On 8/10/2015 2:15 PM, Laurent Bourgès wrote:
Jim,
Here is the new webrev including your proposals to use 32.31 fixed-point
maths and double-precision in addLine() as it is definitely better:
http://cr.openjdk.java.net/~lbourges/marlin/marlin-s3.3/
Why are ceil_int and floor_int implemented differently?
For CHECK_NAN, how fast is Math.isNaN? Is it worth first comparing the
intpart to the value that NaN converts to before calling isNaN? The
majority case, by a wide margin, is that the value is not NaN.
Helpers.java, line 214 - why move the declaration into the for loop?
That is confusing to me since "j" is not used in the for() conditions.
MarlinConst - It looks like the subpixel sizes are now properties. Did
that have performance consequences? The values are no longer true
literal constants when compiling the code now.
That's as far as I got this pass...
...jim
