Skip to content

Commit 1c76bfc

Browse files
committed
wrap up timings in performance notebook
1 parent 1cacaef commit 1c76bfc

File tree

1 file changed

+15
-23
lines changed

1 file changed

+15
-23
lines changed

notebooks/03-uxarray-vis/06-performance.ipynb

Lines changed: 15 additions & 23 deletions
Original file line numberDiff line numberDiff line change
@@ -80,20 +80,11 @@
8080
{
8181
"cell_type": "markdown",
8282
"source": [
83-
"### Data Processing (Subsequent Runs)"
84-
],
85-
"metadata": {
86-
"collapsed": false
87-
}
88-
},
89-
{
90-
"cell_type": "markdown",
91-
"source": [
92-
"| Visualization Method / Grid Resolution | 30km | 15km | 7.5km | 3.75km |\n",
93-
"|-----------------------------------------|-------------|-------------|-------------|--------------|\n",
94-
"| Polygon Raster (Including Antimeridian) | 0 | 0 | 0 | 0 |\n",
95-
"| Polygon Raster (Excluding Antimeridian) | 0.30 (0.00) | 1.02 (0.36) | 3.46 (0.01) | 13.60 (0.08) |\n",
96-
"| Point Raster | 0.13 (0.03) | 0.16 (0.01) | 0.35 (0.00) | 1.08 (0.07) |"
83+
"We can see that Point Rasters are the best in terms of performance, about 86 times faster than Polygon Rasters (excluding AM). \n",
84+
"\n",
85+
"Both polygon plots scale linearly with an increase in resolution. A doubling in resolution leads to about a 4x increase in the number of polygons (a.k.a polygons), which is observed in the timings.\n",
86+
"\n",
87+
"Including antimeridian polygons leads to about a 20x slowdown across all resolutions, so it's suggested to keep `exclude_antimeridian=True` when working with larger datasets."
9788
],
9889
"metadata": {
9990
"collapsed": false
@@ -102,7 +93,7 @@
10293
{
10394
"cell_type": "markdown",
10495
"source": [
105-
"### Visualization "
96+
"### Data Processing (Subsequent Runs)"
10697
],
10798
"metadata": {
10899
"collapsed": false
@@ -111,14 +102,15 @@
111102
{
112103
"cell_type": "markdown",
113104
"source": [
114-
"| Visualization Method / Grid Resolution | 30km | 15km | 7.5km | 3.75km |\n",
115-
"|----------------------------------------|------|------|-------|--------|\n",
116-
"| Polygon Raster (0.5 Pixel Ratio) | 0 | 0 | 0 | 0 |\n",
117-
"| Polygon Raster (1.0 Pixel Ratio) | 0 | 0 | 0 | 0 |\n",
118-
"| Polygon Raster (4.0 Pixel Ratio) | 0 | 0 | 0 | 0 |\n",
119-
"| Point Raster (0.5 Pixel Ratio) | 0 | 0 | 0 | 0 |\n",
120-
"| Point Raster (1.0 Pixel Ratio) | 0 | 0 | 0 | 0 |\n",
121-
"| Point Raster (4.0 Pixel Ratio) | 0 | 0 | 0 | 0 |"
105+
"| Visualization Method / Grid Resolution | 30km | 15km | 7.5km | 3.75km |\n",
106+
"|-----------------------------------------|-------------|-------------|-------------|--------------|\n",
107+
"| Polygon Raster (Including Antimeridian) | 0.31 (0.00) | 1.32 (0.31) | 3.85 (0.06) | 14.36 (0.13) |\n",
108+
"| Polygon Raster (Excluding Antimeridian) | 0.30 (0.00) | 1.02 (0.36) | 3.46 (0.01) | 13.60 (0.08) |\n",
109+
"| Point Raster | 0.13 (0.03) | 0.16 (0.01) | 0.35 (0.00) | 1.08 (0.07) |\n",
110+
"\n",
111+
"For subsequent runs (i.e. we have already run one plotting instance, which computes and caches the necessary data structures), performance for both Polygon methods is essentially identical.\n",
112+
"\n",
113+
"There is no caching currently implemented for Point Rasters, so the performance for each run is consistent with the initial run.\n"
122114
],
123115
"metadata": {
124116
"collapsed": false

0 commit comments

Comments
 (0)