! Parallel Sparse BLAS GPU plugin ! (C) Copyright 2013 ! ! Salvatore Filippone ! Alessandro Fanfarillo ! ! Redistribution and use in source and binary forms, with or without ! modification, are permitted provided that the following conditions ! are met: ! 1. Redistributions of source code must retain the above copyright ! notice, this list of conditions and the following disclaimer. ! 2. Redistributions in binary form must reproduce the above copyright ! notice, this list of conditions, and the following disclaimer in the ! documentation and/or other materials provided with the distribution. ! 3. The name of the PSBLAS group or the names of its contributors may ! not be used to endorse or promote products derived from this ! software without specific written permission. ! ! THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS ! ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED ! TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR ! PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE PSBLAS GROUP OR ITS CONTRIBUTORS ! BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR ! CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF ! SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS ! INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN ! CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ! ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE ! POSSIBILITY OF SUCH DAMAGE. ! subroutine psb_s_cuda_cp_hlg_from_coo(a,b,info) use psb_base_mod use hlldev_mod use psb_vectordev_mod use psb_cuda_env_mod use psb_s_cuda_hlg_mat_mod, psb_protect_name => psb_s_cuda_cp_hlg_from_coo implicit none class(psb_s_cuda_hlg_sparse_mat), intent(inout) :: a class(psb_s_coo_sparse_mat), intent(in) :: b integer(psb_ipk_), intent(out) :: info !locals type(psb_s_coo_sparse_mat) :: tmp integer(psb_ipk_) :: debug_level, debug_unit, hksz integer(psb_ipk_), allocatable :: idisp(:) character(len=20) :: name='hll_from_coo' Integer(Psb_ipk_) :: nza, nr, i,j,irw, idl,err_act, nc, isz,irs integer(psb_ipk_) :: nzm, ir, ic, k, hk, mxrwl, noffs, kc integer(psb_ipk_), allocatable :: irn(:), ja(:), hko(:) real(psb_dpk_), allocatable :: val(:) logical, parameter :: debug=.false. info = psb_success_ debug_unit = psb_get_debug_unit() debug_level = psb_get_debug_level() hksz = max(1,psb_cuda_WarpSize()) if (b%is_by_rows()) then nr = b%get_nrows() nc = b%get_ncols() nza = b%get_nzeros() if (debug) write(0,*) 'Copying through GPU',nza call psi_compute_hckoff_from_coo(a,noffs,isz,hksz,idisp,b,info) if (info /=0) then write(0,*) ' Error from psi_compute_hckoff:',info, noffs,isz return end if if (debug)write(0,*) ' From psi_compute_hckoff:',noffs,isz,a%hkoffs(1:min(10,noffs+1)) if (c_associated(a%deviceMat)) then call freeHllDevice(a%deviceMat) endif info = FallochllDevice(a%deviceMat,hksz,nr,nza,isz,spgpu_type_double,1) if (info == 0) info = psi_CopyCooToHlg(nr,nc,nza, hksz,noffs,isz,& & a%irn,a%hkoffs,idisp,b%ja, b%val, a%deviceMat) call a%set_dev() else ! This is to guarantee tmp%is_by_rows() call b%cp_to_coo(tmp,info) call tmp%fix(info) nr = tmp%get_nrows() nc = tmp%get_ncols() nza = tmp%get_nzeros() if (debug) write(0,*) 'Copying through GPU' call psi_compute_hckoff_from_coo(a,noffs,isz,hksz,idisp,tmp,info) if (info /=0) then write(0,*) ' Error from psi_compute_hckoff:',info, noffs,isz return end if if (debug)write(0,*) ' From psi_compute_hckoff:',noffs,isz,a%hkoffs(1:min(10,noffs+1)) if (c_associated(a%deviceMat)) then call freeHllDevice(a%deviceMat) endif info = FallochllDevice(a%deviceMat,hksz,nr,nza,isz,spgpu_type_double,1) if (info == 0) info = psi_CopyCooToHlg(nr,nc,nza, hksz,noffs,isz,& & a%irn,a%hkoffs,idisp,tmp%ja, tmp%val, a%deviceMat) call tmp%free() call a%set_dev() end if if (info /= 0) goto 9999 return 9999 continue info = psb_err_alloc_dealloc_ return contains subroutine psi_compute_hckoff_from_coo(a,noffs,isz,hksz,idisp,b,info) use psb_base_mod use psi_ext_util_mod implicit none class(psb_s_hll_sparse_mat), intent(inout) :: a class(psb_s_coo_sparse_mat), intent(in) :: b integer(psb_ipk_), allocatable, intent(out) :: idisp(:) integer(psb_ipk_), intent(in) :: hksz integer(psb_ipk_), intent(out) :: info, noffs, isz !locals Integer(Psb_ipk_) :: nza, nr, i,j,irw, idl,err_act, nc, irs integer(psb_ipk_) :: nzm, ir, ic, k, hk, mxrwl, kc logical, parameter :: debug=.false. info = 0 nr = b%get_nrows() nc = b%get_ncols() nza = b%get_nzeros() ! If it is sorted then we can lessen memory impact a%psb_s_base_sparse_mat = b%psb_s_base_sparse_mat if (debug) write(0,*) 'Start compute hckoff_from_coo',nr,nc,nza ! First compute the number of nonzeros in each row. call psb_realloc(nr,a%irn,info) if (info == 0) call psb_realloc(nr+1,idisp,info) if (info /= 0) return a%irn = 0 if (debug) then do i=1, nza if ((1<=b%ia(i)).and.(b%ia(i)<= nr)) then a%irn(b%ia(i)) = a%irn(b%ia(i)) + 1 else write(0,*) 'Out of bouds IA ',i,b%ia(i),nr end if end do else do i=1, nza a%irn(b%ia(i)) = a%irn(b%ia(i)) + 1 end do end if a%nzt = nza ! Second. Figure out the block offsets. call a%set_hksz(hksz) noffs = (nr+hksz-1)/hksz call psb_realloc(noffs+1,a%hkoffs,info) if (debug) write(0,*) ' noffsets ',noffs,info if (info /= 0) return a%hkoffs(1) = 0 j=1 idisp(1) = 0 do i=1,nr,hksz ir = min(hksz,nr-i+1) mxrwl = a%irn(i) idisp(i+1) = idisp(i) + a%irn(i) do k=1,ir-1 idisp(i+k+1) = idisp(i+k) + a%irn(i+k) mxrwl = max(mxrwl,a%irn(i+k)) end do a%hkoffs(j+1) = a%hkoffs(j) + mxrwl*hksz j = j + 1 end do ! ! At this point a%hkoffs(noffs+1) contains the allocation ! size a%ja a%val. ! isz = a%hkoffs(noffs+1) !!$ write(*,*) 'End of psi_comput_hckoff ',info end subroutine psi_compute_hckoff_from_coo end subroutine psb_s_cuda_cp_hlg_from_coo