Use optional instead of infinity

[levinli303]

No description provided.

[375gnu]

clamp to [0, 1]?

[levinli303]

this is to clamp between start time and end time of a phase.

[375gnu]

what's wrong with infinity?

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
1 Code Smell

0.0% Coverage
1.6% Duplication

The version of Java (11.0.17) you have used to run this analysis is deprecated and we will stop accepting it soon. Please update to at least Java 17.
Read more here

[levinli303]

what's wrong with infinity?

i was hoping to enable fastmath sometime. frustum.cpp checks isinf which can easily be avoided. i have isinf checks for startTime/endTime of phases in other frontends id rather not replace with checking against 1e9

[ajtribick]

I have mixed feelings about -ffast-math - some bits of it make a lot more sense than others and I would prefer to rely on the individual sub-flags. Of the ones that aren't switched on by default in C++ on x86_64 (-fno-signaling-nans, -fno-rounding-math, -fexcess-precision=fast), I think -fno-math-errno and -fno-trapping-math are ones we should probably be using: errno for math functions makes sense from the point of view of the C language, which targets machines where floating point representations do not have infinities/NaNs, but is fairly redundant in the case of IEEE floats. -freciprocal-math and -fassociative-math also seem pretty reasonable, in particular the latter seems to be useful for vectorization.

The -ffinite-math-only flag seems much less useful to the point of being potentially dangerous - it makes detecting these cases in user-provided binary-format input files harder (if you've told the system that only finite numbers are present, the compiler is allowed to optimize away checks for these values), and combined with disabling errno makes it really hard to detect the results of calculations doing overflows. -fno-signed-zeros is perhaps less clear-cut, although enabling it prevents roundtripping of certain CMOD files (including the iss.cmod used in the integration tests) from binary to text to binary again.