mirror of
https://github.com/libretro/ppsspp.git
synced 2024-11-27 02:10:34 +00:00
SoftGPU: Move sampler code to a dedicated file.
This commit is contained in:
Unknown W. Brackets
parent
3e76863b8a
commit
d5426e4360
@ -1284,6 +1284,8 @@ set(GPU_SOURCES
|
||||
GPU/Software/Lighting.h
|
||||
GPU/Software/Rasterizer.cpp
|
||||
GPU/Software/Rasterizer.h
|
||||
GPU/Software/Sampler.cpp
|
||||
GPU/Software/Sampler.h
|
||||
GPU/Software/SoftGpu.cpp
|
||||
GPU/Software/SoftGpu.h
|
||||
GPU/Software/TransformUnit.cpp
|
||||
|
||||
@ -250,6 +250,7 @@
|
||||
<ClInclude Include="Software\Clipper.h" />
|
||||
<ClInclude Include="Software\Lighting.h" />
|
||||
<ClInclude Include="Software\Rasterizer.h" />
|
||||
<ClInclude Include="Software\Sampler.h" />
|
||||
<ClInclude Include="Software\SoftGpu.h" />
|
||||
<ClInclude Include="Software\TransformUnit.h" />
|
||||
<ClInclude Include="Common\TextureDecoder.h" />
|
||||
@ -350,6 +351,7 @@
|
||||
<ClCompile Include="Software\Clipper.cpp" />
|
||||
<ClCompile Include="Software\Lighting.cpp" />
|
||||
<ClCompile Include="Software\Rasterizer.cpp" />
|
||||
<ClCompile Include="Software\Sampler.cpp" />
|
||||
<ClCompile Include="Software\SoftGpu.cpp" />
|
||||
<ClCompile Include="Software\TransformUnit.cpp" />
|
||||
<ClCompile Include="Common\TextureDecoder.cpp" />
|
||||
|
||||
@ -259,6 +259,9 @@
|
||||
<ClInclude Include="Common\ShaderTranslation.h">
|
||||
<Filter>Common</Filter>
|
||||
</ClInclude>
|
||||
<ClInclude Include="Software\Sampler.h">
|
||||
<Filter>Software</Filter>
|
||||
</ClInclude>
|
||||
</ItemGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="Math3D.cpp">
|
||||
@ -510,5 +513,8 @@
|
||||
<ClCompile Include="Vulkan\FramebufferVulkan.cpp">
|
||||
<Filter>Vulkan</Filter>
|
||||
</ClCompile>
|
||||
<ClCompile Include="Software\Sampler.cpp">
|
||||
<Filter>Software</Filter>
|
||||
</ClCompile>
|
||||
</ItemGroup>
|
||||
</Project>
|
||||
@ -31,6 +31,7 @@
|
||||
#include "GPU/Common/TextureDecoder.h"
|
||||
#include "GPU/Software/SoftGpu.h"
|
||||
#include "GPU/Software/Rasterizer.h"
|
||||
#include "GPU/Software/Sampler.h"
|
||||
|
||||
#if defined(_M_SSE)
|
||||
#include <emmintrin.h>
|
||||
@ -110,51 +111,6 @@ static inline Vec4<float> Interpolate(const float &c0, const float &c1, const fl
|
||||
return Interpolate(c0, c1, c2, w0.Cast<float>(), w1.Cast<float>(), w2.Cast<float>(), wsum_recip);
|
||||
}
|
||||
|
||||
template <unsigned int texel_size_bits>
|
||||
static inline int GetPixelDataOffset(unsigned int row_pitch_bytes, unsigned int u, unsigned int v)
|
||||
{
|
||||
if (!gstate.isTextureSwizzled())
|
||||
return (v * (row_pitch_bytes * texel_size_bits >> 3)) + (u * texel_size_bits >> 3);
|
||||
|
||||
const int tile_size_bits = 32;
|
||||
const int tiles_in_block_horizontal = 4;
|
||||
const int tiles_in_block_vertical = 8;
|
||||
|
||||
int texels_per_tile = tile_size_bits / texel_size_bits;
|
||||
int tile_u = u / texels_per_tile;
|
||||
int tile_idx = (v % tiles_in_block_vertical) * (tiles_in_block_horizontal) +
|
||||
// TODO: not sure if the *texel_size_bits/8 factor is correct
|
||||
(v / tiles_in_block_vertical) * ((row_pitch_bytes*texel_size_bits/(tile_size_bits))*tiles_in_block_vertical) +
|
||||
(tile_u % tiles_in_block_horizontal) +
|
||||
(tile_u / tiles_in_block_horizontal) * (tiles_in_block_horizontal*tiles_in_block_vertical);
|
||||
|
||||
return tile_idx * (tile_size_bits / 8) + ((u % texels_per_tile) * texel_size_bits) / 8;
|
||||
}
|
||||
|
||||
static inline u32 LookupColor(unsigned int index, unsigned int level)
|
||||
{
|
||||
const bool mipmapShareClut = gstate.isClutSharedForMipmaps();
|
||||
const int clutSharingOffset = mipmapShareClut ? 0 : level * 16;
|
||||
|
||||
switch (gstate.getClutPaletteFormat()) {
|
||||
case GE_CMODE_16BIT_BGR5650:
|
||||
return RGB565ToRGBA8888(reinterpret_cast<u16*>(clut)[index + clutSharingOffset]);
|
||||
|
||||
case GE_CMODE_16BIT_ABGR5551:
|
||||
return RGBA5551ToRGBA8888(reinterpret_cast<u16*>(clut)[index + clutSharingOffset]);
|
||||
|
||||
case GE_CMODE_16BIT_ABGR4444:
|
||||
return RGBA4444ToRGBA8888(reinterpret_cast<u16*>(clut)[index + clutSharingOffset]);
|
||||
|
||||
case GE_CMODE_32BIT_ABGR8888:
|
||||
return clut[index + clutSharingOffset];
|
||||
|
||||
default:
|
||||
ERROR_LOG_REPORT(G3D, "Software: Unsupported palette format: %x", gstate.getClutPaletteFormat());
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
static inline u8 ClampFogDepth(float fogdepth) {
|
||||
if (fogdepth <= 0.0f)
|
||||
return 0;
|
||||
@ -374,123 +330,6 @@ static inline void GetTextureCoordinates(const VertexData& v0, const VertexData&
|
||||
}
|
||||
}
|
||||
|
||||
struct Nearest4 {
|
||||
MEMORY_ALIGNED16(u32 v[4]);
|
||||
|
||||
operator u32() const {
|
||||
return v[0];
|
||||
}
|
||||
};
|
||||
|
||||
template <int N>
|
||||
inline static Nearest4 SampleNearest(int level, int u[N], int v[N], const u8 *srcptr, int texbufw)
|
||||
{
|
||||
Nearest4 res;
|
||||
if (!srcptr) {
|
||||
memset(res.v, 0, sizeof(res.v));
|
||||
return res;
|
||||
}
|
||||
|
||||
GETextureFormat texfmt = gstate.getTextureFormat();
|
||||
|
||||
// TODO: Should probably check if textures are aligned properly...
|
||||
|
||||
switch (texfmt) {
|
||||
case GE_TFMT_4444:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
res.v[i] = RGBA4444ToRGBA8888(*(const u16 *)src);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_5551:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
res.v[i] = RGBA5551ToRGBA8888(*(const u16 *)src);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_5650:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
res.v[i] = RGB565ToRGBA8888(*(const u16 *)src);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_8888:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<32>(texbufw, u[i], v[i]);
|
||||
res.v[i] = *(const u32 *)src;
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT32:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<32>(texbufw, u[i], v[i]);
|
||||
u32 val = src[0] + (src[1] << 8) + (src[2] << 16) + (src[3] << 24);
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), 0);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT16:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
u16 val = src[0] + (src[1] << 8);
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), 0);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT8:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<8>(texbufw, u[i], v[i]);
|
||||
u8 val = *src;
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), 0);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT4:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<4>(texbufw, u[i], v[i]);
|
||||
u8 val = (u[i] & 1) ? (src[0] >> 4) : (src[0] & 0xF);
|
||||
// Only CLUT4 uses separate mipmap palettes.
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), level);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_DXT1:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const DXT1Block *block = (const DXT1Block *)srcptr + (v[i] / 4) * (texbufw / 4) + (u[i] / 4);
|
||||
u32 data[4 * 4];
|
||||
DecodeDXT1Block(data, block, 4, 4, false);
|
||||
res.v[i] = data[4 * (v[i] % 4) + (u[i] % 4)];
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_DXT3:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const DXT3Block *block = (const DXT3Block *)srcptr + (v[i] / 4) * (texbufw / 4) + (u[i] / 4);
|
||||
u32 data[4 * 4];
|
||||
DecodeDXT3Block(data, block, 4, 4);
|
||||
res.v[i] = data[4 * (v[i] % 4) + (u[i] % 4)];
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_DXT5:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const DXT5Block *block = (const DXT5Block *)srcptr + (v[i] / 4) * (texbufw / 4) + (u[i] / 4);
|
||||
u32 data[4 * 4];
|
||||
DecodeDXT5Block(data, block, 4, 4);
|
||||
res.v[i] = data[4 * (v[i] % 4) + (u[i] % 4)];
|
||||
}
|
||||
return res;
|
||||
|
||||
default:
|
||||
ERROR_LOG_REPORT(G3D, "Software: Unsupported texture format: %x", texfmt);
|
||||
memset(res.v, 0, sizeof(res.v));
|
||||
return res;
|
||||
}
|
||||
}
|
||||
|
||||
// NOTE: These likely aren't endian safe
|
||||
static inline u32 GetPixelColor(int x, int y)
|
||||
{
|
||||
@ -1164,38 +1003,6 @@ inline void DrawSinglePixel(const DrawingCoords &p, u16 z, u8 fog, const Vec4<in
|
||||
SetPixelColor(p.x, p.y, new_color);
|
||||
}
|
||||
|
||||
static inline Vec4<int> SampleLinear(int texlevel, int u[4], int v[4], int frac_u, int frac_v, const u8 *tptr, int bufw) {
|
||||
#if defined(_M_SSE)
|
||||
Nearest4 c = SampleNearest<4>(texlevel, u, v, tptr, bufw);
|
||||
|
||||
const __m128i z = _mm_setzero_si128();
|
||||
|
||||
__m128i cvec = _mm_load_si128((const __m128i *)c.v);
|
||||
__m128i tvec = _mm_unpacklo_epi8(cvec, z);
|
||||
tvec = _mm_mullo_epi16(tvec, _mm_set1_epi16(0x100 - frac_v));
|
||||
__m128i bvec = _mm_unpackhi_epi8(cvec, z);
|
||||
bvec = _mm_mullo_epi16(bvec, _mm_set1_epi16(frac_v));
|
||||
|
||||
// This multiplies the left and right sides. We shift right after, although this may round down...
|
||||
__m128i rowmult = _mm_set_epi16(frac_u, frac_u, frac_u, frac_u, 0x100 - frac_u, 0x100 - frac_u, 0x100 - frac_u, 0x100 - frac_u);
|
||||
__m128i tmp = _mm_mulhi_epu16(_mm_add_epi16(tvec, bvec), rowmult);
|
||||
|
||||
// Now we need to add the left and right sides together.
|
||||
__m128i res = _mm_add_epi16(tmp, _mm_shuffle_epi32(tmp, _MM_SHUFFLE(3, 2, 3, 2)));
|
||||
return Vec4<int>(_mm_unpacklo_epi16(res, z));
|
||||
#else
|
||||
Nearest4 nearest = SampleNearest<4>(texlevel, u, v, tptr, bufw);
|
||||
Vec4<int> texcolor_tl = Vec4<int>::FromRGBA(nearest.v[0]);
|
||||
Vec4<int> texcolor_tr = Vec4<int>::FromRGBA(nearest.v[1]);
|
||||
Vec4<int> texcolor_bl = Vec4<int>::FromRGBA(nearest.v[2]);
|
||||
Vec4<int> texcolor_br = Vec4<int>::FromRGBA(nearest.v[3]);
|
||||
// 0x100 causes a slight bias to tl, but without it we'd have to divide by 255 * 255.
|
||||
Vec4<int> t = texcolor_tl * (0x100 - frac_u) + texcolor_tr * frac_u;
|
||||
Vec4<int> b = texcolor_bl * (0x100 - frac_u) + texcolor_br * frac_u;
|
||||
return (t * (0x100 - frac_v) + b * frac_v) / (256 * 256);
|
||||
#endif
|
||||
}
|
||||
|
||||
static inline void ApplyTexturing(Vec4<int> &prim_color, float s, float t, int texlevel, int frac_texlevel, bool bilinear, u8 *texptr[], int texbufw[]) {
|
||||
int u[8] = {0}, v[8] = {0}; // 1.23.8 fixed point
|
||||
int frac_u[2], frac_v[2];
|
||||
@ -1214,9 +1021,9 @@ static inline void ApplyTexturing(Vec4<int> &prim_color, float s, float t, int t
|
||||
GetTexelCoordinates(texlevel + 1, s, t, u[1], v[1]);
|
||||
}
|
||||
|
||||
texcolor0 = Vec4<int>::FromRGBA(SampleNearest<1>(texlevel, u, v, tptr0, bufw0));
|
||||
texcolor0 = Sampler::SampleNearest(texlevel, u[0], v[0], tptr0, bufw0);
|
||||
if (frac_texlevel) {
|
||||
texcolor1 = Vec4<int>::FromRGBA(SampleNearest<1>(texlevel + 1, u + 1, v + 1, tptr1, bufw1));
|
||||
texcolor1 = Sampler::SampleNearest(texlevel + 1, u[1], v[1], tptr1, bufw1);
|
||||
}
|
||||
} else {
|
||||
GetTexelCoordinatesQuad(texlevel, s, t, u, v, frac_u[0], frac_v[0]);
|
||||
@ -1224,9 +1031,9 @@ static inline void ApplyTexturing(Vec4<int> &prim_color, float s, float t, int t
|
||||
GetTexelCoordinatesQuad(texlevel + 1, s, t, u + 4, v + 4, frac_u[1], frac_v[1]);
|
||||
}
|
||||
|
||||
texcolor0 = SampleLinear(texlevel, u, v, frac_u[0], frac_v[0], tptr0, bufw0);
|
||||
texcolor0 = Sampler::SampleLinear(texlevel, u, v, frac_u[0], frac_v[0], tptr0, bufw0);
|
||||
if (frac_texlevel) {
|
||||
texcolor1 = SampleLinear(texlevel + 1, u + 4, v + 4, frac_u[1], frac_v[1], tptr1, bufw1);
|
||||
texcolor1 = Sampler::SampleLinear(texlevel + 1, u + 4, v + 4, frac_u[1], frac_v[1], tptr1, bufw1);
|
||||
}
|
||||
}
|
||||
|
||||
@ -1828,7 +1635,7 @@ bool GetCurrentTexture(GPUDebugBuffer &buffer, int level)
|
||||
u32 *row = (u32 *)buffer.GetData();
|
||||
for (int y = 0; y < h; ++y) {
|
||||
for (int x = 0; x < w; ++x) {
|
||||
row[x] = SampleNearest<1>(level, &x, &y, texptr, texbufw);
|
||||
row[x] = Sampler::SampleNearest(level, x, y, texptr, texbufw).ToRGBA();
|
||||
}
|
||||
row += w;
|
||||
}
|
||||
|
||||
232
GPU/Software/Sampler.cpp
Normal file
232
GPU/Software/Sampler.cpp
Normal file
@ -0,0 +1,232 @@
|
||||
// Copyright (c) 2017- PPSSPP Project.
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU General Public License as published by
|
||||
// the Free Software Foundation, version 2.0 or later versions.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU General Public License 2.0 for more details.
|
||||
|
||||
// A copy of the GPL 2.0 should have been included with the program.
|
||||
// If not, see http://www.gnu.org/licenses/
|
||||
|
||||
// Official git repository and contact information can be found at
|
||||
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
|
||||
|
||||
#include "Common/ColorConv.h"
|
||||
#include "Core/Reporting.h"
|
||||
#include "GPU/Common/TextureDecoder.h"
|
||||
#include "GPU/GPUState.h"
|
||||
#include "GPU/Software/Sampler.h"
|
||||
|
||||
#if defined(_M_SSE)
|
||||
#include <emmintrin.h>
|
||||
#endif
|
||||
|
||||
using namespace Math3D;
|
||||
|
||||
extern u32 clut[4096];
|
||||
|
||||
namespace Sampler {
|
||||
|
||||
template <unsigned int texel_size_bits>
|
||||
static inline int GetPixelDataOffset(unsigned int row_pitch_bytes, unsigned int u, unsigned int v)
|
||||
{
|
||||
if (!gstate.isTextureSwizzled())
|
||||
return (v * (row_pitch_bytes * texel_size_bits >> 3)) + (u * texel_size_bits >> 3);
|
||||
|
||||
const int tile_size_bits = 32;
|
||||
const int tiles_in_block_horizontal = 4;
|
||||
const int tiles_in_block_vertical = 8;
|
||||
|
||||
int texels_per_tile = tile_size_bits / texel_size_bits;
|
||||
int tile_u = u / texels_per_tile;
|
||||
int tile_idx = (v % tiles_in_block_vertical) * (tiles_in_block_horizontal) +
|
||||
// TODO: not sure if the *texel_size_bits/8 factor is correct
|
||||
(v / tiles_in_block_vertical) * ((row_pitch_bytes*texel_size_bits/(tile_size_bits))*tiles_in_block_vertical) +
|
||||
(tile_u % tiles_in_block_horizontal) +
|
||||
(tile_u / tiles_in_block_horizontal) * (tiles_in_block_horizontal*tiles_in_block_vertical);
|
||||
|
||||
return tile_idx * (tile_size_bits / 8) + ((u % texels_per_tile) * texel_size_bits) / 8;
|
||||
}
|
||||
|
||||
static inline u32 LookupColor(unsigned int index, unsigned int level)
|
||||
{
|
||||
const bool mipmapShareClut = gstate.isClutSharedForMipmaps();
|
||||
const int clutSharingOffset = mipmapShareClut ? 0 : level * 16;
|
||||
|
||||
switch (gstate.getClutPaletteFormat()) {
|
||||
case GE_CMODE_16BIT_BGR5650:
|
||||
return RGB565ToRGBA8888(reinterpret_cast<u16*>(clut)[index + clutSharingOffset]);
|
||||
|
||||
case GE_CMODE_16BIT_ABGR5551:
|
||||
return RGBA5551ToRGBA8888(reinterpret_cast<u16*>(clut)[index + clutSharingOffset]);
|
||||
|
||||
case GE_CMODE_16BIT_ABGR4444:
|
||||
return RGBA4444ToRGBA8888(reinterpret_cast<u16*>(clut)[index + clutSharingOffset]);
|
||||
|
||||
case GE_CMODE_32BIT_ABGR8888:
|
||||
return clut[index + clutSharingOffset];
|
||||
|
||||
default:
|
||||
ERROR_LOG_REPORT(G3D, "Software: Unsupported palette format: %x", gstate.getClutPaletteFormat());
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
struct Nearest4 {
|
||||
MEMORY_ALIGNED16(u32 v[4]);
|
||||
|
||||
operator u32() const {
|
||||
return v[0];
|
||||
}
|
||||
};
|
||||
|
||||
template <int N>
|
||||
inline static Nearest4 SampleNearest(int level, int u[N], int v[N], const u8 *srcptr, int texbufw)
|
||||
{
|
||||
Nearest4 res;
|
||||
if (!srcptr) {
|
||||
memset(res.v, 0, sizeof(res.v));
|
||||
return res;
|
||||
}
|
||||
|
||||
GETextureFormat texfmt = gstate.getTextureFormat();
|
||||
|
||||
// TODO: Should probably check if textures are aligned properly...
|
||||
|
||||
switch (texfmt) {
|
||||
case GE_TFMT_4444:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
res.v[i] = RGBA4444ToRGBA8888(*(const u16 *)src);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_5551:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
res.v[i] = RGBA5551ToRGBA8888(*(const u16 *)src);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_5650:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
res.v[i] = RGB565ToRGBA8888(*(const u16 *)src);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_8888:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<32>(texbufw, u[i], v[i]);
|
||||
res.v[i] = *(const u32 *)src;
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT32:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<32>(texbufw, u[i], v[i]);
|
||||
u32 val = src[0] + (src[1] << 8) + (src[2] << 16) + (src[3] << 24);
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), 0);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT16:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<16>(texbufw, u[i], v[i]);
|
||||
u16 val = src[0] + (src[1] << 8);
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), 0);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT8:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<8>(texbufw, u[i], v[i]);
|
||||
u8 val = *src;
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), 0);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_CLUT4:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const u8 *src = srcptr + GetPixelDataOffset<4>(texbufw, u[i], v[i]);
|
||||
u8 val = (u[i] & 1) ? (src[0] >> 4) : (src[0] & 0xF);
|
||||
// Only CLUT4 uses separate mipmap palettes.
|
||||
res.v[i] = LookupColor(gstate.transformClutIndex(val), level);
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_DXT1:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const DXT1Block *block = (const DXT1Block *)srcptr + (v[i] / 4) * (texbufw / 4) + (u[i] / 4);
|
||||
u32 data[4 * 4];
|
||||
DecodeDXT1Block(data, block, 4, 4, false);
|
||||
res.v[i] = data[4 * (v[i] % 4) + (u[i] % 4)];
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_DXT3:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const DXT3Block *block = (const DXT3Block *)srcptr + (v[i] / 4) * (texbufw / 4) + (u[i] / 4);
|
||||
u32 data[4 * 4];
|
||||
DecodeDXT3Block(data, block, 4, 4);
|
||||
res.v[i] = data[4 * (v[i] % 4) + (u[i] % 4)];
|
||||
}
|
||||
return res;
|
||||
|
||||
case GE_TFMT_DXT5:
|
||||
for (int i = 0; i < N; ++i) {
|
||||
const DXT5Block *block = (const DXT5Block *)srcptr + (v[i] / 4) * (texbufw / 4) + (u[i] / 4);
|
||||
u32 data[4 * 4];
|
||||
DecodeDXT5Block(data, block, 4, 4);
|
||||
res.v[i] = data[4 * (v[i] % 4) + (u[i] % 4)];
|
||||
}
|
||||
return res;
|
||||
|
||||
default:
|
||||
ERROR_LOG_REPORT(G3D, "Software: Unsupported texture format: %x", texfmt);
|
||||
memset(res.v, 0, sizeof(res.v));
|
||||
return res;
|
||||
}
|
||||
}
|
||||
|
||||
Vec4<int> SampleNearest(int level, int u, int v, const u8 *tptr, int bufw) {
|
||||
return Vec4<int>::FromRGBA(SampleNearest<1>(level, &u, &v, tptr, bufw));
|
||||
}
|
||||
|
||||
Vec4<int> SampleLinear(int texlevel, int u[4], int v[4], int frac_u, int frac_v, const u8 *tptr, int bufw) {
|
||||
#if defined(_M_SSE)
|
||||
Nearest4 c = SampleNearest<4>(texlevel, u, v, tptr, bufw);
|
||||
|
||||
const __m128i z = _mm_setzero_si128();
|
||||
|
||||
__m128i cvec = _mm_load_si128((const __m128i *)c.v);
|
||||
__m128i tvec = _mm_unpacklo_epi8(cvec, z);
|
||||
tvec = _mm_mullo_epi16(tvec, _mm_set1_epi16(0x100 - frac_v));
|
||||
__m128i bvec = _mm_unpackhi_epi8(cvec, z);
|
||||
bvec = _mm_mullo_epi16(bvec, _mm_set1_epi16(frac_v));
|
||||
|
||||
// This multiplies the left and right sides. We shift right after, although this may round down...
|
||||
__m128i rowmult = _mm_set_epi16(frac_u, frac_u, frac_u, frac_u, 0x100 - frac_u, 0x100 - frac_u, 0x100 - frac_u, 0x100 - frac_u);
|
||||
__m128i tmp = _mm_mulhi_epu16(_mm_add_epi16(tvec, bvec), rowmult);
|
||||
|
||||
// Now we need to add the left and right sides together.
|
||||
__m128i res = _mm_add_epi16(tmp, _mm_shuffle_epi32(tmp, _MM_SHUFFLE(3, 2, 3, 2)));
|
||||
return Vec4<int>(_mm_unpacklo_epi16(res, z));
|
||||
#else
|
||||
Nearest4 nearest = SampleNearest<4>(texlevel, u, v, tptr, bufw);
|
||||
Vec4<int> texcolor_tl = Vec4<int>::FromRGBA(nearest.v[0]);
|
||||
Vec4<int> texcolor_tr = Vec4<int>::FromRGBA(nearest.v[1]);
|
||||
Vec4<int> texcolor_bl = Vec4<int>::FromRGBA(nearest.v[2]);
|
||||
Vec4<int> texcolor_br = Vec4<int>::FromRGBA(nearest.v[3]);
|
||||
// 0x100 causes a slight bias to tl, but without it we'd have to divide by 255 * 255.
|
||||
Vec4<int> t = texcolor_tl * (0x100 - frac_u) + texcolor_tr * frac_u;
|
||||
Vec4<int> b = texcolor_bl * (0x100 - frac_u) + texcolor_br * frac_u;
|
||||
return (t * (0x100 - frac_v) + b * frac_v) / (256 * 256);
|
||||
#endif
|
||||
}
|
||||
|
||||
};
|
||||
27
GPU/Software/Sampler.h
Normal file
27
GPU/Software/Sampler.h
Normal file
@ -0,0 +1,27 @@
|
||||
// Copyright (c) 2017- PPSSPP Project.
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU General Public License as published by
|
||||
// the Free Software Foundation, version 2.0 or later versions.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU General Public License 2.0 for more details.
|
||||
|
||||
// A copy of the GPL 2.0 should have been included with the program.
|
||||
// If not, see http://www.gnu.org/licenses/
|
||||
|
||||
// Official git repository and contact information can be found at
|
||||
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "GPU/Math3D.h"
|
||||
|
||||
namespace Sampler {
|
||||
|
||||
Math3D::Vec4<int> SampleNearest(int level, int u, int v, const u8 *tptr, int bufwbytes);
|
||||
Math3D::Vec4<int> SampleLinear(int level, int u[4], int v[4], int frac_u, int frac_v, const u8 *tptr, int bufwbytes);
|
||||
|
||||
};
|
||||
@ -244,6 +244,7 @@ EXEC_AND_LIB_FILES := \
|
||||
$(SRC)/GPU/Software/Clipper.cpp \
|
||||
$(SRC)/GPU/Software/Lighting.cpp \
|
||||
$(SRC)/GPU/Software/Rasterizer.cpp.arm \
|
||||
$(SRC)/GPU/Software/Sampler.cpp \
|
||||
$(SRC)/GPU/Software/SoftGpu.cpp \
|
||||
$(SRC)/GPU/Software/TransformUnit.cpp \
|
||||
$(SRC)/Core/ELF/ElfReader.cpp \
|
||||
|
||||
Loading…
Reference in New Issue
Block a user