Skip to main content
Jump to: navigation, search

Difference between revisions of "DSDP/RTSC - Performance"

m
m (Build time)
Line 38: Line 38:
  
 
<br><br><br><br><br><br><br><br><br>
 
<br><br><br><br><br><br><br><br><br>
 +
 +
There was no difference between C64P COFF build time with 6.1.12 and 7.0.0.
 +
 
{| cellspacing="1" cellpadding="2" border="1" align="left"
 
{| cellspacing="1" cellpadding="2" border="1" align="left"
 
|+ different configurations, XDCtools 3.20.00.04, 'whole_program' profile
 
|+ different configurations, XDCtools 3.20.00.04, 'whole_program' profile

Revision as of 20:41, 17 February 2010

Benchmarks

Build time

These benchmarks measure the total time to build an application including the generation of configuration package (which only happens the first time a configuration is created).

BIOS memory example built with BIOS 6.21.00.13 for C64P and MSP430 targets, 'whole_program' profile
XDCtools target generating configPkg rebuilding configPkg big C compile partial link opt compile app build Total
3.16.02.32 C64P 3s 25s 13s 28s 7s 2s 78s
3.20.00.04 C64P 3s 17s 13s 28s 7s 2s 70s
3.20.00.04 MSP430 3s 18s 7s 11s 4s 2s 45s
3.20 with -O0 C64P 3s 17s 11s 28s 7s 2s 68s
3.20 with DirectCall C64P 3s 17s 10s 27s 7s 2s 66s
3.20 with DirectCall and -O0 C64P 3s 17s 8s 27s 6s 2s 63s



The last column includes not only -O0, but also -oi0 and --symdebug:none thrown in when compiling the big C config file.










There was no difference between C64P COFF build time with 6.1.12 and 7.0.0.

different configurations, XDCtools 3.20.00.04, 'whole_program' profile
app big C lines modules big C compile time link time
MSP430 local.apps.monitor 8000 18 3s 8s
C64P memory 25000 35 13s 28s
C64P rta 40000 57 24s 55s










Command line tool startup time

These benchmarks measure the time required to start up an java environment, print a message, and shutdown from the command line.

command line tools
XDCtools java tconf xs
3.20.00.04 0.41s 0.8s 1.3s





Various documents

whole_program build flow diagram

Back to the top