nintendo-emulators/desmume
1
0
Fork 0
mirror of https://github.com/TASEmulators/desmume synced 2025-10-05 16:22:49 +02:00
Code Issues Packages Projects Releases Wiki Activity

matrix.h: AltiVec accelerated functions buffer_copy_or_constant_*() and buffer_copy_or_constant_*_fast() now respect the NEEDENDIANSWAP switch.

Browse Source 
- Also do some code cleanup where appropriate.
This commit is contained in:
rogerman
2025-07-19 11:38:06 -07:00
parent e75b67043b
commit 39973c52bc
1 changed files with 94 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

126
desmume/src/matrix.h
View File

@@ -694,10 +694,11 @@ static void buffer_copy_or_constant_s32_fast(void *__restrict dst, const void *_
static void memset_u16(void *dst, const u16 val, const size_t elementCount)
{
CACHE_ALIGN const u16 val_u16 = val;
v128u16 *dst_vec128 = (v128u16 *)dst;
const size_t length_vec128 = elementCount / (sizeof(v128u16) / sizeof(u16));
v128u16 val_vec128 = (v128u16){val,val,val,val,val,val,val,val}; // Don't set as const, since vec_st() cannot store a const vector.
v128u16 val_vec128 = (v128u16)vec_splat( vec_lde(0, &val_u16), 0 ); // Don't set as const, since vec_st() cannot store a const vector.
for (size_t i = 0; i < length_vec128; i++)
vec_st(val_vec128, i*sizeof(v128u16), dst_vec128);
}
@@ -705,18 +706,20 @@ static void memset_u16(void *dst, const u16 val, const size_t elementCount)
template <size_t ELEMENTCOUNT>
static void memset_u16_fast(void *dst, const u16 val)
{
CACHE_ALIGN const u16 val_u16 = val;
v128u16 *dst_vec128 = (v128u16 *)dst;
v128u16 val_vec128 = (v128u16){val,val,val,val,val,val,val,val}; // Don't set as const, since vec_st() cannot store a const vector.
v128u16 val_vec128 = (v128u16)vec_splat( vec_lde(0, &val_u16), 0 ); // Don't set as const, since vec_st() cannot store a const vector.
MACRODO_N(ELEMENTCOUNT / (sizeof(v128u16) / sizeof(u16)), vec_st(val_vec128, (X)*sizeof(v128u16), dst_vec128));
}
static void memset_u32(void *dst, const u32 val, const size_t elementCount)
{
CACHE_ALIGN const u32 val_u32 = val;
v128u32 *dst_vec128 = (v128u32 *)dst;
const size_t length_vec128 = elementCount / (sizeof(v128u32) / sizeof(u32));
v128u32 val_vec128 = (v128u32){val,val,val,val}; // Don't set as const, since vec_st() cannot store a const vector.
v128u32 val_vec128 = (v128u32)vec_splat( vec_lde(0, &val_u32), 0 ); // Don't set as const, since vec_st() cannot store a const vector.
for (size_t i = 0; i < length_vec128; i++)
vec_st(val_vec128, i*sizeof(v128u32), dst_vec128);
}
@@ -724,9 +727,10 @@ static void memset_u32(void *dst, const u32 val, const size_t elementCount)
template <size_t ELEMENTCOUNT>
static void memset_u32_fast(void *dst, const u32 val)
{
CACHE_ALIGN const u32 val_u32 = val;
v128u32 *dst_vec128 = (v128u32 *)dst;
v128u32 val_vec128 = (v128u32){val,val,val,val}; // Don't set as const, since vec_st() cannot store a const vector.
v128u32 val_vec128 = (v128u32)vec_splat( vec_lde(0, &val_u32), 0 ); // Don't set as const, since vec_st() cannot store a const vector.
MACRODO_N(ELEMENTCOUNT / (sizeof(v128u32) / sizeof(u32)), vec_st(val_vec128, (X)*sizeof(v128u32), dst_vec128));
}
@@ -744,31 +748,55 @@ static void stream_copy_fast(void *__restrict dst, const void *__restrict src)
buffer_copy_fast<VECLENGTH>(dst, src);
}
template <class T, size_t VECLENGTH, bool NEEDENDIANSWAP>
template <class T, size_t VECLENGTH, size_t ENDIANSWAPBYTES>
static void __buffer_copy_or_constant_fast(void *__restrict dst, const void *__restrict src, const T &c_vec)
{
MACRODO_N( VECLENGTH / sizeof(v128s8), vec_st(vec_or(vec_ld((X)*sizeof(v128s8),(T*)src),c_vec), (X)*sizeof(v128s8), dst) );
if (ENDIANSWAPBYTES == 4)
{
static const v128u8 permuteVec = ((v128u8){3,2,1,0, 7,6,5,4, 11,10,9,8, 15,14,13,12});
MACRODO_N( VECLENGTH / sizeof(v128s8), vec_st( (T)vec_or(vec_perm(vec_ld((X)*sizeof(T),(v128s8 *__restrict)src), vec_splat_s8(0), permuteVec),(v128s8)c_vec), (X)*sizeof(T), (T *__restrict)dst ) );
}
else if (ENDIANSWAPBYTES == 2)
{
static const v128u8 permuteVec = ((v128u8){1,0, 3,2, 5,4, 7,6, 9,8, 11,10, 13,12, 15,14});
MACRODO_N( VECLENGTH / sizeof(v128s8), vec_st( (T)vec_or(vec_perm(vec_ld((X)*sizeof(T),(v128s8 *__restrict)src), vec_splat_s8(0), permuteVec),(v128s8)c_vec), (X)*sizeof(T), (T *__restrict)dst ) );
}
else
{
MACRODO_N( VECLENGTH / sizeof(v128s8), vec_st( (T)vec_or(vec_ld((X)*sizeof(T),(v128s8 *__restrict)src),(v128s8)c_vec), (X)*sizeof(T), (T *__restrict)dst ) );
}
}
template <class T, bool NEEDENDIANSWAP>
template <class T, size_t ENDIANSWAPBYTES>
static void __buffer_copy_or_constant(void *__restrict dst, const void *__restrict src, const size_t vecLength, const T &c_vec)
{
switch (vecLength)
{
case 128: __buffer_copy_or_constant_fast<T, 128, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 256: __buffer_copy_or_constant_fast<T, 256, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 512: __buffer_copy_or_constant_fast<T, 512, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 768: __buffer_copy_or_constant_fast<T, 768, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 1024: __buffer_copy_or_constant_fast<T, 1024, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 2048: __buffer_copy_or_constant_fast<T, 2048, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 2304: __buffer_copy_or_constant_fast<T, 2304, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 4096: __buffer_copy_or_constant_fast<T, 4096, NEEDENDIANSWAP>(dst, src, c_vec); break;
case 128: __buffer_copy_or_constant_fast<T, 128, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 256: __buffer_copy_or_constant_fast<T, 256, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 512: __buffer_copy_or_constant_fast<T, 512, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 768: __buffer_copy_or_constant_fast<T, 768, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 1024: __buffer_copy_or_constant_fast<T, 1024, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 2048: __buffer_copy_or_constant_fast<T, 2048, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 2304: __buffer_copy_or_constant_fast<T, 2304, ENDIANSWAPBYTES>(dst, src, c_vec); break;
case 4096: __buffer_copy_or_constant_fast<T, 4096, ENDIANSWAPBYTES>(dst, src, c_vec); break;
default:
{
for (size_t i = 0; i < vecLength; i+=sizeof(T))
{
vec_st(vec_or(vec_ld(i,(T*)src),c_vec), i, dst);
if (ENDIANSWAPBYTES == 4)
{
vec_st( (T)vec_or(vec_perm(vec_ld(i,(v128s8 *__restrict)src), vec_splat_s8(0), ((v128u8){3,2,1,0, 7,6,5,4, 11,10,9,8, 15,14,13,12})),(v128s8)c_vec), i, (T *__restrict)dst );
}
else if (ENDIANSWAPBYTES == 2)
{
vec_st( (T)vec_or(vec_perm(vec_ld(i,(v128s8 *__restrict)src), vec_splat_s8(0), ((v128u8){1,0, 3,2, 5,4, 7,6, 9,8, 11,10, 13,12, 15,14})),(v128s8)c_vec), i, (T *__restrict)dst );
}
else
{
vec_st( (T)vec_or(vec_ld(i,(v128s8 *__restrict)src),(v128s8)c_vec), i, (T *__restrict)dst );
}
}
break;
}
@@ -777,47 +805,81 @@ static void __buffer_copy_or_constant(void *__restrict dst, const void *__restri
static void buffer_copy_or_constant_s8(void *__restrict dst, const void *__restrict src, const size_t vecLength, const s8 c)
{
const v128s8 c_vec = {c,c,c,c,c,c,c,c,c,c,c,c,c,c,c,c};
__buffer_copy_or_constant<v128s8, false>(dst, src, vecLength, c_vec);
CACHE_ALIGN const s8 c_8 = c;
const v128s8 c_vec = (v128s8)vec_splat( vec_lde(0, &c_8), 0 );
__buffer_copy_or_constant<v128s8, 0>(dst, src, vecLength, c_vec);
}
template <size_t VECLENGTH>
static void buffer_copy_or_constant_s8_fast(void *__restrict dst, const void *__restrict src, const s8 c)
{
const v128s8 c_vec = {c,c,c,c,c,c,c,c,c,c,c,c,c,c,c,c};
__buffer_copy_or_constant_fast<v128s8, VECLENGTH>(dst, src, c_vec);
CACHE_ALIGN const s8 c_8 = c;
const v128s8 c_vec = (v128s8)vec_splat( vec_lde(0, &c_8), 0 );
__buffer_copy_or_constant_fast<v128s8, VECLENGTH, 0>(dst, src, c_vec);
}
template <bool NEEDENDIANSWAP>
static void buffer_copy_or_constant_s16(void *__restrict dst, const void *__restrict src, const size_t vecLength, const s16 c)
{
const s16 c_16 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_16(c) : c;
const v128s16 c_vec = {c_16, c_16, c_16, c_16, c_16, c_16, c_16, c_16};
__buffer_copy_or_constant<v128s16, NEEDENDIANSWAP>(dst, src, vecLength, c_vec);
CACHE_ALIGN const s16 c_16 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_16(c) : c;
const v128s16 c_vec = (v128s16)vec_splat( vec_lde(0, &c_16), 0 );
if (NEEDENDIANSWAP)
{
__buffer_copy_or_constant<v128s16, 2>(dst, src, vecLength, c_vec);
}
else
{
__buffer_copy_or_constant<v128s16, 0>(dst, src, vecLength, c_vec);
}
}
template <size_t VECLENGTH, bool NEEDENDIANSWAP>
static void buffer_copy_or_constant_s16_fast(void *__restrict dst, const void *__restrict src, const s16 c)
{
const s16 c_16 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_16(c) : c;
const v128s16 c_vec = {c_16, c_16, c_16, c_16, c_16, c_16, c_16, c_16};
__buffer_copy_or_constant_fast<v128s16, VECLENGTH, NEEDENDIANSWAP>(dst, src, c_vec);
CACHE_ALIGN const s16 c_16 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_16(c) : c;
const v128s16 c_vec = (v128s16)vec_splat( vec_lde(0, &c_16), 0 );
if (NEEDENDIANSWAP)
{
__buffer_copy_or_constant_fast<v128s16, VECLENGTH, 2>(dst, src, c_vec);
}
else
{
__buffer_copy_or_constant_fast<v128s16, VECLENGTH, 0>(dst, src, c_vec);
}
}
template <bool NEEDENDIANSWAP>
static void buffer_copy_or_constant_s32(void *__restrict dst, const void *__restrict src, const size_t vecLength, const s32 c)
{
const s32 c_32 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_32(c) : c;
const v128s32 c_vec = {c_32, c_32, c_32, c_32};
__buffer_copy_or_constant<v128s32, NEEDENDIANSWAP>(dst, src, vecLength, c_vec);
CACHE_ALIGN const s32 c_32 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_32(c) : c;
const v128s32 c_vec = (v128s32)vec_splat( vec_lde(0, &c_32), 0 );
if (NEEDENDIANSWAP)
{
__buffer_copy_or_constant<v128s32, 4>(dst, src, vecLength, c_vec);
}
else
{
__buffer_copy_or_constant<v128s32, 0>(dst, src, vecLength, c_vec);
}
}
template <size_t VECLENGTH, bool NEEDENDIANSWAP>
static void buffer_copy_or_constant_s32_fast(void *__restrict dst, const void *__restrict src, const s32 c)
{
const s32 c_32 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_32(c) : c;
const v128s32 c_vec = {c_32, c_32, c_32, c_32};
__buffer_copy_or_constant_fast<v128s32, VECLENGTH, NEEDENDIANSWAP>(dst, src, c_vec);
CACHE_ALIGN const s32 c_32 = (NEEDENDIANSWAP) ? LE_TO_LOCAL_32(c) : c;
const v128s32 c_vec = (v128s32)vec_splat( vec_lde(0, &c_32), 0 );
if (NEEDENDIANSWAP)
{
__buffer_copy_or_constant_fast<v128s32, VECLENGTH, 4>(dst, src, c_vec);
}
else
{
__buffer_copy_or_constant_fast<v128s32, VECLENGTH, 0>(dst, src, c_vec);
}
}
#else // No SIMD