> On May 3, 2016, at 10:56, Boris Zbarsky <[email protected]> wrote: > > On 5/3/16 1:28 PM, Jakob Stoklund Olesen wrote: >> It is not assertions that make unoptimized C++ really slow, it is the lack >> of inlining which is critical to clean up template abstraction goop. > > It depends on the assertions. In the case of the JS engine, the particular > assertions under discussion change the algorithmic complexity of many > algorithms, which makes them much slower in practice. > > In particular, the "release+asserts" configuration (--enable-debug > --enable-optimize) is much slower than the same thing without --enable-debug > in the case of SpiderMonkey.
That makes sense. Would it be worthwhile to put the worst offenders under their own flag so they could be disabled in an optimized build? LLVM had an EXPENSIVE_CHECKS macro for that kind of assertion, but I don’t think they use it any more. People would rarely enable it, so the expensive assertions had a tendency to bit rot. I think if they had been enabled by default, they might have stayed in. /jakob _______________________________________________ dev-tech-js-engine-internals mailing list [email protected] https://lists.mozilla.org/listinfo/dev-tech-js-engine-internals
