Original Post
hi all,
sure this aint a new thing, but the forum search tool is f**ked atm, and won''t help me out 
I''m well aware of both the practical and conceptual issues of limited precision floating point comparison/operators... so this isn''t directly a question regarding them.
From a practical point of view, will the innacuracies caused by such operations be consistant .
e.g.
float a = 1000.456f;
float b = 1000.0f;
float c = a - b;
now, if ''c == 0.45590001f'' (for example), as an "error", would that error be consistant across:
a. The chipset (e.g. that model of pentium/athlon/etc..)
b. The compiler (is such op''s affected by compiler?)
c. The OS (unlikely, but suppose its an option).
From my tests, any errors due to float/double precision have remained constant between runs (not necessarily constant between compilers??) - but I only have 3 compilers and 1 set of hardware 
can''t really call that conclusive research!
any thoughts?
cheers,
Jack
I''m well aware of both the practical and conceptual issues of limited precision floating point comparison/operators... so this isn''t directly a question regarding them.
From a practical point of view, will the innacuracies caused by such operations be consistant .
e.g.
float a = 1000.456f;
float b = 1000.0f;
float c = a - b;
now, if ''c == 0.45590001f'' (for example), as an "error", would that error be consistant across:
a. The chipset (e.g. that model of pentium/athlon/etc..)
b. The compiler (is such op''s affected by compiler?)
c. The OS (unlikely, but suppose its an option).
From my tests, any errors due to float/double precision have remained constant between runs (not necessarily constant between compilers??) - but I only have 3 compilers and 1 set of hardware can''t really call that conclusive research!
any thoughts?
cheers,
Jack
