cosp
(2 files, 848 lines)
rrtm
(2 files, 23 lines)
[root]/branches/Optimisation_LMDZ/libf/phylmd
cosp
(2 files, 848 lines)
rrtm
(2 files, 23 lines)
| Author | Changes | Lines of Code | Lines per Change |
|---|---|---|---|
| adurocher | 70 (100.0%) | 8709 (100.0%) | 124.4 |
Fixed missing xios function in interface
10 lines of code changed in 2 files:
Offload post-processing to XIOS
use expressions in .xml files to offload computations previously done in phys_output_write to XIOS
535 lines of code changed in 1 file:
Use `cvl_comp_threshold` to control whether to use compression or not
cvl_comp_threshold now defaults to 0 (never compress)
17 lines of code changed in 3 files:
Removed some timers in cv3a_driver
3 lines of code changed in 1 file:
Use `cvl_comp_threshold` to control whether to use compression or not
cvl_comp_threshold now defaults to 0 (never compress)
0 lines of code changed in 1 file:
Remove useless zero-initializations
57 lines of code changed in 2 files:
Added a path to execute cv3a_driver without compression at all
+ fix : Compute coef_convective BEFORE cv3a_driver_compressed
280 lines of code changed in 1 file:
Moved compressed part of cv3a_driver to separate subroutine
264 lines of code changed in 1 file:
Use blocks to allocate compressed arrays to the right size
417 lines of code changed in 2 files:
Fixed array size in `cv3_tracer` to match compressed arrays
7 lines of code changed in 1 file:
Refactoring and cleaning cv3a_driver
505 lines of code changed in 3 files:
Create separate subroutines for cv3a_driver and cv4a_driver
1566 lines of code changed in 2 files:
Fixed compilation error without RRTM or without XIOS
11 lines of code changed in 2 files:
Fix conditional compilation to compile without XIOS
107 lines of code changed in 7 files:
Refactoring IOs : added new phys_output_write_xios with XIOS only
This is a combination of 9 commits.
Added phys_output_write_xios
Moved C preprocessing in phys_output_write_xios_mod to a separate file
Added iophy_xios
Removed non-xios code from phys_output_write_xios_mod
Add separate subroutine for debug prints
Cleaning and refactoring iophy_xios
Only one test for vars_defined
Check once for inactive xios fields
Fix bug
2109 lines of code changed in 6 files:
Fix conditional compilation to compile without XIOS
0 lines of code changed in 1 file:
Check if field is active before gather (static check)
Check is still performed at each timestep, but the static check is perfromed (at_current_timestep_arg=.false.). This should have less overhead because the static check is faster.
2 lines of code changed in 1 file:
Refactoring IOs : added new phys_output_write_xios with XIOS only
This is a combination of 9 commits.
Added phys_output_write_xios
Moved C preprocessing in phys_output_write_xios_mod to a separate file
Added iophy_xios
Removed non-xios code from phys_output_write_xios_mod
Add separate subroutine for debug prints
Cleaning and refactoring iophy_xios
Only one test for vars_defined
Check once for inactive xios fields
Fix bug
0 lines of code changed in 2 files:
Check if field is active before omp_gather (dynamic check)
At each timestep and for every field xios_field_is_active( <field>, at_current_timestep_arg=.true.) is called
0 lines of code changed in 1 file:
Check if field is active before omp_gather (dynamic check)
At each timestep and for every field xios_field_is_active( <field>, at_current_timestep_arg=.true.) is called
0 lines of code changed in 1 file:
Fixed missing contiguous in cv3a_compress.f90
3 lines of code changed in 1 file:
cv3p_mixing_new : VLAs removed from block construct (intel 19 bug)
Intel compiler 19 crashes (internal compiler error) when VLA (variable length arrays) are used within `block` constructs
1 lines of code changed in 1 file:
cv3p_mixing_new : modify loops to vectorize math calls
Some local variables have been moved to `block` constructs to be sure there is no dependencies between loops/iterations.
This might (will) create issues with older intel compilers
141 lines of code changed in 1 file:
cv3p_mixing_new : change array temporaries into scalars
with reordered loops, some array temporaries are no longer necessary.
cv3p_mixing_new generates different results than
cv3p_mixing_old with `-fp-model fast` but not with `-fp-model precise`
177 lines of code changed in 1 file:
cv3p_mixing_new : Removed lwork
104 lines of code changed in 1 file:
cv3p_mixing_new : swap loops in first part of function
79 lines of code changed in 1 file:
cv3p_mixing_new : swap loops ij and i
195 lines of code changed in 1 file:
cv3p_mixing_new : merge some loops
155 lines of code changed in 1 file:
cv3p_mixing_new : swapped loops ij and j
183 lines of code changed in 1 file:
Add cv3p_mixing_new for optimization
1257 lines of code changed in 3 files:
Optimization in cv3_tracer
0 lines of code changed in 1 file:
New generic implementation for compression
524 lines of code changed in 5 files:
Optimization in cv3_tracer
0 lines of code changed in 1 file:
New generic implementation for compression
0 lines of code changed in 2 files:
Added timers for physiq and display physic profiling
0 lines of code changed in 1 file:
Added timers for physiq and display physic profiling
0 lines of code changed in 1 file:
Fixed compilation errors
- Fixed some OpenMP syntax errors for scorep
- Fix issues with ifort -check all -warn all -O0
0 lines of code changed in 6 files: