diff --git a/readme.txt b/readme.txt
index 5fbee71f4..59428015d 100644
--- a/readme.txt
+++ b/readme.txt
@@ -76,9 +76,10 @@ Win32-MinGW            - ChibiOS/RT simulator and demo into a WIN32 process,
 
 *** 0.6.7 ***
 - NEW: New chThdCreateFast() API, it is a simplified form of chThdCreate()
-       that allows even faster threads creation. The new API does not support
-       the "mode" and "arg" parameters (still available in the old API).
-- OPT: Removed an unrequired initialization from the chThdCreate().
+  that allows even faster threads creation. The new API does not support
+  the "mode" and "arg" parameters (still available in the old API).
+- OPT: Removed an unrequired initialization and made other small optimizations
+  to the chThdCreate().
 - OPT: Improvements to the test framework, now a virtual timer is used instead
   of software loops into the bechmarks in order to have more stable results.
 - Added the C++ wrapper entries to the documentation.
diff --git a/src/chthreads.c b/src/chthreads.c
index d12bced10..338a65d52 100644
--- a/src/chthreads.c
+++ b/src/chthreads.c
@@ -97,8 +97,8 @@ Thread *chThdCreate(tprio_t prio, tmode_t mode, void *workspace,
 #ifdef CH_USE_DEBUG
   memfill(workspace, wsize, MEM_FILL_PATTERN);
 #endif
-  init_thread(prio, mode, tp);
   SETUP_CONTEXT(workspace, wsize, pf, arg);
+  init_thread(prio, mode, tp);
 #ifdef CH_USE_RESUME
   if (tp->p_flags & P_SUSPENDED)
     tp->p_state = PRSUSPENDED;
@@ -137,8 +137,8 @@ Thread *chThdCreateFast(tprio_t prio, void *workspace,
 #ifdef CH_USE_DEBUG
   memfill(workspace, wsize, MEM_FILL_PATTERN);
 #endif
-  init_thread(prio, 0, tp);
   SETUP_CONTEXT(workspace, wsize, pf, NULL);
+  init_thread(prio, 0, tp);
   chSysLock();
   chSchWakeupS(tp, RDY_OK);
   chSysUnlock();