ShuriZma
/
pcsx2
mirror of https://github.com/PCSX2/pcsx2.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

vif JIT: increment based on sizeof(T) 

Struct on x64 will be 32B so +2 instead of +1
This commit is contained in:
Gregory Hainaut 2016-11-28 09:14:13 +01:00
parent 9862e5d207
commit 4c3e98754e
2 changed files with 6 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
pcsx2/x86/newVif.h
View File

@ -69,8 +69,7 @@ struct __aligned16 nVifBlock {
}; // 16 bytes
#define _hSize 0x4000 // [usn*1:mask*1:upk*4:num*8] hash...
#define _cmpS (sizeof(nVifBlock) - (4))
#define _tParams nVifBlock, _hSize, _cmpS
#define _tParams nVifBlock, _hSize
struct nVifStruct {
__aligned16 nVifBlock block;

15
pcsx2/x86/newVif_HashBucket.h
View File

@ -41,11 +41,10 @@ struct SizeChain
// to perform quick searches.
// T is a struct data type (note: size must be in multiples of 16 bytes!)
// hSize determines the number of buckets HashBucket will use for sorting.
// cmpSize is the size of data to consider 2 structs equal (see find())
// The hash function is determined by taking the first bytes of data and
// performing a modulus the size of hSize. So the most diverse-data should
// be in the first bytes of the struct. (hence why nVifBlock is specifically sorted)
template<typename T, int hSize, int cmpSize>
template<typename T, int hSize>
class HashBucket {
protected:
SizeChain<T> mBucket[hSize];
@ -68,18 +67,14 @@ public:
const __m128i* endpos = (__m128i*)&bucket.Chain[bucket.Size];
const __m128i data128( _mm_load_si128((__m128i*)dataPtr) );
#ifdef __x86_64__
pxAssertMsg(0, "Code is likely not compatible with 64 bits\n"
"Vif structure is 16 bytes in ia32 bits but contains an uptr to the x86 buffer. So 20 bytes in x64\n"
"Code below increments the iterator by 16 bytes, potentially we could put the x86 buffer in the first 2GB\n"
"Another improvement could be a port to std::unordered_map");
#endif
for( const __m128i* chainpos = (__m128i*)bucket.Chain; chainpos<endpos; chainpos+=sizeof(T) / 16u ) {
// Note SSE4/AVX optimization (However it requires to only have the key in the first 16B without the pointer)
// tmp = xor (data128, load(chainpos))
// ptest tmp tmp (zf will be set if tmp == 0, i.e equality)
for( const __m128i* chainpos = (__m128i*)bucket.Chain; chainpos<endpos; ++chainpos ) {
// This inline SSE code is generally faster than using emitter code, since it inlines nicely. --air
int result = _mm_movemask_ps( (cast_m128) _mm_cmpeq_epi32( data128, _mm_load_si128(chainpos) ) );
if( (result&0x7) == 0x7 ) return (T*)chainpos;
}
return NULL;
}