mixed precision: time_manager reals to r8 and clean up #1196

rem1776 · 2023-04-19T21:02:46Z

Description
replaces default reals and double precision values with explicit r8_kinds.

i created wrappers for get_cal_time and real_to_time_type so that they convert passed in r4's to r8 before calling the full routines. This makes all the real calculations done in r8_kind instead of default real precision.

I also cleaned up a lot of the old html comments since some were left in this module and made a small update to the test.

How Has This Been Tested?
oneapi 23.0, amd

Checklist:

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
My changes generate no new warnings
Any dependent changes have been merged and published in downstream modules
New check tests, if applicable, are included
make distcheck passes

mlee03 · 2023-04-20T11:58:18Z

time_manager/time_manager.F90


 if(.not.module_is_initialized) call time_manager_init

 ! Convert time intervals to floating point days; risky for general performance?
-d1 = time1%days * dble(seconds_per_day) + dble(time1%seconds) + time1%ticks/dble(ticks_per_second)
-d2 = time2%days * dble(seconds_per_day) + dble(time2%seconds) + time2%ticks/dble(ticks_per_second)
+d1 = time1%days * real(seconds_per_day, r8_kind) + real(time1%seconds, r8_kind) &


real(time1%days, r8_kind)

@rem1776 this change :)

mlee03 · 2023-04-20T11:58:30Z

time_manager/time_manager.F90

-d2 = time2%days * dble(seconds_per_day) + dble(time2%seconds) + time2%ticks/dble(ticks_per_second)
+d1 = time1%days * real(seconds_per_day, r8_kind) + real(time1%seconds, r8_kind) &
+     + time1%ticks/real(ticks_per_second, r8_kind)
+d2 = time2%days * real(seconds_per_day, r8_kind) + real(time2%seconds, r8_kind) &


real(time2%days, r8_kind)

@rem1776 this one too

mlee03 · 2023-04-20T12:01:32Z

time_manager/time_manager.F90

-end function real_to_time_type
+end function real8_to_time_type
+
+function real4_to_time_type(x, err_msg) result(t)


would this function be needed if all reals are r8_kind in time_manager?

Yeah it's for external routines, if a module is compiled with r4 it'll call this to get a new time type. It gets converted/rounded anyway by safe_rtoi

mlee03 · 2023-04-20T12:03:15Z

time_manager/time_manager.F90

-div = d/dble(n)
+dseconds_per_day  = real(seconds_per_day, r8_kind)
+dticks_per_second = real(ticks_per_second, r8_kind)
+d = time%days*dseconds_per_day*dticks_per_second + real(time%seconds, r8_kind)*dticks_per_second + &


real(time%days,r8_kind)

@rem1776 this one too

mlee03 · 2023-04-20T12:04:29Z

time_manager/time_manager.F90

+dticks_per_second = real(ticks_per_second, r8_kind)
+d = time%days*dseconds_per_day*dticks_per_second + real(time%seconds, r8_kind)*dticks_per_second + &
+    real(time%ticks, r8_kind)
+div = d/real(n, r8_kind)

 days = int(div/(dseconds_per_day*dticks_per_second))
 seconds = int(div/dticks_per_second - days*dseconds_per_day)


real(days,r8_kind)

mlee03 · 2023-04-20T12:10:09Z

time_manager/time_manager.F90

@@ -2528,7 +2049,7 @@ function increment_date_private(Time, years, months, days, hours, minutes, secon
   cmonth = cmonth + months

 ! Adjust year and month number when cmonth falls outside the range 1 to 12
-   cyear = cyear + floor((cmonth-1)/12.)
+   cyear = cyear + floor((cmonth-1)/12.0_r8_kind)


real(cmonth-1,r8_kind)

mlee03 · 2023-04-20T12:16:48Z

time_manager/get_cal_time.F90

@@ -313,34 +322,50 @@ function get_cal_time(time_increment, units, calendar, permit_calendar_conversio
  increment_days = floor(dt/86400)
  increment_seconds = int(dt - increment_days*86400)
 else


lines 292 -323, time_increment and dt are reals doing math with ints

mlee03 · 2023-04-20T12:18:05Z

time_manager/get_cal_time.F90

 !------------------------------------------------------------------------
+
+!> For mixed precision support, just casts to passed in increment to r8
+function get_calendar_time_wrap(time_increment, units, calendar, permit_calendar_conversion)


wondering if this function is needed

It was needed to compile with current mixed mode so I think it'll be needed

do you think using class(*) for time_increment would work? Or is that going to make things messier..

mlee03 · 2023-04-20T14:25:36Z

@rem1776 could you also add mixed precision to the pr title? :)

mlee03 · 2023-05-31T15:25:10Z

@rem1776 ignore recent comments, did not see that they were outdated!

mlee03 · 2023-06-05T12:22:31Z

@J-Lentz @uramirez @mcallic2 don't forget to review this one

J-Lentz · 2023-06-06T19:15:59Z

time_manager/time_manager.F90

-  if (rval .le. big .and. -1.*rval .ge. -1.*big) then
+  real(r8_kind) :: big
+  big = real(huge(ival), r8_kind)
+  if (rval .le. big .and. -1.0_r8_kind*rval .ge. -1.0_r8_kind*big) then


if (rval .le. big .and. -1.0_r8_kind*rval .ge. -1.0_r8_kind*big) then

I'm confused by this if condition. Under what circumstance would the comparisons on the left and right of the .and. ever differ from one another?

BREAKING CHANGE: In coupler_types.F90, `coupler_nd_field_type` and `coupler_nd_values_type` have been renamed to indicate real kind value: `coupler_nd_real4/8_field_type` and `coupler_nd_real4/8_values_type`. The `bc` field within `coupler_nd_bc_type` was modified to use r8_kind within the value and field types, and an additional field added `bc_r4` to use r4_kind values. Includes: * feat: eliminate use of real numbers for mixed precision in `block_control` (#1195) * feat: mixed precision field_manager (#1205) * feat: mixed precision random_numbers_mod (#1191) * feat: mixed precision time_manager reals to r8 and clean up (#1196) * feat: mixed Precision tracer_manager (#1212) * Mixed precision monin_obukhov (#1116) * Mixed precision: `monin_obukhov` unit tests (#1272) * mixed-precision diag_integral_mod (#1217) * mixed precision time_interp (#1252) * mixed precision interpolator_mod (#1305) * Mixed precision astronomy (#1092) * Mixed precision `data_override_mod` (#1323) * mixed precision exchange (#1341) * coupler mixed precision (#1353) * Mixed precision topography_mod (#1250) --------- Co-authored-by: rem1776 <Ryan.Mulhall@noaa.gov> Co-authored-by: MiKyung Lee <58964324+mlee03@users.noreply.github.com> Co-authored-by: mlee03 <Mikyung.Lee@lscamd50-d.gfdl.noaa.gov> Co-authored-by: Caitlyn McAllister <65364559+mcallic2@users.noreply.github.com> Co-authored-by: Jesse Lentz <42011922+J-Lentz@users.noreply.github.com>

mlee03 and others added 13 commits March 8, 2023 11:15

Update mixedmode_base branch to main (NOAA-GFDL#1144)

46d8e44

update the mixedmode_base branch to main (NOAA-GFDL#1154)

dde0b88

Merge branch 'mixedmode' into mixedmode_base

bc9e4d9

rm accidental empty line

2977578

chore: bring mixedmode_base up to date with main (NOAA-GFDL#1173)

b3374f6

clean up comments and any default/double sized reals

b6b12dd

forgot some changes

dcd1969

fix incorrect variable name, whitespace

91590e1

fix line lengths, add missed use

2900f85

add interface for r4

6c2bdad

linelength

3dd1af3

add interface and test for conversions

51f8b6b

whitespace again

0b01311

rem1776 requested review from bensonr and thomas-robinson as code owners April 19, 2023 21:02

rem1776 requested review from J-Lentz, mlee03 and mcallic2 April 19, 2023 21:03

mlee03 reviewed Apr 20, 2023

View reviewed changes

rem1776 changed the title ~~time_manager reals to r8 and clean up~~ mixed precision: time_manager reals to r8 and clean up Apr 20, 2023

rem1776 added 2 commits May 3, 2023 11:37

review fixes

1597338

finish up review changes and any lingering implicit conversions

ef8cb15

mlee03 approved these changes May 31, 2023

View reviewed changes

mcallic2 approved these changes Jun 6, 2023

View reviewed changes

J-Lentz approved these changes Jun 6, 2023

View reviewed changes

rem1776 merged commit f655759 into NOAA-GFDL:mixedmode Jun 7, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

mixed precision: time_manager reals to r8 and clean up #1196

mixed precision: time_manager reals to r8 and clean up #1196

rem1776 commented Apr 19, 2023

mlee03 Apr 20, 2023

mlee03 May 31, 2023

mlee03 Apr 20, 2023

mlee03 May 31, 2023

mlee03 Apr 20, 2023

rem1776 Apr 20, 2023

mlee03 Apr 20, 2023

mlee03 May 31, 2023

mlee03 Apr 20, 2023

mlee03 May 31, 2023

mlee03 Apr 20, 2023

mlee03 Apr 20, 2023

mlee03 Apr 20, 2023

rem1776 Apr 20, 2023

mlee03 May 4, 2023

mlee03 commented Apr 20, 2023

mlee03 commented May 31, 2023

mlee03 commented Jun 5, 2023

J-Lentz Jun 6, 2023

mixed precision: time_manager reals to r8 and clean up #1196

mixed precision: time_manager reals to r8 and clean up #1196

Conversation

rem1776 commented Apr 19, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mlee03 commented Apr 20, 2023

mlee03 commented May 31, 2023

mlee03 commented Jun 5, 2023

Choose a reason for hiding this comment