aboutsummaryrefslogtreecommitdiffstats
path: root/readme.txt
diff options
context:
space:
mode:
authorgdisirio <gdisirio@35acf78f-673a-0410-8e92-d51de3d6d3f4>2007-11-16 13:14:30 +0000
committergdisirio <gdisirio@35acf78f-673a-0410-8e92-d51de3d6d3f4>2007-11-16 13:14:30 +0000
commitda365c95e4460bda1f5b49a2e55ae144da1faf3f (patch)
tree634ff1f5250e36e12c0ebe0707a4792760695cec /readme.txt
parent3fe3d0d6a9ba5b11ebd30f30d5dfcb600cea9e50 (diff)
downloadChibiOS-da365c95e4460bda1f5b49a2e55ae144da1faf3f.tar.gz
ChibiOS-da365c95e4460bda1f5b49a2e55ae144da1faf3f.tar.bz2
ChibiOS-da365c95e4460bda1f5b49a2e55ae144da1faf3f.zip
git-svn-id: svn://svn.code.sf.net/p/chibios/svn/trunk@93 35acf78f-673a-0410-8e92-d51de3d6d3f4
Diffstat (limited to 'readme.txt')
-rw-r--r--readme.txt15
1 files changed, 15 insertions, 0 deletions
diff --git a/readme.txt b/readme.txt
index 36255b42f..291924130 100644
--- a/readme.txt
+++ b/readme.txt
@@ -38,6 +38,21 @@ AVR-AT90CANx-GCC - Port on AVR AT90CAN128, not complete yet.
*** Releases ***
*****************************************************************************
+*** 0.4.1 ***
+- Cleaned up the LPC2148 demo in main.c, it is now well documented and
+ explains everything, I assumed too much stuff to be "obvious".
+- Included a Makefile in the LPC2148 demo that builds in THUMB mode.
+- Added a spreadsheet in the documentation that describes the advantages
+ and disadvantages of the various optimization options (both GCC options and
+ ChibiOS/RT options), very interesting read IMO. No .xls available, ODF only.
+- The GCC option +falign-functions=16 is now default in the Makefile, it is
+ required because of the MAM unit into the LPC chips, without this option
+ the code performance is less predictable and can change of some % points
+ depending on how the code is aligned in the flash memory, unpredictabilty
+ is bad for a RTOS. This option however increases the code size slightly
+ because of alignment gaps.
+- Fine tuning in the scheduler code for improved performance.
+
*** 0.4.0 ***
- NEW, added a benchmark functionality to the test suite. The benchmark
measures the kernel throughput as messages per second and context switches