More work on multi-threaded version, fixed alignment issues in DMA.

BulletMultiThreaded shared implementation works now on both Win32 Threads and PS3 Cell SPU.

Extras/BulletMultiThreaded/SpuFakeDma.cpp

@@ -1,10 +1,54 @@
 
 #include "SpuFakeDma.h"
-
+#include <LinearMath/btScalar.h> //for btAssert
 //Disabling memcpy sometimes helps debugging DMA
 
 #define USE_MEMCPY 1
 
+
+///this unalignedDma should not be frequently used, only for small data. It handles alignment and performs check on size (<16 bytes)
+int	stallingUnalignedDmaSmallGet(void *ls, uint64_t ea, uint32_t size)
+{
+	btAssert(size<16);
+	ATTRIBUTE_ALIGNED16(char	tmpBuffer[32]);
+
+	char* mainMem = (char*)ea;
+	char* localStore = (char*)ls;
+	uint32_t i;
+	
+
+	///make sure last 4 bits are the same, for cellDmaSmallGet
+	uint32_t last4BitsOffset = ea & 0x0f;
+	char* tmpTarget = tmpBuffer + last4BitsOffset;
+#ifdef WIN32
+
+#ifdef USE_MEMCPY
+	memcpy(tmpTarget,mainMem,size);
+#else
+	for ( i=0;i<size;i++)
+	{
+		tmpTarget[i] = mainMem[i];
+	}
+#endif
+#else
+	cellDmaSmallGet(tmpTarget,ea,size,DMA_TAG(1),0,0);
+	//copy into final destination
+#endif
+
+	cellDmaWaitTagStatusAll(DMA_MASK(1));
+
+	//this is slowish, perhaps memcpy on SPU is smarter?
+	for (i=0;i<size;i++)
+	{
+		localStore[i] = tmpTarget[i];
+	}
+
+	return 0;
+}
+
+
+#ifdef WIN32
+
 int	cellDmaLargeGet(void *ls, uint64_t ea, uint32_t size, uint32_t tag, uint32_t tid, uint32_t rid)
 {
 	char* mainMem = (char*)ea;
@@ -58,3 +102,5 @@ void	cellDmaWaitTagStatusAll(int ignore)
 {
 
 }
+
+#endif //WIN32