From faa1f1adb34c4aefa529c8a9d6ab2e8f69200a39 Mon Sep 17 00:00:00 2001
From: Juan Escobar <juan.escobar@aero.obs-mip.fr>
Date: Wed, 19 May 2021 16:33:51 +0200
Subject: [PATCH] Juan 19/05/2021: add loop independent collapse , for all do
conccurent & BR_ routine , or the do loop seq !!!
---
src/MNH/advection_metsv.f90 | 21 ++++--
src/MNH/emoist.f90 | 2 +
src/MNH/gradient_u.f90 | 11 ++-
src/MNH/gradient_v.f90 | 5 ++
src/MNH/mode_prandtl.f90 | 26 ++++---
src/MNH/prandtl.f90 | 11 ++-
src/MNH/rain_ice.f90 | 29 +++++---
src/MNH/rain_ice_nucleation.f90 | 1 +
src/MNH/resolved_cloud.f90 | 1 +
src/MNH/shuman_device.f90 | 16 +++-
src/MNH/tke_eps_sources.f90 | 13 +++-
src/MNH/tridiag_thermo.f90 | 18 ++++-
src/MNH/tridiag_tke.f90 | 10 ++-
src/MNH/tridiag_w.f90 | 16 +++-
src/MNH/tridiag_wind.f90 | 4 +-
src/MNH/turb.f90 | 11 ++-
src/MNH/turb_hor_dyn_corr.f90 | 24 +++++-
src/MNH/turb_hor_thermo_flux.f90 | 122 +++++++++++++++++++------------
src/MNH/turb_hor_tke.f90 | 24 +++++-
src/MNH/turb_hor_uv.f90 | 23 +++++-
src/MNH/turb_hor_uw.f90 | 11 +++
src/MNH/turb_hor_vw.f90 | 11 +++
src/MNH/turb_ver.f90 | 1 +
src/MNH/turb_ver_dyn_flux.f90 | 100 ++++++++++++++++++-------
src/MNH/turb_ver_thermo_corr.f90 | 97 +++++++++++++++---------
src/MNH/turb_ver_thermo_flux.f90 | 19 ++++-
src/Makefile.MESONH.mk | 2 +-
27 files changed, 470 insertions(+), 159 deletions(-)
diff --git a/src/MNH/advection_metsv.f90 b/src/MNH/advection_metsv.f90
index 73c8cd0aa..4a69e687f 100644
--- a/src/MNH/advection_metsv.f90
+++ b/src/MNH/advection_metsv.f90
@@ -527,12 +527,18 @@ IF (.NOT. L1D) THEN
END IF
#else
IF (.NOT. L2D) THEN
- !$acc kernels
- ZCFL(:,:,:) = SQRT(BR_P2(ZCFLU(:,:,:))+BR_P2(ZCFLV(:,:,:))+BR_P2(ZCFLW(:,:,:)))
+ !$acc kernels
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU )
+ ZCFL(JI,JJ,JK) = SQRT(BR_P2(ZCFLU(JI,JJ,JK))+BR_P2(ZCFLV(JI,JJ,JK))+BR_P2(ZCFLW(JI,JJ,JK)))
+ END DO
!$acc end kernels
ELSE
- !$acc kernels
- ZCFL(:,:,:) = SQRT(BR_P2(ZCFLU(:,:,:))+BR_P2(ZCFLW(:,:,:)))
+ !$acc kernels
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU )
+ ZCFL(JI,JJ,JK) = SQRT(BR_P2(ZCFLU(JI,JJ,JK))+BR_P2(ZCFLW(JI,JJ,JK)))
+ END DO
!$acc end kernels
END IF
#endif
@@ -543,7 +549,10 @@ ELSE
#ifndef MNH_BITREP
ZCFL(:,:,:) = SQRT(ZCFLW(:,:,:)**2)
#else
- ZCFL(:,:,:) = SQRT(BR_P2(ZCFLW(:,:,:)))
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU )
+ ZCFL(JI,JJ,JK) = SQRT(BR_P2(ZCFLW(JI,JJ,JK)))
+ END DO
#endif
!$acc end kernels
END IF
@@ -859,12 +868,14 @@ DO JSPL=1,KSPLIT
!$acc end kernels
END IF
!$acc kernels
+ !$acc loop independent collapse(4)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU, JR=1:KRR )
ZR(JI,JJ,JK,JR) = ZR(JI,JJ,JK,JR) + ( ZRRS_PPM(JI,JJ,JK,JR) + ZRRS_OTHER(JI,JJ,JK,JR) + PRRS_CLD(JI,JJ,JK,JR) ) &
* ZTSTEP_PPM / PRHODJ(JI,JJ,JK)
END DO !CONCURRENT
!$acc loop seq
DO JSV = 1, KSV
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZSV(JI,JJ,JK,JSV) = ZSV(JI,JJ,JK,JSV) + ( ZRSVS_PPM(JI,JJ,JK,JSV) + ZRSVS_OTHER(JI,JJ,JK,JSV) + &
PRSVS_CLD(JI,JJ,JK,JSV) ) * ZTSTEP_PPM / PRHODJ(JI,JJ,JK)
diff --git a/src/MNH/emoist.f90 b/src/MNH/emoist.f90
index d08b2042e..7116d12d4 100644
--- a/src/MNH/emoist.f90
+++ b/src/MNH/emoist.f90
@@ -191,6 +191,7 @@ ELSE ! liquid water & ice present
DO JRR=5,KRR
ZRW(1:JIU,1:JJU,1:JKU) = ZRW(1:JIU,1:JJU,1:JKU) + PRM(1:JIU,1:JJU,1:JKU,JRR)
ENDDO
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = 1. + ( & ! Compute A
(1.+ZDELTA) * (PRM(JI,JJ,JK,1) - PRM(JI,JJ,JK,2) - PRM(JI,JJ,JK,4)) &
@@ -218,6 +219,7 @@ ELSE ! liquid water & ice present
DO JRR=3,KRR
ZRW(1:JIU,1:JJU,1:JKU) = ZRW(1:JIU,1:JJU,1:JKU) + PRM(1:JIU,1:JJU,1:JKU,JRR)
ENDDO
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = 1. + ( & ! Compute ZA
(1.+ZDELTA) * (PRM(JI,JJ,JK,1) - PRM(JI,JJ,JK,2)) &
diff --git a/src/MNH/gradient_u.f90 b/src/MNH/gradient_u.f90
index fa743bbd8..a361dcbed 100644
--- a/src/MNH/gradient_u.f90
+++ b/src/MNH/gradient_u.f90
@@ -244,13 +244,15 @@ iztmp3_device = MNH_ALLOCATE_ZT3D( ztmp3_device,JIU,JJU,JKU )
IF (.NOT. LFLAT) THEN
CALL DXF_DEVICE(PA,ZTMP1_DEVICE)
CALL DZM_DEVICE(KKA,KKU,KL,PA,ZTMP2_DEVICE)
- !$acc kernels loop independent collapse(3)
- DO JK=1,JKU ; DO JJ=1,JJU ; DO JI=1,JIU
+ !$acc kernels
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
- END DO ; END DO ; END DO
+ END DO !CONCURRENT
!$acc end kernels
CALL MXF_DEVICE(ZTMP3_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
END DO !CONCURRENT
@@ -434,6 +436,7 @@ IF (.NOT. LFLAT) THEN
CALL DZM_DEVICE(KKA,KKU,KL,PA,ZTMP1_DEVICE)
CALL MXM_DEVICE(PDZZ,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK)/ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -441,6 +444,7 @@ IF (.NOT. LFLAT) THEN
CALL MYM_DEVICE(ZTMP3_DEVICE,ZTMP1_DEVICE)
CALL MXM_DEVICE(PDZY,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -449,6 +453,7 @@ IF (.NOT. LFLAT) THEN
CALL DYM_DEVICE(PA,ZTMP1_DEVICE)
CALL MXM_DEVICE(PDYY,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PGY_U_UV_DEVICE(JI,JJ,JK)= ( ZTMP1_DEVICE(JI,JJ,JK) - ZTMP2_DEVICE(JI,JJ,JK) ) / ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/gradient_v.f90 b/src/MNH/gradient_v.f90
index 0f7192cd8..c1a7810be 100644
--- a/src/MNH/gradient_v.f90
+++ b/src/MNH/gradient_v.f90
@@ -247,12 +247,14 @@ IF (.NOT. LFLAT) THEN
CALL DYF_DEVICE(PA,ZTMP1_DEVICE)
CALL DZM_DEVICE(KKA,KKU,KL,PA,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYF_DEVICE(ZTMP3_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)/PDZZ(JI,JJ,JK)
END DO !CONCURRENT
@@ -435,6 +437,7 @@ IF (.NOT. LFLAT) THEN
CALL MYM_DEVICE(PDZZ,ZTMP2_DEVICE)
CALL DZM_DEVICE(KKA,KKU,KL,PA,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -442,6 +445,7 @@ IF (.NOT. LFLAT) THEN
CALL MXM_DEVICE(ZTMP4_DEVICE,ZTMP2_DEVICE)
CALL MYM_DEVICE(PDZX,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK) *ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -449,6 +453,7 @@ IF (.NOT. LFLAT) THEN
CALL MZF_DEVICE(KKA,KKU,KL,ZTMP4_DEVICE,ZTMP2_DEVICE)
CALL MYM_DEVICE(PDXX,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PGX_V_UV_DEVICE(JI,JJ,JK)= ( ZTMP1_DEVICE(JI,JJ,JK) - ZTMP2_DEVICE(JI,JJ,JK) ) / ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/mode_prandtl.f90 b/src/MNH/mode_prandtl.f90
index 5ac5ddc57..1a714d1c5 100644
--- a/src/MNH/mode_prandtl.f90
+++ b/src/MNH/mode_prandtl.f90
@@ -122,27 +122,35 @@ igphi3logic = MNH_ALLOCATE_GT3D( gphi3logic, JIU,JJU,JKU )
!$acc kernels
IF (HTURBDIM=='3DIM') THEN
!* 3DIM case
- IF (OUSERV) THEN
- ZW1(:,:,:) = 1. + 1.5* (PREDTH1(:,:,:)+PREDR1(:,:,:)) + &
+ IF (OUSERV) THEN
#ifndef MNH_BITREP
+ ZW1(:,:,:) = 1. + 1.5* (PREDTH1(:,:,:)+PREDR1(:,:,:)) + &
( 0.5 * (PREDTH1(:,:,:)**2+PREDR1(:,:,:)**2) &
-#else
- ( 0.5 * (BR_P2(PREDTH1(:,:,:))+BR_P2(PREDR1(:,:,:))) &
-#endif
+ PREDTH1(:,:,:) * PREDR1(:,:,:) &
- )
+ )
+#else
+ DO CONCURRENT (JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZW1(JI,JJ,JK) = 1. + 1.5* (PREDTH1(JI,JJ,JK)+PREDR1(JI,JJ,JK)) + &
+ ( 0.5 * (BR_P2(PREDTH1(JI,JJ,JK))+BR_P2(PREDR1(JI,JJ,JK))) &
+ + PREDTH1(JI,JJ,JK) * PREDR1(JI,JJ,JK) &
+ )
+ END DO
+#endif
ZW2(:,:,:) = 0.5 * (PRED2TH3(:,:,:)-PRED2R3(:,:,:))
PPHI3(:,:,:)= 1. - &
( ( (1.+PREDR1(:,:,:)) * &
(PRED2THR3(:,:,:) + PRED2TH3(:,:,:)) / PREDTH1(:,:,:) &
) + ZW2(:,:,:) &
) / ZW1(:,:,:)
- ELSE
+ ELSE
+#ifndef MNH_BITREP
ZW1(:,:,:) = 1. + 1.5* PREDTH1(:,:,:) + &
-#ifndef MNH_BITREP
0.5* PREDTH1(:,:,:)**2
#else
- 0.5* BR_P2(PREDTH1(:,:,:))
+ DO CONCURRENT (JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZW1(JI,JJ,JK) = 1. + 1.5* PREDTH1(JI,JJ,JK) + &
+ 0.5* BR_P2(PREDTH1(JI,JJ,JK))
+ END DO
#endif
ZW2(:,:,:) = 0.5* PRED2TH3(:,:,:)
PPHI3(:,:,:)= 1. - &
diff --git a/src/MNH/prandtl.f90 b/src/MNH/prandtl.f90
index 0b0815ba5..927314266 100644
--- a/src/MNH/prandtl.f90
+++ b/src/MNH/prandtl.f90
@@ -437,7 +437,7 @@ END WHERE
WHERE (PREDTH1(:,:,:) < -ZMINVAL)
ZW2(:,:,:) = (-ZMINVAL) / (PREDTH1(:,:,:))
END WHERE
-
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZW2(JI,JJ,JK) = MIN( ZW1(JI,JJ,JK),ZW2(JI,JJ,JK) )
END DO
@@ -448,6 +448,7 @@ WHERE (PREDR1(:,:,:)<-ZMINVAL)
END WHERE
!!$ZW1(:,:,:) = MIN(ZW2(:,:,:),ZW1(:,:,:))
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZW1(JI,JJ,JK) = MIN( ZW2(JI,JJ,JK),ZW1(JI,JJ,JK) )
END DO
@@ -550,6 +551,7 @@ ELSE IF (L2D) THEN ! 3D case in a 2D model
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = ZTMP1_DEVICE(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = BR_P2(ZTMP1_DEVICE(JI,JJ,JK))
END DO !CONCURRENT
@@ -587,6 +589,7 @@ END DO !CONCURRENT
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = ZTMP1_DEVICE(:,:,:)**2
#else
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = BR_P2(ZTMP1_DEVICE(JI,JJ,JK))
END DO !CONCURRENT
@@ -623,6 +626,7 @@ END DO !CONCURRENT
CALL GX_M_M_DEVICE(KKA,KKU,KKL,PRM(:,:,:,1),PDXX,PDZZ,PDZX,ZTMP1_DEVICE)
CALL GX_M_M_DEVICE(KKA,KKU,KKL,PTHLM ,PDXX,PDZZ,PDZX,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -635,6 +639,7 @@ END DO
* PEMOIST(:,:,:) * PETHETA(:,:,:) &
* ZTMP2_DEVICE(:,:,:)
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRED2THR3(JI,JJ,JK)= PREDR1(JI,JJ,JK) * PREDTH1(JI,JJ,JK) + BR_P2(XCTV)*BR_P2(PBLL_O_E(JI,JJ,JK)) * &
PEMOIST(JI,JJ,JK) * PETHETA(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
@@ -719,6 +724,7 @@ ELSE ! 3D case in a 3D model
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = ZTMP1_DEVICE(:,:,:)**2 + ZTMP2_DEVICE(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = BR_P2(ZTMP1_DEVICE(JI,JJ,JK)) + BR_P2(ZTMP2_DEVICE(JI,JJ,JK))
END DO
@@ -759,6 +765,7 @@ END DO
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = ZTMP1_DEVICE(:,:,:)**2 + ZTMP2_DEVICE(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = BR_P2(ZTMP1_DEVICE(JI,JJ,JK)) + BR_P2(ZTMP2_DEVICE(JI,JJ,JK))
END DO
@@ -801,6 +808,7 @@ END DO
CALL GY_M_M_DEVICE(KKA,KKU,KKL,PRM(:,:,:,1),PDYY,PDZZ,PDZY,ZTMP3_DEVICE)
CALL GY_M_M_DEVICE(KKA,KKU,KKL,PTHLM ,PDYY,PDZZ,PDZY,ZTMP4_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)+ &
ZTMP3_DEVICE(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
@@ -844,6 +852,7 @@ call Print_msg( NVERB_WARNING, 'GEN', 'PRANDTL', 'OpenACC: L2D=.F. and KRR=0 not
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = ZTMP1_DEVICE(:,:,:)**2 + ZTMP2_DEVICE(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = BR_P2(ZTMP1_DEVICE(JI,JJ,JK)) + BR_P2(ZTMP2_DEVICE(JI,JJ,JK))
END DO
diff --git a/src/MNH/rain_ice.f90 b/src/MNH/rain_ice.f90
index d0de2f3ff..6a7c12bdd 100644
--- a/src/MNH/rain_ice.f90
+++ b/src/MNH/rain_ice.f90
@@ -747,7 +747,7 @@ IZRHODJ = MNH_ALLOCATE_ZT1DP(ZRHODJ,0)
!
!$acc kernels
-!acc loop independent
+!$acc loop independent
DO CONCURRENT ( JL=1:IMICRO )
ZRVT(JL) = PRVT(I1(JL),I2(JL),I3(JL))
ZRCT(JL) = PRCT(I1(JL),I2(JL),I3(JL))
@@ -790,20 +790,26 @@ IZRHODJ = MNH_ALLOCATE_ZT1DP(ZRHODJ,0)
! ZSIGMA_RC(JL) = MAX(PSIGS(I1(JL),I2(JL),I3(JL)) * 2., 1.E-12)
END DO
END IF
-!
- ZZW(:) = ZEXNREF(:)*( XCPD+XCPV*ZRVT(:)+XCL*(ZRCT(:)+ZRRT(:)) &
- +XCI*(ZRIT(:)+ZRST(:)+ZRGT(:)) )
- ZLSFACT(:) = (XLSTT+(XCPV-XCI)*(ZZT(:)-XTT))/ZZW(:) ! L_s/(Pi_ref*C_ph)
- ZLVFACT(:) = (XLVTT+(XCPV-XCL)*(ZZT(:)-XTT))/ZZW(:) ! L_v/(Pi_ref*C_ph)
-
+ !
+ !$acc loop independent
+ DO CONCURRENT ( JL=1:IMICRO )
+ ZZW(JL) = ZEXNREF(JL)*( XCPD+XCPV*ZRVT(JL)+XCL*(ZRCT(JL)+ZRRT(JL)) &
+ +XCI*(ZRIT(JL)+ZRST(JL)+ZRGT(JL)) )
+ ZLSFACT(JL) = (XLSTT+(XCPV-XCI)*(ZZT(JL)-XTT))/ZZW(JL) ! L_s/(Pi_ref*C_ph)
+ ZLVFACT(JL) = (XLVTT+(XCPV-XCL)*(ZZT(JL)-XTT))/ZZW(JL) ! L_v/(Pi_ref*C_ph)
+ END DO
+
#ifndef MNH_BITREP
ZZW(:) = EXP( XALPI - XBETAI/ZZT(:) - XGAMI*ALOG(ZZT(:) ) )
+ ZSSI(:) = ZRVT(:)*( ZPRES(:)-ZZW(:) ) / ( (XMV/XMD) * ZZW(:) ) - 1.0
#else
+ !$acc loop independent
DO CONCURRENT ( JL=1:IMICRO )
ZZW(JL) = BR_EXP( XALPI - XBETAI/ZZT(JL) - XGAMI*BR_LOG(ZZT(JL) ) )
+ ZSSI(JL) = ZRVT(JL)*( ZPRES(JL)-ZZW(JL) ) / ( (XMV/XMD) * ZZW(JL) ) - 1.0
END DO
#endif
- ZSSI(:) = ZRVT(:)*( ZPRES(:)-ZZW(:) ) / ( (XMV/XMD) * ZZW(:) ) - 1.0
+
! Supersaturation over ice
!
IF (LBU_ENABLE .OR. LLES_CALL) THEN
@@ -815,7 +821,10 @@ IZRHODJ = MNH_ALLOCATE_ZT1DP(ZRHODJ,0)
!
!Cloud water split between high and low content part is done here
!according to autoconversion option
- ZRCRAUTC(:) = XCRIAUTC/ZRHODREF(:) ! Autoconversion rc threshold
+ !$acc loop independent
+ DO CONCURRENT ( JL=1:IMICRO )
+ ZRCRAUTC(JL) = XCRIAUTC/ZRHODREF(JL) ! Autoconversion rc threshold
+ END DO
!$acc end kernels
#ifdef MNH_OPENACC
IF (LBU_ENABLE .OR. LLES_CALL) THEN
@@ -1054,6 +1063,7 @@ IZRHODJ = MNH_ALLOCATE_ZT1DP(ZRHODJ,0)
ZLBDAR(:) = 0.
END WHERE
#else
+ !$acc loop independent
DO CONCURRENT ( JL=1:IMICRO )
IF ( ZRRT(JL)>0.0 ) THEN
ZLBDAR(JL) = XLBR * BR_POW( ZRHODREF(JL) * MAX( ZRRT(JL), XRTMIN(3) ), XLBEXR )
@@ -1070,6 +1080,7 @@ END DO ! CONCURRENT
ZLBDAR_RF(:) = 0.
END WHERE
#else
+ !$acc loop independent
DO CONCURRENT ( JL=1:IMICRO )
IF ( ZRRT(JL)>0.0 .AND. ZRF(JL)>0.0 ) THEN
ZLBDAR_RF(JL) = XLBR * BR_POW( ZRHODREF(JL) * MAX( ZRRT(JL)/ZRF(JL), XRTMIN(3) ), XLBEXR )
diff --git a/src/MNH/rain_ice_nucleation.f90 b/src/MNH/rain_ice_nucleation.f90
index 509e7ddab..88858b403 100644
--- a/src/MNH/rain_ice_nucleation.f90
+++ b/src/MNH/rain_ice_nucleation.f90
@@ -203,6 +203,7 @@ IF( INEGT >= 1 ) THEN
( ZSSI(1:INEGT)/ZUSW(1:INEGT) )**XALPHA1 )
END WHERE
#else
+ !$acc loop independent
DO CONCURRENT ( JL=1:INEGT )
IF ( (ZZT(JL)<=XTT-2.0) .AND. (ZZT(JL)>=XTT-5.0) .AND. (ZSSI(JL)>0.0) ) THEN
ZZW(JL) = MAX( XNU20 * BR_EXP( -XBETA2 ),XNU10 * BR_EXP( -XBETA1*(ZZT(JL)-XTT) ) * &
diff --git a/src/MNH/resolved_cloud.f90 b/src/MNH/resolved_cloud.f90
index 5e19a9116..01e131b70 100644
--- a/src/MNH/resolved_cloud.f90
+++ b/src/MNH/resolved_cloud.f90
@@ -722,6 +722,7 @@ ENDIF
! microphysical routines would save
! computing time
!
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
#ifndef MNH_BITREP
ZEXN(JI,JJ,JK) = (PPABST(JI,JJ,JK)/XP00) ** (XRD/XCPD)
diff --git a/src/MNH/shuman_device.f90 b/src/MNH/shuman_device.f90
index 380868f8c..0ca7755d4 100644
--- a/src/MNH/shuman_device.f90
+++ b/src/MNH/shuman_device.f90
@@ -167,6 +167,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PMXF,PA)
+!$acc loop independent collapse(3)
DO JK = 1, IKU
DO JJ = 1, IJU
DO JI = 1 + 1, IIU
@@ -292,6 +293,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PMXM)
+!$acc loop independent collapse(3)
DO JK = 1, IKU
DO JJ = 1, IJU
DO JI = 1 + 1, IIU
@@ -300,6 +302,7 @@ DO JK = 1, IKU
ENDDO
ENDDO
!
+!$acc loop independent collapse(2)
DO JK = 1, IKU
DO JJ=1,IJU
PMXM(1,JJ,JK) = PMXM(IIU-2*JPHEXT+1,JJ,JK) !TODO: voir si ce n'est pas plutot JPHEXT+1
@@ -418,6 +421,7 @@ IKU = SIZE(PA,3)
!
!$acc kernels present(PA,PMYF)
#ifndef _OPT_LINEARIZED_LOOPS
+!$acc loop independent collapse(3)
DO JK=1,IKU
DO JJ=1,IJU-1
DO JI=1,IIU !TODO: remplacer le 1 par JPHEXT ?
@@ -534,6 +538,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PMYM)
+!$acc loop independent collapse(3)
DO JK=1,IKU
DO JJ=2,IJU !TODO: remplacer le 1+1 par 1+JPHEXT ?
DO JI=1,IIU
@@ -870,6 +875,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PDXF)
+!$acc loop independent collapse(3)
DO JK=1,IKU
DO JJ=1,IJU
DO JI=1+1,IIU
@@ -878,6 +884,7 @@ DO JK=1,IKU
END DO
END DO
!
+!$acc loop independent collapse(2)
DO JK=1,IKU
DO JJ=1,IJU
PDXF(IIU,JJ,JK) = PDXF(2*JPHEXT,JJ,JK)
@@ -994,6 +1001,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PDXM)
+!$acc loop independent collapse(3)
DO JK=1,IKU
DO JJ=1,IJU
DO JI=1+1,IIU !TODO: remplacer le 1 par JPHEXT ?
@@ -1002,6 +1010,7 @@ DO JK=1,IKU
END DO
END DO
!
+!$acc loop independent collapse(2)
DO JK=1,IKU
DO JJ=1,IJU
PDXM(1,JJ,JK) = PDXM(IIU-2*JPHEXT+1,JJ,JK) !TODO: remplacer -2*JPHEXT+1 par -JPHEXT ?
@@ -1119,6 +1128,7 @@ IKU = SIZE(PA,3)
!
!$acc kernels present(PA,PDYF)
#ifndef _OPT_LINEARIZED_LOOPS
+!$acc loop independent collapse(3)
DO JK=1,IKU
DO JJ=1,IJU-1 !TODO: remplacer le 1 par JPHEXT ?
DO JI=1,IIU
@@ -1232,6 +1242,7 @@ IKU=SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PDYM)
+!$acc loop independent collapse(3)
DO JK=1,IKU
DO JJ=2,IJU !TODO: remplacer le 2 par JPHEXT+1 ?
DO JI=1,IIU
@@ -1240,9 +1251,6 @@ DO JK=1,IKU
END DO
END DO
!
-DO JJ=1,JPHEXT
- PDYM(:,JJ,:) = PDYM(:,IJU-2*JPHEXT+JJ,:) ! for reprod JPHEXT <> 1
-END DO
#else
JIJKOR = 1 + IIU
JIJKEND = IIU*IJU*IKU
@@ -1345,6 +1353,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PDZF)
+!$acc loop independent collapse(3)
DO JK=1,IKU-1 !TODO: remplacer le 1 par JPHEXT ?
DO JJ=1,IJU
DO JI=1,IIU
@@ -1459,6 +1468,7 @@ IKU = SIZE(PA,3)
!
#ifndef _OPT_LINEARIZED_LOOPS
!$acc kernels present(PA,PDZM)
+!$acc loop independent collapse(3)
DO JK=2,IKU !TODO: remplacer le 1+1 par 1+JPHEXT ?
DO JJ=1,IJU
DO JI=1,IIU
diff --git a/src/MNH/tke_eps_sources.f90 b/src/MNH/tke_eps_sources.f90
index ae9c4b339..17218c59f 100644
--- a/src/MNH/tke_eps_sources.f90
+++ b/src/MNH/tke_eps_sources.f90
@@ -350,7 +350,10 @@ IKE=KKU-JPVEXT_TURB*KKL
#ifndef MNH_BITREP
ZKEFF(:,:,:) = PLM(:,:,:) * SQRT(PTKEM(:,:,:))
#else
-ZKEFF(:,:,:) = PLM(:,:,:) * BR_POW(PTKEM(:,:,:),0.5)
+!$acc loop independent collapse(3)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZKEFF(JI,JJ,JK) = PLM(JI,JJ,JK) * BR_POW(PTKEM(JI,JJ,JK),0.5)
+END DO
#endif
!
!----------------------------------------------------------------------------
@@ -383,8 +386,12 @@ PDP(:,:,IKB) = PDP(:,:,IKB) * (1. + PDZZ(:,:,IKB+KKL)/PDZZ(:,:,IKB))
#ifndef MNH_BITREP
ZFLX(:,:,:) = XCED * SQRT(PTKEM(:,:,:)) / PLEPS(:,:,:)
#else
-ZFLX(:,:,:) = XCED * BR_POW(PTKEM(:,:,:),0.5) / PLEPS(:,:,:)
+!$acc loop independent collapse(3)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZFLX(JI,JJ,JK) = XCED * BR_POW(PTKEM(JI,JJ,JK),0.5) / PLEPS(JI,JJ,JK)
+END DO
#endif
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZSOURCE(JI,JJ,JK) = PRTKES(JI,JJ,JK) / PRHODJ(JI,JJ,JK) + PRTKESM(JI,JJ,JK) / PRHODJ(JI,JJ,JK) &
- PTKEM(JI,JJ,JK) / PTSTEP &
@@ -412,6 +419,7 @@ CALL MZM_DEVICE(PRHODJ,ZTMP2_DEVICE) !Warning: re-used later
#ifndef MNH_BITREP
ZA(:,:,:) = - PTSTEP * XCET * ZTMP1_DEVICE(:,:,:) * ZTMP2_DEVICE(:,:,:) / PDZZ(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = - PTSTEP * XCET * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) / BR_P2(PDZZ(JI,JJ,JK))
END DO !CONCURRENT
@@ -427,6 +435,7 @@ CALL TRIDIAG_TKE(KKA,KKU,KKL,PTKEM,ZA,PTSTEP,PEXPL,PIMPL,PRHODJ,&
CALL GET_HALO(ZRES)
#else
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PTSTEP*ZFLX(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/tridiag_thermo.f90 b/src/MNH/tridiag_thermo.f90
index eefe1b37a..95e86abe0 100644
--- a/src/MNH/tridiag_thermo.f90
+++ b/src/MNH/tridiag_thermo.f90
@@ -266,6 +266,7 @@ CALL MZM_DEVICE(PRHODJ,ZMZM_RHODJ)
#ifndef MNH_BITREP
ZRHODJ_DFDDTDZ_O_DZ2 = ZMZM_RHODJ*PDFDDTDZ/PDZZ**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZRHODJ_DFDDTDZ_O_DZ2(JI,JJ,JK) = ZMZM_RHODJ(JI,JJ,JK)*PDFDDTDZ(JI,JJ,JK)/BR_P2(PDZZ(JI,JJ,JK))
END DO !CONCURRENT
@@ -285,6 +286,7 @@ ZY=0.
! ---------------------------
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,IKB) = PRHODJ(JI,JJ,IKB)*PVARM(JI,JJ,IKB)/PTSTEP &
- ZMZM_RHODJ(JI,JJ,IKB+KKL) * PF(JI,JJ,IKB+KKL)/PDZZ(JI,JJ,IKB+KKL) &
@@ -295,6 +297,7 @@ END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=IKTB+1:IKTE-1)
ZY(JI,JJ,JK) = PRHODJ(JI,JJ,JK)*PVARM(JI,JJ,JK)/PTSTEP &
- ZMZM_RHODJ(JI,JJ,JK+KKL) * PF(JI,JJ,JK+KKL)/PDZZ(JI,JJ,JK+KKL) &
@@ -307,6 +310,7 @@ END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,IKE) = PRHODJ(JI,JJ,IKE)*PVARM(JI,JJ,IKE)/PTSTEP &
- ZMZM_RHODJ(JI,JJ,IKE+KKL) * PF(JI,JJ,IKE+KKL)/PDZZ(JI,JJ,IKE+KKL) &
@@ -327,6 +331,7 @@ IF ( PIMPL > 1.E-10 ) THEN
! --------------
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZB(JI,JJ,IKB) = PRHODJ(JI,JJ,IKB)/PTSTEP &
- ZRHODJ_DFDDTDZ_O_DZ2(JI,JJ,IKB+KKL) * PIMPL
@@ -334,12 +339,14 @@ END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZC(JI,JJ,IKB) = ZRHODJ_DFDDTDZ_O_DZ2(JI,JJ,IKB+KKL) * PIMPL
END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=IKTB+1:IKTE-1)
ZA(JI,JJ,JK) = ZRHODJ_DFDDTDZ_O_DZ2(JI,JJ,JK) * PIMPL
ZB(JI,JJ,JK) = PRHODJ(JI,JJ,JK)/PTSTEP &
@@ -350,6 +357,7 @@ END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZA(JI,JJ,IKE) = ZRHODJ_DFDDTDZ_O_DZ2(JI,JJ,IKE ) * PIMPL
ZB(JI,JJ,IKE) = PRHODJ(JI,JJ,IKE)/PTSTEP &
@@ -364,6 +372,7 @@ END DO !CONCURRENT
! --------
!
!$acc kernels
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZBET(JI,JJ) = ZB(JI,JJ,IKB) ! bet = b(ikb)
PVARP(JI,JJ,IKB) = ZY(JI,JJ,IKB) / ZBET(JI,JJ)
@@ -371,7 +380,8 @@ END DO !CONCURRENT
!
!$acc loop seq
DO JK = IKB+KKL,IKE-KKL,KKL
-!$acc loop gang, vector collapse(2)
+ !$acc loop independent collapse(2)
+ ! acc loop gang, vector collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
!DO JJ=1,JJU
! DO JI=1,JIU
@@ -386,6 +396,7 @@ DO JK = IKB+KKL,IKE-KKL,KKL
END DO !CONCURRENT
END DO
! special treatment for the last level
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZGAM(JI,JJ,IKE) = ZC(JI,JJ,IKE-KKL) / ZBET(JI,JJ)
! gam(k) = c(k-1) / bet
@@ -400,7 +411,8 @@ END DO !CONCURRENT
!
!$acc loop seq
DO JK = IKE-KKL,IKB,-1*KKL
- !$acc loop gang, vector collapse(2)
+ !$acc loop independent collapse(2)
+ ! acc loop gang, vector collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
PVARP(JI,JJ,JK) = PVARP(JI,JJ,JK) - ZGAM(JI,JJ,JK+KKL) * PVARP(JI,JJ,JK+KKL)
END DO !CONCURRENT
@@ -410,6 +422,7 @@ END DO
ELSE
!
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=IKTB:IKTE)
PVARP(JI,JJ,JK) = ZY(JI,JJ,JK) * PTSTEP / PRHODJ(JI,JJ,JK)
END DO !CONCURRENT
@@ -422,6 +435,7 @@ END IF
! ----------------------------------------
!
!$acc kernels
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
PVARP(JI,JJ,KKA)=PVARP(JI,JJ,IKB)
PVARP(JI,JJ,KKU)=PVARP(JI,JJ,IKE)
diff --git a/src/MNH/tridiag_tke.f90 b/src/MNH/tridiag_tke.f90
index 135d4c319..a81715fb4 100644
--- a/src/MNH/tridiag_tke.f90
+++ b/src/MNH/tridiag_tke.f90
@@ -1,3 +1,4 @@
+
!MNH_LIC Copyright 1994-2019 CNRS, Meteo-France and Universite Paul Sabatier
!MNH_LIC This is part of the Meso-NH software governed by the CeCILL-C licence
!MNH_LIC version 1. See LICENSE, CeCILL-C_V1-en.txt and CeCILL-C_V1-fr.txt
@@ -220,12 +221,14 @@ IKE=KKU-JPVEXT_TURB*KKL
!
!
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,IKB) = PVARM(JI,JJ,IKB) + PTSTEP*PSOURCE(JI,JJ,IKB) - &
PEXPL / PRHODJ(JI,JJ,IKB) * PA(JI,JJ,IKB+KKL) * (PVARM(JI,JJ,IKB+KKL) - PVARM(JI,JJ,IKB))
END DO !CONCURRENT
!
DO JK=IKTB+1,IKTE-1
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,JK)= PVARM(JI,JJ,JK) + PTSTEP*PSOURCE(JI,JJ,JK) - &
PEXPL / PRHODJ(JI,JJ,JK) * &
@@ -236,6 +239,7 @@ DO JK=IKTB+1,IKTE-1
END DO !CONCURRENT
END DO
!
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,IKE)= PVARM(JI,JJ,IKE) + PTSTEP*PSOURCE(JI,JJ,IKE) + &
PEXPL / PRHODJ(JI,JJ,IKE) * PA(JI,JJ,IKE) * (PVARM(JI,JJ,IKE)-PVARM(JI,JJ,IKE-KKL))
@@ -250,6 +254,7 @@ IF ( PIMPL > 1.E-10 ) THEN
!
! going up
!
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZBET(JI,JJ) = 1. + PIMPL * (PDIAG(JI,JJ,IKB)-PA(JI,JJ,IKB+KKL) / PRHODJ(JI,JJ,IKB))
! bet = b(ikb)
@@ -276,6 +281,7 @@ IF ( PIMPL > 1.E-10 ) THEN
END DO
END DO
! special treatment for the last level
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZGAM(JI,JJ,IKE) = PIMPL * PA(JI,JJ,IKE) / PRHODJ(JI,JJ,IKE-KKL) / ZBET(JI,JJ)
! gam(k) = c(k-1) / bet
@@ -300,7 +306,8 @@ IF ( PIMPL > 1.E-10 ) THEN
END DO
!
ELSE
- !
+ !
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
PVARP(JI,JJ,IKTB:IKTE) = ZY(JI,JJ,IKTB:IKTE)
END DO !CONCURRENT
@@ -311,6 +318,7 @@ END IF
!* 3. FILL THE UPPER AND LOWER EXTERNAL VALUES
! ----------------------------------------
!
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
PVARP(JI,JJ,KKA)=PVARP(JI,JJ,IKB)
PVARP(JI,JJ,KKU)=PVARP(JI,JJ,IKE)
diff --git a/src/MNH/tridiag_w.f90 b/src/MNH/tridiag_w.f90
index db89be613..b0815bf07 100644
--- a/src/MNH/tridiag_w.f90
+++ b/src/MNH/tridiag_w.f90
@@ -283,6 +283,7 @@ ZY=0.
!!#endif
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,IKB) = ZMZM_RHODJ(JI,JJ,IKB)*PVARM(JI,JJ,IKB)/PTSTEP &
- PRHODJ(JI,JJ,IKB ) * PF(JI,JJ,IKB )/PMZF_DZZ(JI,JJ,IKB ) &
@@ -293,6 +294,7 @@ END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=IKB+1:IKE-1)
ZY(JI,JJ,JK) = ZMZM_RHODJ(JI,JJ,JK)*PVARM(JI,JJ,JK)/PTSTEP &
- PRHODJ(JI,JJ,JK ) * PF(JI,JJ,JK )/PMZF_DZZ(JI,JJ,JK ) &
@@ -305,6 +307,7 @@ END DO !CONCURRENT
!$acc end kernels
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZY(JI,JJ,IKE) = ZMZM_RHODJ(JI,JJ,IKE)*PVARM(JI,JJ,IKE)/PTSTEP &
- PRHODJ(JI,JJ,IKE ) * PF(JI,JJ,IKE )/PMZF_DZZ(JI,JJ,IKE ) &
@@ -329,18 +332,21 @@ END DO !CONCURRENT
!! c(k) = + PRHODJ(k) * PDFDDWDZ(k)/PMZF_DZZ(k)**2
!
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZB(JI,JJ,IKB) = ZMZM_RHODJ(JI,JJ,IKB)/PTSTEP &
- ZRHODJ_DFDDWDZ_O_DZ2(JI,JJ,IKB)
END DO !CONCURRENT
!$acc end kernels
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZC(JI,JJ,IKB) = ZRHODJ_DFDDWDZ_O_DZ2(JI,JJ,IKB)
END DO !CONCURRENT
!$acc end kernels
!$acc kernels ! async
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=IKB+1:IKE-1)
ZA(JI,JJ,JK) = ZRHODJ_DFDDWDZ_O_DZ2(JI,JJ,JK-1)
ZB(JI,JJ,JK) = ZMZM_RHODJ(JI,JJ,JK)/PTSTEP &
@@ -351,11 +357,13 @@ END DO !CONCURRENT
!$acc end kernels
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZA(JI,JJ,IKE) = ZRHODJ_DFDDWDZ_O_DZ2(JI,JJ,IKE-1)
END DO !CONCURRENT
!$acc end kernels
!$acc kernels ! async
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZB(JI,JJ,IKE) = ZMZM_RHODJ(JI,JJ,IKE)/PTSTEP &
- ZRHODJ_DFDDWDZ_O_DZ2(JI,JJ,IKE ) &
@@ -370,12 +378,15 @@ END DO !CONCURRENT
! --------
!
!$acc kernels
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZBET(JI,JJ) = ZB(JI,JJ,IKB) ! bet = b(ikb)
PVARP(JI,JJ,IKB) = ZY(JI,JJ,IKB) / ZBET(JI,JJ)
END DO !CONCURRENT
- !
+!
+!$acc loop seq
DO JK = IKB+1,IKE-1
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZGAM(JI,JJ,JK) = ZC(JI,JJ,JK-1) / ZBET(JI,JJ)
! gam(k) = c(k-1) / bet
@@ -386,6 +397,7 @@ DO JK = IKB+1,IKE-1
END DO !CONCURRENT
END DO
! special treatment for the last level
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZGAM(JI,JJ,IKE) = ZC(JI,JJ,IKE-1) / ZBET(JI,JJ)
! gam(k) = c(k-1) / bet
@@ -398,7 +410,9 @@ END DO !CONCURRENT
!* 3.3 going down
! ----------
!
+!$acc loop seq
DO JK = IKE-1,IKB,-1
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
PVARP(JI,JJ,JK) = PVARP(JI,JJ,JK) - ZGAM(JI,JJ,JK+1) * PVARP(JI,JJ,JK+1)
END DO !CONCURRENT
diff --git a/src/MNH/tridiag_wind.f90 b/src/MNH/tridiag_wind.f90
index 0152538c2..8ab02c872 100644
--- a/src/MNH/tridiag_wind.f90
+++ b/src/MNH/tridiag_wind.f90
@@ -263,6 +263,7 @@ IF ( PIMPL > 1.E-10 ) THEN
! going up
!
!$acc kernels
+ !$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZBET(JI,JJ) = 1. - PIMPL * ( PA(JI,JJ,IKB+KKL) / PRHODJA(JI,JJ,IKB) &
+ PCOEFS(JI,JJ) * PTSTEP ) ! bet = b(ikb)
@@ -271,7 +272,7 @@ IF ( PIMPL > 1.E-10 ) THEN
!
!$acc loop seq
DO JK = IKB+KKL,IKE-KKL,KKL
- !$acc loop gang, vector collapse(2)
+ !$acc loop independent gang, vector collapse(2)
DO CONCURRENT ( JJ=1:JJU , JI=1:JIU )
ZGAM(JI,JJ,JK) = PIMPL * PA(JI,JJ,JK) / PRHODJA(JI,JJ,JK-KKL) / ZBET(JI,JJ)
! gam(k) = c(k-1) / bet
@@ -286,6 +287,7 @@ IF ( PIMPL > 1.E-10 ) THEN
END DO ! CONCURRENT
END DO
! special treatment for the last level
+ !$acc loop independent gang, vector collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU)
ZGAM(JI,JJ,IKE) = PIMPL * PA(JI,JJ,IKE) / PRHODJA(JI,JJ,IKE-KKL) / ZBET(JI,JJ)
! gam(k) = c(k-1) / bet
diff --git a/src/MNH/turb.f90 b/src/MNH/turb.f90
index 81531d768..7b4498dd8 100644
--- a/src/MNH/turb.f90
+++ b/src/MNH/turb.f90
@@ -998,8 +998,11 @@ END IF
ZCDUEFF(:,:) =-SQRT ( (PSFU(:,:)**2 + PSFV(:,:)**2) / &
(XMNH_TINY + ZUSLOPE(:,:)**2 + ZVSLOPE(:,:)**2 ) )
#else
- ZCDUEFF(:,:) =-SQRT ( (BR_P2(PSFU(:,:)) + BR_P2(PSFV(:,:))) / &
- (XMNH_TINY + BR_P2(ZUSLOPE(:,:)) + BR_P2(ZVSLOPE(:,:)) ) )
+ !$acc loop independent collapse(2)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZCDUEFF(JI,JJ) =-SQRT ( (BR_P2(PSFU(JI,JJ)) + BR_P2(PSFV(JI,JJ))) / &
+ (XMNH_TINY + BR_P2(ZUSLOPE(JI,JJ)) + BR_P2(ZVSLOPE(JI,JJ)) ) )
+ END DO
#endif
!$acc end kernels
!
@@ -1728,7 +1731,9 @@ izdrvsatdt = MNH_ALLOCATE_ZT3D( zdrvsatdt, size( pexn, 1 ), size( pexn, 2 ), siz
#ifndef MNH_BITREP
ZRVSAT(:,:,:) = EXP( PALP - PBETA/PT(:,:,:) - PGAM*ALOG( PT(:,:,:) ) )
#else
- ZRVSAT(:,:,:) = BR_EXP( PALP - PBETA/PT(:,:,:) - PGAM*BR_LOG( PT(:,:,:) ) )
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZRVSAT(JI,JJ,JK) = BR_EXP( PALP - PBETA/PT(JI,JJ,JK) - PGAM*BR_LOG( PT(JI,JJ,JK) ) )
+ END DO
#endif
!
!* 1.3 saturation mixing ratio at t
diff --git a/src/MNH/turb_hor_dyn_corr.f90 b/src/MNH/turb_hor_dyn_corr.f90
index 01ee3ad89..1373e2597 100644
--- a/src/MNH/turb_hor_dyn_corr.f90
+++ b/src/MNH/turb_hor_dyn_corr.f90
@@ -403,7 +403,10 @@ IKU = SIZE(PUM,3)
#ifndef MNH_BITREP
ZDIRSINZW(:,:) = SQRT( 1. - PDIRCOSZW(:,:)**2 )
#else
-ZDIRSINZW(:,:) = SQRT( 1. - BR_P2(PDIRCOSZW(:,:)) )
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZDIRSINZW(JI,JJ) = SQRT( 1. - BR_P2(PDIRCOSZW(JI,JJ)) )
+END DO
#endif
!$acc end kernels
!
@@ -437,6 +440,7 @@ CALL ADD3DFIELD_ll( TZFIELDS_ll, ZFLX, 'TURB_HOR_DYN_CORR::ZFLX' )
! Computes the U variance
IF (.NOT. L2D) THEN
!$acc kernels async(2)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK)= (2./3.) * PTKEM(JI,JJ,JK) &
- XCMFS * PK(JI,JJ,JK) *( (4./3.) * GX_U_M_PUM(JI,JJ,JK) &
@@ -613,6 +617,7 @@ ZFLX(:,:,IKB-1) = &
PVSLOPEM(:,:) * PCOSSLOPE(:,:) * PSINSLOPE(:,:) * ZDIRSINZW(:,:) &
- PUSLOPEM(:,:) * PCOSSLOPE(:,:)**2 * ZDIRSINZW(:,:) * PDIRCOSZW(:,:) )
#else
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
ZFLX(JI,JJ,IKB-1) = &
PTAU11M(JI,JJ) * BR_P2(PCOSSLOPE(JI,JJ)) * BR_P2(PDIRCOSZW(JI,JJ)) &
@@ -685,6 +690,7 @@ END IF
#else
CALL MXF_DEVICE(PDXX, ZTMP1_DEVICE)
!$acc kernels async(10)
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) * ZFLX(JI,JJ,JK) / ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -697,18 +703,21 @@ CALL DXM_DEVICE(ZTMP2_DEVICE, ZTMP3_DEVICE)
IF (.NOT. LFLAT) THEN
CALL MZM_DEVICE(PDXX,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) * ZFLX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP4_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP4_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MXM_DEVICE( ZTMP2_DEVICE, ZTMP4_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK) / ZTMP1_DEVICE(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -729,6 +738,7 @@ END IF
IF (KSPLT==1) THEN
! Contribution to the dynamic production of TKE:
!$acc kernels async(2)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZWORK(JI,JJ,JK) = - ZFLX(JI,JJ,JK) * GX_U_M_PUM(JI,JJ,JK)
END DO !CONCURRENT
@@ -779,6 +789,7 @@ END IF
! Computes the V variance
IF (.NOT. L2D) THEN
!$acc kernels async(3)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK)= (2./3.) * PTKEM(JI,JJ,JK) &
- XCMFS * PK(JI,JJ,JK) *( (4./3.) * GY_V_M_PVM(JI,JJ,JK) &
@@ -829,6 +840,7 @@ ZFLX(:,:,IKB-1) = &
PUSLOPEM(:,:) * PSINSLOPE(:,:)**2 * ZDIRSINZW(:,:) * PDIRCOSZW(:,:) &
+ PVSLOPEM(:,:) * PCOSSLOPE(:,:) * PSINSLOPE(:,:) * ZDIRSINZW(:,:) )
#else
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
ZFLX(JI,JJ,IKB-1) = &
PTAU11M(JI,JJ) * BR_P2(PSINSLOPE(JI,JJ)) * BR_P2(PDIRCOSZW(JI,JJ)) &
@@ -896,6 +908,7 @@ IF (.NOT. L2D) THEN
#else
CALL MYF_DEVICE(PDYY, ZTMP1_DEVICE)
!$acc kernels async(10)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) * ZFLX(JI,JJ,JK) / ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -908,18 +921,21 @@ IF (.NOT. L2D) THEN
IF (.NOT. LFLAT) THEN
CALL MZM_DEVICE(PDYY,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) * ZFLX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP4_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP4_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYM_DEVICE( ZTMP2_DEVICE,ZTMP4_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK) / ZTMP1_DEVICE(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -932,6 +948,7 @@ IF (.NOT. L2D) THEN
!$acc end kernels
ELSE
!$acc kernels async(1)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRVS(JI,JJ,JK)=PRVS(JI,JJ,JK) - ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -940,6 +957,7 @@ IF (.NOT. L2D) THEN
! Contribution to the dynamic production of TKE:
IF (KSPLT==1) THEN
!$acc kernels async(2)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZWORK(JI,JJ,JK) = - ZFLX(JI,JJ,JK) * GY_V_M_PVM(JI,JJ,JK)
END DO !CONCURRENT
@@ -995,6 +1013,7 @@ END IF
! Computes the W variance
IF (.NOT. L2D) THEN
!$acc kernels async(2)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = (2./3.) * PTKEM(JI,JJ,JK) &
- XCMFS * PK(JI,JJ,JK) *( (4./3.) * GZ_W_M_PWM(JI,JJ,JK) &
@@ -1039,6 +1058,7 @@ ZFLX(:,:,IKB-1) = &
+ PTAU33M(:,:) * PDIRCOSZW(:,:)**2 &
+2. * PCDUEFF(:,:)* PUSLOPEM(:,:) * ZDIRSINZW(:,:) * PDIRCOSZW(:,:)
#else
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
ZFLX(JI,JJ,IKB-1) = &
PTAU11M(JI,JJ) * BR_P2(ZDIRSINZW(JI,JJ)) &
@@ -1112,6 +1132,7 @@ GZ_W_M_ZWP = GZ_W_M(ZWP,PDZZ)
CALL GZ_W_M_DEVICE(1,IKU,1,ZWP,PDZZ,GZ_W_M_ZWP)
#endif
!$acc kernels async(2)
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=IKB+1:JKU)
ZFLX(JI,JJ,JK)=ZFLX(JI,JJ,JK) &
- XCMFS * PK(JI,JJ,JK) * (4./3.) * (GZ_W_M_ZWP(JI,JJ,JK) - GZ_W_M_PWM(JI,JJ,JK))
@@ -1121,6 +1142,7 @@ END DO !CONCURRENT
IF (KSPLT==1) THEN
!Contribution to the dynamic production of TKE:
!$acc kernels async(2)
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZWORK(JI,JJ,JK) = - ZFLX(JI,JJ,JK) * GZ_W_M_ZWP(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/turb_hor_thermo_flux.f90 b/src/MNH/turb_hor_thermo_flux.f90
index f589aec72..14cba9ab4 100644
--- a/src/MNH/turb_hor_thermo_flux.f90
+++ b/src/MNH/turb_hor_thermo_flux.f90
@@ -342,6 +342,7 @@ ZFLX(:,:,IKE+1) = ZFLX(:,:,IKE)
CALL MXM_DEVICE( PK, ZTMP1_DEVICE )
CALL GX_M_U_DEVICE(1,IKU,1,PTHLM,PDXX,PDZZ,PDZX,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = -XCSHF * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -383,7 +384,10 @@ ZFLX(:,:,IKB-1:IKB-1) = 2. * MXM( SPREAD( PSFTHM(:,:)* PDIRCOSXW(:,:), 3,1) )
- ZFLX(:,:,IKB:IKB)
#else
!$acc kernels
- ZTMP1_DEVICE(:,:,1) = PSFTHM(:,:)* PDIRCOSXW(:,:)
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZTMP1_DEVICE(JI,JJ,1) = PSFTHM(JI,JJ)* PDIRCOSXW(JI,JJ)
+END DO
!$acc end kernels
CALL MXM_DEVICE( ZTMP1_DEVICE(:,:,1:1), ZTMP2_DEVICE(:,:,1:1) )
!$acc kernels
@@ -404,25 +408,29 @@ END IF
#else
IF (.NOT. LFLAT) THEN
CALL MXM_DEVICE(PRHODJ, ZTMP1_DEVICE)
-!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
-END DO
-!$acc end kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
+ END DO
+ !$acc end kernels
CALL DXF_DEVICE(ZTMP2_DEVICE, ZTMP3_DEVICE)
-!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
-END DO
-!$acc end kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
+ END DO
+ !$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP4_DEVICE)
-!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
-END DO
+ !$acc kernels
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL MXF_DEVICE(ZTMP2_DEVICE, ZTMP4_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO
@@ -768,6 +776,7 @@ END IF
CALL MXM_DEVICE( PK, ZTMP1_DEVICE )
CALL GX_M_U_DEVICE(1,IKU,1,PRM(:,:,:,1),PDXX,PDZZ,PDZX,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = -XCHF * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -806,33 +815,38 @@ END DO
IF (.NOT. LFLAT) THEN
CALL MXM_DEVICE(PRHODJ,ZTMP1_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL DXF_DEVICE( ZTMP2_DEVICE, ZTMP3_DEVICE )
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP4_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL MXF_DEVICE(ZTMP2_DEVICE,ZTMP4_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1, ZTMP2_DEVICE, ZTMP4_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- PRRS(JI,JJ,JK,1) = PRRS(JI,JJ,JK,1) - ZTMP3_DEVICE(JI,JJ,JK) + ZTMP4_DEVICE(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ PRRS(JI,JJ,JK,1) = PRRS(JI,JJ,JK,1) - ZTMP3_DEVICE(JI,JJ,JK) + ZTMP4_DEVICE(JI,JJ,JK)
+ END DO
!$acc end kernels
ELSE
CALL MXM_DEVICE(PRHODJ,ZTMP1_DEVICE)
@@ -1140,9 +1154,10 @@ END IF
CALL MYM_DEVICE( PK, ZTMP1_DEVICE )
CALL GY_M_V_DEVICE(1,IKU,1,PTHLM,PDYY,PDZZ,PDZY,ZTMP2_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZFLX(JI,JJ,JK) = -XCSHF * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZFLX(JI,JJ,JK) = -XCSHF * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
+ END DO
ZFLX(:,:,IKE+1) = ZFLX(:,:,IKE)
!$acc end kernels
ELSE
@@ -1186,33 +1201,38 @@ IF (.NOT. L2D) THEN
IF (.NOT. LFLAT) THEN
CALL MYM_DEVICE(PRHODJ, ZTMP1_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL DYF_DEVICE( ZTMP2_DEVICE, ZTMP3_DEVICE )
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP1_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP1_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL MZM_DEVICE(ZTMP1_DEVICE, ZTMP2_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP1_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP1_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL MYF_DEVICE(ZTMP1_DEVICE, ZTMP2_DEVICE)
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- ZTMP1_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP1_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
+ END DO
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1, ZTMP1_DEVICE, ZTMP2_DEVICE )
!$acc kernels
-DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- PRTHLS(JI,JJ,JK) = PRTHLS(JI,JJ,JK) - ZTMP3_DEVICE(JI,JJ,JK) + ZTMP2_DEVICE(JI,JJ,JK)
-END DO
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ PRTHLS(JI,JJ,JK) = PRTHLS(JI,JJ,JK) - ZTMP3_DEVICE(JI,JJ,JK) + ZTMP2_DEVICE(JI,JJ,JK)
+ END DO
!$acc end kernels
ELSE
CALL MYM_DEVICE(PRHODJ, ZTMP1_DEVICE)
@@ -1501,6 +1521,7 @@ IF (KRR/=0) THEN
CALL MYM_DEVICE( PK, ZTMP1_DEVICE )
CALL GY_M_V_DEVICE(1,IKU,1,PRM(:,:,:,1),PDYY,PDZZ,PDZY, ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = -XCHF * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -1546,6 +1567,7 @@ IF (KRR/=0) THEN
IF (.NOT. LFLAT) THEN
CALL MYM_DEVICE(PRHODJ, ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
END DO
@@ -1553,18 +1575,21 @@ IF (KRR/=0) THEN
CALL DYF_DEVICE( ZTMP2_DEVICE, ZTMP3_DEVICE )
!
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZM_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO
!$acc end kernels
CALL MYF_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO
@@ -1572,6 +1597,7 @@ IF (KRR/=0) THEN
CALL DZF_DEVICE(1,IKU,1,ZTMP1_DEVICE,ZTMP2_DEVICE )
!
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRRS(JI,JJ,JK,1) = PRRS(JI,JJ,JK,1) - ZTMP3_DEVICE(JI,JJ,JK) + ZTMP2_DEVICE(JI,JJ,JK)
END DO
diff --git a/src/MNH/turb_hor_tke.f90 b/src/MNH/turb_hor_tke.f90
index 748a65bc3..6a6174eaf 100644
--- a/src/MNH/turb_hor_tke.f90
+++ b/src/MNH/turb_hor_tke.f90
@@ -226,6 +226,7 @@ ZFLX = -XCET * MXM(PK) * GX_M_U(1,IKU,1,PTKEM,PDXX,PDZZ,PDZX) ! < u'e >
CALL MXM_DEVICE(PK,ZTMP1_DEVICE)
CALL GX_M_U_DEVICE(1,IKU,1,PTKEM,PDXX,PDZZ,PDZX,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = -XCET * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) ! < u'e >
END DO !CONCURRENT
@@ -233,9 +234,12 @@ END DO !CONCURRENT
!
! special case near the ground ( uncentred gradient )
!
-ZFLX(:,:,IKB) = ZCOEFF(:,:,IKB+2)*PTKEM(:,:,IKB+2) &
- + ZCOEFF(:,:,IKB+1)*PTKEM(:,:,IKB+1) &
- + ZCOEFF(:,:,IKB )*PTKEM(:,:,IKB )
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZFLX(JI,JJ,IKB) = ZCOEFF(JI,JJ,IKB+2)*PTKEM(JI,JJ,IKB+2) &
+ + ZCOEFF(JI,JJ,IKB+1)*PTKEM(JI,JJ,IKB+1) &
+ + ZCOEFF(JI,JJ,IKB )*PTKEM(JI,JJ,IKB )
+END DO
!$acc end kernels
!
#ifndef MNH_OPENACC
@@ -285,30 +289,35 @@ END IF
IF (.NOT. LFLAT) THEN
CALL MXM_DEVICE(PRHODJ,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DXF_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK)*PINV_PDXX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MXF_DEVICE( ZTMP2_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1,ZTMP2_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PTRH(JI,JJ,JK) =-( ZTMP1_DEVICE(JI,JJ,JK) - ZTMP3_DEVICE(JI,JJ,JK) ) /PRHODJ(JI,JJ,JK)
END DO !CONCURRENT
@@ -352,6 +361,7 @@ IF (.NOT. L2D) THEN
CALL MYM_DEVICE(PK,ZTMP1_DEVICE)
CALL GY_M_V_DEVICE(1,IKU,1,PTKEM,PDYY,PDZZ,PDZY,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) =-XCET * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) ! < v'e >
END DO !CONCURRENT
@@ -415,32 +425,38 @@ IF (.NOT. L2D) THEN
IF (.NOT. LFLAT) THEN
CALL MYM_DEVICE(PRHODJ,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DYF_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK)*PINV_PDYY(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYF_DEVICE(ZTMP2_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1,ZTMP2_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
- PTRH(JI,JJ,JK) = PTRH(JI,JJ,JK) - ( ZTMP1_DEVICE(JI,JJ,JK) - ZTMP3_DEVICE(JI,JJ,JK) ) /PRHODJ(JI,JJ,JK)
+ PTRH(JI,JJ,JK) = PTRH(JI,JJ,JK) - ( ZTMP1_DEVICE(JI,JJ,JK) - ZTMP3_DEVICE(JI,JJ,JK) ) &
+ /PRHODJ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
ELSE
diff --git a/src/MNH/turb_hor_uv.f90 b/src/MNH/turb_hor_uv.f90
index 9866f9adf..d1da623a0 100644
--- a/src/MNH/turb_hor_uv.f90
+++ b/src/MNH/turb_hor_uv.f90
@@ -323,7 +323,10 @@ IKU = SIZE(PUM,3)
#ifndef MNH_BITREP
ZDIRSINZW(:,:) = SQRT( 1. - PDIRCOSZW(:,:)**2 )
#else
-ZDIRSINZW(:,:) = SQRT( 1. - BR_P2(PDIRCOSZW(:,:)) )
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZDIRSINZW(JI,JJ) = SQRT( 1. - BR_P2(PDIRCOSZW(JI,JJ)) )
+END DO
#endif
!$acc end kernels
!
@@ -355,6 +358,7 @@ CALL MXM_DEVICE(PK,ZTMP1_DEVICE)
CALL MYM_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE)
IF (.NOT. L2D) THEN
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK)= - XCMFS * ZTMP2_DEVICE(JI,JJ,JK) * (GY_U_UV_PUM(JI,JJ,JK) + GX_V_UV_PVM(JI,JJ,JK))
END DO !CONCURRENT
@@ -450,6 +454,7 @@ ZFLX(:,:,IKB-1) = &
PDIRCOSZW(:,:) * ZDIRSINZW(:,:) &
+PVSLOPEM(:,:) * (PCOSSLOPE(:,:)**2 - PSINSLOPE(:,:)**2) * ZDIRSINZW(:,:) )
#else
+!$acc loop independent collapse(2)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
ZFLX(JI,JJ,IKB-1) = &
PTAU11M(JI,JJ) * PCOSSLOPE(JI,JJ) * PSINSLOPE(JI,JJ) * BR_P2(PDIRCOSZW(JI,JJ)) &
@@ -507,12 +512,14 @@ END IF
#else
CALL MYM_DEVICE(PRHODJ,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MXM_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -522,30 +529,35 @@ IF (.NOT. LFLAT) THEN
CALL MZM_DEVICE(ZFLX,ZTMP2_DEVICE)
CALL MZM_DEVICE(PDYY,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = PDZY(JI,JJ,JK)/ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MXM_DEVICE(ZTMP4_DEVICE,ZTMP5_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)*ZTMP5_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYF_DEVICE(ZTMP4_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MXM_DEVICE(ZTMP3_DEVICE,ZTMP4_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP5_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1,ZTMP5_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRUS(JI,JJ,JK) = PRUS(JI,JJ,JK) - ZTMP1_DEVICE(JI,JJ,JK) + ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -570,12 +582,14 @@ END IF
#else
CALL MXM_DEVICE(PRHODJ,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYM_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -585,30 +599,35 @@ IF (.NOT. LFLAT) THEN
CALL MZM_DEVICE(ZFLX,ZTMP2_DEVICE)
CALL MZM_DEVICE(PDXX,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = PDZX(JI,JJ,JK)/ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYM_DEVICE(ZTMP4_DEVICE,ZTMP5_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)*ZTMP5_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MXF_DEVICE(ZTMP4_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PMZM_PRHODJ(JI,JJ,JK) * PINV_PDZZ(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MYM_DEVICE(ZTMP3_DEVICE,ZTMP4_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP5_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)*ZTMP4_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1,ZTMP5_DEVICE,ZTMP3_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRVS(JI,JJ,JK) = PRVS(JI,JJ,JK) - ZTMP1_DEVICE(JI,JJ,JK) + ZTMP3_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -635,6 +654,7 @@ IF (KSPLT==1) THEN
#else
IF (.NOT. L2D) THEN
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) * (GY_U_UV_PUM(JI,JJ,JK) + GX_V_UV_PVM(JI,JJ,JK))
END DO !CONCURRENT
@@ -647,6 +667,7 @@ IF (KSPLT==1) THEN
CALL MYF_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE)
CALL MXF_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZWORK(JI,JJ,JK) = - ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/turb_hor_uw.f90 b/src/MNH/turb_hor_uw.f90
index ff27caa32..6b0501fdd 100644
--- a/src/MNH/turb_hor_uw.f90
+++ b/src/MNH/turb_hor_uw.f90
@@ -291,6 +291,7 @@ ZFLX(:,:,:) = &
CALL MZM_DEVICE(PK,ZTMP1_DEVICE)
CALL MXM_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = - XCMFS * ZTMP2_DEVICE(JI,JJ,JK) * GX_W_UW_PWM(JI,JJ,JK)
END DO !CONCURRENT
@@ -331,12 +332,14 @@ PRUS(:,:,:) = PRUS(:,:,:) - DZF( ZFLX* MXM( PMZM_PRHODJ ) / MXM( PDZZ ) )
CALL MXM_DEVICE( PMZM_PRHODJ, ZTMP1_DEVICE )
CALL MXM_DEVICE( PDZZ, ZTMP2_DEVICE )
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK)* ZTMP1_DEVICE(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1, ZTMP3_DEVICE, ZTMP1_DEVICE )
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRUS(JI,JJ,JK) = PRUS(JI,JJ,JK) - ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -355,12 +358,14 @@ END IF
#else
CALL MXM_DEVICE(PRHODJ,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * PINV_PDXX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE, ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK)
END DO !CONCURRENT
@@ -368,12 +373,14 @@ END IF
CALL DXF_DEVICE( ZTMP2_DEVICE,ZTMP1_DEVICE)
IF (.NOT. LFLAT) THEN
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK)*PDZX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZF_DEVICE(1,IKU,1, ZTMP2_DEVICE, ZTMP3_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK)*PINV_PDXX(JI,JJ,JK)
END DO !CONCURRENT
@@ -381,12 +388,14 @@ IF (.NOT. LFLAT) THEN
CALL MXF_DEVICE( ZTMP2_DEVICE, ZTMP3_DEVICE )
CALL MZF_DEVICE(1,IKU,1,PDZZ, ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZM_DEVICE(1,IKU,1, ZTMP4_DEVICE, ZTMP2_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRWS(JI,JJ,JK) = PRWS(JI,JJ,JK) &
- ZTMP1_DEVICE(JI,JJ,JK) &
@@ -410,6 +419,7 @@ IF (KSPLT==1) THEN
#else
CALL GZ_U_UW_DEVICE(1,IKU,1,PUM,PDZZ,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) *( ZTMP1_DEVICE(JI,JJ,JK) + GX_W_UW_PWM(JI,JJ,JK) )
END DO !CONCURRENT
@@ -417,6 +427,7 @@ IF (KSPLT==1) THEN
CALL MXF_DEVICE( ZTMP2_DEVICE,ZTMP1_DEVICE )
CALL MZF_DEVICE(1,IKU,1, ZTMP1_DEVICE, ZTMP2_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZWORK(JI,JJ,JK) = -ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/turb_hor_vw.f90 b/src/MNH/turb_hor_vw.f90
index 6ea12270b..53d816171 100644
--- a/src/MNH/turb_hor_vw.f90
+++ b/src/MNH/turb_hor_vw.f90
@@ -299,6 +299,7 @@ IF (.NOT. L2D) THEN
CALL MZM_DEVICE(PK,ZTMP1_DEVICE)
CALL MYM_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLX(JI,JJ,JK) = - XCMFS * ZTMP2_DEVICE(JI,JJ,JK) * GY_W_VW_PWM(JI,JJ,JK)
END DO !CONCURRENT
@@ -350,12 +351,14 @@ IF (.NOT. L2D) THEN
CALL MYM_DEVICE( PMZM_PRHODJ, ZTMP1_DEVICE )
CALL MYM_DEVICE( PDZZ, ZTMP2_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK)* ZTMP1_DEVICE(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZF_DEVICE(1,IKU,1, ZTMP3_DEVICE, ZTMP1_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRVS(JI,JJ,JK) = PRVS(JI,JJ,JK) - ZTMP1_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
@@ -379,24 +382,28 @@ IF (.NOT. L2D) THEN
IF (.NOT. LFLAT) THEN
CALL MYM_DEVICE(PRHODJ, ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE, ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLX(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DYF_DEVICE( ZTMP2_DEVICE, ZTMP1_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) *PDZY(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL MZF_DEVICE(1,IKU,1,ZTMP2_DEVICE,ZTMP3_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK) * PINV_PDYY(JI,JJ,JK)
END DO !CONCURRENT
@@ -404,12 +411,14 @@ IF (.NOT. L2D) THEN
CALL MYF_DEVICE(ZTMP2_DEVICE,ZTMP3_DEVICE)
CALL MZF_DEVICE(1,IKU,1,PDZZ,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP4_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
!$acc end kernels
CALL DZM_DEVICE(1,IKU,1,ZTMP4_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
PRWS(JI,JJ,JK) = PRWS(JI,JJ,JK) &
- ZTMP1_DEVICE(JI,JJ,JK) &
@@ -443,6 +452,7 @@ IF (KSPLT==1) THEN
#else
CALL GZ_V_VW_DEVICE(1,IKU,1,PVM,PDZZ,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLX(JI,JJ,JK) *( ZTMP1_DEVICE(JI,JJ,JK) + GY_W_VW_PWM(JI,JJ,JK) )
END DO !CONCURRENT
@@ -450,6 +460,7 @@ IF (KSPLT==1) THEN
CALL MYF_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
CALL MZF_DEVICE(1,IKU,1,ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZWORK(JI,JJ,JK) = -ZTMP2_DEVICE(JI,JJ,JK)
END DO !CONCURRENT
diff --git a/src/MNH/turb_ver.f90 b/src/MNH/turb_ver.f90
index 673c4c304..7f2dc537c 100644
--- a/src/MNH/turb_ver.f90
+++ b/src/MNH/turb_ver.f90
@@ -666,6 +666,7 @@ ENDIF
! Denominator factor in 3rd order terms
!
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZD(JI,JJ,JK) = (1.+ZREDTH1(JI,JJ,JK)+ZREDR1(JI,JJ,JK)) * (1.+0.5*(ZREDTH1(JI,JJ,JK)+ZREDR1(JI,JJ,JK)))
END DO
diff --git a/src/MNH/turb_ver_dyn_flux.f90 b/src/MNH/turb_ver_dyn_flux.f90
index 834aa2d85..badf3088d 100644
--- a/src/MNH/turb_ver_dyn_flux.f90
+++ b/src/MNH/turb_ver_dyn_flux.f90
@@ -525,7 +525,10 @@ ZSOURCE(:,:,:) = 0.
#ifndef MNH_BITREP
ZDIRSINZW(:,:) = SQRT(1.-PDIRCOSZW(:,:)**2)
#else
-ZDIRSINZW(:,:) = SQRT(1.-BR_P2(PDIRCOSZW(:,:)))
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZDIRSINZW(JI,JJ) = SQRT(1.-BR_P2(PDIRCOSZW(JI,JJ)))
+END DO
#endif
! compute the coefficients for the uncentred gradient computation near the
! ground
@@ -570,6 +573,7 @@ CALL MXM_DEVICE( PDZZ, ZTMP4_DEVICE )
#ifndef MNH_BITREP
ZA(:,:,:) = -PTSTEP * XCMFS * ZTMP1_DEVICE(:,:,:) * ZTMP3_DEVICE(:,:,:) / ZTMP4_DEVICE(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = -PTSTEP * XCMFS * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) / BR_P2(ZTMP4_DEVICE(JI,JJ,JK))
END DO
@@ -585,21 +589,31 @@ END DO
!$acc kernels
#ifndef MNH_BITREP
ZCOEFFLXU(:,:,1) = PCDUEFF(:,:) * (PDIRCOSZW(:,:)**2 - ZDIRSINZW(:,:)**2) &
+ * PCOSSLOPE(:,:)
#else
-ZCOEFFLXU(:,:,1) = PCDUEFF(:,:) * (BR_P2(PDIRCOSZW(:,:)) - BR_P2(ZDIRSINZW(:,:))) &
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZCOEFFLXU(JI,JJ,1) = PCDUEFF(JI,JJ) * (BR_P2(PDIRCOSZW(JI,JJ)) - BR_P2(ZDIRSINZW(JI,JJ))) &
+ * PCOSSLOPE(JI,JJ)
+END DO
#endif
- * PCOSSLOPE(:,:)
-ZCOEFFLXV(:,:,1) = PCDUEFF(:,:) * PDIRCOSZW(:,:) * PSINSLOPE(:,:)
-
-! prepare the implicit scheme coefficients for the surface flux
-ZCOEFS(:,:,1)= ZCOEFFLXU(:,:,1) * PCOSSLOPE(:,:) * PDIRCOSZW(:,:) &
- +ZCOEFFLXV(:,:,1) * PSINSLOPE(:,:)
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZCOEFFLXV(JI,JJ,1) = PCDUEFF(JI,JJ) * PDIRCOSZW(JI,JJ) * PSINSLOPE(JI,JJ)
+
+ ! prepare the implicit scheme coefficients for the surface flux
+ ZCOEFS(JI,JJ,1)= ZCOEFFLXU(JI,JJ,1) * PCOSSLOPE(JI,JJ) * PDIRCOSZW(JI,JJ) &
+ +ZCOEFFLXV(JI,JJ,1) * PSINSLOPE(JI,JJ)
+END DO
!
! average this flux to be located at the U,W vorticity point
#ifndef MNH_OPENACC
ZCOEFS(:,:,1:1)=MXM(ZCOEFS(:,:,1:1) / PDZZ(:,:,IKB:IKB) )
#else
-ZTMP1_DEVICE(:,:,1) = ZCOEFS(:,:,1) / PDZZ(:,:,IKB)
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZTMP1_DEVICE(JI,JJ,1) = ZCOEFS(JI,JJ,1) / PDZZ(JI,JJ,IKB)
+END DO
!$acc end kernels
CALL MXM_DEVICE(ZTMP1_DEVICE(:,:,1:1),ZCOEFS(:,:,1:1))
#endif
@@ -673,6 +687,7 @@ ZFLXZ(:,:,:) = -XCMFS * MXM(ZKEFF) * &
DZM (PIMPL*ZRES + PEXPL*PUM) / MXM(PDZZ)
#else
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = PIMPL*ZRES(JI,JJ,JK) + PEXPL*PUM(JI,JJ,JK)
END DO
@@ -681,6 +696,7 @@ CALL MXM_DEVICE(ZKEFF,ZTMP1_DEVICE)
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE,ZTMP3_DEVICE)
CALL MXM_DEVICE(PDZZ,ZTMP4_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = -XCMFS * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) / ZTMP4_DEVICE(JI,JJ,JK)
END DO
@@ -738,6 +754,7 @@ PDP(:,:,:) = - MZF( MXF ( ZFLXZ * GZ_U_UW(PUM,PDZZ) ) )
#else
CALL GZ_U_UW_DEVICE(KKA,KKU,KKL,PUM,PDZZ,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU )
ZTMP2_DEVICE(JI,JJ,JK) = ZFLXZ(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK)
END DO
@@ -824,13 +841,15 @@ IF(HTURBDIM=='3DIM') THEN
END IF
#else
CALL MXM_DEVICE(PRHODJ,ZTMP1_DEVICE)
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) /PDXX(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP3_DEVICE )
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK) * ZFLXZ(JI,JJ,JK)
END DO
@@ -839,18 +858,21 @@ IF(HTURBDIM=='3DIM') THEN
IF (.NOT. LFLAT) THEN
CALL MZF_DEVICE(KKA,KKU,KKL,PDZZ,ZTMP2_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZFLXZ(JI,JJ,JK)*PDZX(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL, ZTMP3_DEVICE,ZTMP4_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP4_DEVICE(JI,JJ,JK) / PDXX(JI,JJ,JK)
END DO
!$acc end kernels
CALL MXF_DEVICE( ZTMP3_DEVICE,ZTMP4_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK)
END DO
@@ -872,14 +894,16 @@ IF(HTURBDIM=='3DIM') THEN
ZA(:,:,:)=-MZF( MXF ( ZFLXZ * GX_W_UW( PWM,PDXX,PDZZ,PDZX) ) )
#else
CALL GX_W_UW_DEVICE(KKA,KKU,KKL, PWM,PDXX,PDZZ,PDZX,ZTMP1_DEVICE)
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLXZ(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK)
END DO
!$acc end kernels
CALL MXF_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
CALL MZF_DEVICE(KKA,KKU,KKL, ZTMP1_DEVICE,ZTMP2_DEVICE )
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK)=-ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -1013,6 +1037,7 @@ CALL MYM_DEVICE(ZTMP3_DEVICE,ZTMP4_DEVICE)
#ifndef MNH_BITREP
ZA(:,:,:) = - PTSTEP * XCMFS * ZTMP1_DEVICE(:,:,:) * ZTMP4_DEVICE(:,:,:) / ZTMP2_DEVICE(:,:,:)**2
#else
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = - PTSTEP * XCMFS * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK) / BR_P2(ZTMP2_DEVICE(JI,JJ,JK))
END DO
@@ -1026,21 +1051,31 @@ END DO
! wind following the slope
#ifndef MNH_BITREP
ZCOEFFLXU(:,:,1) = PCDUEFF(:,:) * (PDIRCOSZW(:,:)**2 - ZDIRSINZW(:,:)**2) &
+ * PSINSLOPE(:,:)
#else
-ZCOEFFLXU(:,:,1) = PCDUEFF(:,:) * (BR_P2(PDIRCOSZW(:,:)) - BR_P2(ZDIRSINZW(:,:))) &
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ZCOEFFLXU(JI,JJ,1) = PCDUEFF(JI,JJ) * (BR_P2(PDIRCOSZW(JI,JJ)) - BR_P2(ZDIRSINZW(JI,JJ))) &
+ * PSINSLOPE(JI,JJ)
+END DO
#endif
- * PSINSLOPE(:,:)
ZCOEFFLXV(:,:,1) = PCDUEFF(:,:) * PDIRCOSZW(:,:) * PCOSSLOPE(:,:)
! prepare the implicit scheme coefficients for the surface flux
-ZCOEFS(:,:,1)= ZCOEFFLXU(:,:,1) * PSINSLOPE(:,:) * PDIRCOSZW(:,:) &
- +ZCOEFFLXV(:,:,1) * PCOSSLOPE(:,:)
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ZCOEFS(JI,JJ,1)= ZCOEFFLXU(JI,JJ,1) * PSINSLOPE(JI,JJ) * PDIRCOSZW(JI,JJ) &
+ +ZCOEFFLXV(JI,JJ,1) * PCOSSLOPE(JI,JJ)
+END DO
!
! average this flux to be located at the V,W vorticity point
#ifndef MNH_OPENACC
ZCOEFS(:,:,1:1)=MYM(ZCOEFS(:,:,1:1) / PDZZ(:,:,IKB:IKB) )
#else
-ZTMP1_DEVICE(:,:,1:1) = ZCOEFS(:,:,1:1) / PDZZ(:,:,IKB:IKB)
+!$acc loop independent collapse(2)
+DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZTMP1_DEVICE(JI,JJ,1) = ZCOEFS(JI,JJ,1) / PDZZ(JI,JJ,IKB)
+END DO
!$acc end kernels
CALL MYM_DEVICE(ZTMP1_DEVICE(:,:,1:1),ZCOEFS(:,:,1:1) )
#endif
@@ -1119,14 +1154,16 @@ ZFLXZ(:,:,IKB:IKB) = MYM(PDZZ(:,:,IKB:IKB)) * &
) / 0.5 / ( 1. + MYM(PRHODJ(:,:,KKA:KKA)) / MYM(PRHODJ(:,:,IKB:IKB)) )
#else
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
-ZTMP1_DEVICE(JI,JJ,JK) = PIMPL*ZRES(JI,JJ,JK) + PEXPL*PVM(JI,JJ,JK)
+ ZTMP1_DEVICE(JI,JJ,JK) = PIMPL*ZRES(JI,JJ,JK) + PEXPL*PVM(JI,JJ,JK)
END DO
!$acc end kernels
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE,ZTMP2_DEVICE)
CALL MYM_DEVICE(PDZZ,ZTMP3_DEVICE)
CALL MYM_DEVICE(ZKEFF,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = -XCMFS * ZTMP1_DEVICE(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) / ZTMP3_DEVICE(JI,JJ,JK)
END DO
@@ -1175,6 +1212,7 @@ ZA(:,:,:) = - MZF( MYF ( ZFLXZ * GZ_V_VW(PVM,PDZZ) ) )
#else
CALL GZ_V_VW_DEVICE(KKA,KKU,KKL,PVM,PDZZ,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLXZ(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK)
END DO
@@ -1182,6 +1220,7 @@ END DO
CALL MYF_DEVICE( ZTMP2_DEVICE, ZTMP3_DEVICE )
CALL MZF_DEVICE(KKA,KKU,KKL, ZTMP3_DEVICE, ZTMP1_DEVICE )
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = - ZTMP1_DEVICE(JI,JJ,JK)
END DO
@@ -1261,33 +1300,38 @@ IF(HTURBDIM=='3DIM') THEN
#else
IF (.NOT. L2D) THEN
CALL MYM_DEVICE(PRHODJ,ZTMP1_DEVICE)
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) /PDYY(JI,JJ,JK)
END DO
-!$acc end kernels
+ !$acc end kernels
CALL MZM_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLXZ(JI,JJ,JK)
END DO
-!$acc end kernels
+ !$acc end kernels
CALL DYF_DEVICE( ZTMP2_DEVICE,ZTMP1_DEVICE )
IF (.NOT. LFLAT) THEN
CALL MZF_DEVICE(KKA,KKU,KKL,PDZZ,ZTMP2_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZFLXZ(JI,JJ,JK)*PDZY(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP3_DEVICE,ZTMP4_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP4_DEVICE(JI,JJ,JK) / PDYY(JI,JJ,JK)
END DO
!$acc end kernels
CALL MYF_DEVICE(ZTMP3_DEVICE,ZTMP4_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = PRHODJ(JI,JJ,JK) / ZTMP2_DEVICE(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK)
END DO
@@ -1310,18 +1354,20 @@ IF(HTURBDIM=='3DIM') THEN
ZA(:,:,:) = - MZF( MYF ( ZFLXZ(:,:,:) * GY_W_VW( PWM,PDYY,PDZZ,PDZY) ) )
#else
CALL GY_W_VW_DEVICE(KKA,KKU,KKL, PWM,PDYY,PDZZ,PDZY,ZTMP1_DEVICE)
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZFLXZ(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK)
END DO
-!$acc end kernels
+ !$acc end kernels
CALL MYF_DEVICE(ZTMP2_DEVICE,ZTMP1_DEVICE)
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE,ZTMP2_DEVICE )
-!$acc kernels
+ !$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = - ZTMP2_DEVICE(JI,JJ,JK)
END DO
-!$acc end kernels
+ !$acc end kernels
#endif
!
! evaluate the dynamic production at w(IKB+KKL) in PDP(IKB)
diff --git a/src/MNH/turb_ver_thermo_corr.f90 b/src/MNH/turb_ver_thermo_corr.f90
index c045f7a7d..9227a238d 100644
--- a/src/MNH/turb_ver_thermo_corr.f90
+++ b/src/MNH/turb_ver_thermo_corr.f90
@@ -572,11 +572,15 @@ END IF
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = PPHI3(:,:,:)*PDTH_DZ(:,:,:)**2
#else
- ZTMP1_DEVICE(:,:,:) = PPHI3(:,:,:)*BR_P2(PDTH_DZ(:,:,:))
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP1_DEVICE(JI,JJ,JK) = PPHI3(JI,JJ,JK)*BR_P2(PDTH_DZ(JI,JJ,JK))
+ END DO
#endif
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:),ZTMP2_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZF (JI,JJ,JK) = XCTV*PLM(JI,JJ,JK)*PLEPS(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -718,12 +722,14 @@ END IF
!$acc end kernels
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:),ZTMP2_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP3_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP3_DEVICE(:,:,:),ZTMP4_DEVICE(:,:,:) )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = ZF(JI,JJ,JK) &
+ PIMPL * ZDFDDTDZ(JI,JJ,JK) * ZTMP4_DEVICE(JI,JJ,JK)
@@ -744,17 +750,20 @@ END IF
+ZCOEFF(:,:,IKB )*PTHLP(:,:,IKB ) )**2 &
)
#else
- ZFLXZ(:,:,IKB) = XCTV * PPHI3(:,:,IKB+KKL) * PLM(:,:,IKB) &
- * PLEPS(:,:,IKB) &
+ !$acc loop independent collapse(2)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZFLXZ(JI,JJ,IKB) = XCTV * PPHI3(JI,JJ,IKB+KKL) * PLM(JI,JJ,IKB) &
+ * PLEPS(JI,JJ,IKB) &
*( PEXPL * &
- BR_P2( ZCOEFF(:,:,IKB+2*KKL)*PTHLM(:,:,IKB+2*KKL) &
- +ZCOEFF(:,:,IKB+KKL )*PTHLM(:,:,IKB+KKL ) &
- +ZCOEFF(:,:,IKB )*PTHLM(:,:,IKB ) ) &
+ BR_P2( ZCOEFF(JI,JJ,IKB+2*KKL)*PTHLM(JI,JJ,IKB+2*KKL) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PTHLM(JI,JJ,IKB+KKL ) &
+ +ZCOEFF(JI,JJ,IKB )*PTHLM(JI,JJ,IKB ) ) &
+PIMPL * &
- BR_P2( ZCOEFF(:,:,IKB+2*KKL)*PTHLP(:,:,IKB+2*KKL) &
- +ZCOEFF(:,:,IKB+KKL )*PTHLP(:,:,IKB+KKL ) &
- +ZCOEFF(:,:,IKB )*PTHLP(:,:,IKB ) ) &
- )
+ BR_P2( ZCOEFF(JI,JJ,IKB+2*KKL)*PTHLP(JI,JJ,IKB+2*KKL) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PTHLP(JI,JJ,IKB+KKL ) &
+ +ZCOEFF(JI,JJ,IKB )*PTHLP(JI,JJ,IKB ) ) &
+ )
+ END DO
#endif
!
ZFLXZ(:,:,KKA) = ZFLXZ(:,:,IKB)
@@ -843,6 +852,7 @@ END IF
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:),ZTMP2_DEVICE(:,:,:))
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZF (JI,JJ,JK) = XCTV*PLM(JI,JJ,JK)*PLEPS(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -1031,6 +1041,7 @@ END DO
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:),ZTMP3_DEVICE(:,:,:))
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE(:,:,:),ZTMP4_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP4_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
@@ -1053,6 +1064,7 @@ END DO
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP8_DEVICE(:,:,:),ZTMP1_DEVICE(:,:,:))
!!!
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP7_DEVICE(JI,JJ,JK) = ( ZTMP3_DEVICE(JI,JJ,JK) + ZTMP4_DEVICE(JI,JJ,JK)) * PDR_DZ(JI,JJ,JK) &
* ZTMP2_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK) &
@@ -1068,6 +1080,7 @@ END DO
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:),ZTMP3_DEVICE(:,:,:))
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE(:,:,:),ZTMP4_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP4_DEVICE(JI,JJ,JK) /PDZZ(JI,JJ,JK)
@@ -1078,6 +1091,7 @@ END DO
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:), ZTMP4_DEVICE(:,:,:) )
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE(:,:,:), ZTMP5_DEVICE(:,:,:) )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = ZF(JI,JJ,JK) &
+ PIMPL * XCTV*PLM(JI,JJ,JK)*PLEPS(JI,JJ,JK)*0.5 * ZTMP3_DEVICE(JI,JJ,JK) &
@@ -1087,23 +1101,26 @@ END DO
#endif
!
! special case near the ground ( uncentred gradient )
- ZFLXZ(:,:,IKB) = &
- (XCHT1 * PPHI3(:,:,IKB+KKL) + XCHT2 * PPSI3(:,:,IKB+KKL)) &
+ !$acc loop independent collapse(2)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZFLXZ(JI,JJ,IKB) = &
+ (XCHT1 * PPHI3(JI,JJ,IKB+KKL) + XCHT2 * PPSI3(JI,JJ,IKB+KKL)) &
*( PEXPL * &
- ( ZCOEFF(:,:,IKB+2*KKL)*PTHLM(:,:,IKB+2*KKL) &
- +ZCOEFF(:,:,IKB+KKL )*PTHLM(:,:,IKB+KKL ) &
- +ZCOEFF(:,:,IKB )*PTHLM(:,:,IKB )) &
- *( ZCOEFF(:,:,IKB+2*KKL)*PRM(:,:,IKB+2*KKL,1) &
- +ZCOEFF(:,:,IKB+KKL )*PRM(:,:,IKB+KKL,1 ) &
- +ZCOEFF(:,:,IKB )*PRM(:,:,IKB ,1 )) &
+ ( ZCOEFF(JI,JJ,IKB+2*KKL)*PTHLM(JI,JJ,IKB+2*KKL) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PTHLM(JI,JJ,IKB+KKL ) &
+ +ZCOEFF(JI,JJ,IKB )*PTHLM(JI,JJ,IKB )) &
+ *( ZCOEFF(JI,JJ,IKB+2*KKL)*PRM(JI,JJ,IKB+2*KKL,1) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PRM(JI,JJ,IKB+KKL,1 ) &
+ +ZCOEFF(JI,JJ,IKB )*PRM(JI,JJ,IKB ,1 )) &
+PIMPL * &
- ( ZCOEFF(:,:,IKB+2*KKL)*PTHLP(:,:,IKB+2*KKL) &
- +ZCOEFF(:,:,IKB+KKL )*PTHLP(:,:,IKB+KKL ) &
- +ZCOEFF(:,:,IKB )*PTHLP(:,:,IKB )) &
- *( ZCOEFF(:,:,IKB+2*KKL)*PRP(:,:,IKB+2*KKL ) &
- +ZCOEFF(:,:,IKB+KKL )*PRP(:,:,IKB+KKL ) &
- +ZCOEFF(:,:,IKB )*PRP(:,:,IKB )) &
- )
+ ( ZCOEFF(JI,JJ,IKB+2*KKL)*PTHLP(JI,JJ,IKB+2*KKL) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PTHLP(JI,JJ,IKB+KKL ) &
+ +ZCOEFF(JI,JJ,IKB )*PTHLP(JI,JJ,IKB )) &
+ *( ZCOEFF(JI,JJ,IKB+2*KKL)*PRP(JI,JJ,IKB+2*KKL ) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PRP(JI,JJ,IKB+KKL ) &
+ +ZCOEFF(JI,JJ,IKB )*PRP(JI,JJ,IKB )) &
+ )
+ END DO
!
ZFLXZ(:,:,KKA) = ZFLXZ(:,:,IKB)
!
@@ -1199,11 +1216,15 @@ END DO
#ifndef MNH_BITREP
ZTMP1_DEVICE(:,:,:) = PPSI3(:,:,:)*PDR_DZ(:,:,:)**2
#else
- ZTMP1_DEVICE(:,:,:) = PPSI3(:,:,:)*BR_P2(PDR_DZ(:,:,:))
+ !$acc loop independent collapse(3)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ ZTMP1_DEVICE(JI,JJ,JK) = PPSI3(JI,JJ,JK)*BR_P2(PDR_DZ(JI,JJ,JK))
+ END DO
#endif
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE(:,:,:),ZTMP2_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZF (JI,JJ,JK) = XCTV*PLM(JI,JJ,JK)*PLEPS(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -1349,18 +1370,21 @@ END DO
!$acc end kernels
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE(:,:,:),ZTMP3_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE(:,:,:),ZTMP1_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP3_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL,ZTMP2_DEVICE(:,:,:),ZTMP3_DEVICE(:,:,:))
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = ZF(JI,JJ,JK) &
+ PIMPL * XCTV*PLM(JI,JJ,JK)*PLEPS(JI,JJ,JK) * ZTMP1_DEVICE(JI,JJ,JK) &
@@ -1382,17 +1406,20 @@ END DO
+ZCOEFF(:,:,IKB )*PRP(:,:,IKB ))**2 &
)
#else
- ZFLXZ(:,:,IKB) = XCHV * PPSI3(:,:,IKB+KKL) * PLM(:,:,IKB) &
- * PLEPS(:,:,IKB) &
+ !$acc loop independent collapse(2)
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU )
+ ZFLXZ(JI,JJ,IKB) = XCHV * PPSI3(JI,JJ,IKB+KKL) * PLM(JI,JJ,IKB) &
+ * PLEPS(JI,JJ,IKB) &
*( PEXPL * &
- BR_P2( ZCOEFF(:,:,IKB+2*KKL)*PRM(:,:,IKB+2*KKL,1) &
- +ZCOEFF(:,:,IKB+KKL )*PRM(:,:,IKB+KKL,1 ) &
- +ZCOEFF(:,:,IKB )*PRM(:,:,IKB ,1 )) &
+ BR_P2( ZCOEFF(JI,JJ,IKB+2*KKL)*PRM(JI,JJ,IKB+2*KKL,1) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PRM(JI,JJ,IKB+KKL,1 ) &
+ +ZCOEFF(JI,JJ,IKB )*PRM(JI,JJ,IKB ,1 )) &
+PIMPL * &
- BR_P2( ZCOEFF(:,:,IKB+2*KKL)*PRP(:,:,IKB+2*KKL) &
- +ZCOEFF(:,:,IKB+KKL )*PRP(:,:,IKB+KKL ) &
- +ZCOEFF(:,:,IKB )*PRP(:,:,IKB )) &
- )
+ BR_P2( ZCOEFF(JI,JJ,IKB+2*KKL)*PRP(JI,JJ,IKB+2*KKL) &
+ +ZCOEFF(JI,JJ,IKB+KKL )*PRP(JI,JJ,IKB+KKL ) &
+ +ZCOEFF(JI,JJ,IKB )*PRP(JI,JJ,IKB )) &
+ )
+ END DO
#endif
!
ZFLXZ(:,:,KKA) = ZFLXZ(:,:,IKB)
diff --git a/src/MNH/turb_ver_thermo_flux.f90 b/src/MNH/turb_ver_thermo_flux.f90
index a7f75d118..9d7775365 100644
--- a/src/MNH/turb_ver_thermo_flux.f90
+++ b/src/MNH/turb_ver_thermo_flux.f90
@@ -634,6 +634,7 @@ ZDFDDTDZ(:,:,:) = -XCSHF*ZKEFF(:,:,:)*D_PHI3DTDZ_O_DDTDZ(PPHI3,PREDTH1,PREDR1,PR
#else
CALL DZM_DEVICE(KKA,KKU,KKL,PTHLM,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZF (JI,JJ,JK) = -XCSHF*PPHI3(JI,JJ,JK)*ZKEFF(JI,JJ,JK)*ZTMP1_DEVICE(JI,JJ,JK)/PDZZ(JI,JJ,JK)
END DO
@@ -641,6 +642,7 @@ END DO
!
CALL D_PHI3DTDZ_O_DDTDZ(PPHI3,PREDTH1,PREDR1,PRED2TH3,PRED2THR3,HTURBDIM,GUSERV,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZDFDDTDZ(JI,JJ,JK) = -XCSHF*ZKEFF(JI,JJ,JK)*ZTMP2_DEVICE(JI,JJ,JK)
END DO
@@ -815,6 +817,7 @@ ZTMP1_DEVICE(:,:,:) = PTHLP(:,:,:) - PTHLM(:,:,:)
!$acc end kernels
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = ZF(JI,JJ,JK) + PIMPL * ZDFDDTDZ(JI,JJ,JK) * ZTMP2_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
END DO
@@ -862,6 +865,7 @@ END IF
IF (KRR /= 0) THEN
CALL MZM_DEVICE(PETHETA,ZTMP1_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLXZ(JI,JJ,JK)
END DO
@@ -916,6 +920,7 @@ IF ( KRRL >= 1 ) THEN
!$acc end kernels
CALL MZM_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE )
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)*ZFLXZ(JI,JJ,JK)/PDZZ(JI,JJ,JK)
END DO
@@ -931,6 +936,7 @@ END DO
!$acc end kernels
CALL MZM_DEVICE(ZTMP1_DEVICE,ZTMP2_DEVICE )
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP1_DEVICE(JI,JJ,JK) = ZTMP2_DEVICE(JI,JJ,JK)*ZFLXZ(JI,JJ,JK)/PDZZ(JI,JJ,JK)
END DO
@@ -1007,8 +1013,12 @@ IF (LLES_CALL) THEN
END IF
!* diagnostic of mixing coefficient for heat
CALL DZM_DEVICE(KKA,KKU,KKL,PTHLP,ZA)
-!$acc kernels
- WHERE (ZA(:,:,:)==0.) ZA(:,:,:)=1.E-6
+ !$acc kernels
+ DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
+ IF (ZA(JI,JJ,JK)==0.) THEN
+ ZA(JI,JJ,JK)=1.E-6
+ END IF
+ END DO
ZA(:,:,:) = - ZFLXZ(:,:,:) / ZA(:,:,:) * PDZZ(:,:,:)
ZA(:,:,IKB) = XCSHF*PPHI3(:,:,IKB)*ZKEFF(:,:,IKB)
!$acc end kernels
@@ -1047,6 +1057,7 @@ IF (KRR /= 0) THEN
#else
CALL DZM_DEVICE(KKA,KKU,KKL,PRM(:,:,:,1),ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZF (JI,JJ,JK) = -XCSHF*PPSI3(JI,JJ,JK)*ZKEFF(JI,JJ,JK)*ZTMP1_DEVICE(JI,JJ,JK)/PDZZ(JI,JJ,JK)
END DO
@@ -1054,6 +1065,7 @@ IF (KRR /= 0) THEN
CALL D_PSI3DRDZ_O_DDRDZ(PPSI3,PREDR1,PREDTH1,PRED2R3,PRED2THR3,HTURBDIM,GUSERV,ZTMP1_DEVICE)
!CALL D_PHI3DRDZ_O_DDRDZ_DEVICE(PPSI3,PREDR1,PREDTH1,PRED2R3,PRED2THR3,HTURBDIM,GUSERV,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZDFDDRDZ(JI,JJ,JK) = -XCSHF*ZKEFF(JI,JJ,JK)*ZTMP1_DEVICE(JI,JJ,JK)
END DO
@@ -1227,6 +1239,7 @@ IF (KRR /= 0) THEN
!$acc end kernels
CALL DZM_DEVICE(KKA,KKU,KKL,ZTMP1_DEVICE,ZTMP2_DEVICE)
!$acc kernels
+!$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZFLXZ(JI,JJ,JK) = ZF(JI,JJ,JK) + PIMPL * ZDFDDRDZ(JI,JJ,JK) *ZTMP2_DEVICE(JI,JJ,JK) / PDZZ(JI,JJ,JK)
END DO
@@ -1268,12 +1281,14 @@ END DO
#else
CALL MZM_DEVICE(PEMOIST,ZTMP1_DEVICE)
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZTMP2_DEVICE(JI,JJ,JK) = ZTMP1_DEVICE(JI,JJ,JK) * ZFLXZ(JI,JJ,JK)
END DO
!$acc end kernels
CALL MZF_DEVICE(KKA,KKU,KKL, ZTMP2_DEVICE, ZTMP3_DEVICE )
!$acc kernels
+ !$acc loop independent collapse(3)
DO CONCURRENT ( JI=1:JIU,JJ=1:JJU,JK=1:JKU)
ZA(JI,JJ,JK) = PBETA(JI,JJ,JK) * ZTMP3_DEVICE(JI,JJ,JK)
END DO
diff --git a/src/Makefile.MESONH.mk b/src/Makefile.MESONH.mk
index f7b5994ea..c5ca2e753 100644
--- a/src/Makefile.MESONH.mk
+++ b/src/Makefile.MESONH.mk
@@ -490,7 +490,7 @@ ifeq "$(VER_CDF)" "CDFAUTO"
DIR_CDFC?=${SRC_MESONH}/src/LIB/netcdf-${VERSION_CDFC}
DIR_CDFCXX?=${SRC_MESONH}/src/LIB/netcdf-cxx-${VERSION_CDFCXX}
DIR_CDFF?=${SRC_MESONH}/src/LIB/netcdf-fortran-${VERSION_CDFF}
-CDF_PATH?=${SRC_MESONH}/src/LIB/netcdf-${ARCH}-R${MNH_REAL}I${MNH_INT}
+CDF_PATH?=${OBJDIR_MASTER}/NETCDF-${VERSION_CDFF}
CDF_MOD?=${CDF_PATH}/include/netcdf.mod
#
INC_NETCDF ?= -I${CDF_PATH}/include
--
GitLab