Return to Support Page
 homesearchagentssupportask xilinxmap

Answers Database


PPR 5.0: Methods to reduce PPR runtime


Record #326

Problem Title:
PPR 5.0: Methods to reduce PPR runtime


Problem Description:



Solution 1:




Reducing PPR Run-Time

When you receive a software update, the two main benefits you expect are higher
chip performance and faster run time.  We feel that XACT 5.0 does very well
in meeting the first expectation of better chip performance.  In addition,
faster run time can also be seen over the entire range of designs in our test
suite. While the majority of designs completed sooner, unfortunately a small
number did take longer to process.

The PPR program can be run in many different situations, from a half full
2000 gate part to a 100% full 13,000 gate part.  Much effort was put into
determining the appropriate trade-off between run time and performance which
would give acceptable results for the majority of users.  But since everyone's
situation is different, we have made it easy to adjust the defaults to better
suit individual needs.

In addition, benchmarks and tests are being continuously performed to
determine a better set of defaults.  For XC4000 designs, we have discovered
that the following option can reduce run time by 20 to 30 percent, with a
minimal impact on design performance:

Copy the XACTINIT.DAT file found in your /XACT/DATA directory (or using a
text editor, create an XACTINIT.DAT file), and place it in your current
design directory and add the line


	permute_size = 4

When you rerun PPR, it will now automatically use this option.

To achieve an additional 5 to 7 percent speedup of run time, you can also add
the following lines to your local XACTINIT.DAT file:

	/ppr/improve_routing = false
	/ppr/router_effort = not_in_use

Also, if your 3KA or 4K design does not have stringent performance issues,
and routability is not a problem, you can change the settings to dramatically
reduce run-time:

	PPR <design_name> placer_effort=1 router_effort=1

However, these two changes (especially the last one) are NOT advisable for
general designs.  These setting may be useful for early exploration.  PPR will
complete much sooner, but you are likely to end up with a lower performance
design than if you had used the default options.

On the other end of the spectrum, if you have an XC3000A or XC4000 design
which has trouble routing or achieving desired performance, you should use
the following PPR command line options:

	PPR <design_name> placer_effort=5 router_effort=4

This forces PPR to work harder and produce a better result, but at the
expense of higher run time.


Some "green" PCs have power-down modes that cause the pc to enter a low-power
mode where the processor speed is reduced, the monitor is dimmed or shut off,
and other parts of the pc are placed in low-power modes. This usually via some
bios switches and also by TSRs. For example, a graphics card TSR may send a
command to the monitor asking it to shut down if the graphics data remains
stable for a long enough time.

It is a good idea to turn off power saving, by both removing the TSRs and
disabling power saving in the bios. Power saving may be accidentially
triggered during the periods when PPR is "crunching" and producing no screen
output, such as during the placement phase. This causes the cpu to go into a
low-power mode, which can cause PPR to take extremely longer than it should
to complete a place and route run.




End of Record #326

For the latest news, design tips, and patch information on the Xilinx design environment, check out the Xilinx Expert Journals!

© 1998 Xilinx, Inc. All rights reserved
Trademarks and Patents