[gs-bugs] [Bug 690250] New: Performance slower than other PCL s/w

[bugs.ghostscript.com-bugzilla-daemon at ghostscript.com]

http://bugs.ghostscript.com/show_bug.cgi?id=690250

           Summary: Performance slower than other PCL s/w
           Product: GhostPCL
           Version: trunk
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P4
         Component: PCL raster
        AssignedTo: henry.stiles at artifex.com
        ReportedBy: ray.johnston at artifex.com
         QAContact: gs-bugs at ghostscript.com


This PCL-XL file is claimed by the customer to be 5 times slower than another
vendor's PCL, both on a PC and on a target embedded RIP.

They are running page buffer mode, presumably at 600 dpi (I have asked to make
sure) to a mono 1-bit device.

I ran this with the trunk as of 9387 and saw 400 sec at 1200 dpi. Profiling
showed the time spent in 'pl_mem_node_remove'. When I disabled this with
#define PL_KEEP_GLOBAL_FREE_LIST false in pl/plalloc.h, the time went down
to 29 sec.

Since the customer is probably running 600 dpi, I also repeated the timing
there with 9388 for a non-debug (release) build and a DEBUG build. At 600
dpi, the release build completed in 3.9 sec vs. 260 seconds for the DEBUG
build (where PL_KEEP_GLOBAL_FREE_LIST is 'true').

I will also attach the profile from rev 9388 (PL_KEEP_GLOBAL_FREE_LIST false).



------- You are receiving this mail because: -------
You are the QA contact for the bug, or are watching the QA contact.