void GPULocalMemory::InitVectors()
{
m_xxxa = GSVector4i(0x00008000);
m_xxbx = GSVector4i(0x00007c00);
m_xgxx = GSVector4i(0x000003e0);
m_rxxx = GSVector4i(0x0000001f);
}
void GSDevice11::OMSetRenderTargets(const GSVector2i& rtsize, int count, ID3D11UnorderedAccessView** uav, uint32* counters, const GSVector4i* scissor)
{
m_ctx->OMSetRenderTargetsAndUnorderedAccessViews(0, NULL, NULL, 0, count, uav, counters);
m_state.rtv = NULL;
m_state.dsv = NULL;
if(m_state.viewport != rtsize)
{
m_state.viewport = rtsize;
D3D11_VIEWPORT vp;
memset(&vp, 0, sizeof(vp));
vp.TopLeftX = 0;
vp.TopLeftY = 0;
vp.Width = (float)rtsize.x;
vp.Height = (float)rtsize.y;
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
m_ctx->RSSetViewports(1, &vp);
}
GSVector4i r = scissor ? *scissor : GSVector4i(rtsize).zwxy();
if(!m_state.scissor.eq(r))
{
m_state.scissor = r;
m_ctx->RSSetScissorRects(1, r);
}
}
GSTextureCacheSW::Texture::Texture(GSState* state, uint32 tw0, const GIFRegTEX0& TEX0, const GIFRegTEXA& TEXA)
: m_state(state)
, m_buff(NULL)
, m_tw(tw0)
, m_age(0)
, m_complete(false)
, m_p2t(NULL)
{
m_TEX0 = TEX0;
m_TEXA = TEXA;
if(m_tw == 0)
{
m_tw = std::max<int>(m_TEX0.TW, GSLocalMemory::m_psm[m_TEX0.PSM].pal == 0 ? 3 : 5); // makes one row 32 bytes at least, matches the smallest block size that is allocated for m_buff
}
memset(m_valid, 0, sizeof(m_valid));
m_sharedbits = GSUtil::HasSharedBitsPtr(m_TEX0.PSM);
m_offset = m_state->m_mem.GetOffset(TEX0.TBP0, TEX0.TBW, TEX0.PSM);
m_pages.n = m_offset->GetPages(GSVector4i(0, 0, 1 << TEX0.TW, 1 << TEX0.TH));
memcpy(m_pages.bm, m_offset->GetPagesAsBits(TEX0), sizeof(m_pages.bm));
m_repeating = m_TEX0.IsRepeating(); // repeating mode always works, it is just slightly slower
if(m_repeating)
{
m_p2t = m_state->m_mem.GetPage2TileMap(m_TEX0);
}
}
bool GSTextureOGL::Map(GSMap& m, const GSVector4i* _r)
{
GSVector4i r = _r ? *_r : GSVector4i(0, 0, m_size.x, m_size.y);
// LOTS OF CRAP CODE!!!! PLEASE FIX ME !!!
if (m_type == GSTexture::Offscreen) {
// The fastest way will be to use a PBO to read the data asynchronously. Unfortunately GSdx
// architecture is waiting the data right now.
#if 0
// Maybe it is as good as the code below. I don't know
// With openGL 4.5 you can use glGetTextureSubImage
glGetTextureImage(m_texture_id, GL_TEX_LEVEL_0, m_int_format, m_int_type, 1024*1024*16, m_local_buffer);
#else
// Bind the texture to the read framebuffer to avoid any disturbance
glBindFramebuffer(GL_READ_FRAMEBUFFER, m_fbo_read);
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, m_texture_id, 0);
glPixelStorei(GL_PACK_ALIGNMENT, m_int_alignment);
glReadPixels(r.x, r.y, r.width(), r.height(), m_int_format, m_int_type, m_local_buffer);
glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
#endif
m.bits = m_local_buffer;
m.pitch = m_size.x << m_int_shift;
return true;
} else if (m_type == GSTexture::Texture || m_type == GSTexture::RenderTarget) {
GL_PUSH("Upload Texture %d", m_texture_id); // POP is in Unmap
m_dirty = true;
m_clean = false;
uint32 row_byte = r.width() << m_int_shift;
uint32 map_size = r.height() * row_byte;
m.bits = (uint8*)PboPool::Map(map_size);
m.pitch = row_byte;
#ifdef ENABLE_OGL_DEBUG_MEM_BW
g_real_texture_upload_byte += map_size;
#endif
// Save the area for the unmap
m_r_x = r.x;
m_r_y = r.y;
m_r_w = r.width();
m_r_h = r.height();
return true;
}
return false;
}
void GPUDrawScanline::BeginDraw(const GSRasterizerData* data)
{
memcpy(&m_global, &((const SharedData*)data)->global, sizeof(m_global));
if(m_global.sel.tme && m_global.sel.twin)
{
uint32 u, v;
u = ~(m_global.twin.x << 3) & 0xff; // TWW
v = ~(m_global.twin.y << 3) & 0xff; // TWH
m_local.twin[0].u = GSVector4i((u << 16) | u);
m_local.twin[0].v = GSVector4i((v << 16) | v);
u = m_global.twin.z << 3; // TWX
v = m_global.twin.w << 3; // TWY
m_local.twin[1].u = GSVector4i((u << 16) | u) & ~m_local.twin[0].u;
m_local.twin[1].v = GSVector4i((v << 16) | v) & ~m_local.twin[0].v;
}
m_ds = m_ds_map[m_global.sel];
m_de = NULL;
m_dr = NULL; // TODO
// doesn't need all bits => less functions generated
GPUScanlineSelector sel;
sel.key = 0;
sel.iip = m_global.sel.iip;
sel.tfx = m_global.sel.tfx;
sel.twin = m_global.sel.twin;
sel.sprite = m_global.sel.sprite;
m_sp = m_sp_map[sel];
}
void GSRasterizer::DrawTriangleBottom(GSVertexSW* v, const GSVector4i& scissor)
{
GSVertexSW longest;
longest.p = v[1].p - v[0].p;
int i = longest.p.upl(longest.p == GSVector4::zero()).mask();
if(i & 2) return;
i &= 1;
GSVertexSW& l = v[i];
GSVector4& r = v[1 - i].p;
GSVector4 fscissor(scissor);
GSVector4 tb = l.p.upl(v[2].p).ceil();
GSVector4 tbmax = tb.max(fscissor.yyyy());
GSVector4 tbmin = tb.min(fscissor.wwww());
GSVector4i tbi = GSVector4i(tbmax.zzww(tbmin));
int top = tbi.extract32<0>();
int bottom = tbi.extract32<2>();
if(top >= bottom) return;
longest.t = v[1].t - v[0].t;
longest.c = v[1].c - v[0].c;
GSVertexSW dscan = longest * longest.p.xxxx().rcp();
GSVertexSW vl = v[2] - l;
GSVector4 vr = v[2].p - r;
GSVertexSW dl = vl / vl.p.yyyy();
GSVector4 dr = vr / vr.yyyy();
GSVector4 dy = tbmax.zzzz() - l.p.yyyy();
l.p = l.p.upl(r).xyzw(l.p); // r.x => l.y
dl.p = dl.p.upl(dr).xyzw(dl.p); // dr.x => dl.y
l += dl * dy;
m_dsf.ssp(v, dscan);
DrawTriangleSection(top, bottom, l, dl, dscan, fscissor);
}
// Note: hack is safe, but it could impact the perf a little (normally games do only a couple of clear by frame)
void GSRendererHW::OI_GsMemClear()
{
// Rectangle draw without texture
if ((m_vt.m_primclass == GS_SPRITE_CLASS) && (m_vertex.next == 2) && !PRIM->TME && !PRIM->ABE) {
// 0 clear
if (m_vt.m_eq.rgba == 0xFFFF && m_vt.m_min.c.eq(GSVector4i(0))) {
GL_INS("OI_GsMemClear");
GSOffset* off = m_context->offset.fb;
GSVector4i r = GSVector4i(m_vt.m_min.p.xyxy(m_vt.m_max.p)).rintersect(GSVector4i(m_context->scissor.in));
// Based on WritePixel32
for(int y = r.top; y < r.bottom; y++)
{
uint32* RESTRICT d = &m_mem.m_vm32[off->pixel.row[y]];
int* RESTRICT col = off->pixel.col[0];
for(int x = r.left; x < r.right; x++)
{
d[col[x]] = 0; // Here the constant color
}
}
}
}
void GSDevice11::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVector4i* scissor)
{
ID3D11RenderTargetView* rtv = NULL;
ID3D11DepthStencilView* dsv = NULL;
if (!rt && !ds)
throw GSDXRecoverableError();
if(rt) rtv = *(GSTexture11*)rt;
if(ds) dsv = *(GSTexture11*)ds;
if(m_state.rtv != rtv || m_state.dsv != dsv)
{
m_state.rtv = rtv;
m_state.dsv = dsv;
m_ctx->OMSetRenderTargets(1, &rtv, dsv);
}
GSVector2i size = rt ? rt->GetSize() : ds->GetSize();
if(m_state.viewport != size)
{
bool isNative = theApp.GetConfig("upscale_multiplier", 1) == 1;
m_state.viewport = size;
D3D11_VIEWPORT vp;
memset(&vp, 0, sizeof(vp));
vp.TopLeftX = (spritehack > 0 || isNative) ? 0.0f : -0.01f;
vp.TopLeftY = (spritehack > 0 || isNative) ? 0.0f : -0.01f;
vp.Width = (float)size.x;
vp.Height = (float)size.y;
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
m_ctx->RSSetViewports(1, &vp);
}
GSVector4i r = scissor ? *scissor : GSVector4i(size).zwxy();
if(!m_state.scissor.eq(r))
{
m_state.scissor = r;
m_ctx->RSSetScissorRects(1, r);
}
}
GSVector4i GSWndOGL::GetClientRect()
{
unsigned int h = 480;
unsigned int w = 640;
unsigned int borderDummy;
unsigned int depthDummy;
Window winDummy;
int xDummy;
int yDummy;
if (!m_NativeDisplay) m_NativeDisplay = XOpenDisplay(NULL);
XGetGeometry(m_NativeDisplay, m_NativeWindow, &winDummy, &xDummy, &yDummy, &w, &h, &borderDummy, &depthDummy);
return GSVector4i(0, 0, (int)w, (int)h);
}
void Clear() {
fbo = 0;
viewport = GSVector2i(0, 0);
scissor = GSVector4i(0, 0, 0, 0);
blend = false;
eq_RGB = 0;
eq_A = 0;
f_sRGB = 0;
f_dRGB = 0;
f_sA = 0;
f_dA = 0;
r_msk = true;
g_msk = true;
b_msk = true;
a_msk = true;
bf = 0.0;
depth = false;
depth_func = 0;
depth_mask = false;
stencil = false;
stencil_func = 0;
stencil_pass = 0;
ubo = 0;
ps_ss = 0;
rt = 0;
ds = 0;
tex_unit[0] = 0;
tex_unit[1] = 0;
tex = 0;
tex_handle[0] = 0;
tex_handle[1] = 0;
ps = 0;
gs = 0;
vs = 0;
program = 0;
dirty_prog = false;
dirty_subroutine_vs = false;
dirty_subroutine_ps = false;
dirty_ressources = false;
}
bool GSTextureSW::Map(GSMap& m, const GSVector4i* r)
{
GSVector4i r2 = r != NULL ? *r : GSVector4i(0, 0, m_size.x, m_size.y);
if(m_data != NULL && r2.left >= 0 && r2.right <= m_size.x && r2.top >= 0 && r2.bottom <= m_size.y)
{
if (!m_mapped.test_and_set(std::memory_order_acquire))
{
m.bits = (uint8*)m_data + m_pitch * r2.top + (r2.left << 2);
m.pitch = m_pitch;
return true;
}
}
return false;
}
void GSDevice11::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVector4i* scissor)
{
ID3D11RenderTargetView* rtv = NULL;
ID3D11DepthStencilView* dsv = NULL;
if(rt) rtv = *(GSTexture11*)rt;
if(ds) dsv = *(GSTexture11*)ds;
if(m_state.rtv != rtv || m_state.dsv != dsv)
{
m_state.rtv = rtv;
m_state.dsv = dsv;
m_ctx->OMSetRenderTargets(1, &rtv, dsv);
}
if(m_state.viewport != rt->GetSize())
{
m_state.viewport = rt->GetSize();
D3D11_VIEWPORT vp;
memset(&vp, 0, sizeof(vp));
vp.TopLeftX = 0;
vp.TopLeftY = 0;
vp.Width = (FLOAT)rt->GetWidth();
vp.Height = (FLOAT)rt->GetHeight();
vp.MinDepth = 0.0f;
vp.MaxDepth = 1.0f;
m_ctx->RSSetViewports(1, &vp);
}
GSVector4i r = scissor ? *scissor : GSVector4i(rt->GetSize()).zwxy();
if(!m_state.scissor.eq(r))
{
m_state.scissor = r;
m_ctx->RSSetScissorRects(1, r);
}
}
void GSRasterizer::DrawTriangleSection(int top, int bottom, GSVertexSW& l, const GSVertexSW& dl, GSVector4& r, const GSVector4& dr, const GSVertexSW& dscan, const GSVector4& fscissor)
{
ASSERT(top < bottom);
while(1)
{
do
{
if(IsOneOfMyScanlines(top))
{
GSVector4 lr = l.p.xyxy(r).ceil();
GSVector4 lrmax = lr.max(fscissor.xxxx());
GSVector4 lrmin = lr.min(fscissor.zzzz());
GSVector4i lri = GSVector4i(lrmax.xxzz(lrmin));
int left = lri.extract32<0>();
int right = lri.extract32<2>();
int pixels = right - left;
if(pixels > 0)
{
m_stats.pixels += pixels;
GSVertexSW scan = l + dscan * (lrmax - l.p).xxxx();
m_dsf.ssl(right, left, top, scan);
}
}
}
while(0);
if(++top >= bottom) break;
l += dl;
r += dr;
}
}
void Clear() {
fbo = 0;
viewport = GSVector2i(0, 0);
scissor = GSVector4i(0, 0, 0, 0);
blend = false;
eq_RGB = 0;
f_sRGB = 0;
f_dRGB = 0;
wrgba = 0xF;
bf = 0.0;
depth = false;
depth_func = 0;
depth_mask = false;
stencil = false;
stencil_func = 0;
stencil_pass = 0;
ubo = 0;
ps_ss = 0;
rt = 0;
ds = 0;
for (size_t i = 0; i < countof(tex_unit); i++)
tex_unit[i] = 0;
for (size_t i = 0; i < countof(tex_handle); i++)
tex_handle[i] = 0;
ps = 0;
gs = 0;
vs = 0;
program = 0;
dirty_prog = false;
dirty_ressources = false;
}
void GSDrawScanlineCodeGenerator::InitVectors()
{
#if _M_SSE >= 0x501
GSVector8 log2_coef[4] =
{
GSVector8(0.204446009836232697516f),
GSVector8(-1.04913055217340124191f),
GSVector8(2.28330284476918490682f),
GSVector8(1.0f),
};
for (size_t n = 0; n < countof(log2_coef); ++n)
m_log2_coef[n] = log2_coef[n];
#else
GSVector4i test[8] =
{
GSVector4i::zero(),
GSVector4i(0xffffffff, 0x00000000, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0xffffffff, 0x00000000),
GSVector4i(0x00000000, 0xffffffff, 0xffffffff, 0xffffffff),
GSVector4i(0x00000000, 0x00000000, 0xffffffff, 0xffffffff),
GSVector4i(0x00000000, 0x00000000, 0x00000000, 0xffffffff),
GSVector4i::zero(),
};
GSVector4 log2_coef[4] =
{
GSVector4(0.204446009836232697516f),
GSVector4(-1.04913055217340124191f),
GSVector4(2.28330284476918490682f),
GSVector4(1.0f),
};
for (size_t n = 0; n < countof(test); ++n)
m_test[n] = test[n];
for (size_t n = 0; n < countof(log2_coef); ++n)
m_log2_coef[n] = log2_coef[n];
#endif
}
{
lines = 0;
}
}
else if(m_vt.m_primclass == GS_LINE_CLASS)
{
if(m_vertex.next == lines * 2)
{
// normally, this step would copy the video onto screen with 512 texture mapped horizontal lines,
// but we use the stored video data to create a new texture, and replace the lines with two triangles
m_dev->Recycle(t->m_texture);
t->m_texture = m_dev->CreateTexture(512, 512);
t->m_texture->Update(GSVector4i(0, 0, 448, lines), video, 448 * 4);
m_vertex.buff[2] = m_vertex.buff[m_vertex.next - 2];
m_vertex.buff[3] = m_vertex.buff[m_vertex.next - 1];
m_index.buff[0] = 0;
m_index.buff[1] = 1;
m_index.buff[2] = 2;
m_index.buff[3] = 1;
m_index.buff[4] = 2;
m_index.buff[5] = 3;
m_vertex.head = m_vertex.tail = m_vertex.next = 4;
m_index.tail = 6;
m_vt.Update(m_vertex.buff, m_index.buff, m_index.tail, GS_TRIANGLE_CLASS);
void GSRendererDX9::SetupIA()
{
D3DPRIMITIVETYPE topology;
switch(m_vt.m_primclass)
{
case GS_POINT_CLASS:
topology = D3DPT_POINTLIST;
break;
case GS_LINE_CLASS:
topology = D3DPT_LINELIST;
if(PRIM->IIP == 0)
{
for(size_t i = 0, j = m_index.tail; i < j; i += 2)
{
uint32 tmp = m_index.buff[i + 0];
m_index.buff[i + 0] = m_index.buff[i + 1];
m_index.buff[i + 1] = tmp;
}
}
break;
case GS_TRIANGLE_CLASS:
topology = D3DPT_TRIANGLELIST;
if(PRIM->IIP == 0)
{
for(size_t i = 0, j = m_index.tail; i < j; i += 3)
{
uint32 tmp = m_index.buff[i + 0];
m_index.buff[i + 0] = m_index.buff[i + 2];
m_index.buff[i + 2] = tmp;
}
}
break;
case GS_SPRITE_CLASS:
topology = D3DPT_TRIANGLELIST;
// each sprite converted to quad needs twice the space
while(m_vertex.tail * 2 > m_vertex.maxcount)
{
GrowVertexBuffer();
}
// assume vertices are tightly packed and sequentially indexed (it should be the case)
if(m_vertex.next >= 2)
{
size_t count = m_vertex.next;
int i = (int)count * 2 - 4;
GSVertex* s = &m_vertex.buff[count - 2];
GSVertex* q = &m_vertex.buff[count * 2 - 4];
uint32* RESTRICT index = &m_index.buff[count * 3 - 6];
for(; i >= 0; i -= 4, s -= 2, q -= 4, index -= 6)
{
GSVertex v0 = s[0];
GSVertex v1 = s[1];
v0.RGBAQ = v1.RGBAQ;
v0.XYZ.Z = v1.XYZ.Z;
v0.FOG = v1.FOG;
q[0] = v0;
q[3] = v1;
// swap x, s, u
uint16 x = v0.XYZ.X;
v0.XYZ.X = v1.XYZ.X;
v1.XYZ.X = x;
float s = v0.ST.S;
v0.ST.S = v1.ST.S;
v1.ST.S = s;
uint16 u = v0.U;
v0.U = v1.U;
v1.U = u;
q[1] = v0;
q[2] = v1;
index[0] = i + 0;
index[1] = i + 1;
index[2] = i + 2;
index[3] = i + 1;
index[4] = i + 2;
index[5] = i + 3;
}
m_vertex.head = m_vertex.tail = m_vertex.next = count * 2;
m_index.tail = count * 3;
}
break;
default:
__assume(0);
}
GSDevice9* dev = (GSDevice9*)m_dev;
(*dev)->SetRenderState(D3DRS_SHADEMODE, PRIM->IIP ? D3DSHADE_GOURAUD : D3DSHADE_FLAT); // TODO
void* ptr = NULL;
if(dev->IAMapVertexBuffer(&ptr, sizeof(GSVertexHW9), m_vertex.next))
{
GSVertex* RESTRICT s = (GSVertex*)m_vertex.buff;
GSVertexHW9* RESTRICT d = (GSVertexHW9*)ptr;
for(uint32 i = 0; i < m_vertex.next; i++, s++, d++)
{
GSVector4 p = GSVector4(GSVector4i::load(s->XYZ.u32[0]).upl16());
if(PRIM->TME && !PRIM->FST)
{
p = p.xyxy(GSVector4((float)s->XYZ.Z, s->RGBAQ.Q));
}
else
{
p = p.xyxy(GSVector4::load((float)s->XYZ.Z));
}
GSVector4 t = GSVector4::zero();
if(PRIM->TME)
{
if(PRIM->FST)
{
if(UserHacks_WildHack && !isPackedUV_HackFlag)
{
t = GSVector4(GSVector4i::load(s->UV & 0x3FEF3FEF).upl16());
//printf("GSDX: %08X | D3D9(%d) %s\n", s->UV & 0x3FEF3FEF, m_vertex.next, i == 0 ? "*" : "");
}
else
{
t = GSVector4(GSVector4i::load(s->UV).upl16());
}
}
else
{
t = GSVector4::loadl(&s->ST);
}
}
t = t.xyxy(GSVector4::cast(GSVector4i(s->RGBAQ.u32[0], s->FOG)));
d->p = p;
d->t = t;
}
dev->IAUnmapVertexBuffer();
}
dev->IASetIndexBuffer(m_index.buff, m_index.tail);
dev->IASetPrimitiveTopology(topology);
}
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 GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GSVertexSW& dv, const GSVector4i& scissor, int orientation, int side)
{
// orientation:
// - true: |dv.p.y| > |dv.p.x|
// - false |dv.p.x| > |dv.p.y|
// side:
// - true: top/left edge
// - false: bottom/right edge
// TODO: bit slow and too much duplicated code
// TODO: inner pre-step is still missing (hardly noticable)
GSVector4 fscissor(scissor);
GSVector4 lrtb = v0.p.upl(v1.p).ceil();
if(orientation)
{
GSVector4 tbmax = lrtb.max(fscissor.yyyy());
GSVector4 tbmin = lrtb.min(fscissor.wwww());
GSVector4i tbi = GSVector4i(tbmax.zwzw(tbmin));
int top, bottom;
GSVertexSW edge, dedge;
if((dv.p >= GSVector4::zero()).mask() & 2)
{
top = tbi.extract32<0>();
bottom = tbi.extract32<3>();
if(top >= bottom) return;
edge = v0;
dedge = dv / dv.p.yyyy();
edge += dedge * (tbmax.zzzz() - edge.p.yyyy());
}
else
{
top = tbi.extract32<1>();
bottom = tbi.extract32<2>();
if(top >= bottom) return;
edge = v1;
dedge = dv / dv.p.yyyy();
edge += dedge * (tbmax.wwww() - edge.p.yyyy());
}
GSVector4i p = GSVector4i(edge.p.upl(dedge.p) * 0x10000);
int x = p.extract32<0>();
int dx = p.extract32<1>();
if(side)
{
while(1)
{
do
{
int xi = x >> 16;
int xf = x & 0xffff;
if(scissor.left <= xi && xi < scissor.right && IsOneOfMyScanlines(xi))
{
m_stats.pixels++;
edge.t.u32[3] = (0x10000 - xf) & 0xffff;
m_dsf.ssle(xi + 1, xi, top, edge);
edge.t.u32[3] = 0;
}
}
while(0);
if(++top >= bottom) break;
edge += dedge;
x += dx;
}
}
else
{
while(1)
{
do
{
int xi = (x >> 16) + 1;
int xf = x & 0xffff;
if(scissor.left <= xi && xi < scissor.right && IsOneOfMyScanlines(xi))
{
m_stats.pixels++;
edge.t.u32[3] = xf;
m_dsf.ssle(xi + 1, xi, top, edge);
edge.t.u32[3] = 0;
}
}
while(0);
if(++top >= bottom) break;
edge += dedge;
x += dx;
}
}
}
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA USA.
* http://www.gnu.org/copyleft/gpl.html
*
*/
#include "stdafx.h"
#include "GSVector.h"
const GSVector4i GSVector4i::m_xff[17] =
{
GSVector4i(0x00000000, 0x00000000, 0x00000000, 0x00000000),
GSVector4i(0x000000ff, 0x00000000, 0x00000000, 0x00000000),
GSVector4i(0x0000ffff, 0x00000000, 0x00000000, 0x00000000),
GSVector4i(0x00ffffff, 0x00000000, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0x00000000, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0x000000ff, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0x0000ffff, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0x00ffffff, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0x00000000, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0x000000ff, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0x0000ffff, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0x00ffffff, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0xffffffff, 0x00000000),
GSVector4i(0xffffffff, 0xffffffff, 0xffffffff, 0x000000ff),
GSVector4i(0xffffffff, 0xffffffff, 0xffffffff, 0x0000ffff),
GSVector4i(0xffffffff, 0xffffffff, 0xffffffff, 0x00ffffff),
{
blend(a, b, xmm0);
}
}
void GPUDrawScanlineCodeGenerator::blend(const Xmm& a, const Xmm& b, const Xmm& mask)
{
pand(b, mask);
pandn(mask, a);
por(b, mask);
movdqa(a, b);
}
const GSVector4i GPUDrawScanlineCodeGenerator::m_test[8] =
{
GSVector4i(0xffff0000, 0xffffffff, 0xffffffff, 0xffffffff),
GSVector4i(0x00000000, 0xffffffff, 0xffffffff, 0xffffffff),
GSVector4i(0x00000000, 0xffff0000, 0xffffffff, 0xffffffff),
GSVector4i(0x00000000, 0x00000000, 0xffffffff, 0xffffffff),
GSVector4i(0x00000000, 0x00000000, 0xffff0000, 0xffffffff),
GSVector4i(0x00000000, 0x00000000, 0x00000000, 0xffffffff),
GSVector4i(0x00000000, 0x00000000, 0x00000000, 0xffff0000),
GSVector4i::zero(),
};
alignas(32) const uint16_t GPUDrawScanlineCodeGenerator::m_dither[4][16] =
{
{7, 0, 6, 1, 7, 0, 6, 1, 7, 0, 6, 1, 7, 0, 6, 1},
{2, 5, 3, 4, 2, 5, 3, 4, 2, 5, 3, 4, 2, 5, 3, 4},
{1, 6, 0, 7, 1, 6, 0, 7, 1, 6, 0, 7, 1, 6, 0, 7},
{4, 3, 5, 2, 4, 3, 5, 2, 4, 3, 5, 2, 4, 3, 5, 2},
void GPURendererSW::Draw()
{
GPUDrawScanline::SharedData* sd = new GPUDrawScanline::SharedData();
shared_ptr<GSRasterizerData> data(sd);
GPUScanlineGlobalData& gd = sd->global;
const GPUDrawingEnvironment& env = m_env;
gd.sel.key = 0;
gd.sel.iip = env.PRIM.IIP;
gd.sel.me = env.STATUS.ME;
if(env.PRIM.ABE)
{
gd.sel.abe = env.PRIM.ABE;
gd.sel.abr = env.STATUS.ABR;
}
gd.sel.tge = env.PRIM.TGE;
if(env.PRIM.TME)
{
gd.sel.tme = env.PRIM.TME;
gd.sel.tlu = env.STATUS.TP < 2;
gd.sel.twin = (env.TWIN.u32 & 0xfffff) != 0;
gd.sel.ltf = m_filter == 1 && env.PRIM.TYPE == GPU_POLYGON || m_filter == 2 ? 1 : 0;
const void* t = m_mem.GetTexture(env.STATUS.TP, env.STATUS.TX, env.STATUS.TY);
if(!t) {ASSERT(0); return;}
gd.tex = t;
gd.clut = (uint16*)_aligned_malloc(sizeof(uint16) * 256, 32);
memcpy(gd.clut, m_mem.GetCLUT(env.STATUS.TP, env.CLUT.X, env.CLUT.Y), sizeof(uint16) * (env.STATUS.TP == 0 ? 16 : 256));
gd.twin = GSVector4i(env.TWIN.TWW, env.TWIN.TWH, env.TWIN.TWX, env.TWIN.TWY);
}
gd.sel.dtd = m_dither ? env.STATUS.DTD : 0;
gd.sel.md = env.STATUS.MD;
gd.sel.sprite = env.PRIM.TYPE == GPU_SPRITE;
gd.sel.scalex = m_mem.GetScale().x;
gd.vm = m_mem.GetPixelAddress(0, 0);
data->scissor.left = (int)m_env.DRAREATL.X << m_scale.x;
data->scissor.top = (int)m_env.DRAREATL.Y << m_scale.y;
data->scissor.right = min((int)(m_env.DRAREABR.X + 1) << m_scale.x, m_mem.GetWidth());
data->scissor.bottom = min((int)(m_env.DRAREABR.Y + 1) << m_scale.y, m_mem.GetHeight());
data->buff = (uint8*)_aligned_malloc(sizeof(GSVertexSW) * m_count, 16);
data->vertex = (GSVertexSW*)data->buff;
data->vertex_count = m_count;
memcpy(data->vertex, m_vertices, sizeof(GSVertexSW) * m_count);
data->frame = m_perfmon.GetFrame();
int prims = 0;
switch(env.PRIM.TYPE)
{
case GPU_POLYGON: data->primclass = GS_TRIANGLE_CLASS; prims = data->vertex_count / 3; break;
case GPU_LINE: data->primclass = GS_LINE_CLASS; prims = data->vertex_count / 2; break;
case GPU_SPRITE: data->primclass = GS_SPRITE_CLASS; prims = data->vertex_count / 2; break;
default: __assume(0);
}
// TODO: VertexTrace
GSVector4 tl(+1e10f);
GSVector4 br(-1e10f);
GSVertexSW* v = data->vertex;
for(int i = 0, j = data->vertex_count; i < j; i++)
{
GSVector4 p = v[i].p;
tl = tl.min(p);
br = br.max(p);
}
data->bbox = GSVector4i(tl.xyxy(br));
GSVector4i r = data->bbox.rintersect(data->scissor);
r.left >>= m_scale.x;
r.top >>= m_scale.y;
r.right >>= m_scale.x;
r.bottom >>= m_scale.y;
Invalidate(r);
m_rl->Queue(data);
m_rl->Sync();
m_perfmon.Put(GSPerfMon::Draw, 1);
m_perfmon.Put(GSPerfMon::Prim, prims);
m_perfmon.Put(GSPerfMon::Fillrate, m_rl->GetPixels());
}
void GSRendererDX9::SetupIA(const float& sx, const float& sy)
{
D3DPRIMITIVETYPE topology;
switch(m_vt.m_primclass)
{
case GS_POINT_CLASS:
topology = D3DPT_POINTLIST;
break;
case GS_LINE_CLASS:
topology = D3DPT_LINELIST;
if(PRIM->IIP == 0)
{
for(size_t i = 0, j = m_index.tail; i < j; i += 2)
{
uint32 tmp = m_index.buff[i + 0];
m_index.buff[i + 0] = m_index.buff[i + 1];
m_index.buff[i + 1] = tmp;
}
}
break;
case GS_TRIANGLE_CLASS:
topology = D3DPT_TRIANGLELIST;
if(PRIM->IIP == 0)
{
for(size_t i = 0, j = m_index.tail; i < j; i += 3)
{
uint32 tmp = m_index.buff[i + 0];
m_index.buff[i + 0] = m_index.buff[i + 2];
m_index.buff[i + 2] = tmp;
}
}
break;
case GS_SPRITE_CLASS:
topology = D3DPT_TRIANGLELIST;
// each sprite converted to quad needs twice the space
Lines2Sprites();
break;
default:
__assume(0);
}
GSDevice9* dev = (GSDevice9*)m_dev;
(*dev)->SetRenderState(D3DRS_SHADEMODE, PRIM->IIP ? D3DSHADE_GOURAUD : D3DSHADE_FLAT); // TODO
void* ptr = NULL;
if(dev->IAMapVertexBuffer(&ptr, sizeof(GSVertexHW9), m_vertex.next))
{
GSVertex* RESTRICT s = (GSVertex*)m_vertex.buff;
GSVertexHW9* RESTRICT d = (GSVertexHW9*)ptr;
for(uint32 i = 0; i < m_vertex.next; i++, s++, d++)
{
GSVector4 p = GSVector4(GSVector4i::load(s->XYZ.u32[0]).upl16());
if(PRIM->TME && !PRIM->FST)
{
p = p.xyxy(GSVector4((float)s->XYZ.Z, s->RGBAQ.Q));
}
else
{
p = p.xyxy(GSVector4::load((float)s->XYZ.Z));
}
GSVector4 t = GSVector4::zero();
if(PRIM->TME)
{
if(PRIM->FST)
{
if(UserHacks_WildHack && !isPackedUV_HackFlag)
{
t = GSVector4(GSVector4i::load(s->UV & 0x3FEF3FEF).upl16());
//printf("GSDX: %08X | D3D9(%d) %s\n", s->UV & 0x3FEF3FEF, m_vertex.next, i == 0 ? "*" : "");
}
else
{
t = GSVector4(GSVector4i::load(s->UV).upl16());
}
}
else
{
t = GSVector4::loadl(&s->ST);
}
}
t = t.xyxy(GSVector4::cast(GSVector4i(s->RGBAQ.u32[0], s->FOG)));
d->p = p;
d->t = t;
}
dev->IAUnmapVertexBuffer();
}
dev->IASetIndexBuffer(m_index.buff, m_index.tail);
dev->IASetPrimitiveTopology(topology);
}
bool GSRenderer::Merge(int field)
{
bool en[2];
GSVector4i fr[2];
GSVector4i dr[2];
int baseline = INT_MAX;
for(int i = 0; i < 2; i++)
{
en[i] = IsEnabled(i);
if(en[i])
{
fr[i] = GetFrameRect(i);
dr[i] = GetDisplayRect(i);
baseline = min(dr[i].top, baseline);
//printf("[%d]: %d %d %d %d, %d %d %d %d\n", i, fr[i].x,fr[i].y,fr[i].z,fr[i].w , dr[i].x,dr[i].y,dr[i].z,dr[i].w);
}
}
if(!en[0] && !en[1])
{
return false;
}
// try to avoid fullscreen blur, could be nice on tv but on a monitor it's like double vision, hurts my eyes (persona 4, guitar hero)
//
// NOTE: probably the technique explained in graphtip.pdf (Antialiasing by Supersampling / 4. Reading Odd/Even Scan Lines Separately with the PCRTC then Blending)
bool samesrc =
en[0] && en[1] &&
m_regs->DISP[0].DISPFB.FBP == m_regs->DISP[1].DISPFB.FBP &&
m_regs->DISP[0].DISPFB.FBW == m_regs->DISP[1].DISPFB.FBW &&
m_regs->DISP[0].DISPFB.PSM == m_regs->DISP[1].DISPFB.PSM;
// bool blurdetected = false;
if(samesrc /*&& m_regs->PMODE.SLBG == 0 && m_regs->PMODE.MMOD == 1 && m_regs->PMODE.ALP == 0x80*/)
{
if(fr[0].eq(fr[1] + GSVector4i(0, -1, 0, 0)) && dr[0].eq(dr[1] + GSVector4i(0, 0, 0, 1))
|| fr[1].eq(fr[0] + GSVector4i(0, -1, 0, 0)) && dr[1].eq(dr[0] + GSVector4i(0, 0, 0, 1)))
{
// persona 4:
//
// fr[0] = 0 0 640 448
// fr[1] = 0 1 640 448
// dr[0] = 159 50 779 498
// dr[1] = 159 50 779 497
//
// second image shifted up by 1 pixel and blended over itself
//
// god of war:
//
// fr[0] = 0 1 512 448
// fr[1] = 0 0 512 448
// dr[0] = 127 50 639 497
// dr[1] = 127 50 639 498
//
// same just the first image shifted
int top = min(fr[0].top, fr[1].top);
int bottom = max(dr[0].bottom, dr[1].bottom);
fr[0].top = top;
fr[1].top = top;
dr[0].bottom = bottom;
dr[1].bottom = bottom;
// blurdetected = true;
}
else if(dr[0].eq(dr[1]) && (fr[0].eq(fr[1] + GSVector4i(0, 1, 0, 1)) || fr[1].eq(fr[0] + GSVector4i(0, 1, 0, 1))))
{
// dq5:
//
// fr[0] = 0 1 512 445
// fr[1] = 0 0 512 444
// dr[0] = 127 50 639 494
// dr[1] = 127 50 639 494
int top = min(fr[0].top, fr[1].top);
int bottom = min(fr[0].bottom, fr[1].bottom);
fr[0].top = fr[1].top = top;
fr[0].bottom = fr[1].bottom = bottom;
// blurdetected = true;
}
//printf("samesrc = %d blurdetected = %d\n",samesrc,blurdetected);
}
GSVector2i fs(0, 0);
GSVector2i ds(0, 0);
GSTexture* tex[2] = {NULL, NULL};
if(samesrc && fr[0].bottom == fr[1].bottom)
{
tex[0] = GetOutput(0);
tex[1] = tex[0]; // saves one texture fetch
}
else
{
if(en[0]) tex[0] = GetOutput(0);
if(en[1]) tex[1] = GetOutput(1);
}
GSVector4 src[2];
GSVector4 dst[2];
for(int i = 0; i < 2; i++)
{
if(!en[i] || !tex[i]) continue;
GSVector4i r = fr[i];
// overscan hack
if(dr[i].height() > 512) // hmm
{
int y = GetDeviceSize(i).y;
if(m_regs->SMODE2.INT && m_regs->SMODE2.FFMD) y /= 2;
r.bottom = r.top + y;
}
GSVector4 scale = GSVector4(tex[i]->GetScale()).xyxy();
src[i] = GSVector4(r) * scale / GSVector4(tex[i]->GetSize()).xyxy();
GSVector2 o(0, 0);
if(dr[i].top - baseline >= 4) // 2?
{
o.y = tex[i]->GetScale().y * (dr[i].top - baseline);
if(m_regs->SMODE2.INT && m_regs->SMODE2.FFMD)
{
o.y /= 2;
}
}
dst[i] = GSVector4(o).xyxy() + scale * GSVector4(r.rsize());
fs.x = max(fs.x, (int)(dst[i].z + 0.5f));
fs.y = max(fs.y, (int)(dst[i].w + 0.5f));
}
ds = fs;
if(m_regs->SMODE2.INT && m_regs->SMODE2.FFMD)
{
ds.y *= 2;
}
bool slbg = m_regs->PMODE.SLBG;
bool mmod = m_regs->PMODE.MMOD;
if(tex[0] || tex[1])
{
if(tex[0] == tex[1] && !slbg && (src[0] == src[1] & dst[0] == dst[1]).alltrue())
{
// the two outputs are identical, skip drawing one of them (the one that is alpha blended)
tex[0] = NULL;
}
GSVector4 c = GSVector4((int)m_regs->BGCOLOR.R, (int)m_regs->BGCOLOR.G, (int)m_regs->BGCOLOR.B, (int)m_regs->PMODE.ALP) / 255;
m_dev->Merge(tex, src, dst, fs, slbg, mmod, c);
if(m_regs->SMODE2.INT && m_interlace > 0)
{
if (m_interlace == 7 && m_regs->SMODE2.FFMD == 1) // Auto interlace enabled / Odd frame interlace setting
{
int field2 = 0;
int mode = 2;
m_dev->Interlace(ds, field ^ field2, mode, tex[1] ? tex[1]->GetScale().y : tex[0]->GetScale().y);
}
else
{
int field2 = 1 - ((m_interlace - 1) & 1);
int mode = (m_interlace - 1) >> 1;
m_dev->Interlace(ds, field ^ field2, mode, tex[1] ? tex[1]->GetScale().y : tex[0]->GetScale().y);
}
}
if(m_shadeboost)
{
m_dev->ShadeBoost();
}
if (m_shaderfx)
{
m_dev->ExternalFX();
}
if(m_fxaa)
{
m_dev->FXAA();
}
}
return true;
}
GSVector4i GSVector4i::cast(const GSVector4& v)
{
return GSVector4i(_mm_castps_si128(v.m));
}
void GSRendererHW::Draw()
{
if(m_dev->IsLost() || GSRenderer::IsBadFrame(m_skip, m_userhacks_skipdraw)) {
GL_INS("Warning skipping a draw call (%d)", s_n);
s_n += 3; // Keep it sync with SW renderer
return;
}
GL_PUSH("HW Draw %d", s_n);
GSDrawingEnvironment& env = m_env;
GSDrawingContext* context = m_context;
// It is allowed to use the depth and rt at the same location. However at least 1 must
// be disabled.
// 1/ GoW uses a Cd blending on a 24 bits buffer (no alpha)
// 2/ SuperMan really draws (0,0,0,0) color and a (0) 32-bits depth
// 3/ 50cents really draws (0,0,0,128) color and a (0) 24 bits depth
// Note: FF DoC has both buffer at same location but disable the depth test (write?) with ZTE = 0
const bool no_rt = (context->ALPHA.IsCd() && PRIM->ABE && (context->FRAME.PSM == 1));
const bool no_ds = !no_rt && (
// Depth is always pass (no read) and write are discarded (tekken 5). (Note: DATE is currently implemented with a stencil buffer)
(context->ZBUF.ZMSK && m_context->TEST.ZTST == ZTST_ALWAYS && !m_context->TEST.DATE) ||
// Depth will be written through the RT
(context->FRAME.FBP == context->ZBUF.ZBP && !PRIM->TME && !context->ZBUF.ZMSK && !context->FRAME.FBMSK && context->TEST.ZTE)
);
GIFRegTEX0 TEX0;
TEX0.TBP0 = context->FRAME.Block();
TEX0.TBW = context->FRAME.FBW;
TEX0.PSM = context->FRAME.PSM;
GSTextureCache::Target* rt = no_rt ? NULL : m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::RenderTarget, true);
GSTexture* rt_tex = rt ? rt->m_texture : NULL;
TEX0.TBP0 = context->ZBUF.Block();
TEX0.TBW = context->FRAME.FBW;
TEX0.PSM = context->ZBUF.PSM;
GSTextureCache::Target* ds = no_ds ? NULL : m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::DepthStencil, context->DepthWrite());
GSTexture* ds_tex = ds ? ds->m_texture : NULL;
if(!(rt || no_rt) || !(ds || no_ds))
{
GL_POP();
ASSERT(0);
return;
}
GSTextureCache::Source* tex = NULL;
m_texture_shuffle = false;
if(PRIM->TME)
{
/*
// m_tc->LookupSource will mess with the palette, should not, but we do this after, until it is sorted out
if(GSLocalMemory::m_psm[context->TEX0.PSM].pal > 0)
{
m_mem.m_clut.Read32(context->TEX0, env.TEXA);
}
*/
GSVector4i r;
GetTextureMinMax(r, context->TEX0, context->CLAMP, m_vt.IsLinear());
tex = m_tc->LookupSource(context->TEX0, env.TEXA, r);
if(!tex) {
GL_POP();
return;
}
// FIXME: Could be removed on openGL
if(GSLocalMemory::m_psm[context->TEX0.PSM].pal > 0)
{
m_mem.m_clut.Read32(context->TEX0, env.TEXA);
}
// Hypothesis: texture shuffle is used as a postprocessing effect so texture will be an old target.
// Initially code also tested the RT but it gives too much false-positive
//
// Both input and output are 16 bits and texture was initially 32 bits!
m_texture_shuffle = (context->FRAME.PSM & 0x2) && ((context->TEX0.PSM & 3) == 2) && (m_vt.m_primclass == GS_SPRITE_CLASS) && tex->m_32_bits_fmt;
// Texture shuffle is not yet supported with strange clamp mode
ASSERT(!m_texture_shuffle || (context->CLAMP.WMS < 3 && context->CLAMP.WMT < 3));
}
if (rt) {
// Be sure texture shuffle detection is properly propagated
// Otherwise set or clear the flag (Code in texture cache only set the flag)
// Note: it is important to clear the flag when RT is used as a real 16 bits target.
rt->m_32_bits_fmt = m_texture_shuffle || !(context->FRAME.PSM & 0x2);
}
if(s_dump)
{
uint64 frame = m_perfmon.GetFrame();
string s;
if (s_n >= s_saven) {
// Dump Register state
s = format("%05d_context.txt", s_n);
m_env.Dump(root_hw+s);
m_context->Dump(root_hw+s);
}
if(s_savet && s_n >= s_saven && tex)
{
s = format("%05d_f%lld_tex_%05x_%d_%d%d_%02x_%02x_%02x_%02x.dds",
s_n, frame, (int)context->TEX0.TBP0, (int)context->TEX0.PSM,
(int)context->CLAMP.WMS, (int)context->CLAMP.WMT,
(int)context->CLAMP.MINU, (int)context->CLAMP.MAXU,
(int)context->CLAMP.MINV, (int)context->CLAMP.MAXV);
tex->m_texture->Save(root_hw+s, true);
if(tex->m_palette)
{
s = format("%05d_f%lld_tpx_%05x_%d.dds", s_n, frame, context->TEX0.CBP, context->TEX0.CPSM);
tex->m_palette->Save(root_hw+s, true);
}
}
s_n++;
if(s_save && s_n >= s_saven)
{
s = format("%05d_f%lld_rt0_%05x_%d.bmp", s_n, frame, context->FRAME.Block(), context->FRAME.PSM);
if (rt)
rt->m_texture->Save(root_hw+s);
}
if(s_savez && s_n >= s_saven)
{
s = format("%05d_f%lld_rz0_%05x_%d.bmp", s_n, frame, context->ZBUF.Block(), context->ZBUF.PSM);
if (ds_tex)
ds_tex->Save(root_hw+s);
}
s_n++;
#ifdef ENABLE_OGL_DEBUG
} else {
s_n += 2;
#endif
}
if(m_hacks.m_oi && !(this->*m_hacks.m_oi)(rt_tex, ds_tex, tex))
{
s_n += 1; // keep counter sync
GL_POP();
return;
}
// skip alpha test if possible
GIFRegTEST TEST = context->TEST;
GIFRegFRAME FRAME = context->FRAME;
GIFRegZBUF ZBUF = context->ZBUF;
uint32 fm = context->FRAME.FBMSK;
uint32 zm = context->ZBUF.ZMSK || context->TEST.ZTE == 0 ? 0xffffffff : 0;
if(context->TEST.ATE && context->TEST.ATST != ATST_ALWAYS)
{
if(GSRenderer::TryAlphaTest(fm, zm))
{
context->TEST.ATST = ATST_ALWAYS;
}
}
context->FRAME.FBMSK = fm;
context->ZBUF.ZMSK = zm != 0;
// A couple of hack to avoid upscaling issue. So far it seems to impacts mostly sprite
if ((m_upscale_multiplier > 1) && (m_vt.m_primclass == GS_SPRITE_CLASS)) {
size_t count = m_vertex.next;
GSVertex* v = &m_vertex.buff[0];
// Hack to avoid vertical black line in various games (ace combat/tekken)
if (m_userhacks_align_sprite_X) {
// Note for performance reason I do the check only once on the first
// primitive
int win_position = v[1].XYZ.X - context->XYOFFSET.OFX;
const bool unaligned_position = ((win_position & 0xF) == 8);
const bool unaligned_texture = ((v[1].U & 0xF) == 0) && PRIM->FST; // I'm not sure this check is useful
const bool hole_in_vertex = (count < 4) || (v[1].XYZ.X != v[2].XYZ.X);
if (hole_in_vertex && unaligned_position && (unaligned_texture || !PRIM->FST)) {
// Normaly vertex are aligned on full pixels and texture in half
// pixels. Let's extend the coverage of an half-pixel to avoid
// hole after upscaling
for(size_t i = 0; i < count; i += 2) {
v[i+1].XYZ.X += 8;
// I really don't know if it is a good idea. Neither what to do for !PRIM->FST
if (unaligned_texture)
v[i+1].U += 8;
}
}
}
if (PRIM->FST) {
if ((m_userhacks_round_sprite_offset > 1) || (m_userhacks_round_sprite_offset == 1 && !m_vt.IsLinear())) {
if (m_vt.IsLinear())
RoundSpriteOffset<true>();
else
RoundSpriteOffset<false>();
}
} else {
; // vertical line in Yakuza (note check m_userhacks_align_sprite_X behavior)
}
}
//
DrawPrims(rt_tex, ds_tex, tex);
//
context->TEST = TEST;
context->FRAME = FRAME;
context->ZBUF = ZBUF;
//
GSVector4i r = GSVector4i(m_vt.m_min.p.xyxy(m_vt.m_max.p)).rintersect(GSVector4i(context->scissor.in));
if(fm != 0xffffffff && rt)
{
rt->m_valid = rt->m_valid.runion(r);
m_tc->InvalidateVideoMem(context->offset.fb, r, false);
m_tc->InvalidateVideoMemType(GSTextureCache::DepthStencil, context->FRAME.Block());
}
if(zm != 0xffffffff && ds)
{
ds->m_valid = ds->m_valid.runion(r);
m_tc->InvalidateVideoMem(context->offset.zb, r, false);
m_tc->InvalidateVideoMemType(GSTextureCache::RenderTarget, context->ZBUF.Block());
}
//
if(m_hacks.m_oo)
{
(this->*m_hacks.m_oo)();
}
if(s_dump)
{
uint64 frame = m_perfmon.GetFrame();
string s;
if(s_save && s_n >= s_saven)
{
s = format("%05d_f%lld_rt1_%05x_%d.bmp", s_n, frame, context->FRAME.Block(), context->FRAME.PSM);
if (rt)
rt->m_texture->Save(root_hw+s);
}
if(s_savez && s_n >= s_saven)
{
s = format("%05d_f%lld_rz1_%05x_%d.bmp", s_n, frame, context->ZBUF.Block(), context->ZBUF.PSM);
if (ds_tex)
ds_tex->Save(root_hw+s);
}
s_n++;
if(s_savel > 0 && (s_n - s_saven) > s_savel)
{
s_dump = 0;
}
#ifdef ENABLE_OGL_DEBUG
} else {
s_n += 1;
#endif
}
#ifdef DISABLE_HW_TEXTURE_CACHE
if (rt)
m_tc->Read(rt, r);
#endif
GL_POP();
}
void GSRasterizer::DrawTriangleTopBottom(GSVertexSW* v, const GSVector4i& scissor)
{
GSVertexSW dv[3];
dv[0] = v[1] - v[0];
dv[1] = v[2] - v[0];
GSVertexSW longest = dv[1] * (dv[0].p / dv[1].p).yyyy() - dv[0];
int i = longest.p.upl(longest.p == GSVector4::zero()).mask();
if(i & 2) return;
i &= 1;
GSVertexSW dscan = longest * longest.p.xxxx().rcp();
m_dsf.ssp(v, dscan);
GSVector4 fscissor(scissor);
GSVector4 tb = v[0].p.upl(v[1].p).zwzw(v[1].p.upl(v[2].p)).ceil();
GSVector4 tbmax = tb.max(fscissor.yyyy());
GSVector4 tbmin = tb.min(fscissor.wwww());
GSVector4i tbi = GSVector4i(tbmax.xzyw(tbmin));
int top = tbi.extract32<0>();
int bottom = tbi.extract32<2>();
GSVertexSW& l = v[0];
GSVector4 r = v[0].p;
GSVertexSW dl = dv[i] / dv[i].p.yyyy();
GSVector4 dr = dv[1 - i].p / dv[1 - i].p.yyyy();
GSVector4 dy = tbmax.xxxx() - l.p.yyyy();
l += dl * dy;
r += dr * dy;
if(top < bottom)
{
DrawTriangleSection(top, bottom, l, dl, r, dr, dscan, fscissor);
}
top = tbi.y;
bottom = tbi.w;
if(top < bottom)
{
if(i == 0)
{
l = v[1];
dv[2] = v[2] - v[1];
dl = dv[2] / dv[2].p.yyyy();
}
else
{
r = v[1].p;
dv[2].p = v[2].p - v[1].p;
dr = dv[2].p / dv[2].p.yyyy();
}
l += dl * (tbmax.zzzz() - l.p.yyyy());
r += dr * (tbmax.zzzz() - r.yyyy());
l.p = l.p.upl(r).xyzw(l.p); // r.x => l.y
dl.p = dl.p.upl(dr).xyzw(dl.p); // dr.x => dl.y
DrawTriangleSection(top, bottom, l, dl, dscan, fscissor);
}
}
bool GSRenderer::Merge(int field)
{
bool en[2];
GSVector4i fr[2];
GSVector4i dr[2];
GSVector2i display_baseline = { INT_MAX, INT_MAX };
GSVector2i frame_baseline = { INT_MAX, INT_MAX };
for(int i = 0; i < 2; i++)
{
en[i] = IsEnabled(i);
if(en[i])
{
fr[i] = GetFrameRect(i);
dr[i] = GetDisplayRect(i);
display_baseline.x = min(dr[i].left, display_baseline.x);
display_baseline.y = min(dr[i].top, display_baseline.y);
frame_baseline.x = min(fr[i].left, frame_baseline.x);
frame_baseline.y = min(fr[i].top, frame_baseline.y);
//printf("[%d]: %d %d %d %d, %d %d %d %d\n", i, fr[i].x,fr[i].y,fr[i].z,fr[i].w , dr[i].x,dr[i].y,dr[i].z,dr[i].w);
}
}
if(!en[0] && !en[1])
{
return false;
}
GL_PUSH("Renderer Merge %d (0: enabled %d 0x%x, 1: enabled %d 0x%x)", s_n, en[0], m_regs->DISP[0].DISPFB.Block(), en[1], m_regs->DISP[1].DISPFB.Block());
// try to avoid fullscreen blur, could be nice on tv but on a monitor it's like double vision, hurts my eyes (persona 4, guitar hero)
//
// NOTE: probably the technique explained in graphtip.pdf (Antialiasing by Supersampling / 4. Reading Odd/Even Scan Lines Separately with the PCRTC then Blending)
bool samesrc =
en[0] && en[1] &&
m_regs->DISP[0].DISPFB.FBP == m_regs->DISP[1].DISPFB.FBP &&
m_regs->DISP[0].DISPFB.FBW == m_regs->DISP[1].DISPFB.FBW &&
m_regs->DISP[0].DISPFB.PSM == m_regs->DISP[1].DISPFB.PSM;
if(samesrc /*&& m_regs->PMODE.SLBG == 0 && m_regs->PMODE.MMOD == 1 && m_regs->PMODE.ALP == 0x80*/)
{
// persona 4:
//
// fr[0] = 0 0 640 448
// fr[1] = 0 1 640 448
// dr[0] = 159 50 779 498
// dr[1] = 159 50 779 497
//
// second image shifted up by 1 pixel and blended over itself
//
// god of war:
//
// fr[0] = 0 1 512 448
// fr[1] = 0 0 512 448
// dr[0] = 127 50 639 497
// dr[1] = 127 50 639 498
//
// same just the first image shifted
//
// These kinds of cases are now fixed by the more generic frame_diff code below, as the code here was too specific and has become obsolete.
// NOTE: Persona 4 and God Of War are not rare exceptions, many games have the same(or very similar) offsets.
int topDiff = fr[0].top - fr[1].top;
if (dr[0].eq(dr[1]) && (fr[0].eq(fr[1] + GSVector4i(0, topDiff, 0, topDiff)) || fr[1].eq(fr[0] + GSVector4i(0, topDiff, 0, topDiff))))
{
// dq5:
//
// fr[0] = 0 1 512 445
// fr[1] = 0 0 512 444
// dr[0] = 127 50 639 494
// dr[1] = 127 50 639 494
int top = min(fr[0].top, fr[1].top);
int bottom = min(fr[0].bottom, fr[1].bottom);
fr[0].top = fr[1].top = top;
fr[0].bottom = fr[1].bottom = bottom;
}
}
GSVector2i fs(0, 0);
GSVector2i ds(0, 0);
GSTexture* tex[3] = {NULL, NULL, NULL};
int y_offset[3] = {0, 0, 0};
s_n++;
bool feedback_merge = m_regs->EXTWRITE.WRITE == 1;
if(samesrc && fr[0].bottom == fr[1].bottom && !feedback_merge)
{
tex[0] = GetOutput(0, y_offset[0]);
tex[1] = tex[0]; // saves one texture fetch
y_offset[1] = y_offset[0];
}
else
{
if(en[0]) tex[0] = GetOutput(0, y_offset[0]);
if(en[1]) tex[1] = GetOutput(1, y_offset[1]);
if(feedback_merge) tex[2] = GetFeedbackOutput();
}
GSVector4 src[2];
GSVector4 src_hw[2];
GSVector4 dst[2];
for(int i = 0; i < 2; i++)
{
if(!en[i] || !tex[i]) continue;
GSVector4i r = fr[i];
GSVector4 scale = GSVector4(tex[i]->GetScale()).xyxy();
src[i] = GSVector4(r) * scale / GSVector4(tex[i]->GetSize()).xyxy();
src_hw[i] = (GSVector4(r) + GSVector4 (0, y_offset[i], 0, y_offset[i])) * scale / GSVector4(tex[i]->GetSize()).xyxy();
GSVector2 off(0);
GSVector2i display_diff(dr[i].left - display_baseline.x, dr[i].top - display_baseline.y);
GSVector2i frame_diff(fr[i].left - frame_baseline.x, fr[i].top - frame_baseline.y);
// Time Crisis 2/3 uses two side by side images when in split screen mode.
// Though ignore cases where baseline and display rectangle offsets only differ by 1 pixel, causes blurring and wrong resolution output on FFXII
if(display_diff.x > 2)
{
off.x = tex[i]->GetScale().x * display_diff.x;
}
// If the DX offset is too small then consider the status of frame memory offsets, prevents blurring on Tenchu: Fatal Shadows, Worms 3D
else if(display_diff.x != frame_diff.x)
{
off.x = tex[i]->GetScale().x * frame_diff.x;
}
if(display_diff.y >= 4) // Shouldn't this be >= 2?
{
off.y = tex[i]->GetScale().y * display_diff.y;
if(m_regs->SMODE2.INT && m_regs->SMODE2.FFMD)
{
off.y /= 2;
}
}
else if(display_diff.y != frame_diff.y)
{
off.y = tex[i]->GetScale().y * frame_diff.y;
}
dst[i] = GSVector4(off).xyxy() + scale * GSVector4(r.rsize());
fs.x = max(fs.x, (int)(dst[i].z + 0.5f));
fs.y = max(fs.y, (int)(dst[i].w + 0.5f));
}
ds = fs;
if(m_regs->SMODE2.INT && m_regs->SMODE2.FFMD)
{
ds.y *= 2;
}
m_real_size = ds;
bool slbg = m_regs->PMODE.SLBG;
if(tex[0] || tex[1])
{
if(tex[0] == tex[1] && !slbg && (src[0] == src[1] & dst[0] == dst[1]).alltrue())
{
// the two outputs are identical, skip drawing one of them (the one that is alpha blended)
tex[0] = NULL;
}
GSVector4 c = GSVector4((int)m_regs->BGCOLOR.R, (int)m_regs->BGCOLOR.G, (int)m_regs->BGCOLOR.B, (int)m_regs->PMODE.ALP) / 255;
m_dev->Merge(tex, src_hw, dst, fs, m_regs->PMODE, m_regs->EXTBUF, c);
if(m_regs->SMODE2.INT && m_interlace > 0)
{
if(m_interlace == 7 && m_regs->SMODE2.FFMD) // Auto interlace enabled / Odd frame interlace setting
{
int field2 = 0;
int mode = 2;
m_dev->Interlace(ds, field ^ field2, mode, tex[1] ? tex[1]->GetScale().y : tex[0]->GetScale().y);
}
else
{
int field2 = 1 - ((m_interlace - 1) & 1);
int mode = (m_interlace - 1) >> 1;
m_dev->Interlace(ds, field ^ field2, mode, tex[1] ? tex[1]->GetScale().y : tex[0]->GetScale().y);
}
}
if(m_shadeboost)
{
m_dev->ShadeBoost();
}
if(m_shaderfx)
{
m_dev->ExternalFX();
}
if(m_fxaa)
{
m_dev->FXAA();
}
}
return true;
}
* any later version. * * This Program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with GNU Make; see the file COPYING. If not, write to * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. * http://www.gnu.org/copyleft/gpl.html * */ #include "stdafx.h" #include "GSBlock.h" const GSVector4i GSBlock::m_r16mask(0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15); const GSVector4i GSBlock::m_r8mask(0, 4, 2, 6, 8, 12, 10, 14, 1, 5, 3, 7, 9, 13, 11, 15); const GSVector4i GSBlock::m_r4mask(0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15); const GSVector4i GSBlock::m_xxxa(0x00008000); const GSVector4i GSBlock::m_xxbx(0x00007c00); const GSVector4i GSBlock::m_xgxx(0x000003e0); const GSVector4i GSBlock::m_rxxx(0x0000001f); const GSVector4i GSBlock::m_uw8hmask0 = GSVector4i(0, 0, 0, 0, 1, 1, 1, 1, 8, 8, 8, 8, 9, 9, 9, 9); const GSVector4i GSBlock::m_uw8hmask1 = GSVector4i(2, 2, 2, 2, 3, 3, 3, 3, 10, 10, 10, 10, 11, 11, 11, 11); const GSVector4i GSBlock::m_uw8hmask2 = GSVector4i(4, 4, 4, 4, 5, 5, 5, 5, 12, 12, 12, 12, 13, 13, 13, 13); const GSVector4i GSBlock::m_uw8hmask3 = GSVector4i(6, 6, 6, 6, 7, 7, 7, 7, 14, 14, 14, 14, 15, 15, 15, 15);
void GSRendererHW::Draw()
{
if(m_dev->IsLost()) return;
if(GSRenderer::IsBadFrame(m_skip, m_userhacks_skipdraw)) return;
GSDrawingEnvironment& env = m_env;
GSDrawingContext* context = m_context;
GIFRegTEX0 TEX0;
TEX0.TBP0 = context->FRAME.Block();
TEX0.TBW = context->FRAME.FBW;
TEX0.PSM = context->FRAME.PSM;
GSTextureCache::Target* rt = m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::RenderTarget, true);
TEX0.TBP0 = context->ZBUF.Block();
TEX0.TBW = context->FRAME.FBW;
TEX0.PSM = context->ZBUF.PSM;
GSTextureCache::Target* ds = m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::DepthStencil, context->DepthWrite());
if(!rt || !ds)
{
ASSERT(0);
return;
}
GSTextureCache::Source* tex = NULL;
if(PRIM->TME)
{
/*
// m_tc->LookupSource will mess with the palette, should not, but we do this after, until it is sorted out
if(GSLocalMemory::m_psm[context->TEX0.PSM].pal > 0)
{
m_mem.m_clut.Read32(context->TEX0, env.TEXA);
}
*/
GSVector4i r;
GetTextureMinMax(r, context->TEX0, context->CLAMP, m_vt.IsLinear());
tex = m_tc->LookupSource(context->TEX0, env.TEXA, r);
if(!tex) return;
if(GSLocalMemory::m_psm[context->TEX0.PSM].pal > 0)
{
m_mem.m_clut.Read32(context->TEX0, env.TEXA);
}
}
if(s_dump)
{
uint64 frame = m_perfmon.GetFrame();
string s;
if(s_save && s_n >= s_saven && tex)
{
s = format("c:\\temp2\\_%05d_f%lld_tex_%05x_%d_%d%d_%02x_%02x_%02x_%02x.dds",
s_n, frame, (int)context->TEX0.TBP0, (int)context->TEX0.PSM,
(int)context->CLAMP.WMS, (int)context->CLAMP.WMT,
(int)context->CLAMP.MINU, (int)context->CLAMP.MAXU,
(int)context->CLAMP.MINV, (int)context->CLAMP.MAXV);
tex->m_texture->Save(s, true);
if(tex->m_palette)
{
s = format("c:\\temp2\\_%05d_f%lld_tpx_%05x_%d.dds", s_n, frame, context->TEX0.CBP, context->TEX0.CPSM);
tex->m_palette->Save(s, true);
}
}
s_n++;
if(s_save && s_n >= s_saven)
{
s = format("c:\\temp2\\_%05d_f%lld_rt0_%05x_%d.bmp", s_n, frame, context->FRAME.Block(), context->FRAME.PSM);
rt->m_texture->Save(s);
}
if(s_savez && s_n >= s_saven)
{
s = format("c:\\temp2\\_%05d_f%lld_rz0_%05x_%d.bmp", s_n, frame, context->ZBUF.Block(), context->ZBUF.PSM);
ds->m_texture->Save(s);
}
s_n++;
}
if(m_hacks.m_oi && !(this->*m_hacks.m_oi)(rt->m_texture, ds->m_texture, tex))
{
return;
}
// skip alpha test if possible
GIFRegTEST TEST = context->TEST;
GIFRegFRAME FRAME = context->FRAME;
GIFRegZBUF ZBUF = context->ZBUF;
uint32 fm = context->FRAME.FBMSK;
uint32 zm = context->ZBUF.ZMSK || context->TEST.ZTE == 0 ? 0xffffffff : 0;
if(context->TEST.ATE && context->TEST.ATST != ATST_ALWAYS)
{
if(GSRenderer::TryAlphaTest(fm, zm))
{
context->TEST.ATST = ATST_ALWAYS;
}
}
context->FRAME.FBMSK = fm;
context->ZBUF.ZMSK = zm != 0;
//
DrawPrims(rt->m_texture, ds->m_texture, tex);
//
context->TEST = TEST;
context->FRAME = FRAME;
context->ZBUF = ZBUF;
//
GSVector4i r = GSVector4i(m_vt.m_min.p.xyxy(m_vt.m_max.p)).rintersect(GSVector4i(context->scissor.in));
if(fm != 0xffffffff)
{
rt->m_valid = rt->m_valid.runion(r);
m_tc->InvalidateVideoMem(context->offset.fb, r, false);
}
if(zm != 0xffffffff)
{
ds->m_valid = ds->m_valid.runion(r);
m_tc->InvalidateVideoMem(context->offset.zb, r, false);
}
//
if(m_hacks.m_oo)
{
(this->*m_hacks.m_oo)();
}
if(s_dump)
{
uint64 frame = m_perfmon.GetFrame();
string s;
if(s_save && s_n >= s_saven)
{
s = format("c:\\temp2\\_%05d_f%lld_rt1_%05x_%d.bmp", s_n, frame, context->FRAME.Block(), context->FRAME.PSM);
rt->m_texture->Save(s);
}
if(s_savez && s_n >= s_saven)
{
s = format("c:\\temp2\\_%05d_f%lld_rz1_%05x_%d.bmp", s_n, frame, context->ZBUF.Block(), context->ZBUF.PSM);
ds->m_texture->Save(s);
}
s_n++;
}
#ifdef DISABLE_HW_TEXTURE_CACHE
m_tc->Read(rt, r);
#endif
}