Stream results updated roci results.

doc/sphinx/10_microbenchmarks/M1_STREAM/STREAM.rst

@@ -21,23 +21,35 @@ Problem
 There are four memory operations that the STREAM benchmark measures: Copy, Scale, Add, and Triad.
 
 Copy - Copies data from one array to another:
+
+.. math:: 
+
 b[i] = a[i]
 
 Scale - Multiplies each array element by a constant, a daxpy operation.
+
+.. math::
+
 b[i] = q*a[i]
 
 Add - Adds two arrays element-wise:
+
+.. math::
+
 c[i] = a[i] + b[i]
 
 Triad - Multiply-add operation:
+
+.. math::
+
 a[i] = b[i] + q*c[i]
 
 These operations stress memory and floating point pipelines.They test memory transfer speed, computation speed, and different combinations of these two components of overall performance performance.
 
 Figure of Merit
 ---------------
 
-The primary FOM is the max Triad rate (MB/s).
+The primary FOM is the MAX Triad rate (MB/s).
 
 Run Rules
 ---------
@@ -118,6 +130,17 @@ Example Results
 ATS-3 Rocinante HBM
 -------------------
 
+.. csv-table:: STREAM microbenchmark bandwidth measurement
+   :file: stream-xrds_ats5cce-cray-mpich.csv
+   :align: center
+   :widths: 10, 10, 10
+   :header-rows: 1
+
+.. figure:: stream_cpu_ats3.png
+   :align: center
+   :scale: 50%
+   :alt: STREAM microbenchmark bandwidth measurement
+
 CTS-1 Snow
 -----------
 
@@ -127,7 +150,7 @@ CTS-1 Snow
    :widths: 10, 10, 10
    :header-rows: 1
 
-.. figure:: cpu_cts1.png
+.. figure:: stream_cpu_cts1.png
    :align: center
    :scale: 50%
    :alt: STREAM microbenchmark bandwidth measurement

doc/sphinx/10_microbenchmarks/M1_STREAM/cpu.gp

@@ -1,20 +1,21 @@
 #!/usr/bin/gnuplot
 #STREAM
 set terminal pngcairo enhanced size 1024, 768 dashed font 'Helvetica,18'
-set output "cpu_cts1.png"
+set output "stream_cpu_cts1.png"
 
 set title "STREAM Single node bandwidth" font "serif,22"
-set ylabel "Per core triad (MB/s)"
-set y2label "FOM: Total Triad (MB/s)"
+set ylabel "Per core triad BW (MB/s)"
+set y2label "FOM: Total triad BW (MB/s)"
 
 set xrange [1:40]
-set yrange [3000:15000]
+#set yrange [3000:15000]
 
 # set logscale x 2
-# set logscale y 2
+set logscale y 2
 
 set grid
 show grid
+set key left top
 
 set datafile separator comma
 set key autotitle columnheader
@@ -24,5 +25,9 @@ set style line 2 linetype 1 dashtype 2 linecolor rgb "#FF0000" linewidth 2
 
 plot "stream_cts1.csv" using 1:2 with linespoints linestyle 1 axis x1y1, "" using 1:3 with line linestyle 2 axis x1y2
 
+set output "stream_cpu_ats3.png"
+set xrange [4:115]
+plot "stream-xrds_ats5cce-cray-mpich.csv" using 1:2 with linespoints linestyle 1 axis x1y1, "" using 1:3 with line linestyle 2 axis x1y2
+
 
 

doc/sphinx/10_microbenchmarks/M1_STREAM/stream-xrds_ats5cce-cray-mpich.csv

@@ -0,0 +1,6 @@
+No. Cores,Per Core Bandwidth (MB/s), Total Bandwidth (MB/s)
+8,2.304e+04,1.843e+05
+32,1.883e+04,6.025e+05
+56,1.452e+04,8.131e+05
+88,1.566e+04,1.378e+06
+112,1.414e+04,1.584e+06

doc/sphinx/10_microbenchmarks/M1_STREAM/stream_cts1.csv

@@ -1,4 +1,4 @@
-No. Cores,Bandwidth (MB/s),Total Bandwidth (MB/s)
+No. Cores,Per Core Bandwidth (MB/s),Total Bandwidth (MB/s)
 1,10690.1,10690.1
 2,10701.3,21402.6
 4,9316.5,37266.0

utils/pav_config/tests/stream.yaml

@@ -32,7 +32,7 @@ _base:
         Speed Network backbone.
 
         INSTRUCTIONS:
-       
+
              1) STREAM requires different amounts of memory to run on different
                   systems, depending on both the system cache size(s) and the
                   granularity of the system timer.
@@ -56,7 +56,7 @@ _base:
                       granularity.  20 "ticks" at 10 ms/tic is 200 milliseconds.
                       If the chip is capable of 10 GB/s, it moves 2 GB in 200 msec.
                       This means the each array must be at least 1 GB, or 128M elements.
-       
+
              Version 5.10 increases the default array size from 2 million
                  elements to 10 million elements in response to the increasing
                  size of L3 caches.  The new default size is large enough for caches
@@ -65,7 +65,7 @@ _base:
                  to "ssize_t", which allows array indices >2^32 (4 billion)
                  on properly configured 64-bit systems.  Additional compiler options
                  (such as "-mcmodel=medium") may be required for large memory runs.
-       
+
              Array size can be set at compile time without modifying the source
                  code for the (many) compilers that support preprocessor definitions
                  on the compile line.  E.g.,
@@ -92,14 +92,14 @@ _base:
             CACHE: 105
             SOCKETS: 2
             4 x (105M * 2 ) / 3 ARRAYS / 8 BYTES/ELEMENT = 35 Mi elements = 35000000
-            
+
     scheduler: slurm
     schedule:
         nodes: '10' # 'ALL'
         tasks_per_node: 1
         share_allocation: false
 
-    permute_on: 
+    permute_on:
         - compilers
         - mpis
         - omp_num_threads
@@ -344,6 +344,7 @@ spr_hbm_xrds:
             - 'module load {{compilers.name}}/{{compilers.version}}'
             - 'module load {{mpis.modulefile}}'
 
+
 cts1_ats5:
     inherits_from: cts1_xrds
     subtitle: '{{compilers.name}}-{{compilers.version}}_{{tpn}}_{{mpis.name}}-{{mpis.version}}'
@@ -354,31 +355,64 @@ cts1_ats5:
         - tpn
 
     variables:
-        numnodes: '1'
         tpn: [1, 2, 4, 8, 16, 32, 36]
+        numnodes: '1'
         omp_num_threads: '1'
 
+    schedule:
+        nodes: "{{numnodes}}"
+        share_allocation: true
+        tasks_per_node: "{{tpn}}"
+
     run:
         env:
             GOMP_CPU_AFFINITY: ''
 
+    result_parse:
+        regex:
+            triad_once:
+                regex: '^Triad: *([0-9\.]*) '
+                action: store
+                match_select: last
+                files: '*stream'
+
+    result_evaluate:
+        total_bandwidth: '{{tpn}}*triad_once'
+
 xrds_ats5:
-    inherits_from: cts1_ats5
+    inherits_from: _base
+    subtitle: '{{compilers.name}}-{{compilers.version}}_{{tpn}}_{{mpis.name}}-{{mpis.version}}'
+
+    permute_on:
+        - compilers
+        - mpis
+        - tpn
 
     only_if:
-        "{{sys_name}}": ['crossroads', 'rocinante']
+        "{{sys_os.name}}": [ cos ]
 
     variables:
         tpn: [8, 32, 56, 88, 112]
         arch: "spr"
-        stream_array_size: 35e6
-        omp_places: [cores, sockets]
-        omp_proc_bind: [true]
-
+        target: "xrds-stream.exe"
+        stream_array_size: 35000000
+        #omp_places: [cores, sockets]
+        #omp_proc_bind: [true]
+        numnodes: '1'
+        omp_num_threads: '1'
+
+    chunk: '{{chunk_ids.0}}'
+
     schedule:
-      partition: 'hbm'
+        partition: 'hbm'
+        nodes: "{{numnodes}}"
+        share_allocation: true
+        tasks_per_node: "{{tpn}}"
+        chunking:
+            size: 1
 
     build:
+        on_nodes: false
         preamble:
             #- 'module load friendly-testing' #'module rm craype-hugepages2M'
             - 'module swap PrgEnv-${PE_ENV,,} PrgEnv-{{compilers.pe_env}}'
@@ -394,4 +428,16 @@ xrds_ats5:
             - 'module load {{mpis.name}}/{{mpis.version}}'
 
         env:
-            GOMP_CPU_AFFINITY: ''
+            GOMP_CPU_AFFINITY: ''
+
+    result_parse:
+        regex:
+            triad_once:
+                regex: '^Triad: *([0-9\.]*) '
+                action: store
+                match_select: all
+                files: '*stream'
+
+    result_evaluate:
+        per_proc_bw: 'sum(triad_once)/len(triad_once)'
+        total_bw: 'sum(triad_once)'

utils/pavparse

@@ -185,7 +185,7 @@ if __name__ == '__main__':
                 gdf.index.name = index_title
             if value_title:
                 gdf.columns = value_title
-            grouped_dfs[gtitle] = gdf.squeeze().map(fmt.format)
+            grouped_dfs[gtitle] = gdf.applymap(fmt.format)
             df_mean[gtitle] = gdf.mean().values[0]
 
         plt.legend(plt_legend)