single-precision instead of double precision
Are there any plans to natively support single-precision datasets? (sometimes we store 32-bit instead of 64-bit to save space). If I access my data and then create a covering grid, in the YTCoveringGridBase code, when it makes the grid for the data, it hardcodes in "float64" -- ideally it would keep the precision of the dataset all the way through.
Mike
Hi Mike,
On Wed, Apr 23, 2014 at 12:52 PM, Michael Zingale michael.zingale@stonybrook.edu wrote:
Are there any plans to natively support single-precision datasets? (sometimes we store 32-bit instead of 64-bit to save space). If I access my data and then create a covering grid, in the YTCoveringGridBase code, when it makes the grid for the data, it hardcodes in "float64" -- ideally it would keep the precision of the dataset all the way through.
Yes, there are, with the exception of accumulators like sum, etc. We also won't guarantee that unit conversions will not over/underflow, but that should be okay for your situation. Today I started on the process of doing particles in native bits, and fluids will come a while after that. Mostly it's a problem of getting the Cython routines to use "fused types."
-Matt
Mike
-- Michael Zingale Associate Professor
Dept. of Physics & Astronomy • Stony Brook University • Stony Brook, NY 11794-3800 phone: 631-632-8225 e-mail: Michael.Zingale@stonybrook.edu web: http://www.astro.sunysb.edu/mzingale
yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
ok, great. I'm happy to help test the fluid stuff.
On Wed, Apr 23, 2014 at 1:10 PM, Matthew Turk matthewturk@gmail.com wrote:
Hi Mike,
On Wed, Apr 23, 2014 at 12:52 PM, Michael Zingale michael.zingale@stonybrook.edu wrote:
Are there any plans to natively support single-precision datasets? (sometimes we store 32-bit instead of 64-bit to save space). If I
access my
data and then create a covering grid, in the YTCoveringGridBase code,
when
it makes the grid for the data, it hardcodes in "float64" -- ideally it would keep the precision of the dataset all the way through.
Yes, there are, with the exception of accumulators like sum, etc. We also won't guarantee that unit conversions will not over/underflow, but that should be okay for your situation. Today I started on the process of doing particles in native bits, and fluids will come a while after that. Mostly it's a problem of getting the Cython routines to use "fused types."
-Matt
Mike
-- Michael Zingale Associate Professor
Dept. of Physics & Astronomy * Stony Brook University * Stony Brook, NY 11794-3800 phone: 631-632-8225 e-mail: Michael.Zingale@stonybrook.edu web: http://www.astro.sunysb.edu/mzingale
yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
participants (2)
-
Matthew Turk -
Michael Zingale