// Should be scale free.
void DecodeToTexture(u32 xfbAddr, int srcWidth, int srcHeight, GLuint destTexture)
{
u8* srcAddr = Memory::GetPointer(xfbAddr);
if (!srcAddr)
{
WARN_LOG(VIDEO, "Tried to decode from invalid memory address");
return;
}
g_renderer->ResetAPIState(); // reset any game specific settings
OpenGL_BindAttributelessVAO();
// switch to texture converter frame buffer
// attach destTexture as color destination
FramebufferManager::SetFramebuffer(s_texConvFrameBuffer[1]);
FramebufferManager::FramebufferTexture(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D_ARRAY, destTexture, 0);
// activate source texture
// set srcAddr as data for source texture
glActiveTexture(GL_TEXTURE9);
glBindTexture(GL_TEXTURE_2D, s_srcTexture);
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, srcWidth / 2, srcHeight, 0, GL_BGRA, GL_UNSIGNED_BYTE, srcAddr);
g_sampler_cache->BindNearestSampler(9);
glViewport(0, 0, srcWidth, srcHeight);
s_yuyvToRgbProgram.Bind();
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
FramebufferManager::SetFramebuffer(0);
g_renderer->RestoreAPIState();
}
void FramebufferManager::ReinterpretPixelData(unsigned int convtype)
{
g_renderer->ResetAPIState();
OpenGL_BindAttributelessVAO();
GLuint src_texture = 0;
// We aren't allowed to render and sample the same texture in one draw call,
// so we have to create a new texture and overwrite it completely.
// To not allocate one big texture every time, we've allocated two on
// initialization and just swap them here:
src_texture = m_efbColor;
m_efbColor = m_efbColorSwap;
m_efbColorSwap = src_texture;
FramebufferTexture(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, m_textureType, m_efbColor, 0);
glViewport(0, 0, m_targetWidth, m_targetHeight);
glActiveTexture(GL_TEXTURE9);
glBindTexture(m_textureType, src_texture);
g_sampler_cache->BindNearestSampler(9);
m_pixel_format_shaders[convtype ? 1 : 0].Bind();
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(m_textureType, 0);
g_renderer->RestoreAPIState();
}
static void EncodeToRamUsingShader(GLuint srcTexture,
u8* destAddr, u32 dst_line_size, u32 dstHeight,
u32 writeStride, bool linearFilter)
{
// switch to texture converter frame buffer
// attach render buffer as color destination
FramebufferManager::SetFramebuffer(s_texConvFrameBuffer[0]);
OpenGL_BindAttributelessVAO();
// set source texture
glActiveTexture(GL_TEXTURE9);
glBindTexture(GL_TEXTURE_2D_ARRAY, srcTexture);
if (linearFilter)
g_sampler_cache->BindLinearSampler(9);
else
g_sampler_cache->BindNearestSampler(9);
glViewport(0, 0, (GLsizei)(dst_line_size / 4), (GLsizei)dstHeight);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
int dstSize = dst_line_size * dstHeight;
if ((writeStride != dst_line_size) && (dstHeight > 1))
{
// writing to a texture of a different size
// also copy more then one block line, so the different strides matters
// copy into one pbo first, map this buffer, and then memcpy into GC memory
// in this way, we only have one vram->ram transfer, but maybe a bigger
// CPU overhead because of the pbo
glBindBuffer(GL_PIXEL_PACK_BUFFER, s_PBO);
glBufferData(GL_PIXEL_PACK_BUFFER, dstSize, nullptr, GL_STREAM_READ);
glReadPixels(0, 0, (GLsizei)(dst_line_size / 4), (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE, nullptr);
u8* pbo = (u8*)glMapBufferRange(GL_PIXEL_PACK_BUFFER, 0, dstSize, GL_MAP_READ_BIT);
for (size_t i = 0; i < dstHeight; ++i)
{
memcpy(destAddr, pbo, dst_line_size);
pbo += dst_line_size;
destAddr += writeStride;
}
glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
}
else
{
glReadPixels(0, 0, (GLsizei)(dst_line_size / 4), (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE, destAddr);
}
}
void OpenGLPostProcessing::BlitFromTexture(TargetRectangle src, TargetRectangle dst,
int src_texture, int src_width, int src_height, int layer)
{
ApplyShader();
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
glViewport(dst.left, dst.bottom, dst.GetWidth(), dst.GetHeight());
OpenGL_BindAttributelessVAO();
m_shader.Bind();
glUniform4f(m_uniform_resolution, (float)src_width, (float)src_height, 1.0f / (float)src_width, 1.0f / (float)src_height);
glUniform4f(m_uniform_src_rect, src.left / (float) src_width, src.bottom / (float) src_height,
src.GetWidth() / (float) src_width, src.GetHeight() / (float) src_height);
glUniform1ui(m_uniform_time, (GLuint)m_timer.GetTimeElapsed());
glUniform1i(m_uniform_layer, layer);
if (m_config.IsDirty())
{
for (auto& it : m_config.GetOptions())
{
if (it.second.m_dirty)
{
switch (it.second.m_type)
{
case PostProcessingShaderConfiguration::ConfigurationOption::OptionType::OPTION_BOOL:
glUniform1i(m_uniform_bindings[it.first], it.second.m_bool_value);
break;
case PostProcessingShaderConfiguration::ConfigurationOption::OptionType::OPTION_INTEGER:
switch (it.second.m_integer_values.size())
{
case 1:
glUniform1i(m_uniform_bindings[it.first], it.second.m_integer_values[0]);
break;
case 2:
glUniform2i(m_uniform_bindings[it.first],
it.second.m_integer_values[0],
it.second.m_integer_values[1]);
break;
case 3:
glUniform3i(m_uniform_bindings[it.first],
it.second.m_integer_values[0],
it.second.m_integer_values[1],
it.second.m_integer_values[2]);
break;
case 4:
glUniform4i(m_uniform_bindings[it.first],
it.second.m_integer_values[0],
it.second.m_integer_values[1],
it.second.m_integer_values[2],
it.second.m_integer_values[3]);
break;
}
break;
case PostProcessingShaderConfiguration::ConfigurationOption::OptionType::OPTION_FLOAT:
switch (it.second.m_float_values.size())
{
case 1:
glUniform1f(m_uniform_bindings[it.first], it.second.m_float_values[0]);
break;
case 2:
glUniform2f(m_uniform_bindings[it.first],
it.second.m_float_values[0],
it.second.m_float_values[1]);
break;
case 3:
glUniform3f(m_uniform_bindings[it.first],
it.second.m_float_values[0],
it.second.m_float_values[1],
it.second.m_float_values[2]);
break;
case 4:
glUniform4f(m_uniform_bindings[it.first],
it.second.m_float_values[0],
it.second.m_float_values[1],
it.second.m_float_values[2],
it.second.m_float_values[3]);
break;
}
break;
}
it.second.m_dirty = false;
}
}
m_config.SetDirty(false);
}
glActiveTexture(GL_TEXTURE9);
glBindTexture(GL_TEXTURE_2D_ARRAY, src_texture);
g_sampler_cache->BindLinearSampler(9);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
}
static void EncodeToRamUsingShader(GLuint srcTexture,
u8* destAddr, int dstWidth, int dstHeight, int readStride,
bool linearFilter)
{
// switch to texture converter frame buffer
// attach render buffer as color destination
FramebufferManager::SetFramebuffer(s_texConvFrameBuffer[0]);
OpenGL_BindAttributelessVAO();
// set source texture
glActiveTexture(GL_TEXTURE0+9);
glBindTexture(GL_TEXTURE_2D_ARRAY, srcTexture);
if (linearFilter)
{
glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
}
else
{
glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D_ARRAY, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
}
glViewport(0, 0, (GLsizei)dstWidth, (GLsizei)dstHeight);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
// .. and then read back the results.
// TODO: make this less slow.
int writeStride = bpmem.copyMipMapStrideChannels * 32;
int dstSize = dstWidth*dstHeight*4;
int readHeight = readStride / dstWidth / 4; // 4 bytes per pixel
int readLoops = dstHeight / readHeight;
if (writeStride != readStride && readLoops > 1)
{
// writing to a texture of a different size
// also copy more then one block line, so the different strides matters
// copy into one pbo first, map this buffer, and then memcpy into GC memory
// in this way, we only have one vram->ram transfer, but maybe a bigger
// CPU overhead because of the pbo
glBindBuffer(GL_PIXEL_PACK_BUFFER, s_PBO);
glBufferData(GL_PIXEL_PACK_BUFFER, dstSize, nullptr, GL_STREAM_READ);
glReadPixels(0, 0, (GLsizei)dstWidth, (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE, nullptr);
u8* pbo = (u8*)glMapBufferRange(GL_PIXEL_PACK_BUFFER, 0, dstSize, GL_MAP_READ_BIT);
for (int i = 0; i < readLoops; i++)
{
memcpy(destAddr, pbo, readStride);
pbo += readStride;
destAddr += writeStride;
}
glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
}
else
{
glReadPixels(0, 0, (GLsizei)dstWidth, (GLsizei)dstHeight, GL_BGRA, GL_UNSIGNED_BYTE, destAddr);
}
}
void TextureCache::TCacheEntry::FromRenderTarget(u32 dstAddr, unsigned int dstFormat,
PEControl::PixelFormat srcFormat, const EFBRectangle& srcRect,
bool isIntensity, bool scaleByHalf, unsigned int cbufid,
const float *colmat)
{
g_renderer->ResetAPIState(); // reset any game specific settings
// Make sure to resolve anything we need to read from.
const GLuint read_texture = (srcFormat == PEControl::Z24) ?
FramebufferManager::ResolveAndGetDepthTarget(srcRect) :
FramebufferManager::ResolveAndGetRenderTarget(srcRect);
if (type != TCET_EC_DYNAMIC || g_ActiveConfig.bCopyEFBToTexture)
{
FramebufferManager::SetFramebuffer(framebuffer);
OpenGL_BindAttributelessVAO();
glActiveTexture(GL_TEXTURE0+9);
glBindTexture(GL_TEXTURE_2D_ARRAY, read_texture);
glViewport(0, 0, virtual_width, virtual_height);
GLuint uniform_location;
if (srcFormat == PEControl::Z24)
{
s_DepthMatrixProgram.Bind();
if (s_DepthCbufid != cbufid)
glUniform4fv(s_DepthMatrixUniform, 5, colmat);
s_DepthCbufid = cbufid;
uniform_location = s_DepthCopyPositionUniform;
}
else
{
s_ColorMatrixProgram.Bind();
if (s_ColorCbufid != cbufid)
glUniform4fv(s_ColorMatrixUniform, 7, colmat);
s_ColorCbufid = cbufid;
uniform_location = s_ColorCopyPositionUniform;
}
TargetRectangle R = g_renderer->ConvertEFBRectangle(srcRect);
glUniform4f(uniform_location, static_cast<float>(R.left), static_cast<float>(R.top),
static_cast<float>(R.right), static_cast<float>(R.bottom));
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
}
if (false == g_ActiveConfig.bCopyEFBToTexture)
{
int encoded_size = TextureConverter::EncodeToRamFromTexture(
addr,
read_texture,
srcFormat == PEControl::Z24,
isIntensity,
dstFormat,
scaleByHalf,
srcRect);
u8* dst = Memory::GetPointer(addr);
u64 const new_hash = GetHash64(dst,encoded_size,g_ActiveConfig.iSafeTextureCache_ColorSamples);
// Mark texture entries in destination address range dynamic unless caching is enabled and the texture entry is up to date
if (!g_ActiveConfig.bEFBCopyCacheEnable)
TextureCache::MakeRangeDynamic(addr,encoded_size);
else if (!TextureCache::Find(addr, new_hash))
TextureCache::MakeRangeDynamic(addr,encoded_size);
hash = new_hash;
}
FramebufferManager::SetFramebuffer(0);
if (g_ActiveConfig.bDumpEFBTarget)
{
static int count = 0;
SaveTexture(StringFromFormat("%sefb_frame_%i.png", File::GetUserPath(D_DUMPTEXTURES_IDX).c_str(),
count++), GL_TEXTURE_2D_ARRAY, texture, virtual_width, virtual_height, 0);
}
g_renderer->RestoreAPIState();
}