inline void wildCopy8(UInt8 * dst, const UInt8 * src, UInt8 * dst_end)
{
do
{
copy8(dst, src);
dst += 8;
src += 8;
} while (dst < dst_end);
}
inline void wildCopy8(UInt8 * dst, const UInt8 * src, UInt8 * dst_end)
{
/// Unrolling with clang is doing >10% performance degrade.
#if defined(__clang__)
#pragma nounroll
#endif
do
{
copy8(dst, src);
dst += 8;
src += 8;
} while (dst < dst_end);
}
void PspMemory::copy(byte *dst, const byte *src, uint32 bytes) {
DEBUG_ENTER_FUNC();
#ifdef TEST_MEMORY_COPY
uint32 debugBytes = bytes;
const byte *debugDst = dst, *debugSrc = src;
#endif
PSP_DEBUG_PRINT("copy(): dst[%p], src[%p], bytes[%d]\n", dst, src, bytes);
// align the destination pointer first
uint32 prefixDst = (((uint32)dst) & 0x3);
if (prefixDst) {
prefixDst = 4 - prefixDst; // prefix only if we have address % 4 != 0
PSP_DEBUG_PRINT("prefixDst[%d]\n", prefixDst);
bytes -= prefixDst; // remember we assume bytes >= 4
if (bytes < MIN_AMOUNT_FOR_COMPLEX_COPY) { // check if it's worthwhile to continue
copy8(dst, src, bytes + prefixDst);
#ifdef TEST_MEMORY_COPY
testCopy(debugDst, debugSrc, debugBytes);
#endif
return;
}
while (prefixDst--) {
*dst++ = *src++;
}
}
// check the source pointer alignment now
uint32 alignSrc = (((uint32)src) & 0x3);
if (alignSrc) { // we'll need to realign our reads
copy32Misaligned((uint32 *)dst, src, bytes, alignSrc);
} else {
copy32Aligned((uint32 *)dst, (uint32 *)src, bytes);
}
#ifdef TEST_MEMORY_COPY
testCopy(debugDst, debugSrc, debugBytes);
#endif
}
template <> void inline copy<8>(UInt8 * dst, const UInt8 * src) { copy8(dst, src); }