memcpy vs memmove: memmove correctly handles overlapping source and destination memory buffers, but is slower than memcpy in many implementations. When memory buffers don't overlap, memcpy may be more efficient.

The DS2 SDK is such an implementation, so change many memmoves into memcpys.

1 files changed, 4 insertions, 4 deletions

diff --git a/source/loadzip.cpp b/source/loadzip.cpp
index 4ee3bcb..d3e24a1 100644
--- a/source/loadzip.cpp
+++ b/source/loadzip.cpp
@@ -132,8 +132,7 @@ bool8 LoadZip(const char* zipname,
 	char name[132];
 	unzGetCurrentFileInfo(file, &info, name,128, NULL,0, NULL,0);
 
-	int calc_size = info.uncompressed_size / 0x2000;
-	calc_size *= 0x2000;
+	int calc_size = info.uncompressed_size & ~0x1FFF; // round to lower 0x2000
 	if(!(info.uncompressed_size - calc_size == 512 || info.uncompressed_size == calc_size))
 	{
 	    port = unzGoToNextFile(file);
@@ -190,8 +189,7 @@ bool8 LoadZip(const char* zipname,
 //	assert(info.uncompressed_size <= CMemory::MAX_ROM_SIZE + 512);
 	int FileSize = info.uncompressed_size;
 	
-	int calc_size = FileSize / 0x2000;
-	calc_size *= 0x2000;
+	int calc_size = FileSize & ~0x1FFF; // round to lower 0x2000
 	
 	int l = unzReadCurrentFile(file,ptr,FileSize);
 	if(unzCloseCurrentFile(file) == UNZ_CRCERROR)
@@ -224,6 +222,8 @@ bool8 LoadZip(const char* zipname,
 	if ((FileSize - calc_size == 512 && !Settings.ForceNoHeader) ||
 	    Settings.ForceHeader)
 	{
+	    // memmove required: Overlapping addresses [Neb]
+	    // DS2 DMA notes: Can be split into 512-byte DMA blocks [Neb]
 	    memmove (ptr, ptr + 512, calc_size);
 	    (*headers)++;
 	    FileSize -= 512;