diff --git a/winpr/include/winpr/bitstream.h b/winpr/include/winpr/bitstream.h
index 4f988b6e2..1d1fcf964 100644
--- a/winpr/include/winpr/bitstream.h
+++ b/winpr/include/winpr/bitstream.h
@@ -46,15 +46,88 @@ typedef struct _wBitStream wBitStream;
 extern "C" {
 #endif
 
-WINPR_API void BitStream_Attach(wBitStream* bs, BYTE* buffer, UINT32 capacity);
-WINPR_API void BitStream_Write_Bits(wBitStream* bs, UINT32 bits, UINT32 nbits);
-WINPR_API void BitStream_Flush(wBitStream* bs);
-WINPR_API void BitStream_Prefetch(wBitStream* bs);
-WINPR_API void BitStream_Fetch(wBitStream* bs);
-WINPR_API void BitStream_Shift(wBitStream* bs, UINT32 nbits);
+#define BitStream_Prefetch(_bs) do { \
+	(_bs->prefetch) = 0; \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 4)) \
+		(_bs->prefetch) |= (*(_bs->pointer + 4) << 24); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 5)) \
+		(_bs->prefetch) |= (*(_bs->pointer + 5) << 16); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 6)) \
+		(_bs->prefetch) |= (*(_bs->pointer + 6) << 8); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 7)) \
+		(_bs->prefetch) |= (*(_bs->pointer + 7) << 0); \
+} while(0)
+
+#define BitStream_Fetch(_bs) do { \
+	(_bs->accumulator) = 0; \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 0)) \
+		(_bs->accumulator) |= (*(_bs->pointer + 0) << 24); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 1)) \
+		(_bs->accumulator) |= (*(_bs->pointer + 1) << 16); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 2)) \
+		(_bs->accumulator) |= (*(_bs->pointer + 2) << 8); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 3)) \
+		(_bs->accumulator) |= (*(_bs->pointer + 3) << 0); \
+	BitStream_Prefetch(_bs); \
+} while(0)
+
+#define BitStream_Flush(_bs) do { \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 0)) \
+		*(_bs->pointer + 0) = (_bs->accumulator >> 24); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 1)) \
+		*(_bs->pointer + 1) = (_bs->accumulator >> 16); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 2)) \
+		*(_bs->pointer + 2) = (_bs->accumulator >> 8); \
+	if ((_bs->pointer - _bs->buffer) < (_bs->capacity + 3)) \
+		*(_bs->pointer + 3) = (_bs->accumulator >> 0); \
+} while(0)
+
+#define BitStream_Shift(_bs, _nbits) do { \
+	_bs->accumulator <<= _nbits; \
+	_bs->position += _nbits; \
+	_bs->offset += _nbits; \
+	if (_bs->offset < 32) { \
+		_bs->mask = ((1 << _nbits) - 1); \
+		_bs->accumulator |= ((_bs->prefetch >> (32 - _nbits)) & _bs->mask); \
+		_bs->prefetch <<= _nbits; \
+	} else { \
+		_bs->mask = ((1 << _nbits) - 1); \
+		_bs->accumulator |= ((_bs->prefetch >> (32 - _nbits)) & _bs->mask); \
+		_bs->prefetch <<= _nbits; \
+		_bs->offset -= 32; \
+		_bs->pointer += 4; \
+		BitStream_Prefetch(_bs); \
+		if (_bs->offset) { \
+			_bs->mask = ((1 << _bs->offset) - 1); \
+			_bs->accumulator |= ((_bs->prefetch >> (32 - _bs->offset)) & _bs->mask); \
+			_bs->prefetch <<= _bs->offset; \
+		} \
+	} \
+} while(0)
+
+#define BitStream_Write_Bits(_bs, _bits, _nbits) do { \
+	_bs->position += _nbits; \
+	_bs->offset += _nbits; \
+	if (_bs->offset < 32) { \
+		_bs->accumulator |= (_bits << (32 - _bs->offset)); \
+	} else { \
+		_bs->offset -= 32; \
+		_bs->mask = ((1 << (_nbits - _bs->offset)) - 1); \
+		_bs->accumulator |= ((_bits >> _bs->offset) & _bs->mask); \
+		BitStream_Flush(bs); \
+		_bs->accumulator = 0; \
+		_bs->pointer += 4; \
+		if (_bs->offset) { \
+			_bs->mask = ((1 << _bs->offset) - 1); \
+			_bs->accumulator |= ((_bits & _bs->mask) << (32 - _bs->offset)); \
+		} \
+	} \
+} while(0)
 
 WINPR_API void BitDump(const BYTE* buffer, UINT32 length, UINT32 flags);
 
+WINPR_API void BitStream_Attach(wBitStream* bs, BYTE* buffer, UINT32 capacity);
+
 WINPR_API wBitStream* BitStream_New();
 WINPR_API void BitStream_Free(wBitStream* bs);
 
diff --git a/winpr/libwinpr/utils/collections/BitStream.c b/winpr/libwinpr/utils/collections/BitStream.c
index 2a1b2ccd7..7aa6ab5f5 100644
--- a/winpr/libwinpr/utils/collections/BitStream.c
+++ b/winpr/libwinpr/utils/collections/BitStream.c
@@ -183,28 +183,19 @@ void BitDump(const BYTE* buffer, UINT32 length, UINT32 flags)
 	printf("\n");
 }
 
-void BitStream_Attach(wBitStream* bs, BYTE* buffer, UINT32 capacity)
-{
-	bs->position = 0;
-	bs->buffer = buffer;
-	bs->offset = 0;
-	bs->accumulator = 0;
-	bs->pointer = bs->buffer;
-	bs->capacity = capacity;
-	bs->length = bs->capacity * 8;
-}
+#if 0
 
-void BitStream_Flush(wBitStream* bs)
-{
-	if ((bs->pointer - bs->buffer) < (bs->capacity + 0))
-		*(bs->pointer + 0) = (bs->accumulator >> 24);
-	if ((bs->pointer - bs->buffer) < (bs->capacity + 1))
-		*(bs->pointer + 1) = (bs->accumulator >> 16);
-	if ((bs->pointer - bs->buffer) < (bs->capacity + 2))
-		*(bs->pointer + 2) = (bs->accumulator >> 8);
-	if ((bs->pointer - bs->buffer) < (bs->capacity + 3))
-		*(bs->pointer + 3) = (bs->accumulator >> 0);
-}
+/**
+ * These are the original functions from which the macros are derived.
+ * Since it is much easier to develop and debug functions than macros,
+ * we keep a copy here for later improvements and modifications.
+ */
+
+WINPR_API void BitStream_Prefetch(wBitStream* bs);
+WINPR_API void BitStream_Fetch(wBitStream* bs);
+WINPR_API void BitStream_Flush(wBitStream* bs);
+WINPR_API void BitStream_Shift(wBitStream* bs, UINT32 nbits);
+WINPR_API void BitStream_Write_Bits(wBitStream* bs, UINT32 bits, UINT32 nbits);
 
 void BitStream_Prefetch(wBitStream* bs)
 {
@@ -235,31 +226,16 @@ void BitStream_Fetch(wBitStream* bs)
 	BitStream_Prefetch(bs);
 }
 
-void BitStream_Write_Bits(wBitStream* bs, UINT32 bits, UINT32 nbits)
+void BitStream_Flush(wBitStream* bs)
 {
-	bs->position += nbits;
-	bs->offset += nbits;
-
-	if (bs->offset < 32)
-	{
-		bs->accumulator |= (bits << (32 - bs->offset));
-	}
-	else
-	{
-		bs->offset -= 32;
-		bs->mask = ((1 << (nbits - bs->offset)) - 1);
-		bs->accumulator |= ((bits >> bs->offset) & bs->mask);
-
-		BitStream_Flush(bs);
-		bs->accumulator = 0;
-		bs->pointer += 4;
-
-		if (bs->offset)
-		{
-			bs->mask = ((1 << bs->offset) - 1);
-			bs->accumulator |= ((bits & bs->mask) << (32 - bs->offset));
-		}
-	}
+	if ((bs->pointer - bs->buffer) < (bs->capacity + 0))
+		*(bs->pointer + 0) = (bs->accumulator >> 24);
+	if ((bs->pointer - bs->buffer) < (bs->capacity + 1))
+		*(bs->pointer + 1) = (bs->accumulator >> 16);
+	if ((bs->pointer - bs->buffer) < (bs->capacity + 2))
+		*(bs->pointer + 2) = (bs->accumulator >> 8);
+	if ((bs->pointer - bs->buffer) < (bs->capacity + 3))
+		*(bs->pointer + 3) = (bs->accumulator >> 0);
 }
 
 void BitStream_Shift(wBitStream* bs, UINT32 nbits)
@@ -294,6 +270,46 @@ void BitStream_Shift(wBitStream* bs, UINT32 nbits)
 	}
 }
 
+void BitStream_Write_Bits(wBitStream* bs, UINT32 bits, UINT32 nbits)
+{
+	bs->position += nbits;
+	bs->offset += nbits;
+
+	if (bs->offset < 32)
+	{
+		bs->accumulator |= (bits << (32 - bs->offset));
+	}
+	else
+	{
+		bs->offset -= 32;
+		bs->mask = ((1 << (nbits - bs->offset)) - 1);
+		bs->accumulator |= ((bits >> bs->offset) & bs->mask);
+
+		BitStream_Flush(bs);
+		bs->accumulator = 0;
+		bs->pointer += 4;
+
+		if (bs->offset)
+		{
+			bs->mask = ((1 << bs->offset) - 1);
+			bs->accumulator |= ((bits & bs->mask) << (32 - bs->offset));
+		}
+	}
+}
+
+#endif
+
+void BitStream_Attach(wBitStream* bs, BYTE* buffer, UINT32 capacity)
+{
+	bs->position = 0;
+	bs->buffer = buffer;
+	bs->offset = 0;
+	bs->accumulator = 0;
+	bs->pointer = bs->buffer;
+	bs->capacity = capacity;
+	bs->length = bs->capacity * 8;
+}
+
 wBitStream* BitStream_New()
 {
 	wBitStream* bs = NULL;