% Run the GraphBLAS demo2

% SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2021, All Rights Reserved.
% SPDX-License-Identifier: Apache-2.0

% reset to the default number of threads
clear all
ncores = demo_nproc ;
GrB.clear ;

gbdemo2
 GBDEMO2 Extreme performance differences: GraphBLAS vs built-in methods.
 
  Usage:
 
        gbdemo2             % uses a default bnz = 6000
        gbdemo2 (20000)     % uses bnz = 20000
 
  Many of the GraphBLAS operations used in gbdemo are perhaps 3x to
  50x faster than the corresponding MATLAB operations, depending on how
  many cores your computer has.  Here's an example where GraphBLAS is
  asymptotically far faster than MATLAB R2019a: a simple assignment
  for a large matrix C:
 
        C(I,J) = A
 
  The matrix C is constructed via C = kron (B,B) where nnz (B) is
  roughly the bnz provided on input (with a default of bnz = 6000),
  so that C will have about bnz^2 entries, or 36 million by default.
  I and J are chosen randomly, and A is 5000-by-5000.
 
  When the problem becomes large, MATLAB will take a very long time.
  If you have enough memory, and want to see higher speedups in
  GraphBLAS, increase bnz (and be prepared to wait even longer).
  With the default bnz = 6000, this test takes about 4GB of RAM.
 
  On my Dell XPS 4-core laptop (Intel(R) Core(TM) i7-8565U, 16GB
  RAM), using MATLAB R2019a, when C becomes 9 million by 9 million,
  the computation C(I,J)=A for MATLAB matrices C, I, J, and A takes
  several minutes, whereas GraphBLAS takes less than a second, or
  about 500x faster than MATLAB.  On a desktop with an Intel(R)
  Xeon(R) CPU E5-2698 v4 @ 2.20GHz with 20 hardware cores, the
  speedup over MATLAB is even more dramatic (up to 2,660x has been
  observed).
 
  See also GrB.assign, subsasgn.


# of threads used in GraphBLAS: 20


C(I,J)=A where C is 1 million -by- 1 million
with 35.7126 million entries:

    A is 5000-by-5000 with 49947 entries
    setup time:     0.415349 sec
    GraphBLAS time: 0.264042 sec
    Starting MATLAB ... please wait ... 
    MATLAB time:    0.472303 sec
    Speedup of GraphBLAS over MATLAB: 1.78874
    check time:     0.353221 sec
    all tests passed

C(I,J)=A where C is 4 million -by- 4 million
with 35.8202 million entries:

    A is 5000-by-5000 with 49947 entries
    setup time:     0.454911 sec
    GraphBLAS time: 0.209247 sec
    Starting MATLAB ... please wait ... 
    MATLAB time:    0.482239 sec
    Speedup of GraphBLAS over MATLAB: 2.30464
    check time:     0.239005 sec
    all tests passed

C(I,J)=A where C is 9 million -by- 9 million
with 35.928 million entries:

    A is 5000-by-5000 with 49947 entries
    setup time:     0.536653 sec
    GraphBLAS time: 0.271364 sec
    Starting MATLAB ... please wait ... 
    MATLAB time:    197.428 sec
    Speedup of GraphBLAS over MATLAB: 727.538
    check time:     0.280854 sec
    all tests passed

C(I,J)=A where C is 16 million -by- 16 million
with 35.916 million entries:

    A is 5000-by-5000 with 49947 entries
    setup time:     0.615692 sec
    GraphBLAS time: 0.328814 sec
    Starting MATLAB ... please wait ... 
    MATLAB time:    214.611 sec
    Speedup of GraphBLAS over MATLAB: 652.683
    check time:     0.302003 sec
    all tests passed

C(I,J)=A where C is 25 million -by- 25 million
with 35.964 million entries:

    A is 5000-by-5000 with 49947 entries
    setup time:     0.684777 sec
    GraphBLAS time: 0.295649 sec
    Starting MATLAB ... please wait ... 
    MATLAB time:    237.231 sec
    Speedup of GraphBLAS over MATLAB: 802.407
    check time:     0.340822 sec
    all tests passed

C(I,J)=A where C is 36 million -by- 36 million
with 35.976 million entries:

    A is 5000-by-5000 with 49947 entries
    setup time:     0.77091 sec
    GraphBLAS time: 0.458453 sec
    Starting MATLAB ... please wait ... 
    MATLAB time:    263.984 sec
    Speedup of GraphBLAS over MATLAB: 575.815
    check time:     0.386426 sec
    all tests passed