1! 2! Copyright (c) 2011-2018, NVIDIA CORPORATION. All rights reserved. 3! 4! Licensed under the Apache License, Version 2.0 (the "License"); 5! you may not use this file except in compliance with the License. 6! You may obtain a copy of the License at 7! 8! http://www.apache.org/licenses/LICENSE-2.0 9! 10! Unless required by applicable law or agreed to in writing, software 11! distributed under the License is distributed on an "AS IS" BASIS, 12! WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13! See the License for the specific language governing permissions and 14! limitations under the License. 15! 16 17 18#include "mmul_dir.h" 19 20 21subroutine ftn_gather_real4( a, lda, alpha, buffer, bufrows, bufcols ) 22 implicit none 23 24 integer*8 lda 25 real*4 :: a( lda,* ), alpha 26 integer :: bufrows, bufcols 27 integer i, j, ndx, ndxsave 28 real*4 :: buffer(bufrows * bufcols) 29 30 ! 31 ! This routine gathers the matrix into l1 chunks. The purpose is much as it 32 ! is for the transpose case, and works much like transpose_real8() 33 ! 34 ! What do the parameters mean? 35 ! buffer: buffer array 36 ! a: matrix to be gathered 37 ! bufcols: number of rows in matrix a to gather 38 ! bufrowss: number of cols in matrix a to gather 39 ! lda: number of rows in matrix a 40 ! Note that we don't care what the dimensions of a are. We assume that the 41 ! calling function has done this correctly 42 ! 43 44 ndx = 0 45 do j = 1, bufcols 46 do i = 1, bufrows 47 buffer( ndx + i ) = alpha * a( i, j ) 48 enddo 49 ndx = ndx + bufrows 50 enddo 51 ! write( *, * ) ( a(1, j ), j = 1, bufcols ) 52 ! write( *, * )( buffer( i ), i = 1, bufrows * bufcols ) 53 return 54end subroutine ftn_gather_real4 55