Skip to main content

Iterating faster

Today I spent a few minutes writing a Makefile. I think they may turn out to be some of the most effective minutes I've ever spent.

At work we build using SCons. When I change a single source file, rebuilding takes between 50 and 65 seconds. That's more than long enough to lose my focus and get distracted, so to stay productive I have to work in larger batches. I code in longer bursts and recompile less often, so that the ratio of productive time to down time stays up; unfortunately that also means fewer opportunities to test my changes.

That's not ideal, but it's manageable much of the time. Not at the moment though. Right now I'm developing an algorithm where I need to be able to quickly try out different approaches for each of the steps. In this situation I really want to see the effect of each change in isolation, so I can't batch up my coding. All of a sudden those 50 second build times are a real problem.

I spent a little while looking into ways to speed SCons up and managed to get the rebuild time down to around 45 seconds. Still nowhere near fast enough.

Fortunately I'm working in a very self-contained part of our code-base, so I could limit the amount of files I had to consider. Re-running SCons with verbose output turned on gave me the exact command being used to compile each of them, so I saved those to a shell script. Rebuild time with this shell script was down to about 15 seconds: a huge improvement, but still not quite quick enough. This script was recompiling every file each time I ran it though - surely I could do something a bit smarter than that...?

Enter make. It knows how to check for changed files & I've used it enough that I can put together a Makefile pretty quickly. Factoring out the commands from the shell script into rules and adding enough dependency information to be useful, I had it set up in a couple of minutes. The result: most rebuilds are now around 3 seconds. Perfect.

So now, not only am I saving almost a whole minute of wasted time with each rebuild, I've all but removed the chance to get distracted and also opened the door on a more effective way of working. I can make small changes and rebuild to see the effect of each in isolation. I can develop my algorithm a lot faster now, because I can iterate faster; and that's why I think they'll be some of the most effective minutes I've ever spent.

Comments

Post a Comment

Popular posts from this blog

OpenGL ES and occlusion queries

This is a follow-up to my earlier post "WebGL doesn't have query objects" . Since I wrote that post, the situation has changed a bit. It's still true to say that WebGL doesn't have query objects, but the underlying reason - that OpenGL ES doesn't - is no longer true. For OpenGL ES 2.0 , there's an extension which provides basic query functionality: EXT_occlusion_query_boolean  (which seems to have been based on ARB_occlusion_query2 from regular OpenGL). For OpenGL ES 3.0 , the functionality from that extension appears to have been adopted into the standard. The extension provides two query types, both of which set a boolean value to indicate whether any pixels passed the depth and stencil tests. While this is progress, unfortunately it's still not sufficient to implement the pixel accurate collision detection method I described in an earlier post. For that purpose it's not enough to know whether any  pixels passed the tests; you want to kno...
Post a Comment
Read more

Assert no lock required

This is a technique I learnt about from Jason Gregory's excellent book, Game Engine Architecture (3rd Edition) . If you have a shared resource accessed by multiple threads, where you're fairly certain that it's only ever accessed by one thread at a time, you can use an assert() to check for this at debug time without having to pay the runtime cost of locking a mutex. The implementation is fairly straightforward: class UnnecessaryMutex { public: void lock() { assert(!_locked); _locked = true; } void unlock() { assert(_locked); _locked = false; } private: volatile bool _locked = false; }; #ifdef ENABLE_LOCK_ASSERTS #define BEGIN_ASSERT_LOCK_NOT_REQUIRED(mutex) (mutex).lock() #define END_ASSERT_LOCK_NOT_REQUIRED(mutex) (mutex).unlock() #else #define BEGIN_ASSERT_LOCK_NOT_REQUIRED(mutex) #define END_ASSERT_LOCK_NOT_REQUIRED(mutex) #endif Usage is equally straightforward: UnnecessaryMutex gMutex; void PossiblyOverlappingFunction...
Post a Comment
Read more

Triangle bounding boxes in a single byte

Just thought of a way to store the bounding box for a single triangle in only one byte. It's not really practical or something you'd ever really want to use, but what the hell. Assume we have some kind of indexed mesh structure with a list of vertex positions and a list of triangle indices:   struct Mesh {     std::vector<vec3> verts;     std::vector<uvec3> triangles;   }; We can find the bounding box of a triangle by taking the min and max of all three vertices:   vec3 Mesh::lowerBound(uint32_t tri) const {     vec3 v0 = verts[triangles[tri].x];     vec3 v1 = verts[triangles[tri].y];     vec3 v2 = verts[triangles[tri].z];     return min(min(v0, v1), v2);   }   vec3 Mesh::upperBound(uint32_t tri) const {     vec3 v0 = verts[triangles[tri].x];     vec3 v1 = verts[triangles[tri].y];     vec3 v2 = verts[triangles[tri].z];     return ...
2 comments
Read more