void GSDevice10::StretchRect(Texture& st, const GSVector4& sr, Texture& dt, const GSVector4& dr, ID3D10PixelShader* ps, ID3D10Buffer* ps_cb, ID3D10BlendState* bs, bool linear)
{
BeginScene();
// om
OMSetDepthStencilState(m_convert.dss, 0);
OMSetBlendState(bs, 0);
OMSetRenderTargets(dt, NULL);
// ia
float left = dr.x * 2 / dt.GetWidth() - 1.0f;
float top = 1.0f - dr.y * 2 / dt.GetHeight();
float right = dr.z * 2 / dt.GetWidth() - 1.0f;
float bottom = 1.0f - dr.w * 2 / dt.GetHeight();
GSVertexPT1 vertices[] =
{
{GSVector4(left, top, 0.5f, 1.0f), GSVector2(sr.x, sr.y)},
{GSVector4(right, top, 0.5f, 1.0f), GSVector2(sr.z, sr.y)},
{GSVector4(left, bottom, 0.5f, 1.0f), GSVector2(sr.x, sr.w)},
{GSVector4(right, bottom, 0.5f, 1.0f), GSVector2(sr.z, sr.w)},
};
D3D10_BOX box = {0, 0, 0, sizeof(vertices), 1, 1};
m_dev->UpdateSubresource(m_convert.vb, 0, &box, vertices, 0, 0);
IASetVertexBuffer(m_convert.vb, sizeof(vertices[0]));
IASetInputLayout(m_convert.il);
IASetPrimitiveTopology(D3D10_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP);
// vs
VSSetShader(m_convert.vs, NULL);
// gs
GSSetShader(NULL);
// ps
PSSetShader(ps, ps_cb);
PSSetSamplerState(linear ? m_convert.ln : m_convert.pt, NULL);
PSSetShaderResources(st, NULL);
// rs
RSSet(dt.GetWidth(), dt.GetHeight());
//
DrawPrimitive(countof(vertices));
//
EndScene();
}
void GSDevice11::StretchRect(GSTexture* st, const GSVector4& sr, GSTexture* dt, const GSVector4& dr, ID3D11PixelShader* ps, ID3D11Buffer* ps_cb, ID3D11BlendState* bs, bool linear)
{
BeginScene();
GSVector2i ds = dt->GetSize();
// om
OMSetDepthStencilState(m_convert.dss, 0);
OMSetBlendState(bs, 0);
OMSetRenderTargets(dt, NULL);
// ia
float left = dr.x * 2 / ds.x - 1.0f;
float top = 1.0f - dr.y * 2 / ds.y;
float right = dr.z * 2 / ds.x - 1.0f;
float bottom = 1.0f - dr.w * 2 / ds.y;
GSVertexPT1 vertices[] =
{
{GSVector4(left, top, 0.5f, 1.0f), GSVector2(sr.x, sr.y)},
{GSVector4(right, top, 0.5f, 1.0f), GSVector2(sr.z, sr.y)},
{GSVector4(left, bottom, 0.5f, 1.0f), GSVector2(sr.x, sr.w)},
{GSVector4(right, bottom, 0.5f, 1.0f), GSVector2(sr.z, sr.w)},
};
IASetVertexBuffer(vertices, sizeof(vertices[0]), countof(vertices));
IASetInputLayout(m_convert.il);
IASetPrimitiveTopology(D3D11_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP);
// vs
VSSetShader(m_convert.vs, NULL);
// gs
GSSetShader(NULL);
// ps
PSSetShaderResources(st, NULL);
PSSetSamplerState(linear ? m_convert.ln : m_convert.pt, NULL);
PSSetShader(ps, ps_cb);
//
DrawPrimitive();
//
EndScene();
PSSetShaderResources(NULL, NULL);
}
void GSRendererDX9::UpdateFBA(GSTexture* rt)
{
if (!rt)
return;
GSDevice9* dev = (GSDevice9*)m_dev;
dev->BeginScene();
// om
dev->OMSetDepthStencilState(&m_fba.dss);
dev->OMSetBlendState(&m_fba.bs, 0);
// ia
GSVector4 s = GSVector4(rt->GetScale().x / rt->GetWidth(), rt->GetScale().y / rt->GetHeight());
GSVector4 off = GSVector4(-1.0f, 1.0f);
GSVector4 src = ((m_vt.m_min.p.xyxy(m_vt.m_max.p) + off.xxyy()) * s.xyxy()).sat(off.zzyy());
GSVector4 dst = src * 2.0f + off.xxxx();
GSVertexPT1 vertices[] =
{
{GSVector4(dst.x, -dst.y, 0.5f, 1.0f), GSVector2(0)},
{GSVector4(dst.z, -dst.y, 0.5f, 1.0f), GSVector2(0)},
{GSVector4(dst.x, -dst.w, 0.5f, 1.0f), GSVector2(0)},
{GSVector4(dst.z, -dst.w, 0.5f, 1.0f), GSVector2(0)},
};
dev->IASetVertexBuffer(vertices, sizeof(vertices[0]), countof(vertices));
dev->IASetInputLayout(dev->m_convert.il);
dev->IASetPrimitiveTopology(D3DPT_TRIANGLESTRIP);
// vs
dev->VSSetShader(dev->m_convert.vs, NULL, 0);
// ps
dev->PSSetShader(dev->m_convert.ps[4], NULL, 0);
//
dev->DrawPrimitive();
//
dev->EndScene();
}
void GSDevice10::DoInterlace(Texture& st, Texture& dt, int shader, bool linear, float yoffset)
{
GSVector4 sr(0, 0, 1, 1);
GSVector4 dr(0.0f, yoffset, (float)dt.GetWidth(), (float)dt.GetHeight() + yoffset);
InterlaceConstantBuffer cb;
cb.ZrH = GSVector2(0, 1.0f / dt.GetHeight());
cb.hH = (float)dt.GetHeight() / 2;
m_dev->UpdateSubresource(m_interlace.cb, 0, NULL, &cb, 0, 0);
StretchRect(st, sr, dt, dr, m_interlace.ps[shader], m_interlace.cb, linear);
}
void GSDevice11::DoInterlace(GSTexture* sTex, GSTexture* dTex, int shader, bool linear, float yoffset)
{
GSVector4 s = GSVector4(dTex->GetSize());
GSVector4 sRect(0, 0, 1, 1);
GSVector4 dRect(0.0f, yoffset, s.x, s.y + yoffset);
InterlaceConstantBuffer cb;
cb.ZrH = GSVector2(0, 1.0f / s.y);
cb.hH = s.y / 2;
m_ctx->UpdateSubresource(m_interlace.cb, 0, NULL, &cb, 0, 0);
StretchRect(sTex, sRect, dTex, dRect, m_interlace.ps[shader], m_interlace.cb, linear);
}
void GSDevice11::DoExternalFX(GSTexture* sTex, GSTexture* dTex)
{
GSVector2i s = dTex->GetSize();
GSVector4 sRect(0, 0, 1, 1);
GSVector4 dRect(0, 0, s.x, s.y);
ExternalFXConstantBuffer cb;
InitExternalFX();
cb.xyFrame = GSVector2((float)s.x, (float)s.y);
cb.rcpFrame = GSVector4(1.0f / (float)s.x, 1.0f / (float)s.y, 0.0f, 0.0f);
cb.rcpFrameOpt = GSVector4::zero();
m_ctx->UpdateSubresource(m_shaderfx.cb, 0, NULL, &cb, 0, 0);
StretchRect(sTex, sRect, dTex, dRect, m_shaderfx.ps, m_shaderfx.cb, true);
}
void GSRendererDX::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Source* tex)
{
GSDrawingEnvironment& env = m_env;
GSDrawingContext* context = m_context;
const GSVector2i& rtsize = rt->GetSize();
const GSVector2& rtscale = rt->GetScale();
bool DATE = m_context->TEST.DATE && context->FRAME.PSM != PSM_PSMCT24;
GSTexture* rtcopy = NULL;
ASSERT(m_dev != NULL);
GSDeviceDX* dev = (GSDeviceDX*)m_dev;
if(DATE)
{
if(dev->HasStencil())
{
GSVector4 s = GSVector4(rtscale.x / rtsize.x, rtscale.y / rtsize.y);
GSVector4 o = GSVector4(-1.0f, 1.0f);
GSVector4 src = ((m_vt.m_min.p.xyxy(m_vt.m_max.p) + o.xxyy()) * s.xyxy()).sat(o.zzyy());
GSVector4 dst = src * 2.0f + o.xxxx();
GSVertexPT1 vertices[] =
{
{GSVector4(dst.x, -dst.y, 0.5f, 1.0f), GSVector2(src.x, src.y)},
{GSVector4(dst.z, -dst.y, 0.5f, 1.0f), GSVector2(src.z, src.y)},
{GSVector4(dst.x, -dst.w, 0.5f, 1.0f), GSVector2(src.x, src.w)},
{GSVector4(dst.z, -dst.w, 0.5f, 1.0f), GSVector2(src.z, src.w)},
};
dev->SetupDATE(rt, ds, vertices, m_context->TEST.DATM);
}
else
{
rtcopy = dev->CreateRenderTarget(rtsize.x, rtsize.y, false, rt->GetFormat());
// I'll use VertexTrace when I consider it more trustworthy
dev->CopyRect(rt, rtcopy, GSVector4i(rtsize).zwxy());
}
}
//
dev->BeginScene();
// om
GSDeviceDX::OMDepthStencilSelector om_dssel;
if(context->TEST.ZTE)
{
om_dssel.ztst = context->TEST.ZTST;
om_dssel.zwe = !context->ZBUF.ZMSK;
}
else
{
om_dssel.ztst = ZTST_ALWAYS;
}
if(m_fba)
{
om_dssel.fba = context->FBA.FBA;
}
GSDeviceDX::OMBlendSelector om_bsel;
if(!IsOpaque())
{
om_bsel.abe = PRIM->ABE || PRIM->AA1 && m_vt.m_primclass == GS_LINE_CLASS;
om_bsel.a = context->ALPHA.A;
om_bsel.b = context->ALPHA.B;
om_bsel.c = context->ALPHA.C;
om_bsel.d = context->ALPHA.D;
if(env.PABE.PABE)
{
if(om_bsel.a == 0 && om_bsel.b == 1 && om_bsel.c == 0 && om_bsel.d == 1)
{
// this works because with PABE alpha blending is on when alpha >= 0x80, but since the pixel shader
// cannot output anything over 0x80 (== 1.0) blending with 0x80 or turning it off gives the same result
om_bsel.abe = 0;
}
else
{
//Breath of Fire Dragon Quarter triggers this in battles. Graphics are fine though.
//ASSERT(0);
}
}
}
om_bsel.wrgba = ~GSVector4i::load((int)context->FRAME.FBMSK).eq8(GSVector4i::xffffffff()).mask();
// vs
GSDeviceDX::VSSelector vs_sel;
vs_sel.tme = PRIM->TME;
vs_sel.fst = PRIM->FST;
vs_sel.logz = dev->HasDepth32() ? 0 : m_logz ? 1 : 0;
vs_sel.rtcopy = !!rtcopy;
// The real GS appears to do no masking based on the Z buffer format and writing larger Z values
// than the buffer supports seems to be an error condition on the real GS, causing it to crash.
// We are probably receiving bad coordinates from VU1 in these cases.
if(om_dssel.ztst >= ZTST_ALWAYS && om_dssel.zwe)
{
if(context->ZBUF.PSM == PSM_PSMZ24)
{
if(m_vt.m_max.p.z > 0xffffff)
{
ASSERT(m_vt.m_min.p.z > 0xffffff);
// Fixme :Following conditional fixes some dialog frame in Wild Arms 3, but may not be what was intended.
if (m_vt.m_min.p.z > 0xffffff)
{
vs_sel.bppz = 1;
om_dssel.ztst = ZTST_ALWAYS;
}
}
}
else if(context->ZBUF.PSM == PSM_PSMZ16 || context->ZBUF.PSM == PSM_PSMZ16S)
{
if(m_vt.m_max.p.z > 0xffff)
{
ASSERT(m_vt.m_min.p.z > 0xffff); // sfex capcom logo
// Fixme : Same as above, I guess.
if (m_vt.m_min.p.z > 0xffff)
{
vs_sel.bppz = 2;
om_dssel.ztst = ZTST_ALWAYS;
}
}
}
}
GSDeviceDX::VSConstantBuffer vs_cb;
float sx = 2.0f * rtscale.x / (rtsize.x << 4);
float sy = 2.0f * rtscale.y / (rtsize.y << 4);
float ox = (float)(int)context->XYOFFSET.OFX;
float oy = (float)(int)context->XYOFFSET.OFY;
float ox2 = 2.0f * m_pixelcenter.x / rtsize.x;
float oy2 = 2.0f * m_pixelcenter.y / rtsize.y;
//This hack subtracts around half a pixel from OFX and OFY. (Cannot do this directly,
//because DX10 and DX9 have a different pixel center.)
//
//The resulting shifted output aligns better with common blending / corona / blurring effects,
//but introduces a few bad pixels on the edges.
if(rt->LikelyOffset)
{
// DX9 has pixelcenter set to 0.0, so give it some value here
if(m_pixelcenter.x == 0 && m_pixelcenter.y == 0) { ox2 = -0.0003f; oy2 = -0.0003f; }
ox2 *= rt->OffsetHack_modx;
oy2 *= rt->OffsetHack_mody;
}
vs_cb.VertexScale = GSVector4(sx, -sy, ldexpf(1, -32), 0.0f);
vs_cb.VertexOffset = GSVector4(ox * sx + ox2 + 1, -(oy * sy + oy2 + 1), 0.0f, -1.0f);
// gs
GSDeviceDX::GSSelector gs_sel;
gs_sel.iip = PRIM->IIP;
gs_sel.prim = m_vt.m_primclass;
// ps
GSDeviceDX::PSSelector ps_sel;
GSDeviceDX::PSSamplerSelector ps_ssel;
GSDeviceDX::PSConstantBuffer ps_cb;
if(DATE)
{
if(dev->HasStencil())
{
om_dssel.date = 1;
}
else
{
ps_sel.date = 1 + context->TEST.DATM;
}
}
if(env.COLCLAMP.CLAMP == 0 && /* hack */ !tex && PRIM->PRIM != GS_POINTLIST)
{
ps_sel.colclip = 1;
}
ps_sel.clr1 = om_bsel.IsCLR1();
ps_sel.fba = context->FBA.FBA;
ps_sel.aout = context->FRAME.PSM == PSM_PSMCT16 || context->FRAME.PSM == PSM_PSMCT16S || (context->FRAME.FBMSK & 0xff000000) == 0x7f000000 ? 1 : 0;
if(UserHacks_AlphaHack) ps_sel.aout = 1;
if(PRIM->FGE)
{
ps_sel.fog = 1;
ps_cb.FogColor_AREF = GSVector4::rgba32(env.FOGCOL.u32[0]) / 255;
}
if(context->TEST.ATE)
{
ps_sel.atst = context->TEST.ATST;
switch(ps_sel.atst)
{
case ATST_LESS:
ps_cb.FogColor_AREF.a = (float)((int)context->TEST.AREF - 1);
break;
case ATST_GREATER:
ps_cb.FogColor_AREF.a = (float)((int)context->TEST.AREF + 1);
break;
default:
ps_cb.FogColor_AREF.a = (float)(int)context->TEST.AREF;
break;
}
}
else
{
ps_sel.atst = ATST_ALWAYS;
}
if(tex)
{
ps_sel.wms = context->CLAMP.WMS;
ps_sel.wmt = context->CLAMP.WMT;
ps_sel.fmt = tex->m_fmt;
ps_sel.aem = env.TEXA.AEM;
ps_sel.tfx = context->TEX0.TFX;
ps_sel.tcc = context->TEX0.TCC;
ps_sel.ltf = m_filter == 2 ? m_vt.IsLinear() : m_filter;
ps_sel.rt = tex->m_target;
int w = tex->m_texture->GetWidth();
int h = tex->m_texture->GetHeight();
int tw = (int)(1 << context->TEX0.TW);
int th = (int)(1 << context->TEX0.TH);
GSVector4 WH(tw, th, w, h);
if(PRIM->FST)
{
vs_cb.TextureScale = GSVector4(1.0f / 16) / WH.xyxy();
//Maybe better?
//vs_cb.TextureScale = GSVector4(1.0f / 16) * GSVector4(tex->m_texture->GetScale()).xyxy() / WH.zwzw();
ps_sel.fst = 1;
}
ps_cb.WH = WH;
ps_cb.HalfTexel = GSVector4(-0.5f, 0.5f).xxyy() / WH.zwzw();
ps_cb.MskFix = GSVector4i(context->CLAMP.MINU, context->CLAMP.MINV, context->CLAMP.MAXU, context->CLAMP.MAXV);
GSVector4 clamp(ps_cb.MskFix);
GSVector4 ta(env.TEXA & GSVector4i::x000000ff());
ps_cb.MinMax = clamp / WH.xyxy();
ps_cb.MinF_TA = (clamp + 0.5f).xyxy(ta) / WH.xyxy(GSVector4(255, 255));
ps_ssel.tau = (context->CLAMP.WMS + 3) >> 1;
ps_ssel.tav = (context->CLAMP.WMT + 3) >> 1;
ps_ssel.ltf = ps_sel.ltf;
}
else
{
void GSDevice11::StretchRect(GSTexture* sTex, const GSVector4& sRect, GSTexture* dTex, const GSVector4& dRect, ID3D11PixelShader* ps, ID3D11Buffer* ps_cb, ID3D11BlendState* bs, bool linear)
{
if(!sTex || !dTex)
{
ASSERT(0);
return;
}
BeginScene();
GSVector2i ds = dTex->GetSize();
// om
OMSetDepthStencilState(m_convert.dss, 0);
OMSetBlendState(bs, 0);
OMSetRenderTargets(dTex, NULL);
// ia
float left = dRect.x * 2 / ds.x - 1.0f;
float top = 1.0f - dRect.y * 2 / ds.y;
float right = dRect.z * 2 / ds.x - 1.0f;
float bottom = 1.0f - dRect.w * 2 / ds.y;
GSVertexPT1 vertices[] =
{
{GSVector4(left, top, 0.5f, 1.0f), GSVector2(sRect.x, sRect.y)},
{GSVector4(right, top, 0.5f, 1.0f), GSVector2(sRect.z, sRect.y)},
{GSVector4(left, bottom, 0.5f, 1.0f), GSVector2(sRect.x, sRect.w)},
{GSVector4(right, bottom, 0.5f, 1.0f), GSVector2(sRect.z, sRect.w)},
};
IASetVertexBuffer(vertices, sizeof(vertices[0]), countof(vertices));
IASetInputLayout(m_convert.il);
IASetPrimitiveTopology(D3D11_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP);
// vs
VSSetShader(m_convert.vs, NULL);
// gs
/* NVIDIA HACK!!!!
Not sure why, but having the Geometry shader disabled causes the strange stretching in recent drivers*/
GSSelector sel;
//Don't use shading for stretching, we're just passing through - Note: With Win10 it seems to cause other bugs when shading is off if any of the coords is greater than 0
//I really don't know whats going on there, but this seems to resolve it mostly (if not all, not tester a lot of games, only BIOS, FFXII and VP2)
//sel.iip = (sRect.y > 0.0f || sRect.w > 0.0f) ? 1 : 0;
//sel.prim = 2; //Triangle Strip
//SetupGS(sel);
GSSetShader(NULL, NULL);
/*END OF HACK*/
//
// ps
PSSetShaderResources(sTex, NULL);
PSSetSamplerState(linear ? m_convert.ln : m_convert.pt, NULL);
PSSetShader(ps, ps_cb);
//
DrawPrimitive();
//
EndScene();
PSSetShaderResources(NULL, NULL);
}