===================================================================
RCS file: /cvsrepo/anoncvs/cvs/www/53.html,v
retrieving revision 1.91
retrieving revision 1.92
diff -c -r1.91 -r1.92
*** www/53.html	2019/05/27 22:55:18	1.91
--- www/53.html	2019/06/01 23:12:46	1.92
***************
*** 225,231 ****
      <ul>
      <li>Switch default disk I/O sort algorithm to NSCAN for improved fairness and response times.
      <li>Softdep speedup improvements by the revert of a previously necessary workaround to prevent kva starvation.
!     <li>Large performance and reliability improvements in 
  <a href="https://man.openbsd.org/make.1">make(1)</a>,
  especially in parallel mode. make no longer does any busy waiting, it handles
  concurrent targets correctly, and displays more accurate error messages.
--- 225,231 ----
      <ul>
      <li>Switch default disk I/O sort algorithm to NSCAN for improved fairness and response times.
      <li>Softdep speedup improvements by the revert of a previously necessary workaround to prevent kva starvation.
!     <li>Large performance and reliability improvements in
  <a href="https://man.openbsd.org/make.1">make(1)</a>,
  especially in parallel mode. make no longer does any busy waiting, it handles
  concurrent targets correctly, and displays more accurate error messages.
***************
*** 393,406 ****
      </ul>
  <p>
  
! <li>Over 7,800 ports, major performance and stability improvements in 
  the package build process
      <ul>
      <li>The parallel ports builder is more efficient. The main improvement is
  that dpb consumes much less cpu on busy boxes, but there are lots of small
  optimizations that amount to a large performance increase:
! dpb can now build selected large ports using parallel make, and it 
! has a notion of affinity, so that ports failing on a cluster will be 
  preferentially restarted on the same machine.
      </ul>
  <p>
--- 393,406 ----
      </ul>
  <p>
  
! <li>Over 7,800 ports, major performance and stability improvements in
  the package build process
      <ul>
      <li>The parallel ports builder is more efficient. The main improvement is
  that dpb consumes much less cpu on busy boxes, but there are lots of small
  optimizations that amount to a large performance increase:
! dpb can now build selected large ports using parallel make, and it
! has a notion of affinity, so that ports failing on a cluster will be
  preferentially restarted on the same machine.
      </ul>
  <p>