| 1 |
C $Header$ |
C $Header$ |
| 2 |
|
C $Name$ |
| 3 |
|
|
| 4 |
|
CBOP |
| 5 |
|
C !ROUTINE: CPP_EEOPTIONS.h |
| 6 |
|
C !INTERFACE: |
| 7 |
|
C include "CPP_EEOPTIONS.h" |
| 8 |
C |
C |
| 9 |
C /==========================================================\ |
C !DESCRIPTION: |
| 10 |
C | CPP_EEOPTIONS.h | |
C *==========================================================* |
| 11 |
C |==========================================================| |
C | CPP\_EEOPTIONS.h | |
| 12 |
|
C *==========================================================* |
| 13 |
C | C preprocessor "execution environment" supporting | |
C | C preprocessor "execution environment" supporting | |
| 14 |
C | flags. Use this file to set flags controlling the | |
C | flags. Use this file to set flags controlling the | |
| 15 |
C | execution environment in which a model runs - as opposed | |
C | execution environment in which a model runs - as opposed | |
| 28 |
C | once an experimental configuration has been | |
C | once an experimental configuration has been | |
| 29 |
C | identified, rebuild the code with the appropriate | |
C | identified, rebuild the code with the appropriate | |
| 30 |
C | options set at compile time. | |
C | options set at compile time. | |
| 31 |
C \==========================================================/ |
C *==========================================================* |
| 32 |
|
CEOP |
| 33 |
|
|
| 34 |
#ifndef _CPP_EEOPTIONS_H_ |
#ifndef _CPP_EEOPTIONS_H_ |
| 35 |
#define _CPP_EEOPTIONS_H_ |
#define _CPP_EEOPTIONS_H_ |
| 46 |
C ALWAYS - indicates the choice will be fixed at compile time |
C ALWAYS - indicates the choice will be fixed at compile time |
| 47 |
C so no run-time option will be present |
C so no run-time option will be present |
| 48 |
|
|
| 49 |
C Flag used to indicate whether Fortran formatted write |
C=== Macro related options === |
| 50 |
|
C-- Control storage of floating point operands |
| 51 |
|
C On many systems it improves performance only to use |
| 52 |
|
C 8-byte precision for time stepped variables. |
| 53 |
|
C Constant in time terms ( geometric factors etc.. ) |
| 54 |
|
C can use 4-byte precision, reducing memory utilisation and |
| 55 |
|
C boosting performance because of a smaller working set size. |
| 56 |
|
C However, on vector CRAY systems this degrades performance. |
| 57 |
|
C Enable to switch REAL4_IS_SLOW from genmake2 (with LET_RS_BE_REAL4): |
| 58 |
|
#ifdef LET_RS_BE_REAL4 |
| 59 |
|
#undef REAL4_IS_SLOW |
| 60 |
|
#else /* LET_RS_BE_REAL4 */ |
| 61 |
|
#define REAL4_IS_SLOW |
| 62 |
|
#endif /* LET_RS_BE_REAL4 */ |
| 63 |
|
|
| 64 |
|
C-- Control use of "double" precision constants. |
| 65 |
|
C Use D0 where it means REAL*8 but not where it means REAL*16 |
| 66 |
|
#define D0 d0 |
| 67 |
|
|
| 68 |
|
C-- Enable some old macro conventions for backward compatibility |
| 69 |
|
#undef USE_OLD_MACROS_R4R8toRSRL |
| 70 |
|
|
| 71 |
|
C=== IO related options === |
| 72 |
|
C-- Flag used to indicate whether Fortran formatted write |
| 73 |
C and read are threadsafe. On SGI the routines can be thread |
C and read are threadsafe. On SGI the routines can be thread |
| 74 |
C safe, on Sun it is not possible - if you are unsure then |
C safe, on Sun it is not possible - if you are unsure then |
| 75 |
C undef this option. |
C undef this option. |
| 76 |
#undef FMTFTN_IO_THREADSAFE |
#undef FMTFTN_IO_THREAD_SAFE |
| 77 |
|
|
| 78 |
C Flag used to indicate which flavour of multi-threading |
C-- Flag used to indicate whether Binary write to Local file (i.e., |
| 79 |
C compiler directives to use. Only set one of these. |
C a different file for each tile) and read are thread-safe. |
| 80 |
C USE_SOLARIS_THREADING - Takes directives for SUN Workshop |
#undef LOCBIN_IO_THREAD_SAFE |
| 81 |
C compiler. |
|
| 82 |
C USE_KAP_THREADING - Takes directives for Kuck and |
C-- Flag to turn off the writing of error message to ioUnit zero |
| 83 |
C Associates multi-threading compiler |
#undef DISABLE_WRITE_TO_UNIT_ZERO |
| 84 |
C ( used on Digital platforms ). |
|
| 85 |
C USE_IRIX_THREADING - Takes directives for SGI MIPS |
C-- Alternative formulation of BYTESWAP, faster than |
| 86 |
C Pro Fortran compiler. |
C compiler flag -byteswapio on the Altix. |
| 87 |
C USE_EXEMPLAR_THREADING - Takes directives for HP SPP series |
#undef FAST_BYTESWAP |
| 88 |
C compiler. |
|
| 89 |
C USE_C90_THREADING - Takes directives for CRAY/SGI C90 |
C-- Flag to turn on old default of opening scratch files with the |
| 90 |
C system F90 compiler. |
C STATUS='SCRATCH' option. This method, while perfectly FORTRAN-standard, |
| 91 |
#ifdef TARGET_SUN |
C caused filename conflicts on some multi-node/multi-processor platforms |
| 92 |
#define USE_SOLARIS_THREADING |
C in the past and has been replace by something (hopefully) more robust. |
| 93 |
#endif |
#undef USE_FORTRAN_SCRATCH_FILES |
| 94 |
|
|
| 95 |
#ifdef TARGET_DEC |
C-- Flag defined for eeboot_minimal.F, eeset_parms.F and open_copy_data_file.F |
| 96 |
#define USE_KAP_THREADING |
C to write STDOUT, STDERR and scratch files from process 0 only. |
| 97 |
#endif |
C WARNING: to use only when absolutely confident that the setup is working |
| 98 |
|
C since any message (error/warning/print) from any proc <> 0 will be lost. |
| 99 |
#ifdef TARGET_SGI |
#undef SINGLE_DISK_IO |
| 100 |
#define USE_IRIX_THREADING |
|
| 101 |
#endif |
C=== MPI, EXCH and GLOBAL_SUM related options === |
| 102 |
|
C-- Flag turns off MPI_SEND ready_to_receive polling in the |
| 103 |
#ifdef TARGET_HP |
C gather_* subroutines to speed up integrations. |
| 104 |
#define USE_EXEMPLAR_THREADING |
#undef DISABLE_MPI_READY_TO_RECEIVE |
|
#endif |
|
|
|
|
|
#ifdef TARGET_CRAY_VECTOR |
|
|
#define USE_C90_THREADING |
|
|
#endif |
|
|
|
|
|
C-- Define the mapping for the _BARRIER macro |
|
|
C On some systems low-level hardware support can be accessed through |
|
|
C compiler directives here. |
|
|
#define _BARRIER CALL BARRIER(myThid) |
|
|
|
|
|
C-- Define the mapping for the BEGIN_CRIT() and END_CRIT() macros. |
|
|
C On some systems we simply execute this section only using the |
|
|
C master thread i.e. its not really a critical section. We can |
|
|
C do this because we do not use critical sections in any critical |
|
|
C sections of our code! |
|
|
#define _BEGIN_CRIT(a) _BEGIN_MASTER(a) |
|
|
#define _END_CRIT(a) _END_MASTER(a) |
|
|
|
|
|
C-- Define the mapping for the BEGIN_MASTER_SECTION() and |
|
|
C END_MASTER_SECTION() macros. These are generally implemented by |
|
|
C simply choosing a particular thread to be "the master" and have |
|
|
C it alone execute the BEGIN_MASTER..., END_MASTER.. sections. |
|
|
#define _BEGIN_MASTER(a) IF ( a .EQ. 1 ) THEN |
|
|
#define _END_MASTER(a) ENDIF |
|
| 105 |
|
|
| 106 |
C-- Control MPI based parallel processing |
C-- Control MPI based parallel processing |
| 107 |
#undef ALLOW_USE_MPI |
CXXX We no longer select the use of MPI via this file (CPP_EEOPTIONS.h) |
| 108 |
#undef ALWAYS_USE_MPI |
CXXX To use MPI, use an appropriate genmake2 options file or use |
| 109 |
|
CXXX genmake2 -mpi . |
| 110 |
|
CXXX #undef ALLOW_USE_MPI |
| 111 |
|
|
| 112 |
C-- Control use of communication that might overlap computation. |
C-- Control use of communication that might overlap computation. |
| 113 |
C Under MPI selects/deselects "non-blocking" sends and receives. |
C Under MPI selects/deselects "non-blocking" sends and receives. |
| 114 |
#define ALLOW_ASYNC_COMMUNICATION |
#define ALLOW_ASYNC_COMMUNICATION |
| 119 |
#define ALLOW_SYNC_COMMUNICATION |
#define ALLOW_SYNC_COMMUNICATION |
| 120 |
#undef ALWAYS_USE_SYNC_COMMUNICATION |
#undef ALWAYS_USE_SYNC_COMMUNICATION |
| 121 |
|
|
|
C-- Control storage of floating point operands |
|
|
C On many systems it improves performance only to use |
|
|
C 8-byte precision for time stepped variables. |
|
|
C Constant in time terms ( geometric factors etc.. ) |
|
|
C can use 4-byte precision, reducing memory utilisation and |
|
|
C boosting performance because of a smaller working |
|
|
C set size. However, on vector CRAY systems this degrades |
|
|
C performance. |
|
|
#define REAL4_IS_SLOW |
|
|
|
|
|
#ifdef REAL4_IS_SLOW |
|
|
#define real Real*8 |
|
|
#define REAL Real*8 |
|
|
#define _RS Real*8 |
|
|
#define _RL Real*8 |
|
|
#define RS_IS_REAL8 |
|
|
#define _EXCH_XY_R4(a,b) CALL EXCH_XY_R8 ( a, b ) |
|
|
#define _EXCH_XYZ_R4(a,b) CALL EXCH_XYZ_R8 ( a, b ) |
|
|
#define _GLOBAL_SUM_R4(a,b,c) CALL GLOBAL_SUM_R8( a, b , c) |
|
|
#define _GLOBAL_MAX_R4(a,b,c) CALL GLOBAL_MAX_R8( a, b , c) |
|
|
#endif |
|
|
|
|
|
#ifndef REAL4_IS_SLOW |
|
|
#define real Real*4 |
|
|
#define REAL Real*8 |
|
|
#define _RS Real*4 |
|
|
#define _RL Real*8 |
|
|
#define RS_IS_REAL4 |
|
|
#define _EXCH_XY_R4(a,b) CALL EXCH_XY_R4 ( a, b ) |
|
|
#define _EXCH_XYZ_R4(a,b) CALL EXCH_XYZ_R4 ( a, b ) |
|
|
#define _GLOBAL_SUM_R4(a,b,c) CALL GLOBAL_SUM_R4( a, b , c) |
|
|
#define _GLOBAL_MAX_R4(a,b,c) CALL GLOBAL_MAX_R4( a, b , c) |
|
|
#endif |
|
|
|
|
|
#define _EXCH_XY_R8(a,b) CALL EXCH_XY_R8 ( a, b ) |
|
|
#define _EXCH_XYZ_R8(a,b) CALL EXCH_XYZ_R8 ( a, b ) |
|
|
#define _GLOBAL_SUM_R8(a,b,c) CALL GLOBAL_SUM_R8( a, b , c) |
|
|
#define _GLOBAL_MAX_R8(a,b,c) CALL GLOBAL_MAX_R8( a, b , c) |
|
|
|
|
|
C-- Control use of "double" precision constants. |
|
|
C Use D0 where it means REAL*8 but not where it means REAL*16 |
|
|
#define D0 d0 |
|
|
#ifdef REAL_D0_IS_16BYTES |
|
|
#define D0 |
|
|
#endif |
|
|
|
|
| 122 |
C-- Control XY periodicity in processor to grid mappings |
C-- Control XY periodicity in processor to grid mappings |
| 123 |
C Note: Model code does not need to know whether a domain is |
C Note: Model code does not need to know whether a domain is |
| 124 |
C periodic because it has overlap regions for every box. |
C periodic because it has overlap regions for every box. |
| 125 |
C Model assume that these values have been |
C Model assume that these values have been |
| 126 |
C filled in some way. |
C filled in some way. |
| 129 |
#define CAN_PREVENT_X_PERIODICITY |
#define CAN_PREVENT_X_PERIODICITY |
| 130 |
#define CAN_PREVENT_Y_PERIODICITY |
#define CAN_PREVENT_Y_PERIODICITY |
| 131 |
|
|
| 132 |
C-- Substitue for 1.D variables |
C-- disconnect tiles (no exchange between tiles, just fill-in edges |
| 133 |
C Sun compilers do not use 8-byte precision for literals |
C assuming locally periodic subdomain) |
| 134 |
C unless .Dnn is specified. CRAY vector machines use 16-byte |
#undef DISCONNECTED_TILES |
| 135 |
C precision when they see .Dnn which runs very slowly! |
|
| 136 |
#ifdef REAL_D0_IS_16BYTES |
C-- Always cumulate tile local-sum in the same order by applying MPI allreduce |
| 137 |
#define _d |
C to array of tiles ; can get slower with large number of tiles (big set-up) |
| 138 |
#define _F64( a ) a |
#define GLOBAL_SUM_ORDER_TILES |
| 139 |
#endif |
|
| 140 |
#ifndef REAL_D0_IS_16BYTES |
C-- Alternative way of doing global sum without MPI allreduce call |
| 141 |
#define _d D |
C but instead, explicit MPI send & recv calls. Expected to be slower. |
| 142 |
#define _F64( a ) DFLOAT( a ) |
#undef GLOBAL_SUM_SEND_RECV |
| 143 |
#endif |
|
| 144 |
|
C-- Alternative way of doing global sum on a single CPU |
| 145 |
|
C to eliminate tiling-dependent roundoff errors. Note: This is slow. |
| 146 |
|
#undef CG2D_SINGLECPU_SUM |
| 147 |
|
|
| 148 |
|
C=== Other options (to add/remove pieces of code) === |
| 149 |
|
C-- Flag to turn on checking for errors from all threads and procs |
| 150 |
|
C (calling S/R STOP_IF_ERROR) before stopping. |
| 151 |
|
#define USE_ERROR_STOP |
| 152 |
|
|
| 153 |
|
C-- Control use of communication with other component: |
| 154 |
|
C allow to import and export from/to Coupler interface. |
| 155 |
|
#undef COMPONENT_MODULE |
| 156 |
|
|
| 157 |
#endif /* _CPP_EEOPTIONS_H_ */ |
#endif /* _CPP_EEOPTIONS_H_ */ |
| 158 |
|
|
| 159 |
|
#include "CPP_EEMACROS.h" |
| 160 |
|
|
Parent Directory
| 
Revision Log
| 
Revision Graph
| 
Patch