[Bug target/81906] [7/8 Regression] Calls to rint() wrongly optimized away starting in g++ 6

[joseph at codesourcery dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81906

--- Comment #10 from joseph at codesourcery dot com <joseph at codesourcery dot 
com> ---
A correct (for -frounding-math) SSE sequence would (for arguments with 
absolute value < 2**52) add and subtract 2**52 for a positive operand, 
-2**52 for a negative operand.  Then it would need to copy the sign of the 
original operand to the result to ensure zero results get the correct sign 
in all cases.