The observation was that, for cases 3 and 4, there was very little difference in time.  But, there was significant time save for case 1, compared to 2.  I can move this discussion away.  Thanks. 

Visually, I have not seen the performance from tic/toc numbers you showed.  (I did similar also but not shown in the example.)  patch was displayed once the command was issued.  line ran for some time dependent on n.  I don't know if tic/toc represents the time for graphic handling.

I have to run on software rendering whenever patch is present.  In the other patch bug report, I had indicated that crash or  plot window freeze can be an issue. Os is win10.   I also had difficulties in consistently reproducing these issues.  

When the number of patches is in the 50k-150k range, even with a single facecolor for all patches, 8.4.0 with hardware rendering crashed.  But crash is not what this report is for.  I just provide for your info only.  

I tried to find a workaround.  But it did not work.  Line is significantly slower than patch.  Thanks. 

This example shows the elapsed time to draw similar number of line segments using patch and line are very different. Patch is faster.  The performance of using all lines in a single line command does not show much improvement over drawing the many lines one-by-one. 

% line
n=5000; n=1000; % adjust as needed
tt=reshape(linspace(0, 2*pi, n), [],1); rr=100;
x=rr*sin(tt); y=rr*cos(tt); z=sin(tt);
x_ave=mean(x);y_ave=mean(y); z_ave=mean(z);

x=[x;x_ave]; y=[y;y_ave]; z=[z;z_ave];

figure('name', '1 All-in-one.  Really fast'); axs=axes; hold on; %plot3(x,y,z); hold on;
faces=[ (n+1)*ones(n-1,1) reshape([1:1:n-1],[],1)  reshape([2:1:n],[],1)];
vert =[[1:size(x,1)]'  x y z];
h_tri=patch('Vertices',     vert(:,2:end),'Faces', faces,'FaceColor',[1 0 0],'EdgeColor', [1 0 0],'FaceAlpha', 0.2, 'LineWidth', 1 ); whos h_tri % single handle

figure('name', '2 One-by-one.  Really slow, expected.'); axs=axes;  hold on; % plot3(x,y,z);
vert =[[1:size(x,1)]'  x y z];
for ii=1:n-1,
    faces=[ (n+1) ii  ii+1];
    h_tri=patch('Vertices',     vert(:,2:end),'Faces', faces,'FaceColor',[1 0 0],'EdgeColor', [1 0 0],'FaceAlpha', 0.2, 'LineWidth', 1 ); % mutiple handles
end; whos h_tri

% number of line segments is about 50% of patch.

figure('name', '3 all lines in one is slow, surprisingly.'); plot3(x,y,z); hold on;
xd=[repmat(x_ave, 1, size(x,1)); reshape(x,1,[])];yd=[repmat(y_ave, 1, size(x,1)); reshape(y,1,[])];zd=[repmat(z_ave, 1, size(x,1)); reshape(z,1,[])];
l=line(xd,yd,zd, 'color', 'r'); whos l  % multiple handles

figure('name', '4 separated lines is also slow, as expected.'); axs=axes; plot3(x,y,z); hold on;
for ii=1:n, ll(ii)=line( [x_ave;x(ii)], [y_ave; y(ii)], [z_ave; z(ii)], 'color', 'r'); end; whos ll  % multiple handles

confirming that for both examples in comment #0 and comment #1, Matlab 2023b produces a straight line plot while octave 9.0.91 produces an empty axis.  the patch function does return a handle to a patch object with all the pieces you'd expect to be there, it just doesn't render. updating release to stable.

Hi,

Octave patch does not plot line defined by two points.  Octave does not draw the patch and does not return an error.  The plot is silently empty. Thanks.

Replacing a single patch with a for-loop with many individual line commands is very slow, as you would expect.

% example

pts=[0 0 0;1 1 1]; tri=[1 2]
figure; patch('Vertices', pts,'Faces', tri, 'EdgeColor', 'k');