The-Homebrew-Cloud/dolphin
2
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2025-02-13 15:59:23 +01:00
Code Issues Packages Projects Releases Wiki Activity

VertexLoaderX64: use common code for FORMAT_FLOAT

Browse Source
This commit is contained in:
Tillmann Karras 2015-03-18 11:44:23 +01:00
parent 7030542546
commit 9da86092ae
1 changed files with 19 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
Source/Core/VideoCommon/VertexLoaderX64.cpp
View File

@ -65,7 +65,7 @@ OpArg VertexLoaderX64::GetVertexAddr(int array, u64 attribute)
int VertexLoaderX64::ReadVertex(OpArg data, u64 attribute, int format, int count_in, int count_out, bool dequantize, u8 scaling_exponent, AttributeFormat* native_format) int VertexLoaderX64::ReadVertex(OpArg data, u64 attribute, int format, int count_in, int count_out, bool dequantize, u8 scaling_exponent, AttributeFormat* native_format)
{ {
static const __m128i shuffle_lut[5][3] = { static const __m128i shuffle_lut[4][3] = {
{_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFF00L), // 1x u8 {_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFF00L), // 1x u8
_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFF01L, 0xFFFFFF00L), // 2x u8 _mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFF01L, 0xFFFFFF00L), // 2x u8
_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFF02L, 0xFFFFFF01L, 0xFFFFFF00L)}, // 3x u8 _mm_set_epi32(0xFFFFFFFFL, 0xFFFFFF02L, 0xFFFFFF01L, 0xFFFFFF00L)}, // 3x u8
@ -78,9 +78,6 @@ int VertexLoaderX64::ReadVertex(OpArg data, u64 attribute, int format, int count
{_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFFFFL, 0x0001FFFFL), // 1x s16 {_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFFFFL, 0x0001FFFFL), // 1x s16
_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0x0203FFFFL, 0x0001FFFFL), // 2x s16 _mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0x0203FFFFL, 0x0001FFFFL), // 2x s16
_mm_set_epi32(0xFFFFFFFFL, 0x0405FFFFL, 0x0203FFFFL, 0x0001FFFFL)}, // 3x s16 _mm_set_epi32(0xFFFFFFFFL, 0x0405FFFFL, 0x0203FFFFL, 0x0001FFFFL)}, // 3x s16
{_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0xFFFFFFFFL, 0x00010203L), // 1x float
_mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0x04050607L, 0x00010203L), // 2x float
_mm_set_epi32(0xFFFFFFFFL, 0x08090A0BL, 0x04050607L, 0x00010203L)}, // 3x float
}; };
static const __m128 scale_factors[32] = { static const __m128 scale_factors[32] = {
_mm_set_ps1(1./(1u<< 0)), _mm_set_ps1(1./(1u<< 1)), _mm_set_ps1(1./(1u<< 2)), _mm_set_ps1(1./(1u<< 3)), _mm_set_ps1(1./(1u<< 0)), _mm_set_ps1(1./(1u<< 1)), _mm_set_ps1(1./(1u<< 2)), _mm_set_ps1(1./(1u<< 3)),
@ -110,6 +107,21 @@ int VertexLoaderX64::ReadVertex(OpArg data, u64 attribute, int format, int count
if (attribute == DIRECT) if (attribute == DIRECT)
m_src_ofs += load_bytes; m_src_ofs += load_bytes;
if (format == FORMAT_FLOAT)
{
// Floats don't need to be scaled or converted,
// so we can just load/swap/store them directly
// and return early.
for (int i = 0; i < count_in; i++)
{
LoadAndSwap(32, scratch3, data);
MOV(32, dest, R(scratch3));
data.offset += sizeof(float);
dest.offset += sizeof(float);
}
return load_bytes;
}
if (cpu_info.bSSSE3) if (cpu_info.bSSSE3)
{ {
if (load_bytes > 8) if (load_bytes > 8)
@ -170,29 +182,13 @@ int VertexLoaderX64::ReadVertex(OpArg data, u64 attribute, int format, int count
else else
PSRLD(coords, 16); PSRLD(coords, 16);
break; break;
case FORMAT_FLOAT:
// Floats don't need to be scaled or converted,
// so we can just load/swap/store them directly
// and return early.
// (In SSSE3 we still need to store them.)
for (int i = 0; i < count_in; i++)
{
LoadAndSwap(32, scratch3, data);
MOV(32, dest, R(scratch3));
data.offset += sizeof(float);
dest.offset += sizeof(float);
}
return load_bytes;
} }
} }
if (format != FORMAT_FLOAT) CVTDQ2PS(coords, R(coords));
{
CVTDQ2PS(coords, R(coords));
if (dequantize && scaling_exponent) if (dequantize && scaling_exponent)
MULPS(coords, M(&scale_factors[scaling_exponent])); MULPS(coords, M(&scale_factors[scaling_exponent]));
}
switch (count_out) switch (count_out)
{ {