Documentation/Labs/BuildSystem ImproveCMakeConfigurationTime - Revision history

CarlosLuque: /* Results */

2019-09-17T07:53:42Z

Results

CarlosLuque: /* Results */

2019-09-16T11:46:11Z

Results

CarlosLuque: /* Results */

2019-09-16T11:44:45Z

Results

Show changes

CarlosLuque: /* Conclusions */

2019-09-16T11:39:12Z

Conclusions

CarlosLuque: /* Conclusions */

2019-09-16T11:38:59Z

Conclusions

CarlosLuque at 11:27, 16 September 2019

2019-09-16T11:27:59Z

CarlosLuque: /* Results */

2019-09-16T08:45:30Z

Results

CarlosLuque: /* Enviroment */

2019-09-16T07:53:50Z

Enviroment

CarlosLuque: updating table

2019-09-16T07:47:20Z

updating table

Show changes

JChris.FillionR: /* Results */

2019-09-03T20:01:05Z

Results

@@ Line 103: / Line 103: @@
 We observe several behaviours:
-* When the number of the targets is decreased, the time saving is meaningful in CMake versions. In the desktop 1, the time reduction is around 62% (version 4.14.4) and 52% (version 3.15.20190910-g39c58dd) from four targets to one target. In the desktop2, the time reduction is around 62% (version 4.14.4) and 44% (version 3.15.20190910-g39c58dd) and in the surface, it is  around 63% (version 4.14.4) and 46% (version 3.15.20190910-g39c58dd). Moreover, the time reduction averages from version 4.14 to version 3.15.20190910-g39c58dd are 51%, 54%, and 58% in the desktop 1, desktop 2, and surface, respectivaly.
+* When the number of the targets is decreased, the time saving is meaningful in CMake versions. In the desktop 1, the time reduction is around 62% (version 4.14.4) and 52% (version 3.15.20190910-g39c58dd) from four targets to one target. In the desktop2, the time reduction is around 62% (version 4.14.4) and 44% (version 3.15.20190910-g39c58dd) and in the surface, it is  around 63% (version 4.14.4) and 46% (version 3.15.20190910-g39c58dd). Moreover, the time reduction averages from version 4.14 to version 3.15.20190910-g39c58dd are 51%, 54%, and 58% in the desktop 1, desktop 2, and surface, respectively.
 * The hard disk technologies affect the time reduction. If the hard disk technology is SSD, the time reduction average is 7% (version 4.14.4) and 18% (version 3.15.20190910-g39c58dd).
 * The variable ''Slicer_BUILD_QTLOADABLEMODULES'' has a relevant impact on generate time. The time reduction average is 70% (version 4.14.4), and  65% (version 3.15.20190910-g39c58dd) when the variable ''Slicer_BUILD_QTLOADABLEMODULES'' is not active and four targets. All remaining variables affect lesser in the reduction time.

@@ Line 97: / Line 97: @@
 | cmake -G "Visual Studio 14 2015 Win64"  -DSlicer_BUILD_CLI=1 -DSlicer_BUILD_I18N_SUPPORT=0 -DSlicer_USE_PYTHONQT=1 -DSlicer_BUILD_QTLOADABLEMODULES=1 -DSlicer_BUILD_QTSCRIPTEDMODULES=1 -DSlicer_EP_LABEL_REMOTE_MODULE="" -DCMAKE_CONFIGURATION_TYPES="Debug" . || 117.22 || 119.79 || 111.90 || 68.50 || 69.11 || 56.11
 |}
-The figure below shows the previous information in a graphic mode.
+The below figures show the previous information in a graphic mode.
 [[File:CMakePerformance4144.png | 1000px]]

@@ Line 110: / Line 110: @@
 === Conclusions ===
 The number of the targets, the hard disk technologies and the variable ''Slicer_BUILD_QTLOADABLEMODULES'' have an important impact on the the CMake configuration time in both CMake versions. The hard technologies in the CMake version 4.14.4 affects less than the version nightly (3.15.20190910-g39c58dd).
 The CMake version nightly (3.15.20190910-g39c58dd) performs better than the version 4.14.4 in the configuration time. The time reduction average is 46% in one target.

← Older revision		Revision as of 11:38, 16 September 2019
Line 109:		Line 109:

	=== Conclusions ===		=== Conclusions ===
−	The number of the targets, the hard disk technologies and the variable ''Slicer_BUILD_QTLOADABLEMODULES'' have an important impact on the the CMake configuration time.	+	The number of the targets, the hard disk technologies and the variable ''Slicer_BUILD_QTLOADABLEMODULES'' have an important impact on the the CMake configuration time in both CMake versions. The hard technologies in the CMake version 4.14.4 affects less than the version nightly (3.15.20190910-g39c58dd).
		+	The CMake version nightly (3.15.20190910-g39c58dd) performs better than the version 4.14.4 in the configuration time. The time reduction average is 46% in one target.

@@ Line 99: / Line 99: @@
 The figure below shows the previous information in a graphic mode.
-[[File:CMakePerformance.png]]
+[[File:CMakePerformance4144.png | 1000px]]
 We observe several behaviours:
-* When the number of the targets is decreased, the time saving is meaningful. In the desktop 1, the time reduction is around 65% and is around 64% and 49% in the desktop2 and the surface respectively.
+* When the number of the targets is decreased, the time saving is meaningful in CMake versions. In the desktop 1, the time reduction is around 62% (version 4.14.4) and 52% (version 3.15.20190910-g39c58dd) from four targets to one target. In the desktop2, the time reduction is around 62% (version 4.14.4) and 44% (version 3.15.20190910-g39c58dd) and in the surface, it is  around 63% (version 4.14.4) and 46% (version 3.15.20190910-g39c58dd). Moreover, the time reduction averages from version 4.14 to version 3.15.20190910-g39c58dd are 51%, 54%, and 58% in the desktop 1, desktop 2, and surface, respectivaly.
-* The hard disk technologies affect the time reduction. If the hard disk technology is SSD, the time reduction average is 29 %. Moreover, the time saving is larger test 1 (43%) than tests 4 y 5, because the number of files generated in test 1 is larger than tests 4 y 5 in the configuration four targets.
+* The hard disk technologies affect the time reduction. If the hard disk technology is SSD, the time reduction average is 7% (version 4.14.4) and 18% (version 3.15.20190910-g39c58dd).
-* The variable ''Slicer_BUILD_QTLOADABLEMODULES'' has a relevant impact on generate time. The time reduction is between from 65% to 75% when the variable Slicer_BUILD_QTLOADABLEMODULES is not active and four targets. All remaining variables affect lesser in the reduction time.
+* The variable ''Slicer_BUILD_QTLOADABLEMODULES'' has a relevant impact on generate time. The time reduction average is 70% (version 4.14.4), and  65% (version 3.15.20190910-g39c58dd) when the variable ''Slicer_BUILD_QTLOADABLEMODULES'' is not active and four targets. All remaining variables affect lesser in the reduction time.
 === Conclusions ===
-The number of the targets, the hard disk technologies and the variable ''Slicer_BUILD_QTLOADABLEMODULES'' have an important impact on the the CMake configuration time. The time saving average is 59% when the number of targets is one.  The time reduction average is 29 % if the SSD technology is used.
+The number of the targets, the hard disk technologies and the variable ''Slicer_BUILD_QTLOADABLEMODULES'' have an important impact on the the CMake configuration time.

@@ Line 28: / Line 28: @@
 {| class="wikitable"
 !rowspan="2" | Test name
-!rowspan="2" |Command Line
+!rowspan="2" | Command Line
-!colspan="3" | Cmake 3.14.4
+!colspan="3" | CMake 3.14.4
 !colspan="3" | CMake nightly (3.15.20190910-g39c58dd)
 |-

@@ Line 16: / Line 16: @@
 ==== Enviroment ====
-The test enviroment uses the CMake version 3.14.4 and the master branch of Slicer repository of the fork mt4sd [https://github.com/mt4sd/Slicer link].
+The test enviroment uses the CMake versions ''3.14.4'' and ''nightly (3.15.20190910-g39c58dd)'', and the master branch of Slicer repository (commit: ''e091a2b016eef4a621fce1dfecc6faa7f6433b51'').
-The 21 tests were generated in different configuration variables. All tests were executed in the folder <code>Slicer-SuperBuild/Slicer-build</code> after the Slicer had been built.
+The 21 tests were generated in different configuration variables. All tests were executed ''five times'' in the folder <code>Slicer-SuperBuild/Slicer-build</code> after the Slicer had been built.
 The modified variables are:  ''Slicer_BUILD_CL, Slicer_BUILD_I18N_SUPPORT, Slicer_USE_PYTHONQT, Slicer_BUILD_QTLOADABLEMODULES, Slicer_BUILD_QTSCRIPTEDMODULES, Slicer_EP_LABEL_REMOTE_MODULE, CMAKE_CONFIGURATION_TYPES''. The internationalization(I18n) is not activated any tests.