Profiling Lazy Imports vs importHref on PSK

profiling_data.md

Profiling of lazy-imports vs importHref

Setup

System Specs

• CPU: Intel Core i5
• Memory: 16Gb
• OS: Windows 10
• Polymer CLI: 1.6.0
• Chrome: 64.0.3282.186 (32-bit)

Branches used

• importHref: master code
• lazy-imports: ms/2.0-lazy-imports code
• Modified lazy-imports: ms/refactoring code

Building & Serving

polymer build && polymer serve build/es6-bundled/

Results

Lighthouse Results

These results were generated using Lighthouse from the Audits tab in Chrome DevTools.

Method	Run 1 (ms)	Run 2 (ms)	Run 3 (ms)	Run Avg (ms)	Avg Score
importHref w/ async: true (Default)	3890	3640	3720	3750	78.33
__importLazyGroup w/ async: false (Default)	4100	4180	4220	4166.66	74.66
importHref w/ async: false	5280	4380	4360	4673.33	69.33
__importLazyGroup w/ async: true	3640	3490	3580	3570	80
Modified __importLazyGroup w/ async: false	4510	4190	4170	4290	73.33
Modified __importLazyGroup w/ async: true	3670	3690	3610	3656.66	79.66

These test results represent the initial page load. The next sections will cover post initial load and navigating to pages already loaded.

Loading additional pages

These results were made analyzing function calls in Chrome DevTools Performance tab with a RAW polymer serve. I measured the total time taken in _pageChanged

Method	1	2	3	4	5	6	7	8	9	10	Call Avg (ms)
importHref w/ async: true	1.14	14.67	11.34	12.38	.58	8.76	.83	11.81	11.37	.62	7.35
__importLazyGroup w/ async: true	9.11	10.63	.95	10.74	.27	9.36	.41	14.72	10.2	12.16	8.929
Modified __importLazyGroup w/ async: true	.6	4.05	.64	6.48	.59	.59	10.19	7.28	.58	1.25	3.225

Navigating to existing pages

Measuring repeated calls to import pages that have already been imported. Unfortunately, I couldn't consistently get the _pageChanged function call in the Performance tab. I was however able to more consistently get it to show up when using importHref which generally took ~20-28ms.