Dynamic -ffast-math

Question

Is it possible to selectively enable / disable runtime at runtime? For example, by creating FastMath and AccurateMath classes with a common Math base class so that you can use both implementations at run time? Also for blinking subnormal numbers to zero, etc.

In particular, I don’t know if compiling with -ffast-math will emit an instruction that, after executing it, will affect all numerical calculations in the stream (for example, setting a flag to align trims to zero).

Answer 1

Try the following:

gcc -ffast-math -c first.c gcc -c second.c gcc -o dyn_fast_math first.o second.o 

Entering unique functions in first.c and second.c. That should do the trick. There is rarely any “global” impact of compiler optimization. If this does exist, the binding will most likely fail due to conflict.

I tried a small sample without any problems.

Here is an example.

first.c

 extern double second(); double first () { double dbl; dbl = 1.0; dbl /= 10.0; return dbl; } int main () { printf("first = %f\n", first()); printf("second = %f\n", second()); return 0; } 

second.c

 double second () { double ddbl; ddbl = 1.0; ddbl /= 10.0; return ddbl; } 

compilation

 gcc -S first.c gcc -c first.s gcc -ffast-math -S second.c gcc -ffast-math -c second.s gcc -o prog first.o second.o 

Check the difference between first.s and second.s and you will find the following:

 movapd %xmm1, %xmm2 divsd %xmm0, %xmm2 movapd %xmm2, %xmm0 

changes to this:

 mulsd %xmm1, %xmm0 

Both functions are called, and both return the expected result.

Answer 2

If you do not want to guess with the assembly, you can do the following:

 #pragma fast-math push #pragma fast-math on [..] #pragma fast-math pop 

GCC may have slightly different syntax, but I expect this to be possible too.