Cortex-M3 optimization.

git-svn-id: svn://svn.code.sf.net/p/chibios/svn/trunk@834 35acf78f-673a-0410-8e92-d51de3d6d3f4

docs/reports/STM32F103-72.txt

@@ -27,71 +27,74 @@ Settings: SYSCLK=72, ACR=0x12 (2 wait states)
 --- Test Case 3.3 (Mutexes, priority inheritance, complex case)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 4.1 (CondVar, signal test)
+--- Test Case 3.4 (CondVar, signal test)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 4.2 (CondVar, broadcast test)
+--- Test Case 3.5 (CondVar, broadcast test)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 5.1 (Messages, dispatch test)
+--- Test Case 3.6 (CondVar, inheritance boost test)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 6.1 (Mailboxes, queuing and timeouts)
+--- Test Case 4.1 (Messages, dispatch test)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 7.1 (Events, wait and broadcast)
+--- Test Case 5.1 (Mailboxes, queuing and timeouts)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 8.1 (Heap, allocation and fragmentation test)
+--- Test Case 6.1 (Events, wait and broadcast)
---- Size  : 17220 bytes, not fragmented
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 9.1 (Memory Pools, allocation and enqueuing test)
+--- Test Case 7.1 (Heap, allocation and fragmentation test)
+--- Size  : 17236 bytes, not fragmented
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 10.1 (Dynamic APIs, threads creation from heap)
+--- Test Case 8.1 (Memory Pools, allocation and enqueuing test)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 10.2 (Dynamic APIs, threads creation from memory pool)
+--- Test Case 9.1 (Dynamic APIs, threads creation from heap)
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.1 (Benchmark, context switch #1, optimal)
+--- Test Case 9.2 (Dynamic APIs, threads creation from memory pool)
+--- Result: SUCCESS
+---------------------------------------------------------------------------
+--- Test Case 10.1 (Benchmark, context switch #1, optimal)
 --- Score : 216992 msgs/S, 433984 ctxswc/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.2 (Benchmark, context switch #2, empty ready list)
+--- Test Case 10.2 (Benchmark, context switch #2, empty ready list)
---- Score : 178664 msgs/S, 357328 ctxswc/S
---- Result: SUCCESS
----------------------------------------------------------------------------
---- Test Case 11.3 (Benchmark, context switch #3, 4 threads in ready list)
 --- Score : 178663 msgs/S, 357326 ctxswc/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.4 (Benchmark, threads creation/termination, worst case)
+--- Test Case 10.3 (Benchmark, context switch #3, 4 threads in ready list)
---- Score : 164735 threads/S
+--- Score : 178663 msgs/S, 357326 ctxswc/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.5 (Benchmark, threads creation/termination, optimal)
+--- Test Case 10.4 (Benchmark, threads creation/termination, worst case)
---- Score : 210633 threads/S
+--- Score : 168598 threads/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.6 (Benchmark, mass reschedulation, 5 threads)
+--- Test Case 10.5 (Benchmark, threads creation/termination, optimal)
+--- Score : 216994 threads/S
+--- Result: SUCCESS
+---------------------------------------------------------------------------
+--- Test Case 10.6 (Benchmark, mass reschedulation, 5 threads)
 --- Score : 55551 reschedulations/S, 333306 ctxswc/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.7 (Benchmark, I/O Queues throughput)
+--- Test Case 10.7 (Benchmark, I/O Queues throughput)
 --- Score : 489472 bytes/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.8 (Benchmark, virtual timers set/reset)
+--- Test Case 10.8 (Benchmark, virtual timers set/reset)
 --- Score : 647110 timers/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.9 (Benchmark, semaphores wait/signal)
+--- Test Case 10.9 (Benchmark, semaphores wait/signal)
 --- Score : 823324 wait+signal/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------
---- Test Case 11.10 (Benchmark, mutexes lock/unlock)
+--- Test Case 10.10 (Benchmark, mutexes lock/unlock)
 --- Score : 601124 lock+unlock/S
 --- Result: SUCCESS
 ---------------------------------------------------------------------------

ports/ARMCM3/chcore.c

@@ -42,19 +42,6 @@ void port_halt(void) {
   }
 }
 
-/**
- * Start a thread by invoking its work function.
- * If the work function returns @p chThdExit() is automatically invoked.
- */
-/** @cond never */
-__attribute__((naked, weak))
-/** @endcond */
-void threadstart(void) {
-
-  asm volatile ("blx     r1                                     \n\t" \
-                "bl      chThdExit");
-}
-
 /**
  * System Timer vector.
  * This interrupt is used as system tick.

ports/ARMCM3/chcore.h

@@ -155,8 +155,8 @@ struct context {
   tp->p_ctx.r13->basepri = BASEPRI_USER;                                \
   tp->p_ctx.r13->lr_exc = (regarm_t)0xFFFFFFFD;                         \
   tp->p_ctx.r13->r0 = arg;                                              \
-  tp->p_ctx.r13->r1 = pf;                                               \
+  tp->p_ctx.r13->lr_thd = chThdExit;                                    \
-  tp->p_ctx.r13->pc = threadstart;                                      \
+  tp->p_ctx.r13->pc = pf;                                               \
   tp->p_ctx.r13->xpsr = (regarm_t)0x01000000;                           \
 }
 
@@ -303,7 +303,6 @@ struct context {
 extern "C" {
 #endif
   void port_halt(void);
-  void threadstart(void);
 #ifdef __cplusplus
 }
 #endif

readme.txt

@@ -87,6 +87,8 @@ Win32-MinGW            - ChibiOS/RT simulator and demo into a WIN32 process,
   the macro chSysGetTime() in chTimeNow(), the old names are still recognized
   but marked as deprecated (fixes the bug 2678953 but goes a bit further by
   introducing a new API category "Time").
+- OPT: Small optimization to the Cortex-M3 port code, improved thread
+  related performance scores and smaller code.
 - Removed testcond.c|h and moved the test cases into testmtx.c. Mutexes and
   condvars have to be tested together.
 - Added architecture diagram to the documentation.