void MipsJit::EatInstruction(MIPSOpcode op) {
MIPSInfo info = MIPSGetInfo(op);
if (info & DELAYSLOT) {
ERROR_LOG_REPORT_ONCE(ateDelaySlot, JIT, "Ate a branch op.");
}
if (js.inDelaySlot) {
ERROR_LOG_REPORT_ONCE(ateInDelaySlot, JIT, "Ate an instruction inside a delay slot.");
}
js.numInstructions++;
js.compilerPC += 4;
js.downcountAmount += MIPSGetInstructionCycleEstimate(op);
}
void FramebufferManagerGLES::PackDepthbuffer(VirtualFramebuffer *vfb, int x, int y, int w, int h) {
if (!vfb->fbo) {
ERROR_LOG_REPORT_ONCE(vfbfbozero, SCEGE, "PackDepthbuffer: vfb->fbo == 0");
return;
}
// Pixel size always 4 here because we always request float
const u32 bufSize = vfb->z_stride * (h - y) * 4;
const u32 z_address = (0x04000000) | vfb->z_address;
const int packWidth = std::min(vfb->z_stride, std::min(x + w, (int)vfb->width));
if (!convBuf_ || convBufSize_ < bufSize) {
delete [] convBuf_;
convBuf_ = new u8[bufSize];
convBufSize_ = bufSize;
}
DEBUG_LOG(FRAMEBUF, "Reading depthbuffer to mem at %08x for vfb=%08x", z_address, vfb->fb_address);
draw_->CopyFramebufferToMemorySync(vfb->fbo, Draw::FB_DEPTH_BIT, 0, y, packWidth, h, Draw::DataFormat::D32F, convBuf_, vfb->z_stride);
int dstByteOffset = y * vfb->z_stride * sizeof(u16);
u16 *depth = (u16 *)Memory::GetPointer(z_address + dstByteOffset);
GLfloat *packed = (GLfloat *)convBuf_;
int totalPixels = h == 1 ? packWidth : vfb->z_stride * h;
for (int yp = 0; yp < h; ++yp) {
int row_offset = vfb->z_stride * yp;
for (int xp = 0; xp < packWidth; ++xp) {
const int i = row_offset + xp;
float scaled = FromScaledDepth(packed[i]);
if (scaled <= 0.0f) {
depth[i] = 0;
} else if (scaled >= 65535.0f) {
depth[i] = 65535;
} else {
depth[i] = (int)scaled;
}
}
}
}
// FP only, to suit GL(ES) 2.0
void GenerateDepalShaderFloat(char *buffer, GEBufferFormat pixelFormat, ShaderLanguage lang) {
char *p = buffer;
const char *modFunc = lang == HLSL_DX9 ? "fmod" : "mod";
char lookupMethod[128] = "index.r";
char offset[128] = "";
const GEPaletteFormat clutFormat = gstate.getClutPaletteFormat();
const u32 clutBase = gstate.getClutIndexStartPos();
const int shift = gstate.getClutIndexShift();
const int mask = gstate.getClutIndexMask();
float index_multiplier = 1.0f;
// pixelformat is the format of the texture we are sampling.
bool formatOK = true;
switch (pixelFormat) {
case GE_FORMAT_8888:
if ((mask & (mask + 1)) == 0) {
// If the value has all bits contiguous (bitmask check above), we can mod by it + 1.
const char *rgba = "rrrrrrrrggggggggbbbbbbbbaaaaaaaa";
const u8 rgba_shift = shift & 7;
if (rgba_shift == 0 && mask == 0xFF) {
sprintf(lookupMethod, "index.%c", rgba[shift]);
} else {
sprintf(lookupMethod, "%s(index.%c * %f, %d.0)", modFunc, rgba[shift], 255.99f / (1 << rgba_shift), mask + 1);
index_multiplier = 1.0f / 256.0f;
// Format was OK if there weren't bits from another component.
formatOK = mask <= 255 - (1 << rgba_shift);
}
} else {
formatOK = false;
}
break;
case GE_FORMAT_4444:
if ((mask & (mask + 1)) == 0 && shift < 16) {
const char *rgba = "rrrrggggbbbbaaaa";
const u8 rgba_shift = shift & 3;
if (rgba_shift == 0 && mask == 0xF) {
sprintf(lookupMethod, "index.%c", rgba[shift]);
index_multiplier = 15.0f / 256.0f;
} else {
// Let's divide and mod to get the right bits. A common case is shift=0, mask=01.
sprintf(lookupMethod, "%s(index.%c * %f, %d.0)", modFunc, rgba[shift], 15.99f / (1 << rgba_shift), mask + 1);
index_multiplier = 1.0f / 256.0f;
formatOK = mask <= 15 - (1 << rgba_shift);
}
} else {
formatOK = false;
}
break;
case GE_FORMAT_565:
if ((mask & (mask + 1)) == 0 && shift < 16) {
const u8 shifts[16] = { 0, 1, 2, 3, 4, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4 };
const int multipliers[16] = { 31, 31, 31, 31, 31, 63, 63, 63, 63, 63, 63, 31, 31, 31, 31, 31 };
const char *rgba = "rrrrrggggggbbbbb";
const u8 rgba_shift = shifts[shift];
if (rgba_shift == 0 && mask == multipliers[shift]) {
sprintf(lookupMethod, "index.%c", rgba[shift]);
index_multiplier = multipliers[shift] / 256.0f;
} else {
// We just need to divide the right component by the right value, and then mod against the mask.
// A common case is shift=1, mask=0f.
sprintf(lookupMethod, "%s(index.%c * %f, %d.0)", modFunc, rgba[shift], ((float)multipliers[shift] + 0.99f) / (1 << rgba_shift), mask + 1);
index_multiplier = 1.0f / 256.0f;
formatOK = mask <= multipliers[shift] - (1 << rgba_shift);
}
} else {
formatOK = false;
}
break;
case GE_FORMAT_5551:
if ((mask & (mask + 1)) == 0 && shift < 16) {
const char *rgba = "rrrrrgggggbbbbba";
const u8 rgba_shift = shift % 5;
if (rgba_shift == 0 && mask == 0x1F) {
sprintf(lookupMethod, "index.%c", rgba[shift]);
index_multiplier = 31.0f / 256.0f;
} else if (shift == 15 && mask == 1) {
sprintf(lookupMethod, "index.%c", rgba[shift]);
index_multiplier = 1.0f / 256.0f;
} else {
// A isn't possible here.
sprintf(lookupMethod, "%s(index.%c * %f, %d.0)", modFunc, rgba[shift], 31.99f / (1 << rgba_shift), mask + 1);
index_multiplier = 1.0f / 256.0f;
formatOK = mask <= 31 - (1 << rgba_shift);
}
} else {
formatOK = false;
}
break;
default:
break;
}
float texturePixels = 256.f;
if (clutFormat != GE_CMODE_32BIT_ABGR8888) {
texturePixels = 512.f;
index_multiplier *= 0.5f;
}
// Adjust index_multiplier, similar to the use of 15.99 instead of 16 in the ES 3 path.
// index_multiplier -= 0.01f / texturePixels;
if (!formatOK) {
ERROR_LOG_REPORT_ONCE(depal, G3D, "%i depal unsupported: shift=%i mask=%02x offset=%d", pixelFormat, shift, mask, clutBase);
}
// Offset by half a texel (plus clutBase) to turn NEAREST filtering into FLOOR.
// Technically, the clutBase should be |'d, not added, but that's hard with floats.
float texel_offset = ((float)clutBase + 0.5f) / texturePixels;
sprintf(offset, " + %f", texel_offset);
if (lang == GLSL_140) {
if (gl_extensions.IsGLES) {
WRITE(p, "#version 100\n");
WRITE(p, "precision mediump float;\n");
} else {
WRITE(p, "#version %d\n", gl_extensions.GLSLVersion());
}
WRITE(p, "varying vec2 v_texcoord0;\n");
WRITE(p, "uniform sampler2D tex;\n");
WRITE(p, "uniform sampler2D pal;\n");
WRITE(p, "void main() {\n");
WRITE(p, " vec4 index = texture2D(tex, v_texcoord0);\n");
WRITE(p, " float coord = (%s * %f)%s;\n", lookupMethod, index_multiplier, offset);
WRITE(p, " gl_FragColor = texture2D(pal, vec2(coord, 0.0));\n");
WRITE(p, "}\n");
} else if (lang == HLSL_DX9) {
WRITE(p, "sampler tex: register(s0);\n");
WRITE(p, "sampler pal: register(s1);\n");
WRITE(p, "float4 main(float2 v_texcoord0 : TEXCOORD0) : COLOR0 {\n");
WRITE(p, " float4 index = tex2D(tex, v_texcoord0);\n");
WRITE(p, " float coord = (%s * %f)%s;\n", lookupMethod, index_multiplier, offset);
WRITE(p, " return tex2D(pal, float2(coord, 0.0)).bgra;\n");
WRITE(p, "}\n");
}
}
// TODO: This probably is not the best interface.
// Also, we should try to merge this into the similar function in DrawEngineCommon.
bool TransformUnit::GetCurrentSimpleVertices(int count, std::vector<GPUDebugVertex> &vertices, std::vector<u16> &indices) {
// This is always for the current vertices.
u16 indexLowerBound = 0;
u16 indexUpperBound = count - 1;
if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
const u8 *inds = Memory::GetPointer(gstate_c.indexAddr);
const u16 *inds16 = (const u16 *)inds;
const u32 *inds32 = (const u32 *)inds;
if (inds) {
GetIndexBounds(inds, count, gstate.vertType, &indexLowerBound, &indexUpperBound);
indices.resize(count);
switch (gstate.vertType & GE_VTYPE_IDX_MASK) {
case GE_VTYPE_IDX_8BIT:
for (int i = 0; i < count; ++i) {
indices[i] = inds[i];
}
break;
case GE_VTYPE_IDX_16BIT:
for (int i = 0; i < count; ++i) {
indices[i] = inds16[i];
}
break;
case GE_VTYPE_IDX_32BIT:
WARN_LOG_REPORT_ONCE(simpleIndexes32, G3D, "SimpleVertices: Decoding 32-bit indexes");
for (int i = 0; i < count; ++i) {
// These aren't documented and should be rare. Let's bounds check each one.
if (inds32[i] != (u16)inds32[i]) {
ERROR_LOG_REPORT_ONCE(simpleIndexes32Bounds, G3D, "SimpleVertices: Index outside 16-bit range");
}
indices[i] = (u16)inds32[i];
}
break;
}
} else {
indices.clear();
}
} else {
indices.clear();
}
static std::vector<u32> temp_buffer;
static std::vector<SimpleVertex> simpleVertices;
temp_buffer.resize(65536 * 24 / sizeof(u32));
simpleVertices.resize(indexUpperBound + 1);
VertexDecoder vdecoder;
VertexDecoderOptions options{};
vdecoder.SetVertexType(gstate.vertType, options);
DrawEngineCommon::NormalizeVertices((u8 *)(&simpleVertices[0]), (u8 *)(&temp_buffer[0]), Memory::GetPointer(gstate_c.vertexAddr), &vdecoder, indexLowerBound, indexUpperBound, gstate.vertType);
float world[16];
float view[16];
float worldview[16];
float worldviewproj[16];
ConvertMatrix4x3To4x4(world, gstate.worldMatrix);
ConvertMatrix4x3To4x4(view, gstate.viewMatrix);
Matrix4ByMatrix4(worldview, world, view);
Matrix4ByMatrix4(worldviewproj, worldview, gstate.projMatrix);
vertices.resize(indexUpperBound + 1);
for (int i = indexLowerBound; i <= indexUpperBound; ++i) {
const SimpleVertex &vert = simpleVertices[i];
if (gstate.isModeThrough()) {
if (gstate.vertType & GE_VTYPE_TC_MASK) {
vertices[i].u = vert.uv[0];
vertices[i].v = vert.uv[1];
} else {
vertices[i].u = 0.0f;
vertices[i].v = 0.0f;
}
vertices[i].x = vert.pos.x;
vertices[i].y = vert.pos.y;
vertices[i].z = vert.pos.z;
if (gstate.vertType & GE_VTYPE_COL_MASK) {
memcpy(vertices[i].c, vert.color, sizeof(vertices[i].c));
} else {
memset(vertices[i].c, 0, sizeof(vertices[i].c));
}
} else {
float clipPos[4];
Vec3ByMatrix44(clipPos, vert.pos.AsArray(), worldviewproj);
ScreenCoords screenPos = ClipToScreen(clipPos);
DrawingCoords drawPos = ScreenToDrawing(screenPos);
if (gstate.vertType & GE_VTYPE_TC_MASK) {
vertices[i].u = vert.uv[0] * (float)gstate.getTextureWidth(0);
vertices[i].v = vert.uv[1] * (float)gstate.getTextureHeight(0);
} else {
vertices[i].u = 0.0f;
vertices[i].v = 0.0f;
}
vertices[i].x = drawPos.x;
vertices[i].y = drawPos.y;
vertices[i].z = drawPos.z;
if (gstate.vertType & GE_VTYPE_COL_MASK) {
memcpy(vertices[i].c, vert.color, sizeof(vertices[i].c));
} else {
memset(vertices[i].c, 0, sizeof(vertices[i].c));
}
}
}
// The GE debugger expects these to be set.
gstate_c.curTextureWidth = gstate.getTextureWidth(0);
gstate_c.curTextureHeight = gstate.getTextureHeight(0);
return true;
}
void BaseUpdateUniforms(UB_VS_FS_Base *ub, uint64_t dirtyUniforms, bool flipViewport) {
if (dirtyUniforms & DIRTY_TEXENV) {
Uint8x3ToFloat4(ub->texEnvColor, gstate.texenvcolor);
}
if (dirtyUniforms & DIRTY_ALPHACOLORREF) {
Uint8x3ToInt4_Alpha(ub->alphaColorRef, gstate.getColorTestRef(), gstate.getAlphaTestRef() & gstate.getAlphaTestMask());
}
if (dirtyUniforms & DIRTY_ALPHACOLORMASK) {
Uint8x3ToInt4_Alpha(ub->colorTestMask, gstate.getColorTestMask(), gstate.getAlphaTestMask());
}
if (dirtyUniforms & DIRTY_FOGCOLOR) {
Uint8x3ToFloat4(ub->fogColor, gstate.fogcolor);
}
if (dirtyUniforms & DIRTY_SHADERBLEND) {
Uint8x3ToFloat4(ub->blendFixA, gstate.getFixA());
Uint8x3ToFloat4(ub->blendFixB, gstate.getFixB());
}
if (dirtyUniforms & DIRTY_TEXCLAMP) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
// First wrap xy, then half texel xy (for clamp.)
ub->texClamp[0] = widthFactor;
ub->texClamp[1] = heightFactor;
ub->texClamp[2] = invW * 0.5f;
ub->texClamp[3] = invH * 0.5f;
ub->texClampOffset[0] = gstate_c.curTextureXOffset * invW;
ub->texClampOffset[1] = gstate_c.curTextureYOffset * invH;
}
if (dirtyUniforms & DIRTY_PROJMATRIX) {
Matrix4x4 flippedMatrix;
memcpy(&flippedMatrix, gstate.projMatrix, 16 * sizeof(float));
const bool invertedY = gstate_c.vpHeight < 0;
if (invertedY) {
flippedMatrix[1] = -flippedMatrix[1];
flippedMatrix[5] = -flippedMatrix[5];
flippedMatrix[9] = -flippedMatrix[9];
flippedMatrix[13] = -flippedMatrix[13];
}
const bool invertedX = gstate_c.vpWidth < 0;
if (invertedX) {
flippedMatrix[0] = -flippedMatrix[0];
flippedMatrix[4] = -flippedMatrix[4];
flippedMatrix[8] = -flippedMatrix[8];
flippedMatrix[12] = -flippedMatrix[12];
}
if (flipViewport) {
ConvertProjMatrixToD3D11(flippedMatrix);
} else {
ConvertProjMatrixToVulkan(flippedMatrix);
}
if (g_Config.iRenderingMode == 0 && g_display_rotation != DisplayRotation::ROTATE_0) {
flippedMatrix = flippedMatrix * g_display_rot_matrix;
}
CopyMatrix4x4(ub->proj, flippedMatrix.getReadPtr());
}
if (dirtyUniforms & DIRTY_PROJTHROUGHMATRIX) {
Matrix4x4 proj_through;
if (flipViewport) {
proj_through.setOrthoD3D(0.0f, gstate_c.curRTWidth, gstate_c.curRTHeight, 0, 0, 1);
} else {
proj_through.setOrthoVulkan(0.0f, gstate_c.curRTWidth, 0, gstate_c.curRTHeight, 0, 1);
}
if (g_Config.iRenderingMode == 0 && g_display_rotation != DisplayRotation::ROTATE_0) {
proj_through = proj_through * g_display_rot_matrix;
}
CopyMatrix4x4(ub->proj_through, proj_through.getReadPtr());
}
// Transform
if (dirtyUniforms & DIRTY_WORLDMATRIX) {
ConvertMatrix4x3To3x4Transposed(ub->world, gstate.worldMatrix);
}
if (dirtyUniforms & DIRTY_VIEWMATRIX) {
ConvertMatrix4x3To3x4Transposed(ub->view, gstate.viewMatrix);
}
if (dirtyUniforms & DIRTY_TEXMATRIX) {
ConvertMatrix4x3To3x4Transposed(ub->tex, gstate.tgenMatrix);
}
// Combined two small uniforms
if (dirtyUniforms & (DIRTY_FOGCOEF | DIRTY_STENCILREPLACEVALUE)) {
float fogcoef_stencil[3] = {
getFloat24(gstate.fog1),
getFloat24(gstate.fog2),
(float)gstate.getStencilTestRef()/255.0f
};
if (my_isinf(fogcoef_stencil[1])) {
// not really sure what a sensible value might be.
fogcoef_stencil[1] = fogcoef_stencil[1] < 0.0f ? -10000.0f : 10000.0f;
} else if (my_isnan(fogcoef_stencil[1])) {
// Workaround for https://github.com/hrydgard/ppsspp/issues/5384#issuecomment-38365988
// Just put the fog far away at a large finite distance.
// Infinities and NaNs are rather unpredictable in shaders on many GPUs
// so it's best to just make it a sane calculation.
fogcoef_stencil[0] = 100000.0f;
fogcoef_stencil[1] = 1.0f;
}
#ifndef MOBILE_DEVICE
else if (my_isnanorinf(fogcoef_stencil[1]) || my_isnanorinf(fogcoef_stencil[0])) {
ERROR_LOG_REPORT_ONCE(fognan, G3D, "Unhandled fog NaN/INF combo: %f %f", fogcoef_stencil[0], fogcoef_stencil[1]);
}
#endif
CopyFloat3(ub->fogCoef_stencil, fogcoef_stencil);
}
// Note - this one is not in lighting but in transformCommon as it has uses beyond lighting
if (dirtyUniforms & DIRTY_MATAMBIENTALPHA) {
Uint8x3ToFloat4_AlphaUint8(ub->matAmbient, gstate.materialambient, gstate.getMaterialAmbientA());
}
// Texturing
if (dirtyUniforms & DIRTY_UVSCALEOFFSET) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
if (gstate_c.bezier || gstate_c.spline) {
// When we are generating UV coordinates through the bezier/spline, we need to apply the scaling.
// However, this is missing a check that we're not getting our UV:s supplied for us in the vertices.
ub->uvScaleOffset[0] = gstate_c.uv.uScale * widthFactor;
ub->uvScaleOffset[1] = gstate_c.uv.vScale * heightFactor;
ub->uvScaleOffset[2] = gstate_c.uv.uOff * widthFactor;
ub->uvScaleOffset[3] = gstate_c.uv.vOff * heightFactor;
} else {
ub->uvScaleOffset[0] = widthFactor;
ub->uvScaleOffset[1] = heightFactor;
ub->uvScaleOffset[2] = 0.0f;
ub->uvScaleOffset[3] = 0.0f;
}
}
if (dirtyUniforms & DIRTY_DEPTHRANGE) {
float viewZScale = gstate.getViewportZScale();
float viewZCenter = gstate.getViewportZCenter();
// We had to scale and translate Z to account for our clamped Z range.
// Therefore, we also need to reverse this to round properly.
//
// Example: scale = 65535.0, center = 0.0
// Resulting range = -65535 to 65535, clamped to [0, 65535]
// gstate_c.vpDepthScale = 2.0f
// gstate_c.vpZOffset = -1.0f
//
// The projection already accounts for those, so we need to reverse them.
//
// Additionally, D3D9 uses a range from [0, 1]. We double and move the center.
viewZScale *= (1.0f / gstate_c.vpDepthScale) * 2.0f;
viewZCenter -= 65535.0f * gstate_c.vpZOffset + 32768.5f;
float viewZInvScale;
if (viewZScale != 0.0) {
viewZInvScale = 1.0f / viewZScale;
} else {
viewZInvScale = 0.0;
}
ub->depthRange[0] = viewZScale;
ub->depthRange[1] = viewZCenter;
ub->depthRange[2] = viewZCenter;
ub->depthRange[3] = viewZInvScale;
}
if (dirtyUniforms & DIRTY_BEZIERSPLINE) {
ub->spline_count_u = gstate_c.spline_count_u;
ub->spline_count_v = gstate_c.spline_count_v;
ub->spline_type_u = gstate_c.spline_type_u;
ub->spline_type_v = gstate_c.spline_type_v;
}
}
void ShaderManagerVulkan::BaseUpdateUniforms(int dirtyUniforms) {
if (dirtyUniforms & DIRTY_TEXENV) {
Uint8x3ToFloat4(ub_base.texEnvColor, gstate.texenvcolor);
}
if (dirtyUniforms & DIRTY_ALPHACOLORREF) {
Uint8x3ToInt4_Alpha(ub_base.alphaColorRef, gstate.getColorTestRef(), gstate.getAlphaTestRef() & gstate.getAlphaTestMask());
}
if (dirtyUniforms & DIRTY_ALPHACOLORMASK) {
Uint8x3ToInt4_Alpha(ub_base.colorTestMask, gstate.getColorTestMask(), gstate.getAlphaTestMask());
}
if (dirtyUniforms & DIRTY_FOGCOLOR) {
Uint8x3ToFloat4(ub_base.fogColor, gstate.fogcolor);
}
if (dirtyUniforms & DIRTY_SHADERBLEND) {
Uint8x3ToFloat4(ub_base.blendFixA, gstate.getFixA());
Uint8x3ToFloat4(ub_base.blendFixB, gstate.getFixB());
}
if (dirtyUniforms & DIRTY_TEXCLAMP) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
// First wrap xy, then half texel xy (for clamp.)
ub_base.texClamp[0] = widthFactor;
ub_base.texClamp[1] = heightFactor;
ub_base.texClamp[2] = invW * 0.5f;
ub_base.texClamp[3] = invH * 0.5f;
ub_base.texClampOffset[0] = gstate_c.curTextureXOffset * invW;
ub_base.texClampOffset[1] = gstate_c.curTextureYOffset * invH;
}
if (dirtyUniforms & DIRTY_PROJMATRIX) {
Matrix4x4 flippedMatrix;
memcpy(&flippedMatrix, gstate.projMatrix, 16 * sizeof(float));
const bool invertedY = gstate_c.vpHeight < 0;
if (invertedY) {
flippedMatrix[1] = -flippedMatrix[1];
flippedMatrix[5] = -flippedMatrix[5];
flippedMatrix[9] = -flippedMatrix[9];
flippedMatrix[13] = -flippedMatrix[13];
}
const bool invertedX = gstate_c.vpWidth < 0;
if (invertedX) {
flippedMatrix[0] = -flippedMatrix[0];
flippedMatrix[4] = -flippedMatrix[4];
flippedMatrix[8] = -flippedMatrix[8];
flippedMatrix[12] = -flippedMatrix[12];
}
ConvertProjMatrixToVulkan(flippedMatrix, invertedX, invertedY);
CopyMatrix4x4(ub_base.proj, flippedMatrix.getReadPtr());
}
if (dirtyUniforms & DIRTY_PROJTHROUGHMATRIX) {
Matrix4x4 proj_through;
proj_through.setOrthoVulkan(0.0f, gstate_c.curRTWidth, 0, gstate_c.curRTHeight, 0, 1);
CopyMatrix4x4(ub_base.proj_through, proj_through.getReadPtr());
}
// Transform
if (dirtyUniforms & DIRTY_WORLDMATRIX) {
ConvertMatrix4x3To4x4(ub_base.world, gstate.worldMatrix);
}
if (dirtyUniforms & DIRTY_VIEWMATRIX) {
ConvertMatrix4x3To4x4(ub_base.view, gstate.viewMatrix);
}
if (dirtyUniforms & DIRTY_TEXMATRIX) {
ConvertMatrix4x3To4x4(ub_base.tex, gstate.tgenMatrix);
}
// Combined two small uniforms
if (dirtyUniforms & (DIRTY_FOGCOEF | DIRTY_STENCILREPLACEVALUE)) {
float fogcoef_stencil[3] = {
getFloat24(gstate.fog1),
getFloat24(gstate.fog2),
(float)gstate.getStencilTestRef()
};
if (my_isinf(fogcoef_stencil[1])) {
// not really sure what a sensible value might be.
fogcoef_stencil[1] = fogcoef_stencil[1] < 0.0f ? -10000.0f : 10000.0f;
} else if (my_isnan(fogcoef_stencil[1])) {
// Workaround for https://github.com/hrydgard/ppsspp/issues/5384#issuecomment-38365988
// Just put the fog far away at a large finite distance.
// Infinities and NaNs are rather unpredictable in shaders on many GPUs
// so it's best to just make it a sane calculation.
fogcoef_stencil[0] = 100000.0f;
fogcoef_stencil[1] = 1.0f;
}
#ifndef MOBILE_DEVICE
else if (my_isnanorinf(fogcoef_stencil[1]) || my_isnanorinf(fogcoef_stencil[0])) {
ERROR_LOG_REPORT_ONCE(fognan, G3D, "Unhandled fog NaN/INF combo: %f %f", fogcoef_stencil[0], fogcoef_stencil[1]);
}
#endif
CopyFloat3(ub_base.fogCoef_stencil, fogcoef_stencil);
}
// Texturing
if (dirtyUniforms & DIRTY_UVSCALEOFFSET) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
ub_base.uvScaleOffset[0] = widthFactor;
ub_base.uvScaleOffset[1] = heightFactor;
ub_base.uvScaleOffset[2] = 0.0f;
ub_base.uvScaleOffset[3] = 0.0f;
}
if (dirtyUniforms & DIRTY_DEPTHRANGE) {
float viewZScale = gstate.getViewportZScale();
float viewZCenter = gstate.getViewportZCenter();
float viewZInvScale;
// We had to scale and translate Z to account for our clamped Z range.
// Therefore, we also need to reverse this to round properly.
//
// Example: scale = 65535.0, center = 0.0
// Resulting range = -65535 to 65535, clamped to [0, 65535]
// gstate_c.vpDepthScale = 2.0f
// gstate_c.vpZOffset = -1.0f
//
// The projection already accounts for those, so we need to reverse them.
//
// Additionally, D3D9 uses a range from [0, 1]. We double and move the center.
viewZScale *= (1.0f / gstate_c.vpDepthScale) * 2.0f;
viewZCenter -= 65535.0f * gstate_c.vpZOffset + 32768.5f;
if (viewZScale != 0.0) {
viewZInvScale = 1.0f / viewZScale;
} else {
viewZInvScale = 0.0;
}
ub_base.depthRange[0] = viewZScale;
ub_base.depthRange[1] = viewZCenter;
ub_base.depthRange[2] = viewZCenter;
ub_base.depthRange[3] = viewZInvScale;
}
}
void ShaderManagerVulkan::BaseUpdateUniforms(int dirtyUniforms) {
if (dirtyUniforms & DIRTY_TEXENV) {
Uint8x3ToFloat4(ub_base.texEnvColor, gstate.texenvcolor);
}
if (dirtyUniforms & DIRTY_ALPHACOLORREF) {
Uint8x3ToInt4_Alpha(ub_base.alphaColorRef, gstate.getColorTestRef(), gstate.getAlphaTestRef() & gstate.getAlphaTestMask());
}
if (dirtyUniforms & DIRTY_ALPHACOLORMASK) {
Uint8x3ToInt4_Alpha(ub_base.colorTestMask, gstate.getColorTestMask(), gstate.getAlphaTestMask());
}
if (dirtyUniforms & DIRTY_FOGCOLOR) {
Uint8x3ToFloat4(ub_base.fogColor, gstate.fogcolor);
}
if (dirtyUniforms & DIRTY_SHADERBLEND) {
Uint8x3ToFloat4(ub_base.blendFixA, gstate.getFixA());
Uint8x3ToFloat4(ub_base.blendFixB, gstate.getFixB());
}
if (dirtyUniforms & DIRTY_TEXCLAMP) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
// First wrap xy, then half texel xy (for clamp.)
const float texclamp[4] = {
widthFactor,
heightFactor,
invW * 0.5f,
invH * 0.5f,
};
const float texclampoff[2] = {
gstate_c.curTextureXOffset * invW,
gstate_c.curTextureYOffset * invH,
};
CopyFloat4(ub_base.texClamp, texclamp);
CopyFloat2(ub_base.texClampOffset, texclampoff);
}
if (dirtyUniforms & DIRTY_PROJMATRIX) {
Matrix4x4 flippedMatrix;
memcpy(&flippedMatrix, gstate.projMatrix, 16 * sizeof(float));
const bool invertedY = gstate_c.vpHeight < 0;
if (invertedY) {
flippedMatrix[1] = -flippedMatrix[1];
flippedMatrix[5] = -flippedMatrix[5];
flippedMatrix[9] = -flippedMatrix[9];
flippedMatrix[13] = -flippedMatrix[13];
}
const bool invertedX = gstate_c.vpWidth < 0;
if (invertedX) {
flippedMatrix[0] = -flippedMatrix[0];
flippedMatrix[4] = -flippedMatrix[4];
flippedMatrix[8] = -flippedMatrix[8];
flippedMatrix[12] = -flippedMatrix[12];
}
ConvertProjMatrixToVulkan(flippedMatrix, invertedX, invertedY);
CopyMatrix4x4(ub_base.proj, flippedMatrix.getReadPtr());
}
if (dirtyUniforms & DIRTY_PROJTHROUGHMATRIX) {
Matrix4x4 proj_through;
proj_through.setOrthoVulkan(0.0f, gstate_c.curRTWidth, 0, gstate_c.curRTHeight, 0, 1);
CopyMatrix4x4(ub_base.proj_through, proj_through.getReadPtr());
}
// Transform
if (dirtyUniforms & DIRTY_WORLDMATRIX) {
ConvertMatrix4x3To4x4(ub_base.world, gstate.worldMatrix);
}
if (dirtyUniforms & DIRTY_VIEWMATRIX) {
ConvertMatrix4x3To4x4(ub_base.view, gstate.viewMatrix);
}
if (dirtyUniforms & DIRTY_TEXMATRIX) {
ConvertMatrix4x3To4x4(ub_base.tex, gstate.tgenMatrix);
}
// Combined two small uniforms
if (dirtyUniforms & (DIRTY_FOGCOEF | DIRTY_STENCILREPLACEVALUE)) {
float fogcoef_stencil[3] = {
getFloat24(gstate.fog1),
getFloat24(gstate.fog2),
(float)gstate.getStencilTestRef()
};
if (my_isinf(fogcoef_stencil[1])) {
// not really sure what a sensible value might be.
fogcoef_stencil[1] = fogcoef_stencil[1] < 0.0f ? -10000.0f : 10000.0f;
} else if (my_isnan(fogcoef_stencil[1])) {
// Workaround for https://github.com/hrydgard/ppsspp/issues/5384#issuecomment-38365988
// Just put the fog far away at a large finite distance.
// Infinities and NaNs are rather unpredictable in shaders on many GPUs
// so it's best to just make it a sane calculation.
fogcoef_stencil[0] = 100000.0f;
fogcoef_stencil[1] = 1.0f;
}
#ifndef MOBILE_DEVICE
else if (my_isnanorinf(fogcoef_stencil[1]) || my_isnanorinf(fogcoef_stencil[0])) {
ERROR_LOG_REPORT_ONCE(fognan, G3D, "Unhandled fog NaN/INF combo: %f %f", fogcoef_stencil[0], fogcoef_stencil[1]);
}
#endif
CopyFloat3(ub_base.fogCoef_stencil, fogcoef_stencil);
}
// Texturing
if (dirtyUniforms & DIRTY_UVSCALEOFFSET) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
static const float rescale[4] = { 1.0f, 2 * 127.5f / 128.f, 2 * 32767.5f / 32768.f, 1.0f };
const float factor = rescale[(gstate.vertType & GE_VTYPE_TC_MASK) >> GE_VTYPE_TC_SHIFT];
float uvscaleoff[4];
switch (gstate.getUVGenMode()) {
case GE_TEXMAP_TEXTURE_COORDS:
// Not sure what GE_TEXMAP_UNKNOWN is, but seen in Riviera. Treating the same as GE_TEXMAP_TEXTURE_COORDS works.
case GE_TEXMAP_UNKNOWN:
if (g_Config.bPrescaleUV) {
// We are here but are prescaling UV in the decoder? Let's do the same as in the other case
// except consider *Scale and *Off to be 1 and 0.
uvscaleoff[0] = widthFactor;
uvscaleoff[1] = heightFactor;
uvscaleoff[2] = 0.0f;
uvscaleoff[3] = 0.0f;
} else {
uvscaleoff[0] = gstate_c.uv.uScale * factor * widthFactor;
uvscaleoff[1] = gstate_c.uv.vScale * factor * heightFactor;
uvscaleoff[2] = gstate_c.uv.uOff * widthFactor;
uvscaleoff[3] = gstate_c.uv.vOff * heightFactor;
}
break;
// These two work the same whether or not we prescale UV.
case GE_TEXMAP_TEXTURE_MATRIX:
// We cannot bake the UV coord scale factor in here, as we apply a matrix multiplication
// before this is applied, and the matrix multiplication may contain translation. In this case
// the translation will be scaled which breaks faces in Hexyz Force for example.
// So I've gone back to applying the scale factor in the shader.
uvscaleoff[0] = widthFactor;
uvscaleoff[1] = heightFactor;
uvscaleoff[2] = 0.0f;
uvscaleoff[3] = 0.0f;
break;
case GE_TEXMAP_ENVIRONMENT_MAP:
// In this mode we only use uvscaleoff to scale to the texture size.
uvscaleoff[0] = widthFactor;
uvscaleoff[1] = heightFactor;
uvscaleoff[2] = 0.0f;
uvscaleoff[3] = 0.0f;
break;
default:
ERROR_LOG_REPORT(G3D, "Unexpected UV gen mode: %d", gstate.getUVGenMode());
}
CopyFloat4(ub_base.uvScaleOffset, uvscaleoff);
}
if (dirtyUniforms & DIRTY_DEPTHRANGE) {
float viewZScale = gstate.getViewportZScale();
float viewZCenter = gstate.getViewportZCenter();
float viewZInvScale;
// We had to scale and translate Z to account for our clamped Z range.
// Therefore, we also need to reverse this to round properly.
//
// Example: scale = 65535.0, center = 0.0
// Resulting range = -65535 to 65535, clamped to [0, 65535]
// gstate_c.vpDepthScale = 2.0f
// gstate_c.vpZOffset = -1.0f
//
// The projection already accounts for those, so we need to reverse them.
//
// Additionally, D3D9 uses a range from [0, 1]. We double and move the center.
viewZScale *= (1.0f / gstate_c.vpDepthScale) * 2.0f;
viewZCenter -= 65535.0f * gstate_c.vpZOffset + 32768.5f;
if (viewZScale != 0.0) {
viewZInvScale = 1.0f / viewZScale;
} else {
viewZInvScale = 0.0;
}
float data[4] = { viewZScale, viewZCenter, viewZCenter, viewZInvScale };
CopyFloat4(ub_base.depthRange, data);
}
}
void ShaderManagerDX9::VSUpdateUniforms(int dirtyUniforms) {
// Update any dirty uniforms before we draw
if (dirtyUniforms & DIRTY_PROJMATRIX) {
Matrix4x4 flippedMatrix;
memcpy(&flippedMatrix, gstate.projMatrix, 16 * sizeof(float));
const bool invertedY = gstate_c.vpHeight < 0;
if (!invertedY) {
flippedMatrix[1] = -flippedMatrix[1];
flippedMatrix[5] = -flippedMatrix[5];
flippedMatrix[9] = -flippedMatrix[9];
flippedMatrix[13] = -flippedMatrix[13];
}
const bool invertedX = gstate_c.vpWidth < 0;
if (invertedX) {
flippedMatrix[0] = -flippedMatrix[0];
flippedMatrix[4] = -flippedMatrix[4];
flippedMatrix[8] = -flippedMatrix[8];
flippedMatrix[12] = -flippedMatrix[12];
}
ConvertProjMatrixToD3D(flippedMatrix, invertedX, invertedY);
VSSetMatrix(CONST_VS_PROJ, flippedMatrix.getReadPtr());
}
if (dirtyUniforms & DIRTY_PROJTHROUGHMATRIX) {
Matrix4x4 proj_through;
proj_through.setOrtho(0.0f, gstate_c.curRTWidth, gstate_c.curRTHeight, 0, 0, 1);
ConvertProjMatrixToD3DThrough(proj_through);
VSSetMatrix(CONST_VS_PROJ_THROUGH, proj_through.getReadPtr());
}
// Transform
if (dirtyUniforms & DIRTY_WORLDMATRIX) {
VSSetMatrix4x3_3(CONST_VS_WORLD, gstate.worldMatrix);
}
if (dirtyUniforms & DIRTY_VIEWMATRIX) {
VSSetMatrix4x3_3(CONST_VS_VIEW, gstate.viewMatrix);
}
if (dirtyUniforms & DIRTY_TEXMATRIX) {
VSSetMatrix4x3_3(CONST_VS_TEXMTX, gstate.tgenMatrix);
}
if (dirtyUniforms & DIRTY_FOGCOEF) {
float fogcoef[2] = {
getFloat24(gstate.fog1),
getFloat24(gstate.fog2),
};
if (my_isinf(fogcoef[1])) {
// not really sure what a sensible value might be.
fogcoef[1] = fogcoef[1] < 0.0f ? -10000.0f : 10000.0f;
} else if (my_isnan(fogcoef[1])) {
// Workaround for https://github.com/hrydgard/ppsspp/issues/5384#issuecomment-38365988
// Just put the fog far away at a large finite distance.
// Infinities and NaNs are rather unpredictable in shaders on many GPUs
// so it's best to just make it a sane calculation.
fogcoef[0] = 100000.0f;
fogcoef[1] = 1.0f;
}
#ifndef MOBILE_DEVICE
else if (my_isnanorinf(fogcoef[1]) || my_isnanorinf(fogcoef[0])) {
ERROR_LOG_REPORT_ONCE(fognan, G3D, "Unhandled fog NaN/INF combo: %f %f", fogcoef[0], fogcoef[1]);
}
#endif
VSSetFloatArray(CONST_VS_FOGCOEF, fogcoef, 2);
}
// TODO: Could even set all bones in one go if they're all dirty.
#ifdef USE_BONE_ARRAY
if (u_bone != 0) {
float allBones[8 * 16];
bool allDirty = true;
for (int i = 0; i < numBones; i++) {
if (dirtyUniforms & (DIRTY_BONEMATRIX0 << i)) {
ConvertMatrix4x3To4x4(allBones + 16 * i, gstate.boneMatrix + 12 * i);
} else {
allDirty = false;
}
}
if (allDirty) {
// Set them all with one call
//glUniformMatrix4fv(u_bone, numBones, GL_FALSE, allBones);
} else {
// Set them one by one. Could try to coalesce two in a row etc but too lazy.
for (int i = 0; i < numBones; i++) {
if (dirtyUniforms & (DIRTY_BONEMATRIX0 << i)) {
//glUniformMatrix4fv(u_bone + i, 1, GL_FALSE, allBones + 16 * i);
}
}
}
}
#else
for (int i = 0; i < 8; i++) {
if (dirtyUniforms & (DIRTY_BONEMATRIX0 << i)) {
VSSetMatrix4x3_3(CONST_VS_BONE0 + 3 * i, gstate.boneMatrix + 12 * i);
}
}
#endif
// Texturing
if (dirtyUniforms & DIRTY_UVSCALEOFFSET) {
const float invW = 1.0f / (float)gstate_c.curTextureWidth;
const float invH = 1.0f / (float)gstate_c.curTextureHeight;
const int w = gstate.getTextureWidth(0);
const int h = gstate.getTextureHeight(0);
const float widthFactor = (float)w * invW;
const float heightFactor = (float)h * invH;
float uvscaleoff[4];
switch (gstate.getUVGenMode()) {
case GE_TEXMAP_TEXTURE_COORDS:
// Not sure what GE_TEXMAP_UNKNOWN is, but seen in Riviera. Treating the same as GE_TEXMAP_TEXTURE_COORDS works.
case GE_TEXMAP_UNKNOWN:
if (g_Config.bPrescaleUV) {
// We are here but are prescaling UV in the decoder? Let's do the same as in the other case
// except consider *Scale and *Off to be 1 and 0.
uvscaleoff[0] = widthFactor;
uvscaleoff[1] = heightFactor;
uvscaleoff[2] = 0.0f;
uvscaleoff[3] = 0.0f;
} else {
uvscaleoff[0] = gstate_c.uv.uScale * widthFactor;
uvscaleoff[1] = gstate_c.uv.vScale * heightFactor;
uvscaleoff[2] = gstate_c.uv.uOff * widthFactor;
uvscaleoff[3] = gstate_c.uv.vOff * heightFactor;
}
break;
// These two work the same whether or not we prescale UV.
case GE_TEXMAP_TEXTURE_MATRIX:
// We cannot bake the UV coord scale factor in here, as we apply a matrix multiplication
// before this is applied, and the matrix multiplication may contain translation. In this case
// the translation will be scaled which breaks faces in Hexyz Force for example.
// So I've gone back to applying the scale factor in the shader.
uvscaleoff[0] = widthFactor;
uvscaleoff[1] = heightFactor;
uvscaleoff[2] = 0.0f;
uvscaleoff[3] = 0.0f;
break;
case GE_TEXMAP_ENVIRONMENT_MAP:
// In this mode we only use uvscaleoff to scale to the texture size.
uvscaleoff[0] = widthFactor;
uvscaleoff[1] = heightFactor;
uvscaleoff[2] = 0.0f;
uvscaleoff[3] = 0.0f;
break;
default:
ERROR_LOG_REPORT(G3D, "Unexpected UV gen mode: %d", gstate.getUVGenMode());
}
VSSetFloatArray(CONST_VS_UVSCALEOFFSET, uvscaleoff, 4);
}
if (dirtyUniforms & DIRTY_DEPTHRANGE) {
// Depth is [0, 1] mapping to [minz, maxz], not too hard.
float vpZScale = gstate.getViewportZScale();
float vpZCenter = gstate.getViewportZCenter();
// These are just the reverse of the formulas in GPUStateUtils.
float halfActualZRange = vpZScale / gstate_c.vpDepthScale;
float minz = -((gstate_c.vpZOffset * halfActualZRange) - vpZCenter) - halfActualZRange;
float viewZScale = halfActualZRange * 2.0f;
// Account for the half pixel offset.
float viewZCenter = minz + (DepthSliceFactor() / 256.0f) * 0.5f;
float viewZInvScale;
if (viewZScale != 0.0) {
viewZInvScale = 1.0f / viewZScale;
} else {
viewZInvScale = 0.0;
}
float data[4] = { viewZScale, viewZCenter, viewZCenter, viewZInvScale };
VSSetFloatUniform4(CONST_VS_DEPTHRANGE, data);
}
// Lighting
if (dirtyUniforms & DIRTY_AMBIENT) {
VSSetColorUniform3Alpha(CONST_VS_AMBIENT, gstate.ambientcolor, gstate.getAmbientA());
}
if (dirtyUniforms & DIRTY_MATAMBIENTALPHA) {
VSSetColorUniform3Alpha(CONST_VS_MATAMBIENTALPHA, gstate.materialambient, gstate.getMaterialAmbientA());
}
if (dirtyUniforms & DIRTY_MATDIFFUSE) {
VSSetColorUniform3(CONST_VS_MATDIFFUSE, gstate.materialdiffuse);
}
if (dirtyUniforms & DIRTY_MATEMISSIVE) {
VSSetColorUniform3(CONST_VS_MATEMISSIVE, gstate.materialemissive);
}
if (dirtyUniforms & DIRTY_MATSPECULAR) {
VSSetColorUniform3ExtraFloat(CONST_VS_MATSPECULAR, gstate.materialspecular, getFloat24(gstate.materialspecularcoef));
}
for (int i = 0; i < 4; i++) {
if (dirtyUniforms & (DIRTY_LIGHT0 << i)) {
if (gstate.isDirectionalLight(i)) {
// Prenormalize
float x = getFloat24(gstate.lpos[i * 3 + 0]);
float y = getFloat24(gstate.lpos[i * 3 + 1]);
float z = getFloat24(gstate.lpos[i * 3 + 2]);
float len = sqrtf(x*x + y*y + z*z);
if (len == 0.0f)
len = 1.0f;
else
len = 1.0f / len;
float vec[3] = { x * len, y * len, z * len };
VSSetFloatArray(CONST_VS_LIGHTPOS + i, vec, 3);
} else {
VSSetFloat24Uniform3(CONST_VS_LIGHTPOS + i, &gstate.lpos[i * 3]);
}
VSSetFloat24Uniform3(CONST_VS_LIGHTDIR + i, &gstate.ldir[i * 3]);
VSSetFloat24Uniform3(CONST_VS_LIGHTATT + i, &gstate.latt[i * 3]);
VSSetFloat(CONST_VS_LIGHTANGLE + i, getFloat24(gstate.lcutoff[i]));
VSSetFloat(CONST_VS_LIGHTSPOTCOEF + i, getFloat24(gstate.lconv[i]));
VSSetColorUniform3(CONST_VS_LIGHTAMBIENT + i, gstate.lcolor[i * 3]);
VSSetColorUniform3(CONST_VS_LIGHTDIFFUSE + i, gstate.lcolor[i * 3 + 1]);
VSSetColorUniform3(CONST_VS_LIGHTSPECULAR + i, gstate.lcolor[i * 3 + 2]);
}
}
}
void GenerateDepalShader100(char *buffer, GEBufferFormat pixelFormat) {
char *p = buffer;
char lookupMethod[128] = "index.r";
char offset[128] = "";
const GEPaletteFormat clutFormat = gstate.getClutPaletteFormat();
const u32 clutBase = gstate.getClutIndexStartPos();
int shift = gstate.getClutIndexShift();
int mask = gstate.getClutIndexMask();
float multiplier = 1.0f;
// pixelformat is the format of the texture we are sampling.
bool formatOK = true;
switch (pixelFormat) {
case GE_FORMAT_8888:
if ((mask & 0xF) == 0xF) {
switch (shift) { // bgra?
case 0: strcpy(lookupMethod, "index.r"); break;
case 4: strcpy(lookupMethod, "index.r"); multiplier = (1.0f / 16.0f); break;
case 8: strcpy(lookupMethod, "index.g"); break;
case 12: strcpy(lookupMethod, "index.g"); multiplier = (1.0f / 16.0f); break;
case 16: strcpy(lookupMethod, "index.b"); break;
case 20: strcpy(lookupMethod, "index.b"); multiplier = (1.0f / 16.0f); break;
case 24: strcpy(lookupMethod, "index.a"); break;
case 28: strcpy(lookupMethod, "index.a"); multiplier = (1.0f / 16.0f); break;
default:
formatOK = false;
}
} else {
formatOK = false;
}
break;
case GE_FORMAT_4444:
if ((mask & 0xF) == 0xF) {
switch (shift) { // bgra?
case 0: strcpy(lookupMethod, "index.r"); break;
case 4: strcpy(lookupMethod, "index.g"); break;
case 8: strcpy(lookupMethod, "index.b"); break;
case 12: strcpy(lookupMethod, "index.a"); break;
default:
formatOK = false;
}
multiplier = 1.0f / 16.0f;
} else {
formatOK = false;
}
break;
case GE_FORMAT_565:
if ((mask & 0x3f) == 0x3F) {
switch (shift) { // bgra?
case 0: strcpy(lookupMethod, "index.r"); multiplier = 1.0f / 32.0f; break;
case 5: strcpy(lookupMethod, "index.g"); multiplier = 1.0f / 64.0f; break;
case 11: strcpy(lookupMethod, "index.b"); multiplier = 1.0f / 32.0f; break;
default:
formatOK = false;
}
} else {
formatOK = false;
}
break;
case GE_FORMAT_5551:
if ((mask & 0x1F) == 0x1F) {
switch (shift) { // bgra?
case 0: strcpy(lookupMethod, "index.r"); multiplier = 1.0f / 32.0f; break;
case 5: strcpy(lookupMethod, "index.g"); multiplier = 1.0f / 32.0f; break;
case 10: strcpy(lookupMethod, "index.b"); multiplier = 1.0f / 32.0f; break;
case 15: strcpy(lookupMethod, "index.a"); multiplier = 1.0f / 256.0f; break;
default:
formatOK = false;
}
} else {
formatOK = false;
}
break;
}
float texturePixels = 256.f;
if (clutFormat != GE_CMODE_32BIT_ABGR8888) {
texturePixels = 512.f;
multiplier *= 0.5f;
}
if (!formatOK) {
ERROR_LOG_REPORT_ONCE(depal, G3D, "%i depal unsupported: shift=%i mask=%02x offset=%i", pixelFormat, shift, mask, offset);
}
// Offset by half a texel (plus clutBase) to turn NEAREST filtering into FLOOR.
sprintf(offset, " + %f", (float)clutBase / texturePixels - 0.5f / texturePixels);
#ifdef USING_GLES
WRITE(p, "#version 100\n");
WRITE(p, "precision mediump float;\n");
#else
WRITE(p, "#version 110\n");
#endif
WRITE(p, "varying vec2 v_texcoord0;\n");
WRITE(p, "uniform sampler2D tex;\n");
WRITE(p, "uniform sampler2D pal;\n");
WRITE(p, "void main() {\n");
WRITE(p, " vec4 index = texture2D(tex, v_texcoord0);\n");
WRITE(p, " gl_FragColor = texture2D(pal, vec2((%s * %f)%s, 0.0));\n", lookupMethod, multiplier, offset);
WRITE(p, "}\n");
}
