From 5bd07149d39ed3727e81d63d03924953594a2362 Mon Sep 17 00:00:00 2001 From: Remi Lehe Date: Wed, 21 Aug 2019 08:03:56 -0700 Subject: Cleanup PICSAR reference and fortran files --- Source/Particles/WarpXParticleContainer.cpp | 126 ---------------------------- 1 file changed, 126 deletions(-) (limited to 'Source/Particles/WarpXParticleContainer.cpp') diff --git a/Source/Particles/WarpXParticleContainer.cpp b/Source/Particles/WarpXParticleContainer.cpp index befa5cfed..e82f43ce3 100644 --- a/Source/Particles/WarpXParticleContainer.cpp +++ b/Source/Particles/WarpXParticleContainer.cpp @@ -282,132 +282,6 @@ WarpXParticleContainer::AddNParticles (int lev, Redistribute(); } -/* \brief Current Deposition for thread thread_num using PICSAR - * \param pti : Particle iterator - * \param wp : Array of particle weights - * \param uxp uyp uzp : Array of particle - * \param jx jy jz : Full array of current density - * \param offset : Index of first particle for which current is deposited - * \param np_to_depose: Number of particles for which current is deposited. - Particles [offset,offset+np_tp_depose] deposit their - current - * \param thread_num : Thread number (if tiling) - * \param lev : Level of box that contains particles - * \param depos_lev : Level on which particles deposit (if buffers are used) - * \param dt : Time step for particle level - */ -void -WarpXParticleContainer::DepositCurrentFortran(WarpXParIter& pti, - RealVector& wp, RealVector& uxp, - RealVector& uyp, RealVector& uzp, - MultiFab* jx, MultiFab* jy, MultiFab* jz, - const long offset, const long np_to_depose, - int thread_num, int lev, int depos_lev, - Real dt) -{ - AMREX_ALWAYS_ASSERT_WITH_MESSAGE((depos_lev==(lev-1)) || - (depos_lev==(lev )), - "Deposition buffers only work for lev-1"); - // If no particles, do not do anything - if (np_to_depose == 0) return; - - const long ngJ = jx->nGrow(); - const std::array& dx = WarpX::CellSize(std::max(depos_lev,0)); - const long lvect = 8; - int j_is_nodal = jx->is_nodal() and jy->is_nodal() and jz->is_nodal(); - - BL_PROFILE_VAR_NS("PICSAR::CurrentDeposition", blp_pxr_cd); - BL_PROFILE_VAR_NS("PPC::Evolve::Accumulate", blp_accumulate); - - // Get tile box where current is deposited. - // The tile box is different when depositing in the buffers (depos_lev& xyzmin_tile = WarpX::LowerCorner(tilebox, depos_lev); - const std::array& xyzmin = xyzmin_tile; - -#ifdef AMREX_USE_GPU - // No tiling on GPU: jx_ptr points to the full - // jx array (same for jy_ptr and jz_ptr). - Real* jx_ptr = (*jx)[pti].dataPtr(); - Real* jy_ptr = (*jy)[pti].dataPtr(); - Real* jz_ptr = (*jz)[pti].dataPtr(); - - auto jxntot = (*jx)[pti].length(); - auto jyntot = (*jy)[pti].length(); - auto jzntot = (*jz)[pti].length(); -#else - // Tiling is on: jx_ptr points to local_jx[thread_num] - // (same for jy_ptr and jz_ptr) - tbx.grow(ngJ); - tby.grow(ngJ); - tbz.grow(ngJ); - - local_jx[thread_num].resize(tbx); - local_jy[thread_num].resize(tby); - local_jz[thread_num].resize(tbz); - - Real* jx_ptr = local_jx[thread_num].dataPtr(); - Real* jy_ptr = local_jy[thread_num].dataPtr(); - Real* jz_ptr = local_jz[thread_num].dataPtr(); - - // local_jx[thread_num] is set to zero - local_jx[thread_num].setVal(0.0); - local_jy[thread_num].setVal(0.0); - local_jz[thread_num].setVal(0.0); - - auto jxntot = local_jx[thread_num].length(); - auto jyntot = local_jy[thread_num].length(); - auto jzntot = local_jz[thread_num].length(); -#endif - // GPU, no tiling: deposit directly in jx - // CPU, tiling: deposit into local_jx - // (same for jx and jz) - BL_PROFILE_VAR_START(blp_pxr_cd); - warpx_current_deposition( - jx_ptr, &ngJ, jxntot.getVect(), - jy_ptr, &ngJ, jyntot.getVect(), - jz_ptr, &ngJ, jzntot.getVect(), - &np_to_depose, - m_xp[thread_num].dataPtr() + offset, - m_yp[thread_num].dataPtr() + offset, - m_zp[thread_num].dataPtr() + offset, - uxp.dataPtr() + offset, - uyp.dataPtr() + offset, - uzp.dataPtr() + offset, - m_giv[thread_num].dataPtr() + offset, - wp.dataPtr() + offset, &this->charge, - &xyzmin[0], &xyzmin[1], &xyzmin[2], - &dt, &dx[0], &dx[1], &dx[2], - &WarpX::nox,&WarpX::noy,&WarpX::noz, &j_is_nodal, - &lvect,&WarpX::current_deposition_algo); - - BL_PROFILE_VAR_STOP(blp_pxr_cd); - -#ifndef AMREX_USE_GPU - BL_PROFILE_VAR_START(blp_accumulate); - // CPU, tiling: atomicAdd local_jx into jx - // (same for jx and jz) - (*jx)[pti].atomicAdd(local_jx[thread_num], tbx, tbx, 0, 0, 1); - (*jy)[pti].atomicAdd(local_jy[thread_num], tby, tby, 0, 0, 1); - (*jz)[pti].atomicAdd(local_jz[thread_num], tbz, tbz, 0, 0, 1); - BL_PROFILE_VAR_STOP(blp_accumulate); -#endif -} - /* \brief Current Deposition for thread thread_num * \param pti : Particle iterator * \param wp : Array of particle weights -- cgit v1.2.3 From f61b98426d5cdea5f9dea390bc57d8f281069e3e Mon Sep 17 00:00:00 2001 From: Remi Lehe Date: Wed, 21 Aug 2019 15:14:52 -0700 Subject: Add profilers around current deposition --- Source/Particles/WarpXParticleContainer.cpp | 4 ++++ Tools/performance_tests/functions_perftest.py | 6 +++--- 2 files changed, 7 insertions(+), 3 deletions(-) (limited to 'Source/Particles/WarpXParticleContainer.cpp') diff --git a/Source/Particles/WarpXParticleContainer.cpp b/Source/Particles/WarpXParticleContainer.cpp index b580bdb2c..646644f94 100644 --- a/Source/Particles/WarpXParticleContainer.cpp +++ b/Source/Particles/WarpXParticleContainer.cpp @@ -326,6 +326,8 @@ WarpXParticleContainer::DepositCurrent(WarpXParIter& pti, const Real stagger_shift = j_is_nodal ? 0.0 : 0.5; BL_PROFILE_VAR_NS("PPC::Evolve::Accumulate", blp_accumulate); + BL_PROFILE_VAR_NS("PPC::CurrentDeposition", blp_deposit); + // Get tile box where current is deposited. // The tile box is different when depositing in the buffers (depos_lev( @@ -426,6 +429,7 @@ WarpXParticleContainer::DepositCurrent(WarpXParIter& pti, stagger_shift, q); } } + BL_PROFILE_VAR_STOP(blp_deposit); #ifndef AMREX_USE_GPU BL_PROFILE_VAR_START(blp_accumulate); diff --git a/Tools/performance_tests/functions_perftest.py b/Tools/performance_tests/functions_perftest.py index 90e014f88..1a0767500 100644 --- a/Tools/performance_tests/functions_perftest.py +++ b/Tools/performance_tests/functions_perftest.py @@ -144,9 +144,9 @@ def read_run_perf(filename, n_steps): pattern_list = ['\nParticleContainer::Redistribute().*',\ '\nFabArray::FillBoundary().*',\ '\nFabArray::ParallelCopy().*',\ - '\nPICSAR::CurrentDeposition.*',\ - '\nPICSAR::FieldGather.*',\ - '\nPICSAR::ParticlePush.*',\ + '\nPPC::CurrentDeposition.*',\ + '\nPPC::FieldGather.*',\ + '\nPPC::ParticlePush.*',\ '\nPPC::Evolve::Copy.*',\ '\nWarpX::EvolveEM().*',\ 'Checkpoint().*',\ -- cgit v1.2.3