piwalker/dolphin
1
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2025-07-23 06:09:50 -06:00
Code Issues Packages Projects Releases Wiki Activity

ok, here goes a really experimental commit:

Browse Source 
replace efb to ram implementation by a hybrid approach.
explanation:
when copying from efb to texture, instead of make a copy to a texture or to the ram, copy the data to both, in hi quality to the texture and in native quality to the ram.
then instead of re-decoding the data from ram (very slow) use the data in the texture.
to improve this even more, test if the cpu has modified the data in the ram copy, if so, update the texture in memory and mark it as dynamic to avoid redundant work in future frames.
having all this implemented this is what is archived:
sms: full quality with scaled efb copies and fully functional goop cleaning :)
ztp: efb to texture speed with full map support.
nsmbw: this is a hard to emulate game, as it make a lot of shading and texture modification in cpu. it only have 35 fps in my system with new efb to ram but is 10 fps faster than normal efb to ram.
this game also show me another unimplemented feature, copy efb to multiple textures at the same time (is used to animate coins and other things in the world).
this is a remaining todo in efb to texture.
a lot of games should improve, so please test and let me know any regresion caused by this commit. 
if everyone likes this the next step is, implement efb to multilpe textures and merge efb to ram and efb to texture.
then port to the other plugins.
enjoy.

git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@5846 8ced0084-cf51-0410-be5f-012b33b47a6e
This commit is contained in:
Rodolfo Osvaldo Bogado
2010-07-06 22:27:13 +00:00
parent 464bac82ff
commit 539f63b58b
9 changed files with 383 additions and 200 deletions
Download Patch File Download Diff File Expand all files Collapse all files

80
Source/Plugins/Plugin_VideoDX9/Src/TextureConverter.cpp
View File

@ -197,8 +197,7 @@ void Shutdown()
void EncodeToRamUsingShader(LPDIRECT3DPIXELSHADER9 shader, LPDIRECT3DTEXTURE9 srcTexture, const TargetRectangle& sourceRc,
u8* destAddr, int dstWidth, int dstHeight, int readStride, bool toTexture, bool linearFilter)
{
HRESULT hr;
Renderer::ResetAPIState();
HRESULT hr;
u32 index =0;
while(index < WorkingBuffers && (TrnBuffers[index].Width != dstWidth || TrnBuffers[index].Height != dstHeight))
index++;
@ -271,10 +270,7 @@ void EncodeToRamUsingShader(LPDIRECT3DPIXELSHADER9 shader, LPDIRECT3DTEXTURE9 sr
// Draw...
D3D::drawShadedTexQuad(srcTexture,&SrcRect,1,1,dstWidth,dstHeight,shader,VertexShaderCache::GetSimpleVertexShader(0));
hr = D3D::dev->SetRenderTarget(0, FBManager.GetEFBColorRTSurface());
hr = D3D::dev->SetDepthStencilSurface(FBManager.GetEFBDepthRTSurface());
Renderer::RestoreAPIState();
D3D::drawShadedTexQuad(srcTexture,&SrcRect,1,1,dstWidth,dstHeight,shader,VertexShaderCache::GetSimpleVertexShader(0));
D3D::RefreshSamplerState(0, D3DSAMP_MINFILTER);
// .. and then readback the results.
// TODO: make this less slow.
@ -347,7 +343,7 @@ void EncodeToRam(u32 address, bool bFromZBuffer, bool bIsIntensityFmt, u32 copyf
// Invalidate any existing texture covering this memory range.
// TODO - don't delete the texture if it already exists, just replace the contents.
TextureCache::InvalidateRange(address, size_in_bytes);
TextureCache::InvalidateRange(address, size_in_bytes);
u16 blkW = TexDecoder_GetBlockWidthInTexels(format) - 1;
u16 blkH = TexDecoder_GetBlockHeightInTexels(format) - 1;
@ -385,10 +381,76 @@ void EncodeToRam(u32 address, bool bFromZBuffer, bool bIsIntensityFmt, u32 copyf
if ((format & 0x0f) == 6)
cacheBytes = 64;
int readStride = (expandedWidth * cacheBytes) / TexDecoder_GetBlockWidthInTexels(format);
Renderer::ResetAPIState();
EncodeToRamUsingShader(texconv_shader, source_texture, scaledSource, dest_ptr, expandedWidth / samples, expandedHeight,readStride, true, bScaleByHalf > 0);
D3D::dev->SetRenderTarget(0, FBManager.GetEFBColorRTSurface());
D3D::dev->SetDepthStencilSurface(FBManager.GetEFBDepthRTSurface());
Renderer::RestoreAPIState();
}
void EncodeToRamFromTexture(u32 address,LPDIRECT3DTEXTURE9 source_texture,u32 SourceW, u32 SourceH,float MValueX,float MValueY,float Xstride, float Ystride , bool bFromZBuffer, bool bIsIntensityFmt, u32 copyfmt, int bScaleByHalf, const EFBRectangle& source)
{
u32 format = copyfmt;
if (bFromZBuffer)
{
format |= _GX_TF_ZTF;
if (copyfmt == 11)
format = GX_TF_Z16;
else if (format < GX_TF_Z8 || format > GX_TF_Z24X8)
format |= _GX_TF_CTF;
}
else
if (copyfmt > GX_TF_RGBA8 || (copyfmt < GX_TF_RGB565 && !bIsIntensityFmt))
format |= _GX_TF_CTF;
LPDIRECT3DPIXELSHADER9 texconv_shader = GetOrCreateEncodingShader(format);
if (!texconv_shader)
return;
u8 *dest_ptr = Memory_GetPtr(address);
int width = (source.right - source.left) >> bScaleByHalf;
int height = (source.bottom - source.top) >> bScaleByHalf;
int size_in_bytes = TexDecoder_GetTextureSizeInBytes(width, height, format);
u16 blkW = TexDecoder_GetBlockWidthInTexels(format) - 1;
u16 blkH = TexDecoder_GetBlockHeightInTexels(format) - 1;
u16 samples = TextureConversionShader::GetEncodedSampleCount(format);
// only copy on cache line boundaries
// extra pixels are copied but not displayed in the resulting texture
s32 expandedWidth = (width + blkW) & (~blkW);
s32 expandedHeight = (height + blkH) & (~blkH);
float sampleStride = bScaleByHalf?2.0f:1.0f;
TextureConversionShader::SetShaderParameters(
(float)expandedWidth,
expandedHeight * MValueY,
source.left * MValueX + Xstride ,
source.top * MValueY + Ystride,
sampleStride * MValueX,
sampleStride * MValueY,
(float)SourceW,
(float)SourceH);
TargetRectangle scaledSource;
scaledSource.top = 0;
scaledSource.bottom = expandedHeight;
scaledSource.left = 0;
scaledSource.right = expandedWidth / samples;
int cacheBytes = 32;
if ((format & 0x0f) == 6)
cacheBytes = 64;
int readStride = (expandedWidth * cacheBytes) / TexDecoder_GetBlockWidthInTexels(format);
EncodeToRamUsingShader(texconv_shader, source_texture, scaledSource, dest_ptr, expandedWidth / samples, expandedHeight,readStride, true, bScaleByHalf > 0);
}
void EncodeToRamYUYV(LPDIRECT3DTEXTURE9 srcTexture, const TargetRectangle& sourceRc,u8* destAddr, int dstWidth, int dstHeight)
{
TextureConversionShader::SetShaderParameters(
@ -400,7 +462,11 @@ void EncodeToRamYUYV(LPDIRECT3DTEXTURE9 srcTexture, const TargetRectangle& sourc
1.0f,
(float)Renderer::GetFullTargetWidth(),
(float)Renderer::GetFullTargetHeight());
Renderer::ResetAPIState();
EncodeToRamUsingShader(s_rgbToYuyvProgram, srcTexture, sourceRc, destAddr, dstWidth / 2, dstHeight, 0, false, false);
D3D::dev->SetRenderTarget(0, FBManager.GetEFBColorRTSurface());
D3D::dev->SetDepthStencilSurface(FBManager.GetEFBDepthRTSurface());
Renderer::RestoreAPIState();
}