Add early-clobbers to the output variables in CPUID

Without those early-clobbers, the compiler might decide to schedule a
register that is also used as output. The existing early clobber in
the tmp variable was there so the compiler wouldn't use a register
scheduled as input (especially EAX).

To be honest, I'm not convinced that the compiler should be allowed to
do this. That means that two output variables are scheduled to the
same register... still, this fixes a problem found with GCC 4.2 (at
least the Mac one).

Change-Id: I6cd4676284e9a83d6aac4b439c6e58e347c40106
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>

diff --git a/src/corelib/tools/qsimd.cpp b/src/corelib/tools/qsimd.cpp
index f13009a..fb62192 100644 (file)
--- a/src/corelib/tools/qsimd.cpp
+++ b/src/corelib/tools/qsimd.cpp
@@ -168,7 +168,7 @@ static inline uint detectProcessorFeatures()
         asm ("xchg %%ebx, %2\n"
              "cpuid\n"
              "xchg %%ebx, %2\n"
-            : "=c" (feature_result), "=d" (result), "=&r" (tmp1)
+            : "=&c" (feature_result), "=d" (result), "=&r" (tmp1)
             : "a" (1));
 
         asm ("xchg %%ebx, %1\n"
@@ -182,7 +182,7 @@ static inline uint detectProcessorFeatures()
              "cpuid\n"
              "2:\n"
              "xchg %%ebx, %1\n"
-            : "=d" (extended_result), "=&r" (tmp1)
+            : "=&d" (extended_result), "=&r" (tmp1)
             : "a" (0x80000000)
             : "%ecx"
             );
@@ -284,7 +284,7 @@ static inline uint detectProcessorFeatures()
     asm ("xchg %%rbx, %1\n"
          "cpuid\n"
          "xchg %%rbx, %1\n"
-        : "=c" (feature_result), "=&r" (tmp)
+        : "=&c" (feature_result), "=&r" (tmp)
         : "a" (1)
         : "%edx"
         );