Merge pull request #2369 from Armada651/depth-percision

VideoBackends: Use proper floating point depth precision.
This commit is contained in:
Dolphin Bot 2015-05-08 15:20:57 +02:00
commit ca7fe09f1a
9 changed files with 46 additions and 40 deletions

View File

@ -93,10 +93,10 @@ FramebufferManager::FramebufferManager()
D3D::SetDebugObjectName((ID3D11DeviceChild*)m_efb.color_staging_buf, "EFB color staging texture (used for Renderer::AccessEFB)");
// EFB depth buffer - primary depth buffer
texdesc = CD3D11_TEXTURE2D_DESC(DXGI_FORMAT_R24G8_TYPELESS, m_target_width, m_target_height, m_efb.slices, 1, D3D11_BIND_DEPTH_STENCIL | D3D11_BIND_SHADER_RESOURCE, D3D11_USAGE_DEFAULT, 0, sample_desc.Count, sample_desc.Quality);
texdesc = CD3D11_TEXTURE2D_DESC(DXGI_FORMAT_R32_TYPELESS, m_target_width, m_target_height, m_efb.slices, 1, D3D11_BIND_DEPTH_STENCIL | D3D11_BIND_SHADER_RESOURCE, D3D11_USAGE_DEFAULT, 0, sample_desc.Count, sample_desc.Quality);
hr = D3D::device->CreateTexture2D(&texdesc, nullptr, &buf);
CHECK(hr==S_OK, "create EFB depth texture (size: %dx%d; hr=%#x)", m_target_width, m_target_height, hr);
m_efb.depth_tex = new D3DTexture2D(buf, (D3D11_BIND_FLAG)(D3D11_BIND_DEPTH_STENCIL|D3D11_BIND_SHADER_RESOURCE), DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN, (sample_desc.Count > 1));
m_efb.depth_tex = new D3DTexture2D(buf, (D3D11_BIND_FLAG)(D3D11_BIND_DEPTH_STENCIL|D3D11_BIND_SHADER_RESOURCE), DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_D32_FLOAT, DXGI_FORMAT_UNKNOWN, (sample_desc.Count > 1));
SAFE_RELEASE(buf);
D3D::SetDebugObjectName((ID3D11DeviceChild*)m_efb.depth_tex->GetTex(), "EFB depth texture");
D3D::SetDebugObjectName((ID3D11DeviceChild*)m_efb.depth_tex->GetDSV(), "EFB depth texture depth stencil view");

View File

@ -146,7 +146,7 @@ const char depth_matrix_program[] = {
" in float4 pos : SV_Position,\n"
" in float3 uv0 : TEXCOORD0){\n"
" float4 texcol = Tex0.Sample(samp0,uv0);\n"
" int depth = int(round(texcol.x * float(0xFFFFFF)));\n"
" int depth = clamp(int(texcol.x * 16777216.0), 0, 0xFFFFFF);\n"
// Convert to Z24 format
" int4 workspace;\n"
@ -180,7 +180,7 @@ const char depth_matrix_program_msaa[] = {
" for(int i = 0; i < SAMPLES; ++i)\n"
" texcol += Tex0.Load(int3(uv0.x*(width), uv0.y*(height), uv0.z), i);\n"
" texcol /= SAMPLES;\n"
" int depth = int(round(texcol.x * float(0xFFFFFF)));\n"
" int depth = clamp(int(texcol.x * 16777216.0), 0, 0xFFFFFF);\n"
// Convert to Z24 format
" int4 workspace;\n"

View File

@ -8,6 +8,7 @@
#include <strsafe.h>
#include <unordered_map>
#include "Common/MathUtil.h"
#include "Common/Timer.h"
#include "Core/ConfigManager.h"
@ -419,11 +420,11 @@ u32 Renderer::AccessEFB(EFBAccessType type, u32 x, u32 y, u32 poke_data)
if (bpmem.zcontrol.pixel_format == PEControl::RGB565_Z16)
{
// if Z is in 16 bit format you must return a 16 bit integer
ret = ((u32)(val * 0xffff));
ret = MathUtil::Clamp<u32>((u32)(val * 65536.0f), 0, 0xFFFF);
}
else
{
ret = ((u32)(val * 0xffffff));
ret = MathUtil::Clamp<u32>((u32)(val * 16777216.0f), 0, 0xFFFFFF);
}
D3D::context->Unmap(read_tex, 0);
@ -548,7 +549,7 @@ void Renderer::ClearScreen(const EFBRectangle& rc, bool colorEnable, bool alphaE
// Color is passed in bgra mode so we need to convert it to rgba
u32 rgbaColor = (color & 0xFF00FF00) | ((color >> 16) & 0xFF) | ((color << 16) & 0xFF0000);
D3D::drawClearQuad(rgbaColor, (z & 0xFFFFFF) / float(0xFFFFFF));
D3D::drawClearQuad(rgbaColor, (z & 0xFFFFFF) / 16777216.0f);
D3D::stateman->PopDepthState();
D3D::stateman->PopBlendState();

View File

@ -89,7 +89,7 @@ FramebufferManager::FramebufferManager(int targetWidth, int targetHeight, int ms
glTexParameteri(m_textureType, GL_TEXTURE_MAX_LEVEL, 0);
glTexParameteri(m_textureType, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(m_textureType, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTexImage3D(m_textureType, 0, GL_DEPTH_COMPONENT24, m_targetWidth, m_targetHeight, m_EFBLayers, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, nullptr);
glTexImage3D(m_textureType, 0, GL_DEPTH_COMPONENT32F, m_targetWidth, m_targetHeight, m_EFBLayers, 0, GL_DEPTH_COMPONENT, GL_FLOAT, nullptr);
glBindTexture(m_textureType, m_efbColorSwap);
glTexParameteri(m_textureType, GL_TEXTURE_MAX_LEVEL, 0);
@ -111,7 +111,7 @@ FramebufferManager::FramebufferManager(int targetWidth, int targetHeight, int ms
glTexImage3DMultisample(m_textureType, m_msaaSamples, GL_RGBA, m_targetWidth, m_targetHeight, m_EFBLayers, false);
glBindTexture(m_textureType, m_efbDepth);
glTexImage3DMultisample(m_textureType, m_msaaSamples, GL_DEPTH_COMPONENT24, m_targetWidth, m_targetHeight, m_EFBLayers, false);
glTexImage3DMultisample(m_textureType, m_msaaSamples, GL_DEPTH_COMPONENT32F, m_targetWidth, m_targetHeight, m_EFBLayers, false);
glBindTexture(m_textureType, m_efbColorSwap);
glTexImage3DMultisample(m_textureType, m_msaaSamples, GL_RGBA, m_targetWidth, m_targetHeight, m_EFBLayers, false);
@ -125,7 +125,7 @@ FramebufferManager::FramebufferManager(int targetWidth, int targetHeight, int ms
glTexImage2DMultisample(m_textureType, m_msaaSamples, GL_RGBA, m_targetWidth, m_targetHeight, false);
glBindTexture(m_textureType, m_efbDepth);
glTexImage2DMultisample(m_textureType, m_msaaSamples, GL_DEPTH_COMPONENT24, m_targetWidth, m_targetHeight, false);
glTexImage2DMultisample(m_textureType, m_msaaSamples, GL_DEPTH_COMPONENT32F, m_targetWidth, m_targetHeight, false);
glBindTexture(m_textureType, m_efbColorSwap);
glTexImage2DMultisample(m_textureType, m_msaaSamples, GL_RGBA, m_targetWidth, m_targetHeight, false);
@ -150,7 +150,7 @@ FramebufferManager::FramebufferManager(int targetWidth, int targetHeight, int ms
glTexParameteri(resolvedType, GL_TEXTURE_MAX_LEVEL, 0);
glTexParameteri(resolvedType, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(resolvedType, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTexImage3D(resolvedType, 0, GL_DEPTH_COMPONENT24, m_targetWidth, m_targetHeight, m_EFBLayers, 0, GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, nullptr);
glTexImage3D(resolvedType, 0, GL_DEPTH_COMPONENT32F, m_targetWidth, m_targetHeight, m_EFBLayers, 0, GL_DEPTH_COMPONENT, GL_FLOAT, nullptr);
// Bind resolved textures to resolved framebuffer.
glGenFramebuffers(m_EFBLayers, m_resolvedFramebuffer);

View File

@ -12,6 +12,7 @@
#include "Common/Atomic.h"
#include "Common/CommonPaths.h"
#include "Common/FileUtil.h"
#include "Common/MathUtil.h"
#include "Common/StringUtil.h"
#include "Common/Thread.h"
#include "Common/Timer.h"
@ -914,7 +915,7 @@ void ClearEFBCache()
}
}
void Renderer::UpdateEFBCache(EFBAccessType type, u32 cacheRectIdx, const EFBRectangle& efbPixelRc, const TargetRectangle& targetPixelRc, const u32* data)
void Renderer::UpdateEFBCache(EFBAccessType type, u32 cacheRectIdx, const EFBRectangle& efbPixelRc, const TargetRectangle& targetPixelRc, const void* data)
{
u32 cacheType = (type == PEEK_Z ? 0 : 1);
@ -936,7 +937,18 @@ void Renderer::UpdateEFBCache(EFBAccessType type, u32 cacheRectIdx, const EFBRec
u32 xEFB = efbPixelRc.left + xCache;
u32 xPixel = (EFBToScaledX(xEFB) + EFBToScaledX(xEFB + 1)) / 2;
u32 xData = xPixel - targetPixelRc.left;
s_efbCache[cacheType][cacheRectIdx][yCache * EFB_CACHE_RECT_SIZE + xCache] = data[yData * targetPixelRcWidth + xData];
u32 value;
if (type == PEEK_Z)
{
float* ptr = (float*)data;
value = MathUtil::Clamp<u32>((u32)(ptr[yData * targetPixelRcWidth + xData] * 16777216.0f), 0, 0xFFFFFF);
}
else
{
u32* ptr = (u32*)data;
value = ptr[yData * targetPixelRcWidth + xData];
}
s_efbCache[cacheType][cacheRectIdx][yCache * EFB_CACHE_RECT_SIZE + xCache] = value;
}
}
@ -1005,10 +1017,10 @@ u32 Renderer::AccessEFB(EFBAccessType type, u32 x, u32 y, u32 poke_data)
g_renderer->RestoreAPIState();
}
u32* depthMap = new u32[targetPixelRcWidth * targetPixelRcHeight];
float* depthMap = new float[targetPixelRcWidth * targetPixelRcHeight];
glReadPixels(targetPixelRc.left, targetPixelRc.bottom, targetPixelRcWidth, targetPixelRcHeight,
GL_DEPTH_COMPONENT, GL_UNSIGNED_INT, depthMap);
GL_DEPTH_COMPONENT, GL_FLOAT, depthMap);
UpdateEFBCache(type, cacheRectIdx, efbPixelRc, targetPixelRc, depthMap);
@ -1019,18 +1031,10 @@ u32 Renderer::AccessEFB(EFBAccessType type, u32 x, u32 y, u32 poke_data)
u32 yRect = y % EFB_CACHE_RECT_SIZE;
z = s_efbCache[0][cacheRectIdx][yRect * EFB_CACHE_RECT_SIZE + xRect];
// Scale the 32-bit value returned by glReadPixels to a 24-bit
// value (GC uses a 24-bit Z-buffer).
// TODO: in RE0 this value is often off by one, which causes lighting to disappear
if (bpmem.zcontrol.pixel_format == PEControl::RGB565_Z16)
{
// if Z is in 16 bit format you must return a 16 bit integer
z = z >> 16;
}
else
{
if (bpmem.zcontrol.pixel_format == PEControl::RGB565_Z16)
z = z >> 8;
}
return z;
}
@ -1135,7 +1139,7 @@ u32 Renderer::AccessEFB(EFBAccessType type, u32 x, u32 y, u32 poke_data)
ResetAPIState();
glDepthMask(GL_TRUE);
glClearDepthf(float(poke_data & 0xFFFFFF) / float(0xFFFFFF));
glClearDepthf(float(poke_data & 0xFFFFFF) / 16777216.0f);
glEnable(GL_SCISSOR_TEST);
glScissor(targetPixelRc.left, targetPixelRc.bottom, targetPixelRc.GetWidth(), targetPixelRc.GetHeight());
@ -1269,7 +1273,7 @@ void Renderer::ClearScreen(const EFBRectangle& rc, bool colorEnable, bool alphaE
// depth
glDepthMask(zEnable ? GL_TRUE : GL_FALSE);
glClearDepthf(float(z & 0xFFFFFF) / float(0xFFFFFF));
glClearDepthf(float(z & 0xFFFFFF) / 16777216.0f);
// Update rect for clearing the picture
glEnable(GL_SCISSOR_TEST);

View File

@ -91,7 +91,7 @@ public:
int GetMaxTextureSize() override;
private:
void UpdateEFBCache(EFBAccessType type, u32 cacheRectIdx, const EFBRectangle& efbPixelRc, const TargetRectangle& targetPixelRc, const u32* data);
void UpdateEFBCache(EFBAccessType type, u32 cacheRectIdx, const EFBRectangle& efbPixelRc, const TargetRectangle& targetPixelRc, const void* data);
void BlitScreen(TargetRectangle src, TargetRectangle dst, GLuint src_texture, int src_width, int src_height);
};

View File

@ -294,7 +294,7 @@ void TextureCache::CompileShaders()
"\n"
"void main(){\n"
" vec4 texcol = texture(samp9, vec3(f_uv0.xy, %s));\n"
" int depth = int(round(texcol.x * float(0xFFFFFF)));\n"
" int depth = clamp(int(texcol.x * 16777216.0), 0, 0xFFFFFF);\n"
// Convert to Z24 format
" ivec4 workspace;\n"

View File

@ -563,13 +563,14 @@ static inline void GeneratePixelShader(T& out, DSTALPHA_MODE dstAlphaMode, API_T
// The performance impact of this additional calculation doesn't matter, but it prevents
// the host GPU driver from performing any early depth test optimizations.
if (g_ActiveConfig.bFastDepthCalc)
out.Write("\tint zCoord = iround(rawpos.z * float(0xFFFFFF));\n");
out.Write("\tint zCoord = int(rawpos.z * 16777216.0);\n");
else
{
out.SetConstantsUsed(C_ZBIAS+1, C_ZBIAS+1);
// the screen space depth value = far z + (clip z / clip w) * z range
out.Write("\tint zCoord = " I_ZBIAS"[1].x + iround((clipPos.z / clipPos.w) * float(" I_ZBIAS"[1].y));\n");
out.Write("\tint zCoord = " I_ZBIAS"[1].x + int((clipPos.z / clipPos.w) * float(" I_ZBIAS"[1].y));\n");
}
out.Write("\tzCoord = clamp(zCoord, " I_ZBIAS"[1].x - " I_ZBIAS"[1].y, " I_ZBIAS"[1].x);\n");
// depth texture can safely be ignored if the result won't be written to the depth buffer (early_ztest) and isn't used for fog either
const bool skip_ztexture = !per_pixel_depth && !bpmem.fog.c_proj_fsel.fsel;
@ -1169,13 +1170,13 @@ static inline void WriteFog(T& out, pixel_shader_uid_data* uid_data)
// TODO: Verify that we want to drop lower bits here! (currently taken over from software renderer)
// Maybe we want to use "ze = (A << B_SHF)/((B << B_SHF) - Zs)" instead?
// That's equivalent, but keeps the lower bits of Zs.
out.Write("\tfloat ze = (" I_FOGF"[1].x * 16777215.0) / float(" I_FOGI".y - (zCoord >> " I_FOGI".w));\n");
out.Write("\tfloat ze = (" I_FOGF"[1].x * 16777216.0) / float(" I_FOGI".y - (zCoord >> " I_FOGI".w));\n");
}
else
{
// orthographic
// ze = a*Zs (here, no B_SHF)
out.Write("\tfloat ze = " I_FOGF"[1].x * float(zCoord) / 16777215.0;\n");
out.Write("\tfloat ze = " I_FOGF"[1].x * float(zCoord) / 16777216.0;\n");
}
// x_adjust = sqrt((x-center)^2 + k^2)/k
@ -1224,11 +1225,11 @@ static inline void WritePerPixelDepth(T& out, pixel_shader_uid_data* uid_data, A
if (ApiType == API_OPENGL)
out.Write("\tscreenpos.y = %i - screenpos.y;\n", EFB_HEIGHT);
out.Write("\tdepth = float(" I_ZSLOPE".z + " I_ZSLOPE".x * screenpos.x + " I_ZSLOPE".y * screenpos.y) / float(0xFFFFFF);\n");
out.Write("\tdepth = float(" I_ZSLOPE".z + " I_ZSLOPE".x * screenpos.x + " I_ZSLOPE".y * screenpos.y) / 16777216.0;\n");
}
else
{
out.Write("\tdepth = float(zCoord) / float(0xFFFFFF);\n");
out.Write("\tdepth = float(zCoord) / 16777216.0;\n");
}
}

View File

@ -487,7 +487,7 @@ static void WriteZ16Encoder(char*& p,API_TYPE ApiType)
WriteSampleColor(p, "r", "depth", 0, ApiType);
WRITE(p, " depth *= 16777215.0;\n");
WRITE(p, " depth = clamp(depth * 16777216.0, 0.0, float(0xFFFFFF));\n");
WRITE(p, " expanded.r = floor(depth / (256.0 * 256.0));\n");
WRITE(p, " depth -= expanded.r * 256.0 * 256.0;\n");
WRITE(p, " expanded.g = floor(depth / 256.0);\n");
@ -497,7 +497,7 @@ static void WriteZ16Encoder(char*& p,API_TYPE ApiType)
WriteSampleColor(p, "r", "depth", 1, ApiType);
WRITE(p, " depth *= 16777215.0;\n");
WRITE(p, " depth = clamp(depth * 16777216.0, 0.0, float(0xFFFFFF));\n");
WRITE(p, " expanded.r = floor(depth / (256.0 * 256.0));\n");
WRITE(p, " depth -= expanded.r * 256.0 * 256.0;\n");
WRITE(p, " expanded.g = floor(depth / 256.0);\n");
@ -519,7 +519,7 @@ static void WriteZ16LEncoder(char*& p,API_TYPE ApiType)
WriteSampleColor(p, "r", "depth", 0, ApiType);
WRITE(p, " depth *= 16777215.0;\n");
WRITE(p, " depth = clamp(depth * 16777216.0, 0.0, float(0xFFFFFF));\n");
WRITE(p, " expanded.r = floor(depth / (256.0 * 256.0));\n");
WRITE(p, " depth -= expanded.r * 256.0 * 256.0;\n");
WRITE(p, " expanded.g = floor(depth / 256.0);\n");
@ -531,7 +531,7 @@ static void WriteZ16LEncoder(char*& p,API_TYPE ApiType)
WriteSampleColor(p, "r", "depth", 1, ApiType);
WRITE(p, " depth *= 16777215.0;\n");
WRITE(p, " depth = clamp(depth * 16777216.0, 0.0, float(0xFFFFFF));\n");
WRITE(p, " expanded.r = floor(depth / (256.0 * 256.0));\n");
WRITE(p, " depth -= expanded.r * 256.0 * 256.0;\n");
WRITE(p, " expanded.g = floor(depth / 256.0);\n");
@ -558,7 +558,7 @@ static void WriteZ24Encoder(char*& p, API_TYPE ApiType)
for (int i = 0; i < 2; i++)
{
WRITE(p, " depth%i *= 16777215.0;\n", i);
WRITE(p, " depth%i = clamp(depth%i * 16777216.0, 0.0, float(0xFFFFFF));\n", i, i);
WRITE(p, " expanded%i.r = floor(depth%i / (256.0 * 256.0));\n", i, i);
WRITE(p, " depth%i -= expanded%i.r * 256.0 * 256.0;\n", i, i);