MMW GWLF-E Comparison

This repo is compares the performance of GWLF-E at 0.6.3 against the improvements made by Drexel in their fork.

Prerequisites

• Python 2.7
• Pipenv

Running the Comparison

Check out this branch and run make. This will:

1. Install GWLF-E 0.6.3 in the wikiwatershed directory
2. Install GWLF-E drexeleds#develop in the drexeleds directory
3. Install dependencies for result interpretation in the main directory
4. Run both versions against the tests in tests directory
   • These tests were compiled using stored data from a number of MapShed runs on a local database using the extra/tests.sql query
5. Compare the timings for both and print the results
6. Compare the results for both with a 0.000000000001% tolerance and print the results

Results

Tolerance

Given the latest code, the numbers match:

$ make -s tolerance

huc08__1751.json matches to a tolerance of 1e-14
huc12__60101.json matches to a tolerance of 1e-14
huc08__1750.json matches to a tolerance of 1e-14
huc10__9392.json matches to a tolerance of 1e-14
huc08__1748.json matches to a tolerance of 1e-14
huc12__53956.json matches to a tolerance of 1e-14
huc10__2208.json matches to a tolerance of 1e-14
huc08__1773.json matches to a tolerance of 1e-14
huc10__38.json matches to a tolerance of 1e-14
huc10__1340.json matches to a tolerance of 1e-14
huc10__3176.json matches to a tolerance of 1e-14
huc12__60099.json matches to a tolerance of 1e-14
huc08__1747.json matches to a tolerance of 1e-14
huc10__12968.json matches to a tolerance of 1e-14
huc10__3928.json matches to a tolerance of 1e-14
huc10__11444.json matches to a tolerance of 1e-14
huc08__92.json matches to a tolerance of 1e-14
huc12__55174.json matches to a tolerance of 1e-14
huc10__3179.json matches to a tolerance of 1e-14
huc10__1341.json matches to a tolerance of 1e-14
huc12__60100.json matches to a tolerance of 1e-14

Timing

On my computer, a 2015 i5 13" MacBook Pro, there is significant speedup in the GWLF-E model, and slight slowdowns in reading and writing files:

$ make -s timing

huc08
* Average read_input_file speedup: -11.26%
* Average prepare_input_gms speedup: 0.00%
* Average read_input_gms speedup: 6.97%
* Average run_gwlfe speedup: 82.80%
* Average write_output_file speedup: -42.99%
* Average total speedup: 78.57%

huc10
* Average read_input_file speedup: -19.21%
* Average prepare_input_gms speedup: 0.00%
* Average read_input_gms speedup: 1.46%
* Average run_gwlfe speedup: 79.11%
* Average write_output_file speedup: 9.03%
* Average total speedup: 74.17%

huc12
* Average read_input_file speedup: -12.52%
* Average prepare_input_gms speedup: 0.00%
* Average read_input_gms speedup: 1.57%
* Average run_gwlfe speedup: 79.07%
* Average write_output_file speedup: 1.25%
* Average total speedup: 73.99%

Total
* Average read_input_file speedup: -15.34%
* Average prepare_input_gms speedup: 0.00%
* Average read_input_gms speedup: 3.06%
* Average run_gwlfe speedup: 80.15%
* Average write_output_file speedup: -7.69%
* Average total speedup: 75.38%