From c60697970bf9ff7ad23ea3816b1ca321bf6d3bb7 Mon Sep 17 00:00:00 2001 From: Juan ESCOBAR <juan.escobar@aero.obs-mip.fr> Date: Wed, 30 Mar 2022 14:35:38 +0200 Subject: [PATCH] Juan 30/03/2022:ZSOLVER/dotprod.f90, Cray GPU Opt: exchange seq/do concuurent for better/single kernels optimization --- src/ZSOLVER/dotprod.f90 | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) diff --git a/src/ZSOLVER/dotprod.f90 b/src/ZSOLVER/dotprod.f90 index 48aa61fa0..e473801b8 100644 --- a/src/ZSOLVER/dotprod.f90 +++ b/src/ZSOLVER/dotprod.f90 @@ -184,20 +184,16 @@ CALL MNH_MEM_POSITION_PIN() CALL MNH_MEM_GET(ZDOTPROD, ILBXB,ILBXE ,ILBYB,ILBYE ) #endif !$acc kernels present(ZDOTPROD) -ZDOTPROD = 0. -!$acc loop seq -DO JK = IKB-1,IKE+1 +ZDOTPROD(:,:) = 0. #ifdef MNH_COMPILER_NVHPC !$acc loop independent collapse(2) #endif -!dir$ doconcurrent - DO JJ = ILBYB,ILBYE - !dir$ doconcurrent - DO JI = ILBXB,ILBXE + DO CONCURRENT (JI=ILBXB:ILBXE,JJ=ILBYB:ILBYE) + !$acc loop seq + DO JK = IKB-1,IKE+1 ZDOTPROD(JI,JJ) = ZDOTPROD(JI,JJ) + PA(JI,JJ,JK) * PB(JI,JJ,JK) END DO END DO -END DO !$acc end kernels !$acc update host(ZDOTPROD) PDOTPROD = SUM_DD_R2_ll(ZDOTPROD) -- GitLab