void CRenderTarget::phase_vol_accumulator()
{
if (!m_bHasActiveVolumetric)
{
m_bHasActiveVolumetric = true;
if( !RImplementation.o.dx10_msaa )
u_setrt (rt_Generic_2, NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Generic_2, NULL,NULL,RImplementation.Target->rt_MSAADepth->pZRT);
//u32 clr4clearVol = color_rgba(0,0,0,0); // 0x00
//CHK_DX (HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_TARGET, clr4clearVol, 1.0f, 0L));
FLOAT ColorRGBA[4] = {0.0f, 0.0f, 0.0f, 0.0f};
HW.pDevice->ClearRenderTargetView( rt_Generic_2->pRT, ColorRGBA);
}
else
{
if( !RImplementation.o.dx10_msaa )
u_setrt (rt_Generic_2, NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Generic_2, NULL,NULL,RImplementation.Target->rt_MSAADepth->pZRT);
}
RCache.set_Stencil (FALSE);
RCache.set_CullMode (CULL_NONE);
RCache.set_ColorWriteEnable ();
}
// startup
void CRenderTarget::phase_scene_prepare ()
{
// Clear depth & stencil
//u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB );
//CHK_DX ( HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_ZBUFFER|D3DCLEAR_STENCIL, 0x0, 1.0f, 0L) );
// Igor: soft particles
CEnvDescriptor& E = *g_pGamePersistent->Environment().CurrentEnv;
float fValue = E.m_fSunShaftsIntensity;
// TODO: add multiplication by sun color here
//if (fValue<0.0001) FlagSunShafts = 0;
if ( RImplementation.o.advancedpp &&
(
ps_r2_ls_flags.test(R2FLAG_SOFT_PARTICLES|R2FLAG_DOF) ||
( (ps_r_sun_shafts>0) && (fValue>=0.0001) ) ||
(ps_r_ssao>0)
)
)
{
u_setrt ( Device.dwWidth,Device.dwHeight,rt_Position->pRT,NULL,NULL,HW.pBaseZB );
CHK_DX ( HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_TARGET|D3DCLEAR_ZBUFFER|D3DCLEAR_STENCIL, 0x0, 1.0f, 0L) );
}
else
{
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB );
CHK_DX ( HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_ZBUFFER|D3DCLEAR_STENCIL, 0x0, 1.0f, 0L) );
}
// Igor: for volumetric lights
m_bHasActiveVolumetric = false;
// Clear later if try to draw volumetric
}
void CRenderTarget::phase_rain()
{
if( !RImplementation.o.dx10_msaa )
u_setrt (rt_Color,NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Color,NULL,NULL,rt_MSAADepth->pZRT);
//u_setrt (rt_Normal,NULL,NULL,HW.pBaseZB);
RImplementation.rmNormal();
}
void CRenderTarget::phase_occq ()
{
if( !RImplementation.o.dx10_msaa )
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
else
u_setrt ( Device.dwWidth,Device.dwHeight,NULL,NULL,NULL,rt_MSAADepth->pZRT);
RCache.set_Shader ( s_occq );
RCache.set_CullMode ( CULL_CCW );
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00);
RCache.set_ColorWriteEnable (FALSE );
}
void CRenderTarget::accum_direct_blend ()
{
// blend-copy
if (!RImplementation.o.fp16_blend) {
u_setrt (rt_Accumulator,NULL,NULL,HW.pBaseZB);
// Common calc for quad-rendering
u32 Offset;
u32 C = color_rgba (255,255,255,255);
float _w = float (Device.dwWidth);
float _h = float (Device.dwHeight);
Fvector2 p0,p1;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
float d_Z = EPS_S, d_W = 1.f;
// Fill vertex buffer
FVF::TL2uv* pv = (FVF::TL2uv*) RCache.Vertex.Lock (4,g_combine_2UV->vb_stride,Offset);
pv->set (EPS, float(_h+EPS), d_Z, d_W, C, p0.x, p1.y, p0.x, p1.y); pv++;
pv->set (EPS, EPS, d_Z, d_W, C, p0.x, p0.y, p0.x, p0.y); pv++;
pv->set (float(_w+EPS), float(_h+EPS), d_Z, d_W, C, p1.x, p1.y, p1.x, p1.y); pv++;
pv->set (float(_w+EPS), EPS, d_Z, d_W, C, p1.x, p0.y, p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine_2UV->vb_stride);
RCache.set_Geometry (g_combine_2UV);
RCache.set_Element (s_accum_mask->E[SE_MASK_ACCUM_2D] );
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,dwLightMarkerID,0xff,0x00);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2 );
}
dwLightMarkerID += 2;
}
// begin
void CRenderTarget::phase_scene_begin ()
{
// Enable ANISO
for (u32 i=0; i<HW.Caps.raster.dwStages; i++)
CHK_DX(HW.pDevice->SetSamplerState( i, D3DSAMP_MAXANISOTROPY, ps_r__tf_Anisotropic ));
// Targets, use accumulator for temporary storage
if (RImplementation.o.albedo_wo) u_setrt (rt_Position, rt_Normal, rt_Accumulator, HW.pBaseZB);
else u_setrt (rt_Position, rt_Normal, rt_Color, HW.pBaseZB);
// Stencil - write 0x1 at pixel pos
RCache.set_Stencil ( TRUE,D3DCMP_ALWAYS,0x01,0xff,0xff,D3DSTENCILOP_KEEP,D3DSTENCILOP_REPLACE,D3DSTENCILOP_KEEP);
// Misc - draw only front-faces
CHK_DX(HW.pDevice->SetRenderState ( D3DRS_TWOSIDEDSTENCILMODE,FALSE ));
RCache.set_CullMode ( CULL_CCW );
RCache.set_ColorWriteEnable ( );
}
void CRenderTarget::phase_smap_direct (light* L, u32 sub_phase)
{
// TODO: DX10: Check thst we will never need old SMap implementation
// Targets
if (RImplementation.o.HW_smap) u_setrt (rt_smap_surf, NULL, NULL, rt_smap_depth->pZRT);
//else u_setrt (rt_smap_surf, NULL, NULL, rt_smap_ZB);
else VERIFY(!"Use HW SMap only for DX10!");
// Don't have rect clear for DX10
// TODO: DX9: Full clear must be faster for the near phase for SLI
// inobody clears this buffer _this_ frame.
// Clear
//if (SE_SUN_NEAR==sub_phase) {
// optimized clear
// D3DRECT R;
// R.x1 = L->X.D.minX;
// R.x2 = L->X.D.maxX;
// R.y1 = L->X.D.minY;
// R.y2 = L->X.D.maxY;
// CHK_DX (HW.pDevice->Clear( 1L, &R, D3DCLEAR_ZBUFFER, 0xFFFFFFFF, 1.0f, 0L));
//} else {
// full-clear
// CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_ZBUFFER, 0xFFFFFFFF, 1.0f, 0L));
//}
HW.pContext->ClearDepthStencilView(rt_smap_depth->pZRT, D3D_CLEAR_DEPTH, 1.0f, 0L);
// Prepare viewport for shadow map rendering
if (sub_phase!=SE_SUN_RAIN_SMAP )
RImplementation.rmNormal();
else
{
D3D_VIEWPORT VP = {(float)L->X.D.minX,(float)L->X.D.minY,
(float)(L->X.D.maxX - L->X.D.minX) ,
(float)(L->X.D.maxY - L->X.D.minY) ,
0,1 };
//CHK_DX (HW.pDevice->SetViewport(&VP));
HW.pContext->RSSetViewports(1, &VP);
}
// Stencil - disable
RCache.set_Stencil ( FALSE );
// TODO: DX10: Implement culling reverse for DX10
// Misc - draw only front/back-faces
/*
if (SE_SUN_NEAR==sub_phase) RCache.set_CullMode ( CULL_CCW ); // near
else {
if (RImplementation.o.HW_smap) RCache.set_CullMode ( CULL_CW ); // far, reversed
else RCache.set_CullMode ( CULL_CCW ); // far, front-faces
}
if (RImplementation.o.HW_smap) RCache.set_ColorWriteEnable ( FALSE );
else RCache.set_ColorWriteEnable ( );
*/
}
void CRenderTarget::phase_pp ()
{
// combination/postprocess
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
// Element 0 for for normal post-process
// Element 4 for color map post-process
bool bCMap = u_need_CM();
//RCache.set_Element (s_postprocess->E[bCMap ? 4 : 0]);
if( !RImplementation.o.dx10_msaa )
{
// RCache.set_Shader (s_postprocess );
RCache.set_Element (s_postprocess->E[bCMap ? 4 : 0]);
}
else
{
// RCache.set_Shader( s_postprocess_msaa );
RCache.set_Element (s_postprocess_msaa->E[bCMap ? 4 : 0]);
}
int gblend = clampr (iFloor((1-param_gray)*255.f),0,255);
int nblend = clampr (iFloor((1-param_noise)*255.f),0,255);
u32 p_color = subst_alpha (param_color_base,nblend);
u32 p_gray = subst_alpha (param_color_gray,gblend);
Fvector p_brightness = param_color_add ;
// Msg ("param_gray:%f(%d),param_noise:%f(%d)",param_gray,gblend,param_noise,nblend);
// Msg ("base: %d,%d,%d", color_get_R(p_color), color_get_G(p_color), color_get_B(p_color));
// Msg ("gray: %d,%d,%d", color_get_R(p_gray), color_get_G(p_gray), color_get_B(p_gray));
// Msg ("add: %d,%d,%d", color_get_R(p_brightness), color_get_G(p_brightness), color_get_B(p_brightness));
// Draw full-screen quad textured with our scene image
u32 Offset;
float _w = float(Device.dwWidth);
float _h = float(Device.dwHeight);
Fvector2 n0,n1,r0,r1,l0,l1;
u_calc_tc_duality_ss (r0,r1,l0,l1);
u_calc_tc_noise (n0,n1);
// Fill vertex buffer
float du = ps_r1_pps_u, dv = ps_r1_pps_v;
TL_2c3uv* pv = (TL_2c3uv*) RCache.Vertex.Lock (4,g_postprocess.stride(),Offset);
pv->set(du+0, dv+float(_h), p_color, p_gray, r0.x, r1.y, l0.x, l1.y, n0.x, n1.y); pv++;
pv->set(du+0, dv+0, p_color, p_gray, r0.x, r0.y, l0.x, l0.y, n0.x, n0.y); pv++;
pv->set(du+float(_w), dv+float(_h), p_color, p_gray, r1.x, r1.y, l1.x, l1.y, n1.x, n1.y); pv++;
pv->set(du+float(_w), dv+0, p_color, p_gray, r1.x, r0.y, l1.x, l0.y, n1.x, n0.y); pv++;
RCache.Vertex.Unlock (4,g_postprocess.stride());
// Actual rendering
static shared_str s_brightness = "c_brightness";
static shared_str s_colormap = "c_colormap";
RCache.set_c ( s_brightness, p_brightness.x, p_brightness.y, p_brightness.z, 0 );
RCache.set_c (s_colormap, param_color_map_influence,param_color_map_interpolate,0,0);
RCache.set_Geometry (g_postprocess);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
void CRenderTarget::phase_wallmarks ()
{
// Targets
RCache.set_RT(NULL,2);
RCache.set_RT(NULL,1);
u_setrt (rt_Color,NULL,NULL,HW.pBaseZB);
// Stencil - draw only where stencil >= 0x1
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00);
RCache.set_CullMode (CULL_CCW);
RCache.set_ColorWriteEnable (D3DCOLORWRITEENABLE_RED|D3DCOLORWRITEENABLE_GREEN|D3DCOLORWRITEENABLE_BLUE);
}
void CRenderTarget::phase_downsamp ()
{
// DON'T DO THIS!!!
//IDirect3DSurface9 *source, *dest;
//rt_Position->pSurface->GetSurfaceLevel(0, &source);
//rt_half_depth->pSurface->GetSurfaceLevel(0, &dest);
//HW.pDevice->StretchRect(source, NULL, dest, NULL, D3DTEXF_POINT);
//Fvector2 p0,p1;
u32 Offset = 0;
u_setrt( rt_half_depth,0,0,0/*HW.pBaseZB*/ );
FLOAT ColorRGBA[4] = {0.0f, 0.0f, 0.0f, 0.0f};
HW.pContext->ClearRenderTargetView(rt_half_depth->pRT, ColorRGBA);
u32 w = Device.dwWidth;
u32 h = Device.dwHeight;
if (RImplementation.o.ssao_half_data)
{
set_viewport(HW.pDevice, Device.dwWidth/2, Device.dwHeight/2);
w /= 2;
h /= 2;
}
RCache.set_Stencil (FALSE);
{
Fmatrix m_v2w; m_v2w.invert (Device.mView );
// Fill VB
float scale_X = float(w) / float(TEX_jitter);
float scale_Y = float(h) / float(TEX_jitter);
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set ( -1, 1, 0, 1, 0, 0, scale_Y ); pv++;
pv->set ( -1, -1, 0, 0, 0, 0, 0 ); pv++;
pv->set ( 1, 1, 1, 1, 0, scale_X, scale_Y ); pv++;
pv->set ( 1, -1, 1, 0, 0, scale_X, 0 ); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
// Draw
RCache.set_Element (s_ssao->E[1] );
RCache.set_Geometry (g_combine );
RCache.set_c ("m_v2w", m_v2w );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
if (RImplementation.o.ssao_half_data)
set_viewport(HW.pDevice, Device.dwWidth, Device.dwHeight);
}
void CRenderTarget::phase_accumulator()
{
// Targets
if (dwAccumulatorClearMark==Device.dwFrame) {
// normal operation - setup
if( !RImplementation.o.dx10_msaa )
{
if (RImplementation.o.fp16_blend) u_setrt (rt_Accumulator, NULL,NULL,HW.pBaseZB);
else u_setrt (rt_Accumulator_temp, NULL,NULL,HW.pBaseZB);
}
else
{
if (RImplementation.o.fp16_blend) u_setrt (rt_Accumulator, NULL,NULL, rt_MSAADepth->pZRT);
else u_setrt (rt_Accumulator_temp, NULL,NULL, rt_MSAADepth->pZRT);
}
} else {
// initial setup
dwAccumulatorClearMark = Device.dwFrame;
// clear
if( !RImplementation.o.dx10_msaa )
u_setrt (rt_Accumulator, NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Accumulator, NULL,NULL,rt_MSAADepth->pZRT);
//dwLightMarkerID = 5; // start from 5, increment in 2 units
reset_light_marker();
// Igor: AMD bug workaround. Should be fixed in 8.7 catalyst
// Need for MSAA to work correctly.
if( RImplementation.o.dx10_msaa )
{
HW.pDevice->OMSetRenderTargets(1, &(rt_Accumulator->pRT), 0);
}
// u32 clr4clear = color_rgba(0,0,0,0); // 0x00
//CHK_DX (HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_TARGET, clr4clear, 1.0f, 0L));
FLOAT ColorRGBA[4] = {0.0f, 0.0f, 0.0f, 0.0f};
HW.pDevice->ClearRenderTargetView( rt_Accumulator->pRT, ColorRGBA);
// render this after sun to avoid troubles with sun
/*
// Render emissive geometry, stencil - write 0x0 at pixel pos
RCache.set_xform_project (Device.mProject);
RCache.set_xform_view (Device.mView);
// Stencil - write 0x1 at pixel pos -
RCache.set_Stencil ( TRUE,D3DCMP_ALWAYS,0x01,0xff,0xff,D3DSTENCILOP_KEEP,D3DSTENCILOP_REPLACE,D3DSTENCILOP_KEEP);
//RCache.set_Stencil (TRUE,D3DCMP_ALWAYS,0x00,0xff,0xff,D3DSTENCILOP_KEEP,D3DSTENCILOP_REPLACE,D3DSTENCILOP_KEEP);
RCache.set_CullMode (CULL_CCW);
RCache.set_ColorWriteEnable ();
RImplementation.r_dsgraph_render_emissive ();
*/
// Stencil - draw only where stencil >= 0x1
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00);
RCache.set_CullMode (CULL_NONE);
RCache.set_ColorWriteEnable ();
}
// Restore viewport after shadow map rendering
RImplementation.rmNormal();
}
void CRenderTarget::phase_smap_direct (light* L, u32 sub_phase)
{
// Targets
if (RImplementation.o.HW_smap) u_setrt (rt_smap_surf, NULL, NULL, rt_smap_depth->pRT);
else u_setrt (rt_smap_surf, NULL, NULL, rt_smap_ZB);
// Clear
if (SE_SUN_NEAR==sub_phase) {
// optimized clear
D3DRECT R;
R.x1 = L->X.D.minX;
R.x2 = L->X.D.maxX;
R.y1 = L->X.D.minY;
R.y2 = L->X.D.maxY;
CHK_DX (HW.pDevice->Clear( 1L, &R, D3DCLEAR_ZBUFFER, 0xFFFFFFFF, 1.0f, 0L));
} else {
// full-clear
CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_ZBUFFER, 0xFFFFFFFF, 1.0f, 0L));
}
// Stencil - disable
RCache.set_Stencil ( FALSE );
// Misc - draw only front/back-faces
/*
if (SE_SUN_NEAR==sub_phase) RCache.set_CullMode ( CULL_CCW ); // near
else {
if (RImplementation.o.HW_smap) RCache.set_CullMode ( CULL_CW ); // far, reversed
else RCache.set_CullMode ( CULL_CCW ); // far, front-faces
}
*/
// Cull always CCW. If you want to revert to previouse solution, please, revert bias setup/
RCache.set_CullMode ( CULL_CCW ); // near
if (RImplementation.o.HW_smap) RCache.set_ColorWriteEnable ( FALSE );
else RCache.set_ColorWriteEnable ( );
}
void CRenderTarget::phase_smap_spot (light* L)
{
// Targets + viewport
// TODO: DX10: CHeck if we don't need old-style SMAP
if (RImplementation.o.HW_smap) u_setrt (rt_smap_surf, NULL, NULL, rt_smap_depth->pZRT);
//else u_setrt (rt_smap_surf, NULL, NULL, rt_smap_ZB);
else VERIFY(!"Use HW SMap only for DX10!");
D3D_VIEWPORT VP = {(float)L->X.S.posX, (float)L->X.S.posY, (float)L->X.S.size, (float)L->X.S.size, 0, 1};
//CHK_DX (HW.pDevice->SetViewport(&VP));
HW.pContext->RSSetViewports(1, &VP);
// Misc - draw only front-faces //back-faces
RCache.set_CullMode ( CULL_CCW );
RCache.set_Stencil ( FALSE );
// no transparency
#pragma todo("can optimize for multi-lights covering more than say 50%...")
if (RImplementation.o.HW_smap) RCache.set_ColorWriteEnable (FALSE);
//CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_ZBUFFER, 0xffffffff, 1.0f, 0L));
// Do it once per smap generation pass in phase_smap_spot_clear
//HW.pContext->ClearDepthStencilView( rt_smap_depth->pZRT, D3D_CLEAR_DEPTH, 1.0f, 0L);
}
// end
void CRenderTarget::phase_scene_end ()
{
disable_aniso ();
if (!RImplementation.o.albedo_wo) return;
// transfer from "rt_Accumulator" into "rt_Color"
u_setrt ( rt_Color, 0, 0, HW.pBaseZB );
RCache.set_CullMode ( CULL_NONE );
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00); // stencil should be >= 1
if (RImplementation.o.nvstencil) u_stencil_optimize (FALSE);
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00); // stencil should be >= 1
RCache.set_ColorWriteEnable ();
// common calc for quad-rendering
u32 Offset;
u32 C = color_rgba (255,255,255,255);
float _w = float (Device.dwWidth);
float _h = float (Device.dwHeight);
Fvector2 p0,p1;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
float d_Z = EPS_S, d_W = 1.f;
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set (EPS, float(_h+EPS), d_Z, d_W, C, p0.x, p1.y); pv++;
pv->set (EPS, EPS, d_Z, d_W, C, p0.x, p0.y); pv++;
pv->set (float(_w+EPS), float(_h+EPS), d_Z, d_W, C, p1.x, p1.y); pv++;
pv->set (float(_w+EPS), EPS, d_Z, d_W, C, p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
// if (stencil>=1 && aref_pass) stencil = light_id
RCache.set_Element (s_accum_mask->E[SE_MASK_ALBEDO]); // masker
RCache.set_Geometry (g_combine);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
void CRenderTarget::mark_msaa_edges()
{
u32 Offset;
float d_Z = EPS_S, d_W = 1.f;
u32 C = color_rgba (255,255,255,255);
// Fill vertex buffer
FVF::TL2uv* pv = (FVF::TL2uv*) RCache.Vertex.Lock (4,g_combine_2UV->vb_stride,Offset);
pv->set (-1, -1, 0, d_W, C, 0, 1, 0, 0); pv++;
pv->set (-1, 1, d_Z, d_W, C, 0, 0, 0, 0); pv++;
pv->set (1, -1, d_Z, d_W, C, 1, 1, 0, 0); pv++;
pv->set (1, 1, d_Z, d_W, C, 1, 0, 0, 0); pv++;
RCache.Vertex.Unlock (4,g_combine_2UV->vb_stride);
u_setrt (NULL,NULL,NULL,rt_MSAADepth->pZRT);
RCache.set_Element (s_mark_msaa_edges->E[0]);
RCache.set_Geometry (g_combine_2UV);
StateManager.SetStencil( TRUE, D3DCMP_ALWAYS, 0x80, 0xFF, 0x80, D3DSTENCILOP_KEEP, D3DSTENCILOP_REPLACE, D3DSTENCILOP_KEEP );
StateManager.SetColorWriteEnable( 0 );
StateManager.SetDepthFunc( D3DCMP_ALWAYS );
StateManager.SetDepthEnable( FALSE );
StateManager.SetCullMode( D3DCULL_NONE );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
StateManager.SetColorWriteEnable( D3D10_COLOR_WRITE_ENABLE_ALL );
}
CRenderTarget::CRenderTarget ()
{
param_blur = 0.f;
param_gray = 0.f;
param_noise = 0.f;
param_duality_h = 0.f;
param_duality_v = 0.f;
param_noise_fps = 25.f;
param_noise_scale = 1.f;
im_noise_time = 1/100;
im_noise_shift_w = 0;
im_noise_shift_h = 0;
param_color_base = color_rgba(127,127,127, 0);
param_color_gray = color_rgba(85,85,85, 0);
param_color_add.set( 0.0f, 0.0f, 0.0f );
dwAccumulatorClearMark = 0;
dxRenderDeviceRender::Instance().Resources->Evict ();
// Blenders
b_occq = new CBlender_light_occq();
b_accum_mask = new CBlender_accum_direct_mask();
b_accum_direct = new CBlender_accum_direct();
b_accum_point = new CBlender_accum_point();
b_accum_spot = new CBlender_accum_spot();
b_accum_reflected = new CBlender_accum_reflected();
b_bloom = new CBlender_bloom_build();
b_ssao = new CBlender_SSAO();
b_luminance = new CBlender_luminance();
b_combine = new CBlender_combine();
// NORMAL
{
u32 w=Device.dwWidth, h=Device.dwHeight;
rt_Position.create (r2_RT_P, w,h,D3DFMT_A16B16G16R16F);
rt_Normal.create (r2_RT_N, w,h,D3DFMT_A16B16G16R16F);
// select albedo & accum
if (RImplementation.o.mrtmixdepth)
{
// NV50
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8 );
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F);
}
else
{
// can't - mix-depth
if (RImplementation.o.fp16_blend) {
// NV40
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A16B16G16R16F); // expand to full
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F);
} else {
// R4xx, no-fp-blend,-> albedo_wo
VERIFY (RImplementation.o.albedo_wo);
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8 ); // normal
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F);
rt_Accumulator_temp.create (r2_RT_accum_temp, w,h,D3DFMT_A16B16G16R16F);
}
}
// generic(LDR) RTs
rt_Generic_0.create (r2_RT_generic0,w,h,D3DFMT_A8R8G8B8 );
rt_Generic_1.create (r2_RT_generic1,w,h,D3DFMT_A8R8G8B8 );
// Igor: for volumetric lights
//rt_Generic_2.create (r2_RT_generic2,w,h,D3DFMT_A8R8G8B8 );
// temp: for higher quality blends
if (RImplementation.o.advancedpp)
rt_Generic_2.create (r2_RT_generic2,w,h,D3DFMT_A16B16G16R16F);
}
// OCCLUSION
s_occq.create (b_occq, "r2\\occq");
// DIRECT (spot)
D3DFORMAT depth_format = (D3DFORMAT)RImplementation.o.HW_smap_FORMAT;
if (RImplementation.o.HW_smap)
{
D3DFORMAT nullrt = D3DFMT_R5G6B5;
if (RImplementation.o.nullrt) nullrt = (D3DFORMAT)MAKEFOURCC('N','U','L','L');
u32 size =RImplementation.o.smapsize ;
rt_smap_depth.create (r2_RT_smap_depth, size,size,depth_format );
rt_smap_surf.create (r2_RT_smap_surf, size,size,nullrt );
rt_smap_ZB = NULL;
s_accum_mask.create (b_accum_mask, "r2\\accum_mask");
s_accum_direct.create (b_accum_direct, "r2\\accum_direct");
if (RImplementation.o.advancedpp)
s_accum_direct_volumetric.create("accum_volumetric_sun");
}
else
{
u32 size =RImplementation.o.smapsize ;
rt_smap_surf.create (r2_RT_smap_surf, size,size,D3DFMT_R32F);
rt_smap_depth = NULL;
R_CHK (HW.pDevice->CreateDepthStencilSurface (size,size,D3DFMT_D24X8,D3DMULTISAMPLE_NONE,0,TRUE,&rt_smap_ZB,NULL));
s_accum_mask.create (b_accum_mask, "r2\\accum_mask");
s_accum_direct.create (b_accum_direct, "r2\\accum_direct");
if (RImplementation.o.advancedpp)
s_accum_direct_volumetric.create("accum_volumetric_sun");
}
// POINT
{
s_accum_point.create (b_accum_point, "r2\\accum_point_s");
accum_point_geom_create ();
g_accum_point.create (D3DFVF_XYZ, g_accum_point_vb, g_accum_point_ib);
accum_omnip_geom_create ();
g_accum_omnipart.create (D3DFVF_XYZ, g_accum_omnip_vb, g_accum_omnip_ib);
}
// SPOT
{
s_accum_spot.create (b_accum_spot, "r2\\accum_spot_s", "lights\\lights_spot01");
accum_spot_geom_create ();
g_accum_spot.create (D3DFVF_XYZ, g_accum_spot_vb, g_accum_spot_ib);
}
{
s_accum_volume.create("accum_volumetric", "lights\\lights_spot01");
accum_volumetric_geom_create();
g_accum_volumetric.create( D3DFVF_XYZ, g_accum_volumetric_vb, g_accum_volumetric_ib);
}
// REFLECTED
{
s_accum_reflected.create (b_accum_reflected, "r2\\accum_refl");
}
// BLOOM
{
D3DFORMAT fmt = D3DFMT_A8R8G8B8; //; // D3DFMT_X8R8G8B8
u32 w=BLOOM_size_X, h=BLOOM_size_Y;
u32 fvf_build = D3DFVF_XYZRHW|D3DFVF_TEX4|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3);
u32 fvf_filter = (u32)D3DFVF_XYZRHW|D3DFVF_TEX8|D3DFVF_TEXCOORDSIZE4(0)|D3DFVF_TEXCOORDSIZE4(1)|D3DFVF_TEXCOORDSIZE4(2)|D3DFVF_TEXCOORDSIZE4(3)|D3DFVF_TEXCOORDSIZE4(4)|D3DFVF_TEXCOORDSIZE4(5)|D3DFVF_TEXCOORDSIZE4(6)|D3DFVF_TEXCOORDSIZE4(7);
rt_Bloom_1.create (r2_RT_bloom1, w,h, fmt);
rt_Bloom_2.create (r2_RT_bloom2, w,h, fmt);
g_bloom_build.create (fvf_build, RCache.Vertex.Buffer(), RCache.QuadIB);
g_bloom_filter.create (fvf_filter, RCache.Vertex.Buffer(), RCache.QuadIB);
s_bloom_dbg_1.create ("effects\\screen_set", r2_RT_bloom1);
s_bloom_dbg_2.create ("effects\\screen_set", r2_RT_bloom2);
s_bloom.create (b_bloom, "r2\\bloom");
f_bloom_factor = 0.5f;
}
//HBAO
if (RImplementation.o.ssao_opt_data)
{
u32 w = 0;
u32 h = 0;
if (RImplementation.o.ssao_half_data)
{
w = Device.dwWidth / 2;
h = Device.dwHeight / 2;
}
else
{
w = Device.dwWidth;
h = Device.dwHeight;
}
D3DFORMAT fmt = HW.Caps.id_vendor==0x10DE?D3DFMT_R32F:D3DFMT_R16F;
rt_half_depth.create (r2_RT_half_depth, w, h, fmt);
s_ssao.create (b_ssao, "r2\\ssao");
}
//SSAO
if (RImplementation.o.ssao_blur_on)
{
u32 w = Device.dwWidth, h = Device.dwHeight;
rt_ssao_temp.create (r2_RT_ssao_temp, w, h, D3DFMT_G16R16F);
s_ssao.create (b_ssao, "r2\\ssao");
}
// TONEMAP
{
rt_LUM_64.create (r2_RT_luminance_t64, 64, 64, D3DFMT_A16B16G16R16F );
rt_LUM_8.create (r2_RT_luminance_t8, 8, 8, D3DFMT_A16B16G16R16F );
s_luminance.create (b_luminance, "r2\\luminance");
f_luminance_adapt = 0.5f;
t_LUM_src.create (r2_RT_luminance_src);
t_LUM_dest.create (r2_RT_luminance_cur);
// create pool
for (u32 it=0; it<HW.Caps.iGPUNum*2; it++) {
string256 name;
sprintf (name,"%s_%d", r2_RT_luminance_pool,it );
rt_LUM_pool[it].create (name, 1, 1, D3DFMT_R32F );
u_setrt (rt_LUM_pool[it], 0, 0, 0 );
CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_TARGET, 0x7f7f7f7f, 1.0f, 0L));
}
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
}
// COMBINE
{
static D3DVERTEXELEMENT9 dwDecl[] =
{
{ 0, 0, D3DDECLTYPE_FLOAT4, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0 }, // pos+uv
{ 0, 16, D3DDECLTYPE_D3DCOLOR, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_COLOR, 0 },
{ 0, 20, D3DDECLTYPE_FLOAT2, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_TEXCOORD, 0 },
D3DDECL_END()
};
s_combine.create (b_combine, "r2\\combine");
s_combine_volumetric.create ("combine_volumetric");
s_combine_dbg_0.create ("effects\\screen_set", r2_RT_smap_surf );
s_combine_dbg_1.create ("effects\\screen_set", r2_RT_luminance_t8 );
s_combine_dbg_Accumulator.create ("effects\\screen_set", r2_RT_accum );
g_combine_VP.create (dwDecl, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine.create (FVF::F_TL, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine_2UV.create (FVF::F_TL2uv, RCache.Vertex.Buffer(), RCache.QuadIB);
u32 fvf_aa_blur = D3DFVF_XYZRHW|D3DFVF_TEX4|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3);
g_aa_blur.create (fvf_aa_blur, RCache.Vertex.Buffer(), RCache.QuadIB);
u32 fvf_aa_AA = D3DFVF_XYZRHW|D3DFVF_TEX7|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3)|D3DFVF_TEXCOORDSIZE2(4)|D3DFVF_TEXCOORDSIZE4(5)|D3DFVF_TEXCOORDSIZE4(6);
g_aa_AA.create (fvf_aa_AA, RCache.Vertex.Buffer(), RCache.QuadIB);
t_envmap_0.create (r2_T_envs0);
t_envmap_1.create (r2_T_envs1);
}
// Build textures
{
// Build material(s)
{
// Surface
R_CHK (D3DXCreateVolumeTexture(HW.pDevice,TEX_material_LdotN,TEX_material_LdotH,4,1,0,D3DFMT_A8L8,D3DPOOL_MANAGED,&t_material_surf));
t_material = dxRenderDeviceRender::Instance().Resources->_CreateTexture(r2_material);
t_material->surface_set (t_material_surf);
// Fill it (addr: x=dot(L,N),y=dot(L,H))
D3DLOCKED_BOX R;
R_CHK (t_material_surf->LockBox (0,&R,0,0));
for (u32 slice=0; slice<4; slice++)
{
for (u32 y=0; y<TEX_material_LdotH; y++)
{
for (u32 x=0; x<TEX_material_LdotN; x++)
{
u16* p = (u16*) (LPBYTE (R.pBits) + slice*R.SlicePitch + y*R.RowPitch + x*2);
float ld = float(x) / float (TEX_material_LdotN-1);
float ls = float(y) / float (TEX_material_LdotH-1) + EPS_S;
ls *= powf(ld,1/32.f);
float fd,fs;
switch (slice)
{
case 0: { // looks like OrenNayar
fd = powf(ld,0.75f); // 0.75
fs = powf(ls,16.f)*.5f;
} break;
case 1: {// looks like Blinn
fd = powf(ld,0.90f); // 0.90
fs = powf(ls,24.f);
} break;
case 2: { // looks like Phong
fd = ld; // 1.0
fs = powf(ls*1.01f,128.f );
} break;
case 3: { // looks like Metal
float s0 = _abs (1-_abs (0.05f*_sin(33.f*ld)+ld-ls));
float s1 = _abs (1-_abs (0.05f*_cos(33.f*ld*ls)+ld-ls));
float s2 = _abs (1-_abs (ld-ls));
fd = ld; // 1.0
fs = powf (_max(_max(s0,s1),s2), 24.f);
fs *= powf (ld,1/7.f);
} break;
default:
fd = fs = 0;
}
s32 _d = clampr (iFloor (fd*255.5f), 0,255);
s32 _s = clampr (iFloor (fs*255.5f), 0,255);
if ((y==(TEX_material_LdotH-1)) && (x==(TEX_material_LdotN-1))) { _d = 255; _s=255; }
*p = u16 (_s*256 + _d);
}
}
}
R_CHK (t_material_surf->UnlockBox (0));
// #ifdef DEBUG
// R_CHK (D3DXSaveTextureToFile ("x:\\r2_material.dds",D3DXIFF_DDS,t_material_surf,0));
// #endif
}
// Build noise table
if (1)
{
// Surfaces
D3DLOCKED_RECT R[TEX_jitter_count];
for (int it1=0; it1<TEX_jitter_count-1; it1++)
{
string_path name;
sprintf (name,"%s%d",r2_jitter,it1);
R_CHK (D3DXCreateTexture (HW.pDevice,TEX_jitter,TEX_jitter,1,0,D3DFMT_Q8W8V8U8,D3DPOOL_MANAGED,&t_noise_surf[it1]));
t_noise[it1] = dxRenderDeviceRender::Instance().Resources->_CreateTexture (name);
t_noise[it1]->surface_set (t_noise_surf[it1]);
R_CHK (t_noise_surf[it1]->LockRect (0,&R[it1],0,0));
}
// Fill it,
for (u32 y=0; y<TEX_jitter; y++)
{
for (u32 x=0; x<TEX_jitter; x++)
{
DWORD data [TEX_jitter_count-1];
generate_jitter (data,TEX_jitter_count-1);
for (u32 it2=0; it2<TEX_jitter_count-1; it2++)
{
u32* p = (u32*) (LPBYTE (R[it2].pBits) + y*R[it2].Pitch + x*4);
*p = data [it2];
}
}
}
for (int it3=0; it3<TEX_jitter_count-1; it3++) {
R_CHK (t_noise_surf[it3]->UnlockRect(0));
}
// generate HBAO jitter texture (last)
int it = TEX_jitter_count - 1;
string_path name;
sprintf (name,"%s%d",r2_jitter,it);
R_CHK (D3DXCreateTexture (HW.pDevice,TEX_jitter,TEX_jitter,1,0,D3DFMT_A32B32G32R32F,D3DPOOL_MANAGED,&t_noise_surf[it]));
t_noise[it] = dxRenderDeviceRender::Instance().Resources->_CreateTexture (name);
t_noise[it]->surface_set (t_noise_surf[it]);
R_CHK (t_noise_surf[it]->LockRect (0,&R[it],0,0));
// Fill it,
for (u32 y=0; y<TEX_jitter; y++)
{
for (u32 x=0; x<TEX_jitter; x++)
{
float numDir = 1.0f;
switch (ps_r_ssao)
{
case 1: numDir = 4.0f; break;
case 2: numDir = 6.0f; break;
case 3: numDir = 8.0f; break;
}
float angle = 2 * PI * ::Random.randF(0.0f, 1.0f) / numDir;
float dist = ::Random.randF(0.0f, 1.0f);
//float dest[4];
float* p = (float*) (LPBYTE (R[it].pBits) + y*R[it].Pitch + x*4*sizeof(float));
*p = (float)(_cos(angle));
*(p+1) = (float)(_sin(angle));
*(p+2) = (float)(dist);
*(p+3) = 0;
//generate_hbao_jitter (data,TEX_jitter*TEX_jitter);
}
}
R_CHK (t_noise_surf[it]->UnlockRect(0));
}
}
// PP
s_postprocess.create ("postprocess");
g_postprocess.create (D3DFVF_XYZRHW|D3DFVF_DIFFUSE|D3DFVF_SPECULAR|D3DFVF_TEX3,RCache.Vertex.Buffer(),RCache.QuadIB);
// Menu
s_menu.create ("distort");
g_menu.create (FVF::F_TL,RCache.Vertex.Buffer(),RCache.QuadIB);
// Igor: TMP
// Create an RT for online screenshot makining
//u32 w = Device.dwWidth, h = Device.dwHeight;
//HW.pDevice->CreateOffscreenPlainSurface(Device.dwWidth,Device.dwHeight,D3DFMT_A8R8G8B8,D3DPOOL_SYSTEMMEM,&pFB,NULL);
//HW.pDevice->CreateOffscreenPlainSurface(Device.dwWidth,Device.dwHeight,rt_Color->fmt,D3DPOOL_SYSTEMMEM,&pFB,NULL);
D3DSURFACE_DESC desc;
HW.pBaseRT->GetDesc(&desc);
HW.pDevice->CreateOffscreenPlainSurface(Device.dwWidth,Device.dwHeight,desc.Format,D3DPOOL_SYSTEMMEM,&pFB,NULL);
//
dwWidth = Device.dwWidth;
dwHeight = Device.dwHeight;
}
void CRenderTarget::phase_bloom ()
{
PIX_EVENT(phase_bloom);
u32 Offset;
// Targets
u_setrt ( rt_Bloom_1,NULL,NULL,NULL ); // No need for ZBuffer at all
//RImplementation.rmNormal();
// Clear - don't clear - it's stupid here :)
// Stencil - disable
// Misc - draw everything (no culling)
//CHK_DX (HW.pDevice->SetRenderState ( D3DRS_ZENABLE, FALSE ));
RCache.set_Z(FALSE);
// Transfer into Bloom1
{
float _w = float(Device.dwWidth);
float _h = float(Device.dwHeight);
float _2w = _w/2; float tw = BLOOM_size_X;
float _2h = _h/2; float th = BLOOM_size_Y;
float _aspect_w = _2w/tw;
float _aspect_h = _2h/th;
Fvector2 one = { 1.f/_w, 1.f/_h }; one.x*=_aspect_w; one.y*=_aspect_h;
Fvector2 half = { .5f/_w, .5f/_h };
Fvector2 a_0 = { half.x + 0, half.y + 0 };
Fvector2 a_1 = { half.x + one.x, half.y + 0 };
Fvector2 a_2 = { half.x + 0, half.y + one.y };
Fvector2 a_3 = { half.x + one.x, half.y + one.y };
Fvector2 b_0 = { 1 + a_0.x, 1 + a_0.y };
Fvector2 b_1 = { 1 + a_1.x, 1 + a_1.y };
Fvector2 b_2 = { 1 + a_2.x, 1 + a_2.y };
Fvector2 b_3 = { 1 + a_3.x, 1 + a_3.y };
// Fill vertex buffer
v_build* pv = (v_build*) RCache.Vertex.Lock (4,g_bloom_build->vb_stride,Offset);
pv->p.set (EPS, float(th+EPS), EPS,1.f);
pv->uv0.set (a_0.x,b_0.y); pv->uv1.set (a_1.x,b_1.y); pv->uv2.set (a_2.x,b_2.y); pv->uv3.set (a_3.x,b_3.y);
pv++;
pv->p.set (EPS, EPS, EPS,1.f);
pv->uv0.set (a_0.x,a_0.y); pv->uv1.set (a_1.x,a_1.y); pv->uv2.set (a_2.x,a_2.y); pv->uv3.set (a_3.x,a_3.y);
pv++;
pv->p.set (float(tw+EPS),float(th+EPS), EPS,1.f);
pv->uv0.set (b_0.x,b_0.y); pv->uv1.set (b_1.x,b_1.y); pv->uv2.set (b_2.x,b_2.y); pv->uv3.set (b_3.x,b_3.y);
pv++;
pv->p.set (float(tw+EPS),EPS, EPS,1.f);
pv->uv0.set (b_0.x,a_0.y); pv->uv1.set (b_1.x,a_1.y); pv->uv2.set (b_2.x,a_2.y); pv->uv3.set (b_3.x,a_3.y);
pv++;
RCache.Vertex.Unlock (4,g_bloom_build->vb_stride);
// Perform combine (all scalers must account for 4 samples + final diffuse multiply);
float s = ps_r2_ls_bloom_threshold; // scale
f_bloom_factor = .9f*f_bloom_factor + .1f*ps_r2_ls_bloom_speed*Device.fTimeDelta; // speed
if( !RImplementation.o.dx10_msaa )
RCache.set_Element (s_bloom->E[0]);
else
RCache.set_Element (s_bloom_msaa->E[0]);
RCache.set_c ("b_params", s,s,s, f_bloom_factor);
RCache.set_Geometry (g_bloom_build );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// Capture luminance values
phase_luminance ( );
if (ps_r2_ls_flags.test(R2FLAG_FASTBLOOM))
{
// FAST FILTER
float _w = BLOOM_size_X;
float _h = BLOOM_size_Y;
float ddw = (1.f/_w)*ps_r2_ls_bloom_kernel_b;
float ddh = (1.f/_h)*ps_r2_ls_bloom_kernel_b;
Fvector2 p0; p0.set (.5f/_w, .5f/_h);
Fvector2 p1; p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
v_build* pv = (v_build*) RCache.Vertex.Lock (4,g_bloom_build->vb_stride,Offset);
pv->p.set(EPS, float(_h+EPS), EPS,1.f); pv->uv0.set(p0.x-ddw,p1.y-ddh);pv->uv1.set(p0.x+ddw,p1.y+ddh);pv->uv2.set(p0.x+ddw,p1.y-ddh);pv->uv3.set(p0.x-ddw,p1.y+ddh);pv++;
pv->p.set(EPS, EPS, EPS,1.f); pv->uv0.set(p0.x-ddw,p0.y-ddh);pv->uv1.set(p0.x+ddw,p0.y+ddh);pv->uv2.set(p0.x+ddw,p0.y-ddh);pv->uv3.set(p0.x-ddw,p0.y+ddh);pv++;
pv->p.set(float(_w+EPS),float(_h+EPS), EPS,1.f); pv->uv0.set(p1.x-ddw,p1.y-ddh);pv->uv1.set(p1.x+ddw,p1.y+ddh);pv->uv2.set(p1.x+ddw,p1.y-ddh);pv->uv3.set(p1.x-ddw,p1.y+ddh);pv++;
pv->p.set(float(_w+EPS),EPS, EPS,1.f); pv->uv0.set(p1.x-ddw,p0.y-ddh);pv->uv1.set(p1.x+ddw,p0.y+ddh);pv->uv2.set(p1.x+ddw,p0.y-ddh);pv->uv3.set(p1.x-ddw,p0.y+ddh);pv++;
RCache.Vertex.Unlock (4,g_bloom_build->vb_stride);
RCache.set_Geometry (g_bloom_build);
// P0
u_setrt (rt_Bloom_2,NULL,NULL,NULL); // No need for ZBuffer at all
RCache.set_Element (s_bloom->E[3]);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
// P1
u_setrt (rt_Bloom_1,NULL,NULL,NULL); // No need for ZBuffer at all
RCache.set_Element (s_bloom->E[4]);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
} else {
// SLOW FILTER
// Transfer into Bloom2, horizontal filter
{
float _w = BLOOM_size_X;
float _h = BLOOM_size_Y;
Fvector2 two = { 2.f/_w, 2.f/_h };
Fvector2 one = { 1.f/_w, 1.f/_h };
Fvector2 half = { .5f/_w, .5f/_h };
Fvector4 a_0 = { half.x, half.y, half.y, half.x }; // center
Fvector4 a_1 = { a_0.x - one.x - half.x, half.y, half.y, a_0.w + one.x + half.x }; // -1,+1i
Fvector4 a_2 = { a_1.x - two.x, half.y, half.y, a_1.w + two.x }; // -2,+2i
Fvector4 a_3 = { a_2.x - two.x, half.y, half.y, a_2.w + two.x }; // -3,+3i
Fvector4 a_4 = { a_3.x - two.x, half.y, half.y, a_3.w + two.x }; // -4,+4i
Fvector4 a_5 = { a_4.x - two.x, half.y, half.y, a_4.w + two.x }; // -5,+5i
Fvector4 a_6 = { a_5.x - two.x, half.y, half.y, a_5.w + two.x }; // -6,+6i
Fvector4 a_7 = { a_6.x - two.x, half.y, half.y, a_6.w + two.x }; // -7,+7i
// Fill vertex buffer
v_filter* pv = (v_filter*) RCache.Vertex.Lock (4,g_bloom_filter->vb_stride,Offset);
// 0 - LB
pv->p.set (EPS, float(_h+EPS), EPS,1.f);
pv->uv0.set (a_0.x,1+a_0.y,0,0);
pv->uv1.set (a_1.x,1+a_1.y,1+a_1.z,a_1.w);
pv->uv2.set (a_2.x,1+a_2.y,1+a_2.z,a_2.w);
pv->uv3.set (a_3.x,1+a_3.y,1+a_3.z,a_3.w);
pv->uv4.set (a_4.x,1+a_4.y,1+a_4.z,a_4.w);
pv->uv5.set (a_5.x,1+a_5.y,1+a_5.z,a_5.w);
pv->uv6.set (a_6.x,1+a_6.y,1+a_6.z,a_6.w);
pv->uv7.set (a_7.x,1+a_7.y,1+a_7.z,a_7.w);
pv++;
// 1 - LT
pv->p.set (EPS, EPS, EPS,1.f);
pv->uv0.set (a_0.x,a_0.y,0,0);
pv->uv1.set (a_1.x,a_1.y,a_1.z,a_1.w);
pv->uv2.set (a_2.x,a_2.y,a_2.z,a_2.w);
pv->uv3.set (a_3.x,a_3.y,a_3.z,a_3.w);
pv->uv4.set (a_4.x,a_4.y,a_4.z,a_4.w);
pv->uv5.set (a_5.x,a_5.y,a_5.z,a_5.w);
pv->uv6.set (a_6.x,a_6.y,a_6.z,a_6.w);
pv->uv7.set (a_7.x,a_7.y,a_7.z,a_7.w);
pv++;
// 2 - RB
pv->p.set (float(_w+EPS), float(_h+EPS), EPS,1.f);
pv->uv0.set (1+a_0.x,1+a_0.y,0,0);
pv->uv1.set (1+a_1.x,1+a_1.y,1+a_1.z,1+a_1.w);
pv->uv2.set (1+a_2.x,1+a_2.y,1+a_2.z,1+a_2.w);
pv->uv3.set (1+a_3.x,1+a_3.y,1+a_3.z,1+a_3.w);
pv->uv4.set (1+a_4.x,1+a_4.y,1+a_4.z,1+a_4.w);
pv->uv5.set (1+a_5.x,1+a_5.y,1+a_5.z,1+a_5.w);
pv->uv6.set (1+a_6.x,1+a_6.y,1+a_6.z,1+a_6.w);
pv->uv7.set (1+a_7.x,1+a_7.y,1+a_7.z,1+a_7.w);
pv++;
// 3 - RT
pv->p.set (float(_w+EPS), EPS, EPS,1.f);
pv->uv0.set (1+a_0.x,a_0.y,0,0);
pv->uv1.set (1+a_1.x,a_1.y,a_1.z,1+a_1.w);
pv->uv2.set (1+a_2.x,a_2.y,a_2.z,1+a_2.w);
pv->uv3.set (1+a_3.x,a_3.y,a_3.z,1+a_3.w);
pv->uv4.set (1+a_4.x,a_4.y,a_4.z,1+a_4.w);
pv->uv5.set (1+a_5.x,a_5.y,a_5.z,1+a_5.w);
pv->uv6.set (1+a_6.x,a_6.y,a_6.z,1+a_6.w);
pv->uv7.set (1+a_7.x,a_7.y,a_7.z,1+a_7.w);
pv++;
RCache.Vertex.Unlock (4,g_bloom_filter->vb_stride);
// Perform filtering
Fvector4 w0,w1;
float kernel = ps_r2_ls_bloom_kernel_g;
CalcGauss_wave (w0,w1,kernel,kernel/3.f,ps_r2_ls_bloom_kernel_scale);
u_setrt (rt_Bloom_2,NULL,NULL,NULL); // No need for ZBuffer at all
RCache.set_Element (s_bloom->E[1]);
RCache.set_ca ("weight", 0, w0);
RCache.set_ca ("weight", 1, w1);
RCache.set_Geometry (g_bloom_filter);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// Transfer into Bloom1, vertical filter
{
float _w = BLOOM_size_X;
float _h = BLOOM_size_Y;
Fvector2 two = { 2.f/_w, 2.f/_h };
Fvector2 one = { 1.f/_w, 1.f/_h };
Fvector2 half = { .5f/_w, .5f/_h };
Fvector4 a_0 = { half.x, half.y, half.y, half.x }; // center
Fvector4 a_1 = { half.x, a_0.y - one.y - half.y, half.y + one.y + a_0.z, half.x }; // -1,+1i
Fvector4 a_2 = { half.x, a_1.y - two.y, two.y + a_1.z, half.x }; // -2,+2i
Fvector4 a_3 = { half.x, a_2.y - two.y, two.y + a_2.z, half.x }; // -3,+3i
Fvector4 a_4 = { half.x, a_3.y - two.y, two.y + a_3.z, half.x }; // -4,+4i
Fvector4 a_5 = { half.x, a_4.y - two.y, two.y + a_4.z, half.x }; // -5,+5i
Fvector4 a_6 = { half.x, a_5.y - two.y, two.y + a_5.z, half.x }; // -6,+6i
Fvector4 a_7 = { half.x, a_6.y - two.y, two.y + a_6.z, half.x }; // -7,+7i
// Fill vertex buffer
v_filter* pv = (v_filter*) RCache.Vertex.Lock (4,g_bloom_filter->vb_stride,Offset);
// 0 - LB
pv->p.set (EPS, float(_h+EPS), EPS,1.f);
pv->uv0.set (a_0.x,1+a_0.y,0,0);
pv->uv1.set (a_1.x,1+a_1.y,1+a_1.z,a_1.w);
pv->uv2.set (a_2.x,1+a_2.y,1+a_2.z,a_2.w);
pv->uv3.set (a_3.x,1+a_3.y,1+a_3.z,a_3.w);
pv->uv4.set (a_4.x,1+a_4.y,1+a_4.z,a_4.w);
pv->uv5.set (a_5.x,1+a_5.y,1+a_5.z,a_5.w);
pv->uv6.set (a_6.x,1+a_6.y,1+a_6.z,a_6.w);
pv->uv7.set (a_7.x,1+a_7.y,1+a_7.z,a_7.w);
pv++;
// 1 - LT
pv->p.set (EPS, EPS, EPS,1.f);
pv->uv0.set (a_0.x,a_0.y,0,0);
pv->uv1.set (a_1.x,a_1.y,a_1.z,a_1.w);
pv->uv2.set (a_2.x,a_2.y,a_2.z,a_2.w);
pv->uv3.set (a_3.x,a_3.y,a_3.z,a_3.w);
pv->uv4.set (a_4.x,a_4.y,a_4.z,a_4.w);
pv->uv5.set (a_5.x,a_5.y,a_5.z,a_5.w);
pv->uv6.set (a_6.x,a_6.y,a_6.z,a_6.w);
pv->uv7.set (a_7.x,a_7.y,a_7.z,a_7.w);
pv++;
// 2 - RB
pv->p.set (float(_w+EPS), float(_h+EPS), EPS,1.f);
pv->uv0.set (1+a_0.x,1+a_0.y,0,0);
pv->uv1.set (1+a_1.x,1+a_1.y,1+a_1.z,1+a_1.w);
pv->uv2.set (1+a_2.x,1+a_2.y,1+a_2.z,1+a_2.w);
pv->uv3.set (1+a_3.x,1+a_3.y,1+a_3.z,1+a_3.w);
pv->uv4.set (1+a_4.x,1+a_4.y,1+a_4.z,1+a_4.w);
pv->uv5.set (1+a_5.x,1+a_5.y,1+a_5.z,1+a_5.w);
pv->uv6.set (1+a_6.x,1+a_6.y,1+a_6.z,1+a_6.w);
pv->uv7.set (1+a_7.x,1+a_7.y,1+a_7.z,1+a_7.w);
pv++;
// 3 - RT
pv->p.set (float(_w+EPS), EPS, EPS,1.f);
pv->uv0.set (1+a_0.x,a_0.y,0,0);
pv->uv1.set (1+a_1.x,a_1.y,a_1.z,1+a_1.w);
pv->uv2.set (1+a_2.x,a_2.y,a_2.z,1+a_2.w);
pv->uv3.set (1+a_3.x,a_3.y,a_3.z,1+a_3.w);
pv->uv4.set (1+a_4.x,a_4.y,a_4.z,1+a_4.w);
pv->uv5.set (1+a_5.x,a_5.y,a_5.z,1+a_5.w);
pv->uv6.set (1+a_6.x,a_6.y,a_6.z,1+a_6.w);
pv->uv7.set (1+a_7.x,a_7.y,a_7.z,1+a_7.w);
pv++;
RCache.Vertex.Unlock (4,g_bloom_filter->vb_stride);
// Perform filtering
Fvector4 w0,w1;
float kernel = ps_r2_ls_bloom_kernel_g * float(Device.dwHeight)/float(Device.dwWidth);
CalcGauss_wave (w0,w1,kernel,kernel/3.f,ps_r2_ls_bloom_kernel_scale);
u_setrt (rt_Bloom_1,NULL,NULL,NULL); // No need for ZBuffer at all
RCache.set_Element (s_bloom->E[2]);
RCache.set_ca ("weight", 0, w0);
RCache.set_ca ("weight", 1, w1);
RCache.set_Geometry (g_bloom_filter);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
}
// we are left here with bloom-target setup as primary one
// for FP16-BLEND capable HW we can blend flares into smaller target, because they are smooth
//if (RImplementation.o.fp16_blend) g_pGamePersistent->Environment().RenderFlares (); // lens-flares
bool _menu_pp = g_pGamePersistent?g_pGamePersistent->OnRenderPPUI_query():false;
if (_menu_pp)
{
//CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_TARGET, 0, 1.0f, 0L));
FLOAT ColorRGBA[4] = {0.0f, 0.0f, 0.0f, 0.0f};
HW.pDevice->ClearRenderTargetView( RCache.get_RT(), ColorRGBA);
};
// re-enable z-buffer
//CHK_DX (HW.pDevice->SetRenderState ( D3DRS_ZENABLE, TRUE ));
RCache.set_Z(TRUE);
}
void CRenderTarget::phase_luminance()
{
u32 Offset = 0;
float eps = EPS_S;
// Targets
RCache.set_Stencil (FALSE);
RCache.set_CullMode (CULL_NONE);
RCache.set_ColorWriteEnable ();
CHK_DX (HW.pDevice->SetRenderState (D3DRS_ZENABLE,FALSE));
// 000: Perform LUM-SAT, pass 0, 256x256 => 64x64
u_setrt (rt_LUM_64,NULL,NULL,NULL);
{
float ts = 64;
float _w = float(BLOOM_size_X);
float _h = float(BLOOM_size_Y);
Fvector2 one = { 2.f/_w, 2.f/_h }; // two, infact
Fvector2 half = { 1.f/_w, 1.f/_h }; // one, infact
Fvector2 a_0 = { half.x + 0, half.y + 0 };
Fvector2 a_1 = { half.x + one.x, half.y + 0 };
Fvector2 a_2 = { half.x + 0, half.y + one.y };
Fvector2 a_3 = { half.x + one.x, half.y + one.y };
Fvector2 b_0 = { 1 + a_0.x, 1 + a_0.y };
Fvector2 b_1 = { 1 + a_1.x, 1 + a_1.y };
Fvector2 b_2 = { 1 + a_2.x, 1 + a_2.y };
Fvector2 b_3 = { 1 + a_3.x, 1 + a_3.y };
// Fill vertex buffer
v_build* pv = (v_build*) RCache.Vertex.Lock (4,g_bloom_build->vb_stride,Offset);
pv->p.set (eps, float(ts+eps), eps,1.f); pv->uv0.set (a_0.x,b_0.y); pv->uv1.set (a_1.x,b_1.y); pv->uv2.set (a_2.x,b_2.y); pv->uv3.set (a_3.x,b_3.y); pv++;
pv->p.set (eps, eps, eps,1.f); pv->uv0.set (a_0.x,a_0.y); pv->uv1.set (a_1.x,a_1.y); pv->uv2.set (a_2.x,a_2.y); pv->uv3.set (a_3.x,a_3.y); pv++;
pv->p.set (float(ts+eps), float(ts+eps), eps,1.f); pv->uv0.set (b_0.x,b_0.y); pv->uv1.set (b_1.x,b_1.y); pv->uv2.set (b_2.x,b_2.y); pv->uv3.set (b_3.x,b_3.y); pv++;
pv->p.set (float(ts+eps), eps, eps,1.f); pv->uv0.set (b_0.x,a_0.y); pv->uv1.set (b_1.x,a_1.y); pv->uv2.set (b_2.x,a_2.y); pv->uv3.set (b_3.x,a_3.y); pv++;
RCache.Vertex.Unlock (4,g_bloom_build->vb_stride);
RCache.set_Element (s_luminance->E[0]);
RCache.set_Geometry (g_bloom_build );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// 111: Perform LUM-SAT, pass 1, 64x64 => 8x8
u_setrt (rt_LUM_8,NULL,NULL,NULL);
{
// Build filter-kernel
float _ts = 8;
float _src = float(64);
Fvector2 a[16],b[16];
for (int k=0; k<16; k++) {
int _x = (k*2+1)%8; // 1,3,5,7
int _y = ((k/4)*2+1); // 1,1,1,1 ~ 3,3,3,3 ~...etc...
a[k].set(_x,_y).div (_src);
b[k].set(a[k]).add (1);
}
// Fill vertex buffer
v_filter* pv = (v_filter*) RCache.Vertex.Lock (4,g_bloom_filter->vb_stride,Offset);
pv->p.set (eps, float(_ts+eps), eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(a[t].x,b[t].y, b[t+8].y,a[t+8].x); // xy/yx - left+down
pv++;
pv->p.set (eps, eps, eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(a[t].x,a[t].y, a[t+8].y,a[t+8].x); // xy/yx - left+up
pv++;
pv->p.set (float(_ts+eps),float(_ts+eps), eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(b[t].x,b[t].y, b[t+8].y,b[t+8].x); // xy/yx - right+down
pv++;
pv->p.set (float(_ts+eps),eps, eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(b[t].x,a[t].y, a[t+8].y,b[t+8].x); // xy/yx - right+up
pv++;
RCache.Vertex.Unlock (4,g_bloom_filter->vb_stride);
RCache.set_Element (s_luminance->E[1] );
RCache.set_Geometry (g_bloom_filter );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// 222: Perform LUM-SAT, pass 2, 8x8 => 1x1
u32 gpu_id = Device.dwFrame%2;
u_setrt (rt_LUM_pool[gpu_id*2+1],NULL,NULL,NULL);
{
// Build filter-kernel
float _ts = 1;
float _src = float(8);
Fvector2 a[16],b[16];
for (int k=0; k<16; k++) {
int _x = (k*2+1)%8; // 1,3,5,7
int _y = ((k/4)*2+1); // 1,1,1,1 ~ 3,3,3,3 ~...etc...
a[k].set(_x,_y).div (_src);
b[k].set(a[k]).add (1);
}
// Fill vertex buffer
v_filter* pv = (v_filter*) RCache.Vertex.Lock (4,g_bloom_filter->vb_stride,Offset);
pv->p.set (eps, float(_ts+eps), eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(a[t].x,b[t].y, b[t+8].y,a[t+8].x); // xy/yx - left+down
pv++;
pv->p.set (eps, eps, eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(a[t].x,a[t].y, a[t+8].y,a[t+8].x); // xy/yx - left+up
pv++;
pv->p.set (float(_ts+eps),float(_ts+eps), eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(b[t].x,b[t].y, b[t+8].y,b[t+8].x); // xy/yx - right+down
pv++;
pv->p.set (float(_ts+eps),eps, eps,1.f); for (int t=0; t<8; t++) pv->uv[t].set(b[t].x,a[t].y, a[t+8].y,b[t+8].x); // xy/yx - right+up
pv++;
RCache.Vertex.Unlock (4,g_bloom_filter->vb_stride);
f_luminance_adapt = .9f*f_luminance_adapt + .1f*Device.fTimeDelta*ps_r2_tonemap_adaptation;
float amount = ps_r2_ls_flags.test(R2FLAG_TONEMAP)?ps_r2_tonemap_amount:0;
Fvector3 _none, _full, _result;
_none.set (1, 0, 1 );
_full.set (ps_r2_tonemap_middlegray, 1.f, ps_r2_tonemap_low_lum );
_result.lerp (_none, _full, amount );
RCache.set_Element (s_luminance->E[2] );
RCache.set_Geometry (g_bloom_filter );
RCache.set_c("MiddleGray", _result.x,_result.y,_result.z,f_luminance_adapt );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// Cleanup states
CHK_DX (HW.pDevice->SetRenderState(D3DRS_ZENABLE,TRUE));
}
CRenderTarget::CRenderTarget ()
{
u32 SampleCount = 1;
if (ps_r_ssao_mode!=2/*hdao*/)
ps_r_ssao = _min(ps_r_ssao, 3);
if( RImplementation.o.dx10_msaa )
SampleCount = RImplementation.o.dx10_msaa_samples;
#ifdef DEBUG
Msg ("MSAA samples = %d", SampleCount );
if( RImplementation.o.dx10_msaa_opt )
Msg ("dx10_MSAA_opt = on" );
if( RImplementation.o.dx10_gbuffer_opt )
Msg ("dx10_gbuffer_opt = on" );
#endif // DEBUG
param_blur = 0.f;
param_gray = 0.f;
param_noise = 0.f;
param_duality_h = 0.f;
param_duality_v = 0.f;
param_noise_fps = 25.f;
param_noise_scale = 1.f;
im_noise_time = 1/100;
im_noise_shift_w = 0;
im_noise_shift_h = 0;
param_color_base = color_rgba(127,127,127, 0);
param_color_gray = color_rgba(85,85,85, 0);
//param_color_add = color_rgba(0,0,0, 0);
param_color_add.set( 0.0f, 0.0f, 0.0f );
dwAccumulatorClearMark = 0;
dxRenderDeviceRender::Instance().Resources->Evict ();
// Blenders
b_occq = xr_new<CBlender_light_occq> ();
b_accum_mask = xr_new<CBlender_accum_direct_mask> ();
b_accum_direct = xr_new<CBlender_accum_direct> ();
b_accum_point = xr_new<CBlender_accum_point> ();
b_accum_spot = xr_new<CBlender_accum_spot> ();
b_accum_reflected = xr_new<CBlender_accum_reflected> ();
b_bloom = xr_new<CBlender_bloom_build> ();
if( RImplementation.o.dx10_msaa )
{
b_bloom_msaa = xr_new<CBlender_bloom_build_msaa> ();
b_postprocess_msaa = xr_new<CBlender_postprocess_msaa> ();
}
b_luminance = xr_new<CBlender_luminance> ();
b_combine = xr_new<CBlender_combine> ();
b_ssao = xr_new<CBlender_SSAO_noMSAA> ();
if( RImplementation.o.dx10_msaa )
{
int bound = RImplementation.o.dx10_msaa_samples;
if( RImplementation.o.dx10_msaa_opt )
bound = 1;
for( int i = 0; i < bound; ++i )
{
static LPCSTR SampleDefs[] = { "0","1","2","3","4","5","6","7" };
b_combine_msaa[i] = xr_new<CBlender_combine_msaa> ();
b_accum_mask_msaa[i] = xr_new<CBlender_accum_direct_mask_msaa> ();
b_accum_direct_msaa[i] = xr_new<CBlender_accum_direct_msaa> ();
b_accum_direct_volumetric_msaa[i] = xr_new<CBlender_accum_direct_volumetric_msaa> ();
//b_accum_direct_volumetric_sun_msaa[i] = xr_new<CBlender_accum_direct_volumetric_sun_msaa> ();
b_accum_spot_msaa[i] = xr_new<CBlender_accum_spot_msaa> ();
b_accum_volumetric_msaa[i] = xr_new<CBlender_accum_volumetric_msaa> ();
b_accum_point_msaa[i] = xr_new<CBlender_accum_point_msaa> ();
b_accum_reflected_msaa[i] = xr_new<CBlender_accum_reflected_msaa> ();
b_ssao_msaa[i] = xr_new<CBlender_SSAO_MSAA> ();
static_cast<CBlender_accum_direct_mask_msaa*>( b_accum_mask_msaa[i] )->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_accum_direct_volumetric_msaa*>(b_accum_direct_volumetric_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
//static_cast<CBlender_accum_direct_volumetric_sun_msaa*>(b_accum_direct_volumetric_sun_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_accum_direct_msaa*>(b_accum_direct_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_accum_volumetric_msaa*>(b_accum_volumetric_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_accum_spot_msaa*>(b_accum_spot_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_accum_point_msaa*>(b_accum_point_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_accum_reflected_msaa*>(b_accum_reflected_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_combine_msaa*>(b_combine_msaa[i])->SetDefine( "ISAMPLE", SampleDefs[i]);
static_cast<CBlender_SSAO_MSAA*>(b_ssao_msaa[i])->SetDefine("ISAMPLE", SampleDefs[i]);
}
}
// NORMAL
{
u32 w=Device.dwWidth, h=Device.dwHeight;
rt_Position.create (r2_RT_P, w,h,D3DFMT_A16B16G16R16F, SampleCount );
if( RImplementation.o.dx10_msaa )
rt_MSAADepth.create( r2_RT_MSAAdepth, w, h, D3DFMT_D24S8, SampleCount );
if( !RImplementation.o.dx10_gbuffer_opt )
rt_Normal.create (r2_RT_N, w,h,D3DFMT_A16B16G16R16F, SampleCount );
// select albedo & accum
if (RImplementation.o.mrtmixdepth)
{
// NV50
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8 , SampleCount );
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F, SampleCount );
}
else
{
// can't - mix-depth
if (RImplementation.o.fp16_blend) {
// NV40
if( !RImplementation.o.dx10_gbuffer_opt )
{
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A16B16G16R16F, SampleCount ); // expand to full
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F, SampleCount );
}
else
{
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8, SampleCount ); // expand to full
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F, SampleCount );
}
} else {
// R4xx, no-fp-blend,-> albedo_wo
VERIFY (RImplementation.o.albedo_wo);
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8 , SampleCount ); // normal
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F, SampleCount );
rt_Accumulator_temp.create (r2_RT_accum_temp, w,h,D3DFMT_A16B16G16R16F, SampleCount );
}
}
// generic(LDR) RTs
rt_Generic_0.create (r2_RT_generic0,w,h,D3DFMT_A8R8G8B8, 1 );
rt_Generic_1.create (r2_RT_generic1,w,h,D3DFMT_A8R8G8B8, 1 );
if( RImplementation.o.dx10_msaa )
{
rt_Generic_0_r.create(r2_RT_generic0_r,w,h,D3DFMT_A8R8G8B8, SampleCount );
rt_Generic_1_r.create(r2_RT_generic1_r,w,h,D3DFMT_A8R8G8B8, SampleCount );
rt_Generic.create (r2_RT_generic,w,h, D3DFMT_A8R8G8B8, 1 );
}
// Igor: for volumetric lights
//rt_Generic_2.create (r2_RT_generic2,w,h,D3DFMT_A8R8G8B8 );
// temp: for higher quality blends
if (RImplementation.o.advancedpp)
rt_Generic_2.create (r2_RT_generic2,w,h,D3DFMT_A16B16G16R16F, SampleCount );
}
// OCCLUSION
s_occq.create (b_occq, "r2\\occq");
// DIRECT (spot)
D3DFORMAT depth_format = (D3DFORMAT)RImplementation.o.HW_smap_FORMAT;
if (RImplementation.o.HW_smap)
{
D3DFORMAT nullrt = D3DFMT_R5G6B5;
if (RImplementation.o.nullrt) nullrt = (D3DFORMAT)MAKEFOURCC('N','U','L','L');
u32 size =RImplementation.o.smapsize ;
rt_smap_depth.create (r2_RT_smap_depth, size,size,depth_format );
if (RImplementation.o.dx10_minmax_sm)
{
rt_smap_depth_minmax.create( r2_RT_smap_depth_minmax, size/4,size/4, D3DFMT_R32F );
CBlender_createminmax TempBlender;
s_create_minmax_sm.create( &TempBlender, "null" );
}
//rt_smap_surf.create (r2_RT_smap_surf, size,size,nullrt );
//rt_smap_ZB = NULL;
s_accum_mask.create (b_accum_mask, "r3\\accum_mask");
s_accum_direct.create (b_accum_direct, "r3\\accum_direct");
if( RImplementation.o.dx10_msaa )
{
int bound = RImplementation.o.dx10_msaa_samples;
if( RImplementation.o.dx10_msaa_opt )
bound = 1;
for( int i = 0; i < bound; ++i )
{
s_accum_direct_msaa[i].create (b_accum_direct_msaa[i], "r3\\accum_direct");
s_accum_mask_msaa[i].create (b_accum_mask_msaa[i], "r3\\accum_direct");
}
}
if (RImplementation.o.advancedpp)
{
s_accum_direct_volumetric.create("accum_volumetric_sun_nomsaa");
if (RImplementation.o.dx10_minmax_sm)
s_accum_direct_volumetric_minmax.create("accum_volumetric_sun_nomsaa_minmax");
if( RImplementation.o.dx10_msaa )
{
static LPCSTR snames[] = { "accum_volumetric_sun_msaa0",
"accum_volumetric_sun_msaa1",
"accum_volumetric_sun_msaa2",
"accum_volumetric_sun_msaa3",
"accum_volumetric_sun_msaa4",
"accum_volumetric_sun_msaa5",
"accum_volumetric_sun_msaa6",
"accum_volumetric_sun_msaa7"
};
int bound = RImplementation.o.dx10_msaa_samples;
if( RImplementation.o.dx10_msaa_opt )
bound = 1;
for( int i = 0; i < bound; ++i )
{
//s_accum_direct_volumetric_msaa[i].create (b_accum_direct_volumetric_sun_msaa[i], "r3\\accum_direct");
s_accum_direct_volumetric_msaa[i].create (snames[i]);
}
}
}
}
else
{
// TODO: DX10: Check if we need old-style SMap
VERIFY(!"Use HW SMAPs only!");
//u32 size =RImplementation.o.smapsize ;
//rt_smap_surf.create (r2_RT_smap_surf, size,size,D3DFMT_R32F);
//rt_smap_depth = NULL;
//R_CHK (HW.pDevice->CreateDepthStencilSurface (size,size,D3DFMT_D24X8,D3DMULTISAMPLE_NONE,0,TRUE,&rt_smap_ZB,NULL));
//s_accum_mask.create (b_accum_mask, "r2\\accum_mask");
//s_accum_direct.create (b_accum_direct, "r2\\accum_direct");
//if (RImplementation.o.advancedpp)
// s_accum_direct_volumetric.create("accum_volumetric_sun");
}
// RAIN
// TODO: DX10: Create resources only when DX10 rain is enabled.
// Or make DX10 rain switch dynamic?
{
CBlender_rain TempBlender;
s_rain.create( &TempBlender, "null");
if( RImplementation.o.dx10_msaa )
{
static LPCSTR SampleDefs[] = { "0","1","2","3","4","5","6","7" };
CBlender_rain_msaa TempBlender[8];
int bound = RImplementation.o.dx10_msaa_samples;
if( RImplementation.o.dx10_msaa_opt )
bound = 1;
for( int i = 0; i < bound; ++i )
{
TempBlender[i].SetDefine( "ISAMPLE", SampleDefs[i] );
s_rain_msaa[i].create( &TempBlender[i], "null");
s_accum_spot_msaa[i].create (b_accum_spot_msaa[i], "r2\\accum_spot_s", "lights\\lights_spot01");
s_accum_point_msaa[i].create (b_accum_point_msaa[i], "r2\\accum_point_s");
//s_accum_volume_msaa[i].create(b_accum_direct_volumetric_msaa[i], "lights\\lights_spot01");
s_accum_volume_msaa[i].create(b_accum_volumetric_msaa[i], "lights\\lights_spot01");
s_combine_msaa[i].create (b_combine_msaa[i], "r2\\combine");
}
}
}
if( RImplementation.o.dx10_msaa )
{
CBlender_msaa TempBlender;
s_mark_msaa_edges.create( &TempBlender, "null" );
}
// POINT
{
s_accum_point.create (b_accum_point, "r2\\accum_point_s");
accum_point_geom_create ();
g_accum_point.create (D3DFVF_XYZ, g_accum_point_vb, g_accum_point_ib);
accum_omnip_geom_create ();
g_accum_omnipart.create (D3DFVF_XYZ, g_accum_omnip_vb, g_accum_omnip_ib);
}
// SPOT
{
s_accum_spot.create (b_accum_spot, "r2\\accum_spot_s", "lights\\lights_spot01");
accum_spot_geom_create ();
g_accum_spot.create (D3DFVF_XYZ, g_accum_spot_vb, g_accum_spot_ib);
}
{
s_accum_volume.create("accum_volumetric", "lights\\lights_spot01");
accum_volumetric_geom_create();
g_accum_volumetric.create( D3DFVF_XYZ, g_accum_volumetric_vb, g_accum_volumetric_ib);
}
// REFLECTED
{
s_accum_reflected.create (b_accum_reflected, "r2\\accum_refl");
if( RImplementation.o.dx10_msaa )
{
int bound = RImplementation.o.dx10_msaa_samples;
if( RImplementation.o.dx10_msaa_opt )
bound = 1;
for( int i = 0; i < bound; ++i )
{
s_accum_reflected_msaa[i].create( b_accum_reflected_msaa[i], "null");
}
}
}
// BLOOM
{
D3DFORMAT fmt = D3DFMT_A8R8G8B8; //; // D3DFMT_X8R8G8B8
u32 w=BLOOM_size_X, h=BLOOM_size_Y;
u32 fvf_build = D3DFVF_XYZRHW|D3DFVF_TEX4|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3);
u32 fvf_filter = (u32)D3DFVF_XYZRHW|D3DFVF_TEX8|D3DFVF_TEXCOORDSIZE4(0)|D3DFVF_TEXCOORDSIZE4(1)|D3DFVF_TEXCOORDSIZE4(2)|D3DFVF_TEXCOORDSIZE4(3)|D3DFVF_TEXCOORDSIZE4(4)|D3DFVF_TEXCOORDSIZE4(5)|D3DFVF_TEXCOORDSIZE4(6)|D3DFVF_TEXCOORDSIZE4(7);
rt_Bloom_1.create (r2_RT_bloom1, w,h, fmt);
rt_Bloom_2.create (r2_RT_bloom2, w,h, fmt);
g_bloom_build.create (fvf_build, RCache.Vertex.Buffer(), RCache.QuadIB);
g_bloom_filter.create (fvf_filter, RCache.Vertex.Buffer(), RCache.QuadIB);
s_bloom_dbg_1.create ("effects\\screen_set", r2_RT_bloom1);
s_bloom_dbg_2.create ("effects\\screen_set", r2_RT_bloom2);
s_bloom.create (b_bloom, "r2\\bloom");
if( RImplementation.o.dx10_msaa )
{
s_bloom_msaa.create (b_bloom_msaa, "r2\\bloom");
s_postprocess_msaa.create(b_postprocess_msaa, "r2\\post");
}
f_bloom_factor = 0.5f;
}
// TONEMAP
{
rt_LUM_64.create (r2_RT_luminance_t64, 64, 64, D3DFMT_A16B16G16R16F );
rt_LUM_8.create (r2_RT_luminance_t8, 8, 8, D3DFMT_A16B16G16R16F );
s_luminance.create (b_luminance, "r2\\luminance");
f_luminance_adapt = 0.5f;
t_LUM_src.create (r2_RT_luminance_src);
t_LUM_dest.create (r2_RT_luminance_cur);
// create pool
for (u32 it=0; it<HW.Caps.iGPUNum*2; it++) {
string256 name;
xr_sprintf (name,"%s_%d", r2_RT_luminance_pool,it );
rt_LUM_pool[it].create (name, 1, 1, D3DFMT_R32F );
//u_setrt (rt_LUM_pool[it], 0, 0, 0 );
//CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_TARGET, 0x7f7f7f7f, 1.0f, 0L));
FLOAT ColorRGBA[4] = { 127.0f/255.0f, 127.0f/255.0f, 127.0f/255.0f, 127.0f/255.0f};
HW.pDevice->ClearRenderTargetView(rt_LUM_pool[it]->pRT, ColorRGBA);
}
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
}
// HBAO
if (RImplementation.o.ssao_opt_data)
{
u32 w = 0;
u32 h = 0;
if (RImplementation.o.ssao_half_data)
{
w = Device.dwWidth / 2;
h = Device.dwHeight / 2;
}
else
{
w = Device.dwWidth;
h = Device.dwHeight;
}
D3DFORMAT fmt = HW.Caps.id_vendor==0x10DE?D3DFMT_R32F:D3DFMT_R16F;
rt_half_depth.create (r2_RT_half_depth, w, h, fmt);
s_ssao.create (b_ssao, "r2\\ssao");
}
if (RImplementation.o.ssao_blur_on)
{
u32 w = Device.dwWidth, h = Device.dwHeight;
rt_ssao_temp.create (r2_RT_ssao_temp, w, h, D3DFMT_G16R16F, SampleCount);
s_ssao.create (b_ssao, "r2\\ssao");
if( RImplementation.o.dx10_msaa )
{
int bound = RImplementation.o.dx10_msaa_opt ? 1 : RImplementation.o.dx10_msaa_samples;
for( int i = 0; i < bound; ++i )
{
s_ssao_msaa[i].create( b_ssao_msaa[i], "null");
}
}
}
// COMBINE
{
static D3DVERTEXELEMENT9 dwDecl[] =
{
{ 0, 0, D3DDECLTYPE_FLOAT4, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0 }, // pos+uv
D3DDECL_END()
};
s_combine.create (b_combine, "r2\\combine");
s_combine_volumetric.create ("combine_volumetric");
s_combine_dbg_0.create ("effects\\screen_set", r2_RT_smap_surf );
s_combine_dbg_1.create ("effects\\screen_set", r2_RT_luminance_t8 );
s_combine_dbg_Accumulator.create ("effects\\screen_set", r2_RT_accum );
g_combine_VP.create (dwDecl, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine.create (FVF::F_TL, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine_2UV.create (FVF::F_TL2uv, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine_cuboid.create (dwDecl, RCache.Vertex.Buffer(), RCache.Index.Buffer());
u32 fvf_aa_blur = D3DFVF_XYZRHW|D3DFVF_TEX4|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3);
g_aa_blur.create (fvf_aa_blur, RCache.Vertex.Buffer(), RCache.QuadIB);
u32 fvf_aa_AA = D3DFVF_XYZRHW|D3DFVF_TEX7|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3)|D3DFVF_TEXCOORDSIZE2(4)|D3DFVF_TEXCOORDSIZE4(5)|D3DFVF_TEXCOORDSIZE4(6);
g_aa_AA.create (fvf_aa_AA, RCache.Vertex.Buffer(), RCache.QuadIB);
t_envmap_0.create (r2_T_envs0);
t_envmap_1.create (r2_T_envs1);
}
// Build textures
{
// Testure for async sreenshots
{
D3D10_TEXTURE2D_DESC desc;
desc.Width = Device.dwWidth;
desc.Height = Device.dwHeight;
desc.MipLevels = 1;
desc.ArraySize = 1;
desc.SampleDesc.Count = 1;
desc.SampleDesc.Quality = 0;
desc.Format = DXGI_FORMAT_R8G8B8A8_SNORM;
desc.Usage = D3D10_USAGE_STAGING;
desc.BindFlags = 0;
desc.CPUAccessFlags = D3D10_CPU_ACCESS_READ;
desc.MiscFlags = 0;
R_CHK( HW.pDevice->CreateTexture2D(&desc, 0, &t_ss_async) );
}
// Build material(s)
{
// Create immutable texture.
// So we need to init data _before_ the creation.
// Surface
//R_CHK (D3DXCreateVolumeTexture(HW.pDevice,TEX_material_LdotN,TEX_material_LdotH,4,1,0,D3DFMT_A8L8,D3DPOOL_MANAGED,&t_material_surf));
//t_material = dxRenderDeviceRender::Instance().Resources->_CreateTexture(r2_material);
//t_material->surface_set (t_material_surf);
// Use DXGI_FORMAT_R8G8_UNORM
u16 tempData[TEX_material_LdotN*TEX_material_LdotH*TEX_material_Count];
D3D10_TEXTURE3D_DESC desc;
desc.Width = TEX_material_LdotN;
desc.Height = TEX_material_LdotH;
desc.Depth = TEX_material_Count;
desc.MipLevels = 1;
desc.Format = DXGI_FORMAT_R8G8_UNORM;
desc.Usage = D3D10_USAGE_IMMUTABLE;
desc.BindFlags = D3D10_BIND_SHADER_RESOURCE;
desc.CPUAccessFlags = 0;
desc.MiscFlags = 0;
D3D10_SUBRESOURCE_DATA subData;
subData.pSysMem = tempData;
subData.SysMemPitch = desc.Width*2;
subData.SysMemSlicePitch = desc.Height*subData.SysMemPitch;
// Fill it (addr: x=dot(L,N),y=dot(L,H))
//D3DLOCKED_BOX R;
//R_CHK (t_material_surf->LockBox (0,&R,0,0));
for (u32 slice=0; slice<TEX_material_Count; slice++)
{
for (u32 y=0; y<TEX_material_LdotH; y++)
{
for (u32 x=0; x<TEX_material_LdotN; x++)
{
u16* p = (u16*)
(LPBYTE (subData.pSysMem)
+ slice*subData.SysMemSlicePitch
+ y*subData.SysMemPitch + x*2);
float ld = float(x) / float (TEX_material_LdotN-1);
float ls = float(y) / float (TEX_material_LdotH-1) + EPS_S;
ls *= powf(ld,1/32.f);
float fd,fs;
switch (slice)
{
case 0: { // looks like OrenNayar
fd = powf(ld,0.75f); // 0.75
fs = powf(ls,16.f)*.5f;
}
break;
case 1: {// looks like Blinn
fd = powf(ld,0.90f); // 0.90
fs = powf(ls,24.f);
}
break;
case 2: { // looks like Phong
fd = ld; // 1.0
fs = powf(ls*1.01f,128.f );
}
break;
case 3: { // looks like Metal
float s0 = _abs (1-_abs (0.05f*_sin(33.f*ld)+ld-ls));
float s1 = _abs (1-_abs (0.05f*_cos(33.f*ld*ls)+ld-ls));
float s2 = _abs (1-_abs (ld-ls));
fd = ld; // 1.0
fs = powf (_max(_max(s0,s1),s2), 24.f);
fs *= powf (ld,1/7.f);
}
break;
default:
fd = fs = 0;
}
s32 _d = clampr (iFloor (fd*255.5f), 0,255);
s32 _s = clampr (iFloor (fs*255.5f), 0,255);
if ((y==(TEX_material_LdotH-1)) && (x==(TEX_material_LdotN-1))) {
_d = 255;
_s=255;
}
*p = u16 (_s*256 + _d);
}
}
}
//R_CHK (t_material_surf->UnlockBox (0));
R_CHK(HW.pDevice->CreateTexture3D(&desc, &subData, &t_material_surf));
t_material = dxRenderDeviceRender::Instance().Resources->_CreateTexture(r2_material);
t_material->surface_set (t_material_surf);
//R_CHK (D3DXCreateVolumeTexture(HW.pDevice,TEX_material_LdotN,TEX_material_LdotH,4,1,0,D3DFMT_A8L8,D3DPOOL_MANAGED,&t_material_surf));
//t_material = dxRenderDeviceRender::Instance().Resources->_CreateTexture(r2_material);
//t_material->surface_set (t_material_surf);
// #ifdef DEBUG
// R_CHK (D3DXSaveTextureToFile ("x:\\r2_material.dds",D3DXIFF_DDS,t_material_surf,0));
// #endif
}
// Build noise table
if (1)
{
// Surfaces
//D3DLOCKED_RECT R[TEX_jitter_count];
//for (int it=0; it<TEX_jitter_count; it++)
//{
// string_path name;
// xr_sprintf (name,"%s%d",r2_jitter,it);
// R_CHK (D3DXCreateTexture (HW.pDevice,TEX_jitter,TEX_jitter,1,0,D3DFMT_Q8W8V8U8,D3DPOOL_MANAGED,&t_noise_surf[it]));
// t_noise[it] = dxRenderDeviceRender::Instance().Resources->_CreateTexture (name);
// t_noise[it]->surface_set (t_noise_surf[it]);
// R_CHK (t_noise_surf[it]->LockRect (0,&R[it],0,0));
//}
// Use DXGI_FORMAT_R8G8B8A8_SNORM
static const int sampleSize = 4;
u32 tempData[TEX_jitter_count][TEX_jitter*TEX_jitter];
D3D10_TEXTURE2D_DESC desc;
desc.Width = TEX_jitter;
desc.Height = TEX_jitter;
desc.MipLevels = 1;
desc.ArraySize = 1;
desc.SampleDesc.Count = 1;
desc.SampleDesc.Quality = 0;
desc.Format = DXGI_FORMAT_R8G8B8A8_SNORM;
//desc.Usage = D3D10_USAGE_IMMUTABLE;
desc.Usage = D3D10_USAGE_DEFAULT;
desc.BindFlags = D3D10_BIND_SHADER_RESOURCE;
desc.CPUAccessFlags = 0;
desc.MiscFlags = 0;
D3D10_SUBRESOURCE_DATA subData[TEX_jitter_count];
for (int it=0; it<TEX_jitter_count-1; it++)
{
subData[it].pSysMem = tempData[it];
subData[it].SysMemPitch = desc.Width*sampleSize;
}
// Fill it,
for (u32 y=0; y<TEX_jitter; y++)
{
for (u32 x=0; x<TEX_jitter; x++)
{
DWORD data [TEX_jitter_count-1];
generate_jitter (data,TEX_jitter_count-1);
for (u32 it=0; it<TEX_jitter_count-1; it++)
{
u32* p = (u32*)
(LPBYTE (subData[it].pSysMem)
+ y*subData[it].SysMemPitch
+ x*4);
*p = data [it];
}
}
}
//for (int it=0; it<TEX_jitter_count; it++) {
// R_CHK (t_noise_surf[it]->UnlockRect(0));
//}
for (int it=0; it<TEX_jitter_count-1; it++)
{
string_path name;
xr_sprintf (name,"%s%d",r2_jitter,it);
//R_CHK (D3DXCreateTexture (HW.pDevice,TEX_jitter,TEX_jitter,1,0,D3DFMT_Q8W8V8U8,D3DPOOL_MANAGED,&t_noise_surf[it]));
R_CHK( HW.pDevice->CreateTexture2D(&desc, &subData[it], &t_noise_surf[it]) );
t_noise[it] = dxRenderDeviceRender::Instance().Resources->_CreateTexture (name);
t_noise[it]->surface_set (t_noise_surf[it]);
//R_CHK (t_noise_surf[it]->LockRect (0,&R[it],0,0));
}
float tempDataHBAO[TEX_jitter*TEX_jitter*4];
// generate HBAO jitter texture (last)
D3D10_TEXTURE2D_DESC descHBAO;
descHBAO.Width = TEX_jitter;
descHBAO.Height = TEX_jitter;
descHBAO.MipLevels = 1;
descHBAO.ArraySize = 1;
descHBAO.SampleDesc.Count = 1;
descHBAO.SampleDesc.Quality = 0;
descHBAO.Format = DXGI_FORMAT_R32G32B32A32_FLOAT;
//desc.Usage = D3D10_USAGE_IMMUTABLE;
descHBAO.Usage = D3D10_USAGE_DEFAULT;
descHBAO.BindFlags = D3D10_BIND_SHADER_RESOURCE;
descHBAO.CPUAccessFlags = 0;
descHBAO.MiscFlags = 0;
it = TEX_jitter_count-1;
subData[it].pSysMem = tempDataHBAO;
subData[it].SysMemPitch = descHBAO.Width*sampleSize * sizeof(float);
// Fill it,
for (u32 y=0; y<TEX_jitter; y++)
{
for (u32 x=0; x<TEX_jitter; x++)
{
float numDir = 1.0f;
switch (ps_r_ssao)
{
case 1:
numDir = 4.0f;
break;
case 2:
numDir = 6.0f;
break;
case 3:
numDir = 8.0f;
break;
case 4:
numDir = 8.0f;
break;
}
float angle = 2 * PI * ::Random.randF(0.0f, 1.0f) / numDir;
float dist = ::Random.randF(0.0f, 1.0f);
float *p = (float*)
(LPBYTE (subData[it].pSysMem)
+ y*subData[it].SysMemPitch
+ x*4*sizeof(float));
*p = (float)(_cos(angle));
*(p+1) = (float)(_sin(angle));
*(p+2) = (float)(dist);
*(p+3) = 0;
}
}
string_path name;
xr_sprintf (name,"%s%d",r2_jitter,it);
//R_CHK (D3DXCreateTexture (HW.pDevice,TEX_jitter,TEX_jitter,1,0,D3DFMT_Q8W8V8U8,D3DPOOL_MANAGED,&t_noise_surf[it]));
R_CHK( HW.pDevice->CreateTexture2D(&descHBAO, &subData[it], &t_noise_surf[it]) );
t_noise[it] = dxRenderDeviceRender::Instance().Resources->_CreateTexture (name);
t_noise[it]->surface_set (t_noise_surf[it]);
// Create noise mipped
{
// Autogen mipmaps
desc.MipLevels = 0;
R_CHK( HW.pDevice->CreateTexture2D(&desc, 0, &t_noise_surf_mipped) );
t_noise_mipped = dxRenderDeviceRender::Instance().Resources->_CreateTexture(r2_jitter_mipped);
t_noise_mipped->surface_set(t_noise_surf_mipped);
// Update texture. Generate mips.
HW.pDevice->CopySubresourceRegion( t_noise_surf_mipped, 0, 0, 0, 0, t_noise_surf[0], 0, 0 );
D3DX10FilterTexture(t_noise_surf_mipped, 0, D3DX10_FILTER_POINT);
}
}
}
// PP
s_postprocess.create ("postprocess");
g_postprocess.create (D3DFVF_XYZRHW|D3DFVF_DIFFUSE|D3DFVF_SPECULAR|D3DFVF_TEX3,RCache.Vertex.Buffer(),RCache.QuadIB);
// Menu
s_menu.create ("distort");
g_menu.create (FVF::F_TL,RCache.Vertex.Buffer(),RCache.QuadIB);
//
dwWidth = Device.dwWidth;
dwHeight = Device.dwHeight;
}
void CRenderTarget::phase_ssao ()
{
u32 Offset = 0;
FLOAT ColorRGBA[4] = {0.0f, 0.0f, 0.0f, 0.0f};
HW.pDevice->ClearRenderTargetView(rt_ssao_temp->pRT, ColorRGBA);
// low/hi RTs
if( !RImplementation.o.dx10_msaa )
{
u_setrt ( rt_ssao_temp,0,0,0/*HW.pBaseZB*/ );
}
else
{
u_setrt ( rt_ssao_temp, 0, 0, 0/*RImplementation.Target->rt_MSAADepth->pZRT*/ );
}
RCache.set_Stencil (FALSE);
/*RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00); // stencil should be >= 1
if (RImplementation.o.nvstencil) {
u_stencil_optimize (CRenderTarget::SO_Combine);
RCache.set_ColorWriteEnable ();
}*/
// Compute params
Fmatrix m_v2w; m_v2w.invert (Device.mView );
float fSSAONoise = 2.0f;
fSSAONoise *= tan(deg2rad(67.5f));
fSSAONoise /= tan(deg2rad(Device.fFOV));
float fSSAOKernelSize = 150.0f;
fSSAOKernelSize *= tan(deg2rad(67.5f));
fSSAOKernelSize /= tan(deg2rad(Device.fFOV));
// Fill VB
float scale_X = float(Device.dwWidth) * 0.5f / float(TEX_jitter);
float scale_Y = float(Device.dwHeight) * 0.5f / float(TEX_jitter);
u32 _w = Device.dwWidth/2;
u32 _h = Device.dwHeight/2;
set_viewport(HW.pDevice, _w, _h);
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set ( -1, 1, 0, 1, 0, 0, scale_Y ); pv++;
pv->set ( -1, -1, 0, 0, 0, 0, 0 ); pv++;
pv->set ( 1, 1, 1, 1, 0, scale_X, scale_Y ); pv++;
pv->set ( 1, -1, 1, 0, 0, scale_X, 0 ); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
// Draw
RCache.set_Element (s_ssao->E[0] );
RCache.set_Geometry (g_combine );
RCache.set_c ("m_v2w", m_v2w );
RCache.set_c ("ssao_noise_tile_factor", fSSAONoise );
RCache.set_c ("ssao_kernel_size", fSSAOKernelSize );
RCache.set_c ("resolution", float(_w), float(_h), 1.0f / float(_w), 1.0f / float(_h) );
if( !RImplementation.o.dx10_msaa )
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
else
{
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
/*RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
if( RImplementation.o.dx10_msaa_opt )
{
RCache.set_Element( s_ssao_msaa[0]->E[0] );
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
for( u32 i = 0; i < RImplementation.o.dx10_msaa_samples; ++i )
{
RCache.set_Element ( s_ssao_msaa[i]->E[0] );
StateManager.SetSampleMask ( u32(1) << i );
RCache.set_Stencil ( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
StateManager.SetSampleMask( 0xffffffff );
}*/
//RCache.set_Stencil( FALSE, D3DCMP_EQUAL, 0x01, 0xff, 0 );
}
set_viewport(HW.pDevice, Device.dwWidth, Device.dwHeight);
RCache.set_Stencil (FALSE);
}
void CRenderTarget::accum_direct_f (u32 sub_phase)
{
// Select target
if (SE_SUN_LUMINANCE==sub_phase) {
accum_direct_lum ();
return ;
}
phase_accumulator ();
u_setrt (rt_Generic_0,NULL,NULL,HW.pBaseZB);
// *** assume accumulator setted up ***
light* fuckingsun = (light*)RImplementation.Lights.sun_adapted._get() ;
// Common calc for quad-rendering
u32 Offset;
u32 C = color_rgba (255,255,255,255);
float _w = float (Device.dwWidth);
float _h = float (Device.dwHeight);
Fvector2 p0,p1;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
float d_Z = EPS_S, d_W = 1.f;
// Common constants (light-related)
Fvector L_dir,L_clr; float L_spec;
L_clr.set (fuckingsun->color.r,fuckingsun->color.g,fuckingsun->color.b);
L_spec = u_diffuse2s (L_clr);
Device.mView.transform_dir (L_dir,fuckingsun->direction);
L_dir.normalize ();
// Perform masking (only once - on the first/near phase)
RCache.set_CullMode (CULL_NONE );
if (SE_SUN_NEAR==sub_phase) //.
{
// For sun-filter - clear to zero
CHK_DX (HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_TARGET, 0, 1.0f, 0L));
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set (EPS, float(_h+EPS), d_Z, d_W, C, p0.x, p1.y); pv++;
pv->set (EPS, EPS, d_Z, d_W, C, p0.x, p0.y); pv++;
pv->set (float(_w+EPS), float(_h+EPS), d_Z, d_W, C, p1.x, p1.y); pv++;
pv->set (float(_w+EPS), EPS, d_Z, d_W, C, p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
RCache.set_Geometry (g_combine);
// setup
float intensity = 0.3f*fuckingsun->color.r + 0.48f*fuckingsun->color.g + 0.22f*fuckingsun->color.b;
Fvector dir = L_dir;
dir.normalize().mul (- _sqrt(intensity+EPS));
RCache.set_Element (s_accum_mask->E[SE_MASK_DIRECT]); // masker
RCache.set_c ("Ldynamic_dir", dir.x,dir.y,dir.z,0 );
// if (stencil>=1 && aref_pass) stencil = light_id
RCache.set_ColorWriteEnable (FALSE );
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,dwLightMarkerID,0x01,0xff,D3DSTENCILOP_KEEP,D3DSTENCILOP_REPLACE,D3DSTENCILOP_KEEP);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// recalculate d_Z, to perform depth-clipping
Fvector center_pt; center_pt.mad (Device.vCameraPosition,Device.vCameraDirection,ps_r2_sun_near);
Device.mFullTransform.transform(center_pt) ;
d_Z = center_pt.z ;
// nv-stencil recompression
if (RImplementation.o.nvstencil && (SE_SUN_NEAR==sub_phase)) u_stencil_optimize(); //. driver bug?
// Perform lighting
{
u_setrt (rt_Generic_0,NULL,NULL,HW.pBaseZB); // enshure RT setup
RCache.set_CullMode (CULL_NONE );
RCache.set_ColorWriteEnable ();
// texture adjustment matrix
float fTexelOffs = (.5f / float(RImplementation.o.smapsize));
float fRange = (SE_SUN_NEAR==sub_phase)?ps_r2_sun_depth_near_scale:ps_r2_sun_depth_far_scale;
float fBias = (SE_SUN_NEAR==sub_phase)?ps_r2_sun_depth_near_bias:ps_r2_sun_depth_far_bias;
Fmatrix m_TexelAdjust =
{
0.5f, 0.0f, 0.0f, 0.0f,
0.0f, -0.5f, 0.0f, 0.0f,
0.0f, 0.0f, fRange, 0.0f,
0.5f + fTexelOffs, 0.5f + fTexelOffs, fBias, 1.0f
};
// compute xforms
Fmatrix m_shadow;
{
FPU::m64r ();
Fmatrix xf_invview; xf_invview.invert (Device.mView) ;
Fmatrix xf_project; xf_project.mul (m_TexelAdjust,fuckingsun->X.D.combine);
m_shadow.mul (xf_project, xf_invview);
// tsm-bias
if (SE_SUN_FAR == sub_phase)
{
Fvector bias; bias.mul (L_dir,ps_r2_sun_tsm_bias);
Fmatrix bias_t; bias_t.translate(bias);
m_shadow.mulB_44 (bias_t);
}
FPU::m24r ();
}
// Make jitter texture
Fvector2 j0,j1;
float scale_X = float(Device.dwWidth) / float(TEX_jitter);
//float scale_Y = float(Device.dwHeight)/ float(TEX_jitter);
float offset = (.5f / float(TEX_jitter));
j0.set (offset,offset);
j1.set (scale_X,scale_X).add(offset);
// Fill vertex buffer
FVF::TL2uv* pv = (FVF::TL2uv*) RCache.Vertex.Lock (4,g_combine_2UV->vb_stride,Offset);
pv->set (EPS, float(_h+EPS), d_Z, d_W, C, p0.x, p1.y, j0.x, j1.y); pv++;
pv->set (EPS, EPS, d_Z, d_W, C, p0.x, p0.y, j0.x, j0.y); pv++;
pv->set (float(_w+EPS), float(_h+EPS), d_Z, d_W, C, p1.x, p1.y, j1.x, j1.y); pv++;
pv->set (float(_w+EPS), EPS, d_Z, d_W, C, p1.x, p0.y, j1.x, j0.y); pv++;
RCache.Vertex.Unlock (4,g_combine_2UV->vb_stride);
RCache.set_Geometry (g_combine_2UV);
// setup
RCache.set_Element (s_accum_direct->E[sub_phase]);
RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
RCache.set_c ("Ldynamic_color", L_clr.x,L_clr.y,L_clr.z,L_spec );
RCache.set_c ("m_shadow", m_shadow );
// setup stencil
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,dwLightMarkerID,0xff,0x00);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
}
// startup
void CRenderTarget::phase_scene_prepare ()
{
// Clear depth & stencil
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB );
CHK_DX ( HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_ZBUFFER|D3DCLEAR_STENCIL, 0x0, 1.0f, 0L) );
}
CRenderTarget::CRenderTarget ()
{
param_blur = 0.f;
param_gray = 0.f;
param_noise = 0.f;
param_duality_h = 0.f;
param_duality_v = 0.f;
param_noise_fps = 25.f;
param_noise_scale = 1.f;
im_noise_time = 1/100;
im_noise_shift_w = 0;
im_noise_shift_h = 0;
param_color_base = color_rgba(127,127,127, 0);
param_color_gray = color_rgba(85,85,85, 0);
param_color_add = color_rgba(0,0,0, 0);
dwAccumulatorClearMark = 0;
Device.Resources->Evict ();
// Blenders
b_occq = xr_new<CBlender_light_occq> ();
b_accum_mask = xr_new<CBlender_accum_direct_mask> ();
b_accum_direct = xr_new<CBlender_accum_direct> ();
b_accum_point = xr_new<CBlender_accum_point> ();
b_accum_spot = xr_new<CBlender_accum_spot> ();
b_accum_reflected = xr_new<CBlender_accum_reflected> ();
b_bloom = xr_new<CBlender_bloom_build> ();
b_luminance = xr_new<CBlender_luminance> ();
b_combine = xr_new<CBlender_combine> ();
// NORMAL
{
u32 w=Device.dwWidth, h=Device.dwHeight;
rt_Position.create (r2_RT_P, w,h,D3DFMT_A16B16G16R16F);
rt_Normal.create (r2_RT_N, w,h,D3DFMT_A16B16G16R16F);
// select albedo & accum
if (RImplementation.o.mrtmixdepth)
{
// NV50
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8 );
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F);
}
else
{
// can't - mix-depth
if (RImplementation.o.fp16_blend) {
// NV40
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A16B16G16R16F); // expand to full
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F);
} else {
// R4xx, no-fp-blend,-> albedo_wo
VERIFY (RImplementation.o.albedo_wo);
rt_Color.create (r2_RT_albedo, w,h,D3DFMT_A8R8G8B8 ); // normal
rt_Accumulator.create (r2_RT_accum, w,h,D3DFMT_A16B16G16R16F);
rt_Accumulator_temp.create (r2_RT_accum_temp, w,h,D3DFMT_A16B16G16R16F);
}
}
// generic(LDR) RTs
rt_Generic_0.create (r2_RT_generic0,w,h,D3DFMT_A8R8G8B8 );
rt_Generic_1.create (r2_RT_generic1,w,h,D3DFMT_A8R8G8B8 );
}
// OCCLUSION
s_occq.create (b_occq, "r2\\occq");
// DIRECT (spot)
D3DFORMAT depth_format = (D3DFORMAT)RImplementation.o.HW_smap_FORMAT;
if (RImplementation.o.HW_smap)
{
D3DFORMAT nullrt = D3DFMT_R5G6B5;
if (RImplementation.o.nullrt) nullrt = (D3DFORMAT)MAKEFOURCC('N','U','L','L');
u32 size =RImplementation.o.smapsize ;
rt_smap_depth.create (r2_RT_smap_depth, size,size,depth_format );
rt_smap_surf.create (r2_RT_smap_surf, size,size,nullrt );
rt_smap_ZB = NULL;
s_accum_mask.create (b_accum_mask, "r2\\accum_mask");
s_accum_direct.create (b_accum_direct, "r2\\accum_direct");
}
else
{
u32 size =RImplementation.o.smapsize ;
rt_smap_surf.create (r2_RT_smap_surf, size,size,D3DFMT_R32F);
rt_smap_depth = NULL;
R_CHK (HW.pDevice->CreateDepthStencilSurface (size,size,D3DFMT_D24X8,D3DMULTISAMPLE_NONE,0,TRUE,&rt_smap_ZB,NULL));
s_accum_mask.create (b_accum_mask, "r2\\accum_mask");
s_accum_direct.create (b_accum_direct, "r2\\accum_direct");
}
// POINT
{
s_accum_point.create (b_accum_point, "r2\\accum_point_s");
accum_point_geom_create ();
g_accum_point.create (D3DFVF_XYZ, g_accum_point_vb, g_accum_point_ib);
accum_omnip_geom_create ();
g_accum_omnipart.create (D3DFVF_XYZ, g_accum_omnip_vb, g_accum_omnip_ib);
}
// SPOT
{
s_accum_spot.create (b_accum_spot, "r2\\accum_spot_s", "lights\\lights_spot01");
accum_spot_geom_create ();
g_accum_spot.create (D3DFVF_XYZ, g_accum_spot_vb, g_accum_spot_ib);
}
// REFLECTED
{
s_accum_reflected.create (b_accum_reflected, "r2\\accum_refl");
}
// BLOOM
{
D3DFORMAT fmt = D3DFMT_A8R8G8B8; //; // D3DFMT_X8R8G8B8
u32 w=BLOOM_size_X, h=BLOOM_size_Y;
u32 fvf_build = D3DFVF_XYZRHW|D3DFVF_TEX4|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3);
u32 fvf_filter = (u32)D3DFVF_XYZRHW|D3DFVF_TEX8|D3DFVF_TEXCOORDSIZE4(0)|D3DFVF_TEXCOORDSIZE4(1)|D3DFVF_TEXCOORDSIZE4(2)|D3DFVF_TEXCOORDSIZE4(3)|D3DFVF_TEXCOORDSIZE4(4)|D3DFVF_TEXCOORDSIZE4(5)|D3DFVF_TEXCOORDSIZE4(6)|D3DFVF_TEXCOORDSIZE4(7);
rt_Bloom_1.create (r2_RT_bloom1, w,h, fmt);
rt_Bloom_2.create (r2_RT_bloom2, w,h, fmt);
g_bloom_build.create (fvf_build, RCache.Vertex.Buffer(), RCache.QuadIB);
g_bloom_filter.create (fvf_filter, RCache.Vertex.Buffer(), RCache.QuadIB);
s_bloom_dbg_1.create ("effects\\screen_set", r2_RT_bloom1);
s_bloom_dbg_2.create ("effects\\screen_set", r2_RT_bloom2);
s_bloom.create (b_bloom, "r2\\bloom");
f_bloom_factor = 0.5f;
}
// TONEMAP
{
rt_LUM_64.create (r2_RT_luminance_t64, 64, 64, D3DFMT_A16B16G16R16F );
rt_LUM_8.create (r2_RT_luminance_t8, 8, 8, D3DFMT_A16B16G16R16F );
s_luminance.create (b_luminance, "r2\\luminance");
f_luminance_adapt = 0.5f;
t_LUM_src.create (r2_RT_luminance_src);
t_LUM_dest.create (r2_RT_luminance_cur);
// create pool
for (u32 it=0; it<4; it++) {
string256 name;
sprintf (name,"%s_%d", r2_RT_luminance_pool,it );
rt_LUM_pool[it].create (name, 1, 1, D3DFMT_R32F );
u_setrt (rt_LUM_pool[it], 0, 0, 0 );
CHK_DX (HW.pDevice->Clear( 0L, NULL, D3DCLEAR_TARGET, 0x7f7f7f7f, 1.0f, 0L));
}
u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
}
// COMBINE
{
static D3DVERTEXELEMENT9 dwDecl[] =
{
{ 0, 0, D3DDECLTYPE_FLOAT4, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0 }, // pos+uv
D3DDECL_END()
};
s_combine.create (b_combine, "r2\\combine");
s_combine_dbg_0.create ("effects\\screen_set", r2_RT_smap_surf );
s_combine_dbg_1.create ("effects\\screen_set", r2_RT_luminance_t8 );
s_combine_dbg_Accumulator.create ("effects\\screen_set", r2_RT_accum );
g_combine_VP.create (dwDecl, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine.create (FVF::F_TL, RCache.Vertex.Buffer(), RCache.QuadIB);
g_combine_2UV.create (FVF::F_TL2uv, RCache.Vertex.Buffer(), RCache.QuadIB);
u32 fvf_aa_blur = D3DFVF_XYZRHW|D3DFVF_TEX4|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3);
g_aa_blur.create (fvf_aa_blur, RCache.Vertex.Buffer(), RCache.QuadIB);
u32 fvf_aa_AA = D3DFVF_XYZRHW|D3DFVF_TEX7|D3DFVF_TEXCOORDSIZE2(0)|D3DFVF_TEXCOORDSIZE2(1)|D3DFVF_TEXCOORDSIZE2(2)|D3DFVF_TEXCOORDSIZE2(3)|D3DFVF_TEXCOORDSIZE2(4)|D3DFVF_TEXCOORDSIZE4(5)|D3DFVF_TEXCOORDSIZE4(6);
g_aa_AA.create (fvf_aa_AA, RCache.Vertex.Buffer(), RCache.QuadIB);
t_envmap_0.create (r2_T_envs0);
t_envmap_1.create (r2_T_envs1);
}
// Build textures
{
// Build material(s)
{
// Surface
R_CHK (D3DXCreateVolumeTexture(HW.pDevice,TEX_material_LdotN,TEX_material_LdotH,4,1,0,D3DFMT_A8L8,D3DPOOL_MANAGED,&t_material_surf));
t_material = Device.Resources->_CreateTexture(r2_material);
t_material->surface_set (t_material_surf);
// Fill it (addr: x=dot(L,N),y=dot(L,H))
D3DLOCKED_BOX R;
R_CHK (t_material_surf->LockBox (0,&R,0,0));
for (u32 slice=0; slice<4; slice++)
{
for (u32 y=0; y<TEX_material_LdotH; y++)
{
for (u32 x=0; x<TEX_material_LdotN; x++)
{
u16* p = (u16*) (LPBYTE (R.pBits) + slice*R.SlicePitch + y*R.RowPitch + x*2);
float ld = float(x) / float (TEX_material_LdotN-1);
float ls = float(y) / float (TEX_material_LdotH-1) + EPS_S;
ls *= powf(ld,1/32.f);
float fd,fs;
switch (slice)
{
case 0: { // looks like OrenNayar
fd = powf(ld,0.75f); // 0.75
fs = powf(ls,16.f)*.5f;
} break;
case 1: {// looks like Blinn
fd = powf(ld,0.90f); // 0.90
fs = powf(ls,24.f);
} break;
case 2: { // looks like Phong
fd = ld; // 1.0
fs = powf(ls*1.01f,128.f );
} break;
case 3: { // looks like Metal
float s0 = _abs (1-_abs (0.05f*_sin(33.f*ld)+ld-ls));
float s1 = _abs (1-_abs (0.05f*_cos(33.f*ld*ls)+ld-ls));
float s2 = _abs (1-_abs (ld-ls));
fd = ld; // 1.0
fs = powf (_max(_max(s0,s1),s2), 24.f);
fs *= powf (ld,1/7.f);
} break;
default:
fd = fs = 0;
}
s32 _d = clampr (iFloor (fd*255.5f), 0,255);
s32 _s = clampr (iFloor (fs*255.5f), 0,255);
if ((y==(TEX_material_LdotH-1)) && (x==(TEX_material_LdotN-1))) { _d = 255; _s=255; }
*p = u16 (_s*256 + _d);
}
}
}
R_CHK (t_material_surf->UnlockBox (0));
// #ifdef DEBUG
// R_CHK (D3DXSaveTextureToFile ("x:\\r2_material.dds",D3DXIFF_DDS,t_material_surf,0));
// #endif
}
// Build noise table
if (1)
{
// Surfaces
D3DLOCKED_RECT R[TEX_jitter_count];
for (int it=0; it<TEX_jitter_count; it++)
{
string_path name;
sprintf (name,"%s%d",r2_jitter,it);
R_CHK (D3DXCreateTexture (HW.pDevice,TEX_jitter,TEX_jitter,1,0,D3DFMT_Q8W8V8U8,D3DPOOL_MANAGED,&t_noise_surf[it]));
t_noise[it] = Device.Resources->_CreateTexture (name);
t_noise[it]->surface_set (t_noise_surf[it]);
R_CHK (t_noise_surf[it]->LockRect (0,&R[it],0,0));
}
// Fill it,
for (u32 y=0; y<TEX_jitter; y++)
{
for (u32 x=0; x<TEX_jitter; x++)
{
DWORD data [TEX_jitter_count];
generate_jitter (data,TEX_jitter_count);
for (u32 it=0; it<TEX_jitter_count; it++)
{
u32* p = (u32*) (LPBYTE (R[it].pBits) + y*R[it].Pitch + x*4);
*p = data [it];
}
}
}
for (int it=0; it<TEX_jitter_count; it++) {
R_CHK (t_noise_surf[it]->UnlockRect(0));
}
}
}
// PP
s_postprocess.create ("postprocess");
g_postprocess.create (D3DFVF_XYZRHW|D3DFVF_DIFFUSE|D3DFVF_SPECULAR|D3DFVF_TEX3,RCache.Vertex.Buffer(),RCache.QuadIB);
// Menu
s_menu.create ("distort");
g_menu.create (FVF::F_TL,RCache.Vertex.Buffer(),RCache.QuadIB);
//
dwWidth = Device.dwWidth;
dwHeight = Device.dwHeight;
}
void CRenderTarget::draw_rain( light &RainSetup )
{
float fRainFactor = g_pGamePersistent->Environment().CurrentEnv->rain_density;
// Common calc for quad-rendering
u32 Offset;
u32 C = color_rgba (255,255,255,255);
float _w = float (Device.dwWidth);
float _h = float (Device.dwHeight);
Fvector2 p0,p1;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
float d_Z = EPS_S, d_W = 1.f;
// Common constants (light-related)
Fvector L_dir;
Device.mView.transform_dir (L_dir,RainSetup.direction);
L_dir.normalize ();
Fvector W_dirX;
Device.mView.transform_dir (W_dirX,Fvector().set(1.0f, 0.0f, 0.0f));
W_dirX.normalize ();
Fvector W_dirZ;
Device.mView.transform_dir (W_dirZ,Fvector().set(0.0f, 0.0f, 1.0f));
W_dirZ.normalize ();
// Perform masking (only once - on the first/near phase)
//RCache.set_CullMode (CULL_NONE );
//if (SE_SUN_NEAR==sub_phase) //.
{
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set (EPS, float(_h+EPS), d_Z, d_W, C, p0.x, p1.y); pv++;
pv->set (EPS, EPS, d_Z, d_W, C, p0.x, p0.y); pv++;
pv->set (float(_w+EPS), float(_h+EPS), d_Z, d_W, C, p1.x, p1.y); pv++;
pv->set (float(_w+EPS), EPS, d_Z, d_W, C, p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
RCache.set_Geometry (g_combine);
// setup
// float intensity = 0.3f*fuckingsun->color.r + 0.48f*fuckingsun->color.g + 0.22f*fuckingsun->color.b;
// Fvector dir = L_dir;
// dir.normalize().mul (- _sqrt(intensity+EPS));
// RCache.set_Element (s_accum_mask->E[SE_MASK_DIRECT]); // masker
// RCache.set_c ("Ldynamic_dir", dir.x,dir.y,dir.z,0 );
// if (stencil>=1 && aref_pass) stencil = light_id
// Done in blender!
//RCache.set_ColorWriteEnable (FALSE );
// RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,dwLightMarkerID,0x01,0xff,D3DSTENCILOP_KEEP,D3DSTENCILOP_REPLACE,D3DSTENCILOP_KEEP);
// RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// recalculate d_Z, to perform depth-clipping
const float fRainFar = ps_r3_dyn_wet_surf_far;
Fvector center_pt; center_pt.mad (Device.vCameraPosition,Device.vCameraDirection,fRainFar);
Device.mFullTransform.transform(center_pt) ;
d_Z = center_pt.z ;
// nv-stencil recompression
//if (RImplementation.o.nvstencil && (SE_SUN_NEAR==sub_phase)) u_stencil_optimize(); //. driver bug?
// Perform lighting
{
// phase_accumulator () ;
// RCache.set_CullMode (CULL_NONE);
// RCache.set_ColorWriteEnable () ;
// texture adjustment matrix
//float fRange = (SE_SUN_NEAR==sub_phase)?ps_r2_sun_depth_near_scale:ps_r2_sun_depth_far_scale;
float fRange = 1;
//float fBias = (SE_SUN_NEAR==sub_phase)?ps_r2_sun_depth_near_bias:ps_r2_sun_depth_far_bias;
//float fBias = 0.00001;
float fBias = -0.0001;
float smapsize = float(RImplementation.o.smapsize);
float fTexelOffs = (.5f / smapsize);
// float view_dimX = float(RainSetup.X.D.maxX-RainSetup.X.D.minX-2)/smapsize;
// float view_dimY = float(RainSetup.X.D.maxX-RainSetup.X.D.minX-2)/smapsize;
// float view_sx = float(RainSetup.X.D.minX+1)/smapsize;
// float view_sy = float(RainSetup.X.D.minY+1)/smapsize;
float view_dimX = float(RainSetup.X.D.maxX-RainSetup.X.D.minX)/smapsize;
float view_dimY = float(RainSetup.X.D.maxX-RainSetup.X.D.minX)/smapsize;
float view_sx = float(RainSetup.X.D.minX)/smapsize;
float view_sy = float(RainSetup.X.D.minY)/smapsize;
Fmatrix m_TexelAdjust =
{
view_dimX/2.f, 0.0f, 0.0f, 0.0f,
0.0f, -view_dimY/2.f, 0.0f, 0.0f,
0.0f, 0.0f, fRange, 0.0f,
view_dimX/2.f + view_sx + fTexelOffs, view_dimY/2.f + view_sy + fTexelOffs, fBias, 1.0f
};
// compute xforms
FPU::m64r ();
Fmatrix xf_invview; xf_invview.invert (Device.mView) ;
// shadow xform
Fmatrix m_shadow;
{
Fmatrix xf_project; xf_project.mul (m_TexelAdjust,RainSetup.X.D.combine);
m_shadow.mul (xf_project, xf_invview);
FPU::m24r ();
}
/*
// texture adjustment matrix
//float fRange = (SE_SUN_NEAR==sub_phase)?ps_r2_sun_depth_near_scale:ps_r2_sun_depth_far_scale;
float fRange = 1;
//float fBias = (SE_SUN_NEAR==sub_phase)?ps_r2_sun_depth_near_bias:ps_r2_sun_depth_far_bias;
// TODO: DX10: Remove this when fix inverse culling for far region
float fBias = 0;
Fmatrix m_TexelAdjust =
{
0.5f, 0.0f, 0.0f, 0.0f,
0.0f, -0.5f, 0.0f, 0.0f,
0.0f, 0.0f, fRange, 0.0f,
0.5f, 0.5f, fBias, 1.0f
};
// compute xforms
FPU::m64r ();
Fmatrix xf_invview; xf_invview.invert (Device.mView) ;
// shadow xform
Fmatrix m_shadow;
{
Fmatrix xf_project; xf_project.mul (m_TexelAdjust,RainSetup.X.D.combine);
m_shadow.mul (xf_project, xf_invview);
FPU::m24r ();
}
*/
// clouds xform
Fmatrix m_clouds_shadow;
{
static float w_shift = 0;
Fmatrix m_xform;
//Fvector direction = RainSetup.direction ;
Fvector normal ; normal.setHP(1,0);
//w_shift += 0.003f*Device.fTimeDelta;
//Fvector position; position.set(0,0,0);
//m_xform.build_camera_dir (position,direction,normal) ;
m_xform.identity();
Fvector localnormal;m_xform.transform_dir(localnormal,normal); localnormal.normalize();
m_clouds_shadow.mul (m_xform,xf_invview) ;
//m_xform.scale (0.002f,0.002f,1.f) ;
m_xform.scale (1.f,1.f,1.f) ;
m_clouds_shadow.mulA_44 (m_xform) ;
m_xform.translate (localnormal.mul(w_shift)) ;
m_clouds_shadow.mulA_44 (m_xform) ;
}
// Make jitter texture
Fvector2 j0,j1;
float scale_X = float(Device.dwWidth) / float(TEX_jitter);
//float scale_Y = float(Device.dwHeight)/ float(TEX_jitter);
float offset = (.5f / float(TEX_jitter));
j0.set (offset,offset);
j1.set (scale_X,scale_X).add(offset);
// Fill vertex buffer
FVF::TL2uv* pv = (FVF::TL2uv*) RCache.Vertex.Lock (4,g_combine_2UV->vb_stride,Offset);
pv->set (-1, -1, d_Z, d_W, C, 0, 1, 0, scale_X); pv++;
pv->set (-1, 1, d_Z, d_W, C, 0, 0, 0, 0); pv++;
pv->set (1, -1, d_Z, d_W, C, 1, 1, scale_X, scale_X); pv++;
pv->set (1, 1, d_Z, d_W, C, 1, 0, scale_X, 0); pv++;
RCache.Vertex.Unlock (4,g_combine_2UV->vb_stride);
RCache.set_Geometry (g_combine_2UV);
// setup
//RCache.set_Element (s_accum_direct->E[sub_phase]);
//u_setrt (rt_Normal,NULL,NULL,HW.pBaseZB);
//RCache.set_Element (s_rain->E[0]);
//RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
// RCache.set_c ("Ldynamic_color", L_clr.x,L_clr.y,L_clr.z,L_spec );
//RCache.set_c ("m_shadow", m_shadow );
//RCache.set_c ("m_sunmask", m_clouds_shadow );
/*
// nv-DBT
float zMin,zMax;
if (SE_SUN_NEAR==sub_phase) {
zMin = 0;
zMax = ps_r2_sun_near;
} else {
extern float OLES_SUN_LIMIT_27_01_07;
zMin = ps_r2_sun_near;
zMax = OLES_SUN_LIMIT_27_01_07;
}
center_pt.mad(Device.vCameraPosition,Device.vCameraDirection,zMin); Device.mFullTransform.transform (center_pt);
zMin = center_pt.z ;
center_pt.mad(Device.vCameraPosition,Device.vCameraDirection,zMax); Device.mFullTransform.transform (center_pt);
zMax = center_pt.z ;
*/
// TODO: DX10: Check if DX10 has analog for NV DBT
// if (u_DBT_enable(zMin,zMax)) {
// z-test always
// HW.pDevice->SetRenderState(D3DRS_ZFUNC, D3DCMP_ALWAYS);
// HW.pDevice->SetRenderState(D3DRS_ZWRITEENABLE, FALSE);
// }
// Fetch4 : enable
// if (RImplementation.o.HW_smap_FETCH4) {
//. we hacked the shader to force smap on S0
//# define FOURCC_GET4 MAKEFOURCC('G','E','T','4')
// HW.pDevice->SetSamplerState ( 0, D3DSAMP_MIPMAPLODBIAS, FOURCC_GET4 );
// }
// setup stencil
// RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,dwLightMarkerID,0xff,0x00);
// RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
// Fetch4 : disable
// if (RImplementation.o.HW_smap_FETCH4) {
//. we hacked the shader to force smap on S0
//# define FOURCC_GET1 MAKEFOURCC('G','E','T','1')
// HW.pDevice->SetSamplerState ( 0, D3DSAMP_MIPMAPLODBIAS, FOURCC_GET1 );
// }
// Use for intermediate results
// Patch normal
if( !RImplementation.o.dx10_msaa )
u_setrt (rt_Accumulator,NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Accumulator,NULL,NULL,rt_MSAADepth->pZRT);
//u_setrt (rt_Normal,NULL,NULL,HW.pBaseZB);
RCache.set_Element (s_rain->E[1]);
RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
RCache.set_c ("WorldX", W_dirX.x,W_dirX.y,W_dirX.z,0 );
RCache.set_c ("WorldZ", W_dirZ.x,W_dirZ.y,W_dirZ.z,0 );
RCache.set_c ("m_shadow", m_shadow );
RCache.set_c ("m_sunmask", m_clouds_shadow );
RCache.set_c ("RainDensity", fRainFactor, 0, 0, 0 );
RCache.set_c ("RainFallof", ps_r3_dyn_wet_surf_near, ps_r3_dyn_wet_surf_far, 0, 0 );
if( !RImplementation.o.dx10_msaa )
{
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x01, 0 );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
// per pixel execution
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
// per sample
if( RImplementation.o.dx10_msaa_opt )
{
RCache.set_Element( s_rain_msaa[0]->E[0]);
RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
RCache.set_c ("WorldX", W_dirX.x,W_dirX.y,W_dirX.z,0 );
RCache.set_c ("WorldZ", W_dirZ.x,W_dirZ.y,W_dirZ.z,0 );
RCache.set_c ("m_shadow", m_shadow );
RCache.set_c ("m_sunmask", m_clouds_shadow );
RCache.set_c ("RainDensity", fRainFactor, 0, 0, 0 );
RCache.set_CullMode(CULL_NONE );
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
for( u32 i = 0; i < RImplementation.o.dx10_msaa_samples; ++i )
{
RCache.set_Element ( s_rain_msaa[i]->E[0]);
RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
RCache.set_c ("WorldX", W_dirX.x,W_dirX.y,W_dirX.z,0 );
RCache.set_c ("WorldZ", W_dirZ.x,W_dirZ.y,W_dirZ.z,0 );
RCache.set_c ("m_shadow", m_shadow );
RCache.set_c ("m_sunmask", m_clouds_shadow );
RCache.set_c ("RainDensity", fRainFactor, 0, 0, 0 );
StateManager.SetSampleMask ( u32(1) << i );
RCache.set_CullMode(CULL_NONE );
RCache.set_Stencil ( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
StateManager.SetSampleMask( 0xffffffff );
}
}
// Apply normal
RCache.set_Element (s_rain->E[2]);
RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
RCache.set_c ("m_shadow", m_shadow );
RCache.set_c ("m_sunmask", m_clouds_shadow );
if( !RImplementation.o.dx10_gbuffer_opt )
{
// Do this in blender!
//StateManager.SetColorWriteEnable( D3Dxx_COLOR_WRITE_ENABLE_RED | D3Dxx_COLOR_WRITE_ENABLE_GREEN | D3Dxx_COLOR_WRITE_ENABLE_BLUE );
if( ! RImplementation.o.dx10_msaa )
u_setrt (rt_Normal,NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Normal,NULL,NULL,rt_MSAADepth->pZRT);
}
else
{
//StateManager.SetColorWriteEnable( D3Dxx_COLOR_WRITE_ENABLE_RED | D3Dxx_COLOR_WRITE_ENABLE_GREEN );
if( ! RImplementation.o.dx10_msaa )
u_setrt (rt_Position,NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Position,NULL,NULL,rt_MSAADepth->pZRT);
}
if( ! RImplementation.o.dx10_msaa )
{
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x01, 0 );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
// per pixel execution
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
// per sample
if( RImplementation.o.dx10_msaa_opt )
{
RCache.set_Element ( s_rain_msaa[0]->E[1]);
RCache.set_Stencil ( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.set_CullMode(CULL_NONE );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
for( u32 i = 0; i < RImplementation.o.dx10_msaa_samples; ++ i )
{
RCache.set_Element (s_rain_msaa[i]->E[1]);
StateManager.SetSampleMask ( u32(1) << i );
RCache.set_Stencil ( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.set_CullMode(CULL_NONE );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
StateManager.SetSampleMask( 0xffffffff );
}
}
// Apply gloss
RCache.set_Element (s_rain->E[3]);
RCache.set_c ("Ldynamic_dir", L_dir.x,L_dir.y,L_dir.z,0 );
RCache.set_c ("m_shadow", m_shadow );
RCache.set_c ("m_sunmask", m_clouds_shadow );
// It is restored automatically by a set_Element call
//StateManager.SetColorWriteEnable( D3Dxx_COLOR_WRITE_ENABLE_ALL );
if( ! RImplementation.o.dx10_msaa )
u_setrt (rt_Color,NULL,NULL,HW.pBaseZB);
else
u_setrt (rt_Color,NULL,NULL,rt_MSAADepth->pZRT);
if( ! RImplementation.o.dx10_msaa )
{
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x01, 0 );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
// per pixel execution
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x01, 0x81, 0 );
RCache.Render ( D3DPT_TRIANGLELIST,Offset,0,4,0,2);
// per sample
if( RImplementation.o.dx10_msaa_opt )
{
RCache.set_Element(s_rain_msaa[0]->E[2]);
RCache.set_Stencil( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
else
{
for( u32 i = 0; i < RImplementation.o.dx10_msaa_samples; ++i )
{
RCache.set_Element (s_rain_msaa[i]->E[2]);
RCache.set_Stencil ( TRUE, D3DCMP_EQUAL, 0x81, 0x81, 0 );
StateManager.SetSampleMask ( u32(1) << i );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
StateManager.SetSampleMask( 0xffffffff );
}
}
// TODO: DX10: Check if DX10 has analog for NV DBT
// disable depth bounds
// u_DBT_disable ();
}
}
void CRenderTarget::phase_combine ()
{
bool _menu_pp = g_pGamePersistent?g_pGamePersistent->OnRenderPPUI_query():false;
u32 Offset = 0;
Fvector2 p0,p1;
//*** exposure-pipeline
u32 gpu_id = Device.dwFrame%2;
{
t_LUM_src->surface_set (rt_LUM_pool[gpu_id*2+0]->pSurface);
t_LUM_dest->surface_set (rt_LUM_pool[gpu_id*2+1]->pSurface);
}
// low/hi RTs
u_setrt ( rt_Generic_0,rt_Generic_1,0,HW.pBaseZB );
RCache.set_CullMode ( CULL_NONE );
RCache.set_Stencil ( FALSE );
BOOL split_the_scene_to_minimize_wait = FALSE;
if (ps_r2_ls_flags.test(R2FLAG_EXP_SPLIT_SCENE)) split_the_scene_to_minimize_wait=TRUE;
// draw skybox
if (1)
{
RCache.set_ColorWriteEnable ();
CHK_DX(HW.pDevice->SetRenderState ( D3DRS_ZENABLE, FALSE ));
g_pGamePersistent->Environment().RenderSky ();
CHK_DX(HW.pDevice->SetRenderState ( D3DRS_ZENABLE, TRUE ));
}
//
//if (RImplementation.o.bug) {
RCache.set_Stencil (TRUE,D3DCMP_LESSEQUAL,0x01,0xff,0x00); // stencil should be >= 1
if (RImplementation.o.nvstencil) {
u_stencil_optimize (FALSE);
RCache.set_ColorWriteEnable ();
}
//}
// calc m-blur matrices
Fmatrix m_previous, m_current;
Fvector2 m_blur_scale;
{
static Fmatrix m_saved_viewproj;
// (new-camera) -> (world) -> (old_viewproj)
Fmatrix m_invview; m_invview.invert (Device.mView);
m_previous.mul (m_saved_viewproj,m_invview);
m_current.set (Device.mProject) ;
m_saved_viewproj.set(Device.mFullTransform) ;
float scale = ps_r2_mblur/2.f;
m_blur_scale.set (scale,-scale).div(12.f);
}
// Draw full-screen quad textured with our scene image
if (!_menu_pp)
{
// Compute params
Fmatrix m_v2w; m_v2w.invert (Device.mView );
CEnvDescriptorMixer& envdesc= g_pGamePersistent->Environment().CurrentEnv ;
const float minamb = 0.001f;
Fvector4 ambclr = { _max(envdesc.ambient.x*2,minamb), _max(envdesc.ambient.y*2,minamb), _max(envdesc.ambient.z*2,minamb), 0 };
ambclr.mul (ps_r2_sun_lumscale_amb);
Fvector4 envclr = { envdesc.sky_color.x*2+EPS, envdesc.sky_color.y*2+EPS, envdesc.sky_color.z*2+EPS, envdesc.weight };
Fvector4 fogclr = { envdesc.fog_color.x, envdesc.fog_color.y, envdesc.fog_color.z, 0 };
envclr.x *= 2*ps_r2_sun_lumscale_hemi;
envclr.y *= 2*ps_r2_sun_lumscale_hemi;
envclr.z *= 2*ps_r2_sun_lumscale_hemi;
Fvector4 sunclr,sundir;
// sun-params
{
light* fuckingsun = (light*)RImplementation.Lights.sun_adapted._get() ;
Fvector L_dir,L_clr; float L_spec;
L_clr.set (fuckingsun->color.r,fuckingsun->color.g,fuckingsun->color.b);
L_spec = u_diffuse2s (L_clr);
Device.mView.transform_dir (L_dir,fuckingsun->direction);
L_dir.normalize ();
sunclr.set (L_clr.x,L_clr.y,L_clr.z,L_spec);
sundir.set (L_dir.x,L_dir.y,L_dir.z,0);
}
// Fill VB
float _w = float(Device.dwWidth);
float _h = float(Device.dwHeight);
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
// Fill vertex buffer
Fvector4* pv = (Fvector4*) RCache.Vertex.Lock (4,g_combine_VP->vb_stride,Offset);
pv->set (hclip(EPS, _w), hclip(_h+EPS, _h), p0.x, p1.y); pv++;
pv->set (hclip(EPS, _w), hclip(EPS, _h), p0.x, p0.y); pv++;
pv->set (hclip(_w+EPS, _w), hclip(_h+EPS, _h), p1.x, p1.y); pv++;
pv->set (hclip(_w+EPS, _w), hclip(EPS, _h), p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine_VP->vb_stride);
// Setup textures
IDirect3DBaseTexture9* e0 = _menu_pp?0:envdesc.sky_r_textures_env[0].second->surface_get();
IDirect3DBaseTexture9* e1 = _menu_pp?0:envdesc.sky_r_textures_env[1].second->surface_get();
t_envmap_0->surface_set (e0); _RELEASE(e0);
t_envmap_1->surface_set (e1); _RELEASE(e1);
// Draw
RCache.set_Element (s_combine->E[0] );
RCache.set_Geometry (g_combine_VP );
RCache.set_c ("m_v2w", m_v2w );
RCache.set_c ("L_ambient", ambclr );
RCache.set_c ("Ldynamic_color", sunclr );
RCache.set_c ("Ldynamic_dir", sundir );
RCache.set_c ("env_color", envclr );
RCache.set_c ("fog_color", fogclr );
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// Forward rendering
{
u_setrt (rt_Generic_0,0,0,HW.pBaseZB); // LDR RT
RCache.set_CullMode (CULL_CCW);
RCache.set_Stencil (FALSE);
RCache.set_ColorWriteEnable ();
g_pGamePersistent->Environment().RenderClouds ();
RImplementation.render_forward ();
if (g_pGamePersistent) g_pGamePersistent->OnRenderPPUI_main() ; // PP-UI
}
// Perform blooming filter and distortion if needed
RCache.set_Stencil (FALSE);
phase_bloom ( ); // HDR RT invalidated here
// Distortion filter
BOOL bDistort = RImplementation.o.distortion_enabled; // This can be modified
{
if ((0==RImplementation.mapDistort.size()) && !_menu_pp) bDistort= FALSE;
if (bDistort) {
u_setrt (rt_Generic_1,0,0,HW.pBaseZB); // Now RT is a distortion mask
RCache.set_CullMode (CULL_CCW);
RCache.set_Stencil (FALSE);
RCache.set_ColorWriteEnable ();
CHK_DX(HW.pDevice->Clear ( 0L, NULL, D3DCLEAR_TARGET, color_rgba(127,127,0,127), 1.0f, 0L));
RImplementation.r_dsgraph_render_distort ();
if (g_pGamePersistent) g_pGamePersistent->OnRenderPPUI_PP() ; // PP-UI
}
}
// PP enabled ?
BOOL PP_Complex = u_need_PP ();
if (_menu_pp) PP_Complex = FALSE;
// Combine everything + perform AA
if (PP_Complex) u_setrt ( rt_Color,0,0,HW.pBaseZB ); // LDR RT
else u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
//. u_setrt ( Device.dwWidth,Device.dwHeight,HW.pBaseRT,NULL,NULL,HW.pBaseZB);
RCache.set_CullMode ( CULL_NONE ) ;
RCache.set_Stencil ( FALSE ) ;
if (1)
{
//
struct v_aa {
Fvector4 p;
Fvector2 uv0;
Fvector2 uv1;
Fvector2 uv2;
Fvector2 uv3;
Fvector2 uv4;
Fvector4 uv5;
Fvector4 uv6;
};
float _w = float(Device.dwWidth);
float _h = float(Device.dwHeight);
float ddw = 1.f/_w;
float ddh = 1.f/_h;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
// Fill vertex buffer
v_aa* pv = (v_aa*) RCache.Vertex.Lock (4,g_aa_AA->vb_stride,Offset);
pv->p.set(EPS, float(_h+EPS), EPS,1.f); pv->uv0.set(p0.x, p1.y);pv->uv1.set(p0.x-ddw,p1.y-ddh);pv->uv2.set(p0.x+ddw,p1.y+ddh);pv->uv3.set(p0.x+ddw,p1.y-ddh);pv->uv4.set(p0.x-ddw,p1.y+ddh);pv->uv5.set(p0.x-ddw,p1.y,p1.y,p0.x+ddw);pv->uv6.set(p0.x,p1.y-ddh,p1.y+ddh,p0.x);pv++;
pv->p.set(EPS, EPS, EPS,1.f); pv->uv0.set(p0.x, p0.y);pv->uv1.set(p0.x-ddw,p0.y-ddh);pv->uv2.set(p0.x+ddw,p0.y+ddh);pv->uv3.set(p0.x+ddw,p0.y-ddh);pv->uv4.set(p0.x-ddw,p0.y+ddh);pv->uv5.set(p0.x-ddw,p0.y,p0.y,p0.x+ddw);pv->uv6.set(p0.x,p0.y-ddh,p0.y+ddh,p0.x);pv++;
pv->p.set(float(_w+EPS),float(_h+EPS), EPS,1.f); pv->uv0.set(p1.x, p1.y);pv->uv1.set(p1.x-ddw,p1.y-ddh);pv->uv2.set(p1.x+ddw,p1.y+ddh);pv->uv3.set(p1.x+ddw,p1.y-ddh);pv->uv4.set(p1.x-ddw,p1.y+ddh);pv->uv5.set(p1.x-ddw,p1.y,p1.y,p1.x+ddw);pv->uv6.set(p1.x,p1.y-ddh,p1.y+ddh,p1.x);pv++;
pv->p.set(float(_w+EPS),EPS, EPS,1.f); pv->uv0.set(p1.x, p0.y);pv->uv1.set(p1.x-ddw,p0.y-ddh);pv->uv2.set(p1.x+ddw,p0.y+ddh);pv->uv3.set(p1.x+ddw,p0.y-ddh);pv->uv4.set(p1.x-ddw,p0.y+ddh);pv->uv5.set(p1.x-ddw,p0.y,p0.y,p1.x+ddw);pv->uv6.set(p1.x,p0.y-ddh,p0.y+ddh,p1.x);pv++;
RCache.Vertex.Unlock (4,g_aa_AA->vb_stride);
// Draw COLOR
if (ps_r2_ls_flags.test(R2FLAG_AA)) RCache.set_Element (s_combine->E[bDistort?3:1]); // look at blender_combine.cpp
else RCache.set_Element (s_combine->E[bDistort?4:2]); // look at blender_combine.cpp
RCache.set_c ("e_barrier", ps_r2_aa_barier.x, ps_r2_aa_barier.y, ps_r2_aa_barier.z, 0);
RCache.set_c ("e_weights", ps_r2_aa_weight.x, ps_r2_aa_weight.y, ps_r2_aa_weight.z, 0);
RCache.set_c ("e_kernel", ps_r2_aa_kernel, ps_r2_aa_kernel, ps_r2_aa_kernel, 0);
RCache.set_c ("m_current", m_current);
RCache.set_c ("m_previous", m_previous);
RCache.set_c ("m_blur", m_blur_scale.x,m_blur_scale.y, 0,0);
RCache.set_Geometry (g_aa_AA);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
RCache.set_Stencil (FALSE);
// if FP16-BLEND !not! supported - draw flares here, overwise they are already in the bloom target
/* if (!RImplementation.o.fp16_blend)*/ g_pGamePersistent->Environment().RenderFlares (); // lens-flares
// PP-if required
if (PP_Complex) {
phase_pp ();
}
// Re-adapt luminance
RCache.set_Stencil (FALSE);
//*** exposure-pipeline-clear
{
std::swap (rt_LUM_pool[gpu_id*2+0],rt_LUM_pool[gpu_id*2+1]);
t_LUM_src->surface_set (NULL);
t_LUM_dest->surface_set (NULL);
}
#ifdef DEBUG
RCache.set_CullMode ( CULL_CCW );
static xr_vector<Fplane> saved_dbg_planes;
if (bDebug) saved_dbg_planes= dbg_planes;
else dbg_planes = saved_dbg_planes;
if (1) for (u32 it=0; it<dbg_planes.size(); it++)
{
Fplane& P = dbg_planes[it];
Fvector zero ;
zero.mul (P.n,P.d);
Fvector L_dir,L_up=P.n,L_right;
L_dir.set (0,0,1); if (_abs(L_up.dotproduct(L_dir))>.99f) L_dir.set(1,0,0);
L_right.crossproduct(L_up,L_dir); L_right.normalize ();
L_dir.crossproduct (L_right,L_up); L_dir.normalize ();
Fvector p0,p1,p2,p3;
float sz = 100.f;
p0.mad (zero,L_right,sz).mad (L_dir,sz);
p1.mad (zero,L_right,sz).mad (L_dir,-sz);
p2.mad (zero,L_right,-sz).mad (L_dir,-sz);
p3.mad (zero,L_right,-sz).mad (L_dir,+sz);
RCache.dbg_DrawTRI (Fidentity,p0,p1,p2,0xffffffff);
RCache.dbg_DrawTRI (Fidentity,p2,p3,p0,0xffffffff);
}
static xr_vector<dbg_line_t> saved_dbg_lines;
if (bDebug) saved_dbg_lines = dbg_lines;
else dbg_lines = saved_dbg_lines;
if (1) for (u32 it=0; it<dbg_lines.size(); it++)
{
RCache.dbg_DrawLINE (Fidentity,dbg_lines[it].P0,dbg_lines[it].P1,dbg_lines[it].color);
}
#endif
// ********************* Debug
/*
if (0) {
u32 C = color_rgba (255,255,255,255);
float _w = float(Device.dwWidth)/3;
float _h = float(Device.dwHeight)/3;
// draw light-spheres
#ifdef DEBUG
if (0) for (u32 it=0; it<dbg_spheres.size(); it++)
{
Fsphere S = dbg_spheres[it].first;
Fmatrix M;
u32 ccc = dbg_spheres[it].second.get();
M.scale (S.R,S.R,S.R);
M.translate_over (S.P);
RCache.dbg_DrawEllipse (M,ccc);
RCache.dbg_DrawAABB (S.P,.05f,.05f,.05f,ccc);
}
#endif
// Draw quater-screen quad textured with our direct-shadow-map-image
if (1)
{
u32 IX=0,IY=1;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set ((IX+0)*_w+EPS, (IY+1)*_h+EPS, EPS, 1.f, C, p0.x, p1.y); pv++;
pv->set ((IX+0)*_w+EPS, (IY+0)*_h+EPS, EPS, 1.f, C, p0.x, p0.y); pv++;
pv->set ((IX+1)*_w+EPS, (IY+1)*_h+EPS, EPS, 1.f, C, p1.x, p1.y); pv++;
pv->set ((IX+1)*_w+EPS, (IY+0)*_h+EPS, EPS, 1.f, C, p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
// Draw COLOR
RCache.set_Shader (s_combine_dbg_0);
RCache.set_Geometry (g_combine);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
// Draw quater-screen quad textured with our accumulator
if (0)
{
u32 IX=1,IY=1;
p0.set (.5f/_w, .5f/_h);
p1.set ((_w+.5f)/_w, (_h+.5f)/_h );
// Fill vertex buffer
FVF::TL* pv = (FVF::TL*) RCache.Vertex.Lock (4,g_combine->vb_stride,Offset);
pv->set ((IX+0)*_w+EPS, (IY+1)*_h+EPS, EPS, 1.f, C, p0.x, p1.y); pv++;
pv->set ((IX+0)*_w+EPS, (IY+0)*_h+EPS, EPS, 1.f, C, p0.x, p0.y); pv++;
pv->set ((IX+1)*_w+EPS, (IY+1)*_h+EPS, EPS, 1.f, C, p1.x, p1.y); pv++;
pv->set ((IX+1)*_w+EPS, (IY+0)*_h+EPS, EPS, 1.f, C, p1.x, p0.y); pv++;
RCache.Vertex.Unlock (4,g_combine->vb_stride);
// Draw COLOR
RCache.set_Shader (s_combine_dbg_1);
RCache.set_Geometry (g_combine);
RCache.Render (D3DPT_TRIANGLELIST,Offset,0,4,0,2);
}
}
*/
#ifdef DEBUG
dbg_spheres.clear ();
dbg_lines.clear ();
dbg_planes.clear ();
#endif
}
