c461 100644
--- a/timeloop.py
+++ b/timeloop.py
@@ -78,6 +78,7 @@ class TimeLoop:
"""
# Run a few time step to get first estimate
if number_of_time_steps_for_estimation == 'auto':
+ self.run(1)
iterations, total_time = self.run_time_span(0.5)
duration_of_time_step = total_time / iterations
else:
--
GitLab