commit fee2baa5c4f6535bb03a738cc2baa1ff9527fe04
parent 511d33ca409b340799fd7e3715daf5f61face8ff
Author: Ivan Gankevich <igankevich@ya.ru>
Date: Tue, 15 Aug 2017 14:28:50 +0300
Update to the latest benchmark repository revision.
Diffstat:
2 files changed, 15 insertions(+), 14 deletions(-)
diff --git a/R/benchmarks.R b/R/benchmarks.R
@@ -10,12 +10,9 @@ arma.load_benchmark_data <- function(attempt, framework, models, tags) {
idx <- 1
for (t in tags) {
values <- arma.load(
- file.path("build", "arma-benchmarks", "output"),
- "gpulab1",
- attempt,
- 10000,
- framework,
- m,
+ file.path(
+ "build", "arma-benchmarks", "output",
+ "gpulab1", attempt, 10000, framework, m),
t,
".*\\s+([0-9]+)us.*"
)
@@ -70,12 +67,9 @@ arma.load_io_benchmark_data <- function(attempt, filesystems, suffix, tags) {
idx <- 1
for (t in tags) {
values <- arma.load(
- file.path("build", "arma-benchmarks", "output"),
- "gpulab1",
- paste(attempt, fs, suffix, sep="-"),
- 10000,
- "openmp",
- "ar",
+ file.path(
+ "build", "arma-benchmarks", "output", "gpulab1",
+ paste(attempt, fs, suffix, sep="-"), 10000, "openmp", "ar"),
t,
".*\\s+([0-9]+)us.*"
)
diff --git a/arma-thesis.org b/arma-thesis.org
@@ -1205,7 +1205,7 @@ fi
cd $dir
git checkout master
git pull
-git checkout 908739d3a21354b86b5397619deca26c84355fb6
+git checkout 1ed6679387f0b79d8495c8bf55a6b0b304347e48
#+end_src
#+RESULTS:
@@ -3603,6 +3603,7 @@ model CPU architecture is superior compared to GPU due to better handling of
complex information dependencies, non-intensive calculations (multiplications
and additions) and complex memory access patterns.
+#+header :results output raw :exports results
#+name: tab-arma-performance
#+begin_src R
source(file.path("R", "benchmarks.R"))
@@ -3673,6 +3674,11 @@ possibility to optimise this routine is to use root finding method with
quadratic convergence rate to reduce the number of transcendental functions that
need to be computed.
+**** I/O performance.
+:PROPERTIES:
+:header-args:R: :results output raw :exports results
+:END:
+
Although, in the current benchmarks writing data to files does not consume much
of the running time, the use of network-mounted file systems may slow down this
stage. To optimise it wavy surface parts were written to file as soon as full
@@ -3687,6 +3693,7 @@ computation so the total running time is decreased
system has the same effect, but the total reduction in execution time is small,
because local file system is more performant.
+#+header :results output raw :exports results
#+name: tab-arma-io-performance
#+begin_src R
source(file.path("R", "benchmarks.R"))
@@ -3765,7 +3772,7 @@ wavy surface.
In order to investigate, how GPGPU computations can be used to speed-up velocity
potential field computation, we benchmarked simplified version of
-eq.\nbsp{}eqref:phi-3d:
+eq.\nbsp{}eqref:eq-phi-3d:
\begin{align}
\label{eq:phi-linear}
\phi(x,y,z,t) &= \InverseFourierY{
