User talk:Bossenne - Revision history

Bossenne: /* Data Processing */

2012-09-14T17:05:41Z

‎Data Processing

Bossenne: /* Optimal size of blocks */

2012-09-14T17:05:17Z

‎Optimal size of blocks

Bossenne: /* Optimal size of blocks */

2012-09-14T17:04:54Z

‎Optimal size of blocks

Bossenne: /* Parallel efficiency */

2012-09-14T17:04:43Z

‎Parallel efficiency

Bossenne: /* Speed-up */

2012-09-14T17:04:34Z

‎Speed-up

Bossenne: /* Performances */

2012-09-14T17:04:06Z

‎Performances

Bossenne: /* Analysis of the call-graph profile */

2012-09-14T17:03:18Z

‎Analysis of the call-graph profile

Bossenne: /* Analysis of the call-graph profile */

2012-09-14T17:01:46Z

‎Analysis of the call-graph profile

Bossenne: /* Analysis */

2012-09-14T17:00:23Z

‎Analysis

Bossenne: /* Analysis */

2012-09-14T16:59:43Z

‎Analysis

@@ Line 1,043: / Line 1,043: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:Efficacite40.png|center|thumb|Parallel efficiency - Weak Scaling|300px]]
@@ Line 1,057: / Line 1,056: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:Acc40.png|center|thumb|Speed-up - Weak Scaling|300px]]

@@ Line 891: / Line 891: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
-|[[File:Efficacité réduite 80.png|center|thumb|Speed-up - Strong Scaling|300px]]
+|[[File:Efficacité réduite 80.png|center|thumb|Reduced efficiency - Strong Scaling|300px]]
 |}

@@ Line 890: / Line 890: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:Efficacité réduite 80.png|center|thumb|Speed-up - Strong Scaling|300px]]

@@ Line 878: / Line 878: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:P3D - eff SS.png|center|thumb|Parallel efficiency - Strong Scaling|300px]]

@@ Line 868: / Line 868: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:P3D - acc SS.png|center|thumb|Speed-up - Strong Scaling|300px]]

@@ Line 116: / Line 116: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:P3D_-_acc_SS.png|center|thumb|Performances of H-Allegro|300px]]

← Older revision		Revision as of 17:03, 14 September 2012
Line 1,288:		Line 1,288:

	This report details the percentage of the total time that was spent in a function and its children. The total of these percentages is not equal to 100%. For instance the run time percentage for the main egals 100% because it is a routine which takes a lot of time, but the routine solving the time stepping has a run time of 97.3%. This report gives also the amount of time that was propagated from the function into the parent (self), the amount of time that was propagated from the children into the function (children), the number of calls from a routine to another.		This report details the percentage of the total time that was spent in a function and its children. The total of these percentages is not equal to 100%. For instance the run time percentage for the main egals 100% because it is a routine which takes a lot of time, but the routine solving the time stepping has a run time of 97.3%. This report gives also the amount of time that was propagated from the function into the parent (self), the amount of time that was propagated from the children into the function (children), the number of calls from a routine to another.
		+
		+	{\| class="wikitable" style="margin: 1em auto 1em auto;"
		+	\|-
		+	\|[[File:routineEchange.png\|center\|thumb\|Call-graph of the routine "Echange"\|300px]]
		+	\|}

	For instance, if we consider the subroutine der_sub_MOD_echange (5), we can see that the parent functions are der_MOD_interpolate2 (16), der_MOD_derivative1 (15), der_MOD_derivative2 (11), der_MOD_interpolate1 (8). This subroutine is called 1014 times by the routines der_MOD_interpolate2 (93 calls), der_MOD_derivative1 (174 calls), der_MOD_derivative2 (327 calls) and der_MOD_interpolate1 (420 calls). The time spent in this function is about 0.01 second.		For instance, if we consider the subroutine der_sub_MOD_echange (5), we can see that the parent functions are der_MOD_interpolate2 (16), der_MOD_derivative1 (15), der_MOD_derivative2 (11), der_MOD_interpolate1 (8). This subroutine is called 1014 times by the routines der_MOD_interpolate2 (93 calls), der_MOD_derivative1 (174 calls), der_MOD_derivative2 (327 calls) and der_MOD_interpolate1 (420 calls). The time spent in this function is about 0.01 second.

@@ Line 1,282: / Line 1,282: @@
 This part is very interesting because it allows knowing which routine calls another one and how long that takes. The routines are indexed by decreasing order of CPU time comsumption.
-This report details the percentage of the total time that was spent in a function and its children. The total of these percentages is not equal to 100%. For instance the run time percentage for the main egals 100% because it is a routine which takes a lot of time, but the routine solving the time stepping has a run time of 97.3%. This report gives also the amount of time that was propagated from the function into the parent (self), the amount of time that was propagated from the children into the function (children), the number of calls from a routine to another.
+{|  class="wikitable" style="margin: 1em auto 1em auto;"
-Figure
+This report details the percentage of the total time that was spent in a function and its children. The total of these percentages is not equal to 100%. For instance the run time percentage for the main egals 100% because it is a routine which takes a lot of time, but the routine solving the time stepping has a run time of 97.3%. This report gives also the amount of time that was propagated from the function into the parent (self), the amount of time that was propagated from the children into the function (children), the number of calls from a routine to another.
 For instance, if we consider the subroutine der_sub_MOD_echange (5), we can see that the parent functions are der_MOD_interpolate2 (16), der_MOD_derivative1 (15), der_MOD_derivative2 (11), der_MOD_interpolate1 (8). This subroutine is called 1014 times by the routines der_MOD_interpolate2 (93 calls), der_MOD_derivative1 (174 calls), der_MOD_derivative2 (327 calls) and der_MOD_interpolate1 (420 calls). The time spent in this function is about 0.01 second.

@@ Line 1,264: / Line 1,264: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:Profilplat4x1x1.png|center|thumb|Flat profile|300px]]
@@ Line 1,274: / Line 1,273: @@
 {|  class="wikitable" style="margin: 1em auto 1em auto;"
 |-
 |[[File:Comparaisoncall.png|center|thumb|Number of calls : 2x2x1 cores|300px]]

@@ Line 1,261: / Line 1,261: @@
 *der_sub_MOD_interpolate2_on_slice
-For instance for a calculation on 4x1x1 processors, the flat profile will be the following :
+For instance for a calculation on 4x1x1 cores, the flat profile will be the following :
-figure
+{|  class="wikitable" style="margin: 1em auto 1em auto;"
 It indicates that the subroutine derivative2_on_slice represents 20 % of total consumed CPU time. Thanks to the cumulative seconds, we can deduce that the total consumed CPU time egals 0.25 second. The CPU time of the subroutine derivative2_on_slice is about 0.05 second. This subroutine is called 545 times. We can say that the average number of milliseconds spent in this subroutine and its children per call is 0.09 ms.
 Remark : If the calculation is run on 2x2x1 instead of 4x1x1, the number of calls of the subroutines derivative and interpolate periodic are divided by 2, while the number of calls of the subroutines derivative and interpolate on slice increases.
 ==== Analysis of the call-graph profile ====