Posts Tagged performance

Fixed-Point: Shrink Data Sizes 4x

Tags: , , ,

Floating-point math is fast these days, but fixed-point still has a purpose: we can use it to store real numbers in less than 32 bits. Saving a measly 16 or 24 bits off a float might not sound appealing, but cutting the data size in half or quarter often does when multiplied across large amounts of real numbers. We can shrink downloads, improve load times, save memory, and fit more into the CPU’s data caches. So today we’ll look at storing numbers in fixed-point formats and see how easy it can be to shrink our data!

Read the rest of this article »

1 Comment

Adding Bit Fields to C#

Tags: ,

Bit fields give us the ability to use less than one byte to store an integer. This can be really useful to fit more data into a CPU cache or to reduce memory usage in general. Unfortunately, bit fields aren’t built into C#. Today we’ll work around that by creating our own!

Read the rest of this article »

3 Comments

Burst-Compiled Generic Algorithms

Tags: , , ,

Many algorithms get used over and over: searching, sorting, filtering, etc. C# makes these available with LINQ and functions like Array.Sort, but these can’t be compiled by Burst because interfaces and delegates aren’t supported. So how do we implement these generic algorithms to avoid re-writing them over and over? Today we’ll explore one technique to solve this problem. Read on to learn how!

Read the rest of this article »

3 Comments

Steal Some More Bits

Tags: , , ,

Today we continue stealing float bits, but in an entirely different way this time. We’ll end up with the ability to switch between float and 21-bit integer modes and to know which mode we’re in. We can do all of this without using any more than four bytes just by exploiting a little knowledge of the float data format. Read on to learn how!

Read the rest of this article »

No Comments

Steal Some Bits

Tags: , ,

With a bit of understanding and some C# trickery, we can exploit how float works to cram in a few more bits and make some big performance gains. Today we’ll see how to steal some of the bits from a float!

Read the rest of this article »

No Comments

Math.Abs vs. Mathf.Abs vs. math.abs vs. Custom

Tags: , ,

A reader recently asked what the fastest way to take an absolute value was. It occurred to me that there are a lot of ways to do this in Unity! So today we’ll try them all out and see which is best.

Read the rest of this article »

7 Comments

NativeHashSet<T>

Tags: , , ,

Unity’s Native Collections package, currently in Preview, provides a hash map but not a hash set. Today we’ll supplement NativeHashMap<TKey, TValue> with our own NativeHashSet<T> as part of the NativeCollections repo. Read on for performance results and to see how to use it!

Read the rest of this article »

No Comments

Assertions in Burst

Tags: , , ,

Assertions are an incredibly handy tool, but do they work in Burst-compiled jobs? Today we’ll find out!

Read the rest of this article »

2 Comments

How IL2CPP Calls Burst

Tags: , ,

Ever wonder how code compiled with IL2CPP can call code compiled by Burst? Today we’ll dive into the details and find out!

Read the rest of this article »

3 Comments

Burst’s FloatPrecision and FloatMode: Don’t Assume

Tags: , ,

Unity 2019.1’s new Burst job compiler has two options to increase performance even further: FloatPrecision and FloatMode. By sacrificing some exactness in our calculations, we should be able to increase speed. Today’s article is about using those options and examining the results to verify the results.

Read the rest of this article »

6 Comments