Happy New Year everybody!
I've encountered some weirdness in *WebAssembly* vs. *asm.js* performance
comparison and I'm wondering if I'm doing something wrong...
1. I've compiled some simple computationally intensive code (calculation of
Pi to 10,000 digits with Rabinowitz & Wagon spigot algorithm) as x64 with
VS2015 Microsoft's compiler and Clang in Windows, then with Clang 4.0 &
libcpp in Ubuntu 14.04 (Windows 10 Ubuntu subsystem) and got very close
results: about 6.7 seconds.
2. Then I compiled this code to *asm.js* with Emscripten latest and -O3
flag and ran it in the latest released Chrome (8.7 sec) and Nightly (7
sec). Impressively close to the native speed!
3. Finally, I installed Emscripten incoming, which uses Clang 3.9, and
compiled my code:
emcc t.cpp -std=c++14 -O3 -s WASM=1 -o index.html
Running the resulting *WebAssembly* code in Nightly produced the following
console output:
trying binaryen method: native-wasml
binaryen method succeeded.
and took almost 15 seconds.
In Canary, it took almost 17 seconds.
The time was measured only for the Pi calculation:
int main(int argc, char** argv)
{
Stopwatch<> sw;
Pi::calculate(10000);
sw.stop();
std::cout << sw.elapsed()/1.0e+6 << " seconds\n";
}
So, any asm.js or wasm translation time shouldn't affect the numbers...
What is wrong here???
Thank you,
Boris
--
You received this message because you are subscribed to the Google Groups
"emscripten-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
