54 lines
10 KiB
TeX
54 lines
10 KiB
TeX
\section{Discussion}
|
|
\label{ch:discussion}
|
|
|
|
In this section the results are interpreted. This section is structured in different questions to answer. First it will be analyzed what the browser differences are. One section will deal with the performance of the pure JavaScript implementations while the next will inspect how Simplify.wasm performs. Then further insights to the performance of the WebAssembly implementation will be given. It will be investigated how long it takes to set up the WebAssembly call and how much time is spent to actually execute the simplification routines. Next the case of Turf.js will be addressed and if its format conversions are reasonable under specific circumstances. Finally, the performance of mobile devices will be evaluated.
|
|
|
|
\subsection{Browser differences for the JavaScript implementations}
|
|
|
|
The first thing to see from the results of chapter \ref{ch:case1} and \ref{ch:case3} is that there is actually a considerable performance difference in the two versions of Simplify.js. So here we take a closer look at the JavaScript performance of the browsers. Interestingly, a clear winner between the similar algorithms cannot be determined as the performance is inconsistent across browsers. While the original version is faster in Firefox and Safari, the altered version is superior in Chrome and Edge. This is regardless of whether the high quality mode is switched on or not. The difference is however more significant when the preprocessing step is disabled.
|
|
|
|
In figure \ref{fig:win_chro_simplify_vs_true} and \ref{fig:win_edge_simplify_vs_true} one can see how similar Chrome and Edge perform with high quality mode enabled. When disabled however, the algorithms perform similar in Edge, while in Chrome the alternative version still improves upon the original.
|
|
|
|
In Firefox the result is very different. Without the high quality mode the original version performs about 2.5 times faster than the alternative. When disabling the preprocessing, the performance gain is even higher. The original performs constantly 3 times faster.
|
|
|
|
The same results can be reproduced under Firefox on macOS with the "Bavarian outline" dataset. Interestingly, under Safari the algorithms perform similarly with a small preference to the original version. This applies to either case tested.
|
|
|
|
With so much variance it is hard to determine the best performing browser regarding the JavaScript implementation. Under the right circumstances Chrome can produce the fastest results with the alternative implementation. Safari is consistently very fast. Even while it falls short to Firefox's results with the original algorithm when high quality is turned on. The greatest discrepancy was produced by Firefox with high quality enabled. There the alternate version produced the slowest results while the results with Simplify.js can compete with Chrome's results with the Simplify.js alternative. Edge lies between these two browsers with not too bad, but also not the fastest results.
|
|
|
|
\subsection{Browser differences for Simplify.wasm}
|
|
|
|
So diverse the results from last chapter are, so monotonous they will be here. The performance of the Simplify.wasm function is consistent across all browsers tested. This is a major benefit brought by WebAssembly often described as predictable performance.
|
|
|
|
The variance is very low when the preprocessing is turned off through the high quality mode. The browsers produce about the same runtimes under the same conditions. When high quality is off the Chrome browser got its nose ahead with a mean run time of 0.66ms. Edge follows with 1.02ms and Firefox takes an average 1.10ms. The results of chapter \ref{ch:case3} show that Safari is a bit faster at the high quality mode than Firefox but slower without.
|
|
|
|
\subsection{Insights into Simplify.wasm}
|
|
\label{ch:discussion-wasm-insights}
|
|
|
|
So far, when the performance of Simplify.wasm was addressed, it meant the time spent for the whole process of preparing memory to running the algorithm as WebAssembly bytecode to loading back the result to JavaScript. This makes sense when comparing it to the JavaScript library with the motive to replace it one by one. It does however not produce meaningful comparisons of WebAssembly performance in contrast to the native JavaScript runtime. Further insights to Simplify.wasm call will be provided here.
|
|
|
|
First the parts where JavaScript is run will be examined. Chapter \ref{ch:case2} shows that there is as good as no variance in the memory initialization. This is obvious due to the fact that this step is not dependent on any other parameter than the polyline length. Initial versions of the library produced in this thesis were not as efficient in flattening the coordinate array as the final version. By replacing the built-in \texttt{Array.prototype.flat}-method with a simple \texttt{for} loop, a good optimization was achieved on the JavaScript side of the Simplify.wasm process. The \texttt{flat} method is a rather new feature of ECMAScript and its performance might be enhanced in future browser versions. This example shows that when writing JavaScript code one can quickly deviate from the "fast path" even when dealing with simple problems.
|
|
|
|
On the other side of the process exists the function \texttt{loadResult}. It is dependent on the size of the resulting polyline. Since the result is often very small, the green bar can rarely be seen in figures \ref{fig:win_edge_simplify_stack_false} and \ref{fig:win_edge_simplify_stack_true}. Merely at low tolerance values the influence is visible. The maximum fraction is at tolerance value 0.05 where the operation takes 4.26\% of the total execution time.
|
|
|
|
Now when comparing the two graphs one can clearly see that the influence of the JavaScript portions is much greater when the high quality mode is turned off. The time taken for preparing the memory in both cases is about 0.67ms. The execution time of the algorithms is so low in the first case, that it comes down to making up only 24.47\% when taking the median values. In case where high quality is enabled the results do not look as drastic. The median value of execution time is 4.31ms and with that much greater than preparation time. Whether JavaScript is at an advantage in the first case, and the high execution time justifies the switch of runtimes in the latter, will be examined in the next chapter.
|
|
|
|
|
|
\subsection{Comparison Simplify.wasm vs Simplify.js}
|
|
|
|
The results from chapter \ref{ch:case1} and \ref{ch:case3} have shown that Simplify.wasm is always faster when the high quality mode is enabled. The Firefox and Chrome browser are able to optimize at least one version of the JavaScript routines in a way that they come close to the performance of the WebAssembly based library. In Edge, the original version is three times, the alternative two times slower. In Safari, both take about twice the time than Simplify.wasm. On the other hand with preprocessing enabled, there is often one JavaScript version that surpasses the Simplify.wasm performance. In Edge and Safari its even both JavaScript versions that perform faster at higher tolerance values.
|
|
|
|
This shows that it is not always ideal to replace a library with a WebAssembly based approach. The cost of the overhead might exceed the performance gain when the execution time is low. In section \ref{ch:discussion-wasm-insights} it is pointed out, that the pure execution time of the simplification algorithm is fastest with WebAssembly. When preparing the geodata beforehand, for example by serializing it in a binary representation, one could immediately call the bytecode. This poses further effort regarding memory management to the web developer. One has to weigh up the complexity overhead to the performance benefit when considering such approaches.
|
|
|
|
\subsection{Analysis of Turf.js implementation}
|
|
|
|
In this section the method used by Turf.js is evaluated. As seen when using the Chrome or Edge browser, the original library is the slower JavaScript method for simplification. There the data transformation is definitely unfavorable. In Safari, where the JavaScript versions perform equally, the overhead will still lead to worse run times. Lastly the Firefox browser will be examined. The results from chapter \ref{ch:case4} show that there are indeed cases where the method from Turf.js achieves better performance than running the alternative Simplify.js library directly. These are the ones where the execution time is large enough to justify the overhead. Namely when high quality is enabled or low tolerance values when high quality is disabled.
|
|
|
|
Those conclusions are similar to the Simplify.wasm results, where overhead costs also played a role. Here however, one has to bear in mind that it is only one browser that is affected by a performance gain in certain circumstances. In the three other browsers the method is definitely disadvantageous.
|
|
|
|
\subsection{Mobile device analysis}
|
|
|
|
Here the results from the iPad benchmarks in chapter \ref{ch:case5} are discussed. It stands out that the two browsers show identical results. This is due to Apple Inc.'s security guidelines concerning the iOS App Store. It is specifically restricted for web browsers to use any other engine than WebKit \parencite[see][section 2.5.6]{apple2019appstore}. Because of this, Firefox for iOS cannot use the Gecko engine developed by Mozilla. This explains why the two browsers perform equally.
|
|
|
|
The results from the two browsers lead to similar conclusions as the benchmarks of Safari under macOS did. Simplify.wasm is always fastest in high quality mode. With preprocessing, the JavaScript implementations outperform the WebAssembly based approach at higher tolerance ranges. Both JavaScript versions perform equally.
|
|
|
|
The mobile device has lower hardware capabilities than the MacBook Pro tested in \ref{ch:case3}. That is why it is not surprising, that the runtimes are higher on the iPad. The concrete results of chapter \ref{ch:case3} and \ref{ch:case5} are not directly comparable as different data sets were used. So the focus lies on the behavior of the algorithms. In the case of high quality enabled, the algorithms take about twice as long on the mobile device. This affects both, JavaScript and WebAssembly, equally. With high quality disabled, one can see that the JavaScript performance of the desktop device gets gradually better where at the mobile device the performance stagnates. |