It's ~3x slower than native on my test (native vs asm.js vs PNaCl): <a href="https://github.com/bkaradzic/bgfx#30hz-test-for-browsers" rel="nofollow">https://github.com/bkaradzic/bgfx#30hz-test-for-browsers</a>
If the entire codebase gets through to this process, it seems like it would be no effort to port to another architecture (arm, mips, etc.)<p>Not saying it is an optimal way to cross compile.