Merge pull request #15978 from unknownbrackets/softgpu-immediate

Correct some of immediate vertex handling

GPU/Common/FramebufferManagerCommon.cpp

@@ -810,7 +810,7 @@ void FramebufferManagerCommon::CopyToColorFromOverlappingFramebuffers(VirtualFra
 		}
 	}
 
-	if (dst != currentRenderVfb_ && tookActions) {
+	if (currentRenderVfb_ && dst != currentRenderVfb_ && tookActions) {
 		// Will probably just change the name of the current renderpass, since one was started by the reinterpret itself.
 		draw_->BindFramebufferAsRenderTarget(currentRenderVfb_->fbo, { Draw::RPAction::KEEP, Draw::RPAction::KEEP, Draw::RPAction::KEEP }, "After Reinterpret");
 	}

GPU/Common/VertexDecoderCommon.h

@@ -28,6 +28,7 @@
 #include "GPU/ge_constants.h"
 #include "GPU/Common/ShaderCommon.h"
 #include "GPU/GPUCommon.h"
+#include "GPU/GPUState.h"
 
 #if PPSSPP_ARCH(ARM)
 #include "Common/ArmEmitter.h"
@@ -299,6 +300,8 @@ class VertexReader {
 	bool hasNormal() const { return decFmt_.nrmfmt != 0; }
 	bool hasUV() const { return decFmt_.uvfmt != 0; }
 	bool isThrough() const { return (vtype_ & GE_VTYPE_THROUGH) != 0; }
+	bool skinningEnabled() const { return vertTypeIsSkinningEnabled(vtype_); }
+	int numBoneWeights() const { return vertTypeGetNumBoneWeights(vtype_); }
 	void Goto(int index) {
 		data_ = base_ + index * decFmt_.stride;
 	}

GPU/Debugger/Debugger.cpp

@@ -73,6 +73,7 @@ void SetBreakNext(BreakNext next) {
 		GPUBreakpoints::AddCmdBreakpoint(GE_CMD_PRIM, true);
 		GPUBreakpoints::AddCmdBreakpoint(GE_CMD_BEZIER, true);
 		GPUBreakpoints::AddCmdBreakpoint(GE_CMD_SPLINE, true);
+		GPUBreakpoints::AddCmdBreakpoint(GE_CMD_VAP, true);
 	} else if (next == BreakNext::CURVE) {
 		GPUBreakpoints::AddCmdBreakpoint(GE_CMD_BEZIER, true);
 		GPUBreakpoints::AddCmdBreakpoint(GE_CMD_SPLINE, true);
@@ -111,7 +112,7 @@ bool NotifyCommand(u32 pc) {
 	}
 
 	bool process = true;
-	if (cmd == GE_CMD_PRIM || cmd == GE_CMD_BEZIER || cmd == GE_CMD_SPLINE) {
+	if (cmd == GE_CMD_PRIM || cmd == GE_CMD_BEZIER || cmd == GE_CMD_SPLINE || cmd == GE_CMD_VAP) {
 		primsThisFrame++;
 
 		if (!restrictPrimRanges.empty()) {

GPU/Debugger/GECommandTable.cpp

@@ -275,9 +275,7 @@ static constexpr GECmdInfo geCmdInfo[] = {
 	{ GE_CMD_VTCT, "immt", GECmdFormat::FLOAT },
 	{ GE_CMD_VTCQ, "immq", GECmdFormat::FLOAT },
 	{ GE_CMD_VCV, "immrgb", GECmdFormat::RGB },
-	// TODO: Confirm if any other bits are used?
 	{ GE_CMD_VAP, "imma_prim", GECmdFormat::ALPHA_PRIM },
-	// TODO: Confirm it's 8 bit?
 	{ GE_CMD_VFC, "immfog", GECmdFormat::DATA8 },
 	{ GE_CMD_VSCV, "immrgb1", GECmdFormat::RGB },
 	{ GE_CMD_UNKNOWN_FA, "unknownfa", GECmdFormat::NONE },

GPU/Debugger/GECommandTable.h

@@ -66,7 +66,7 @@ enum class GECmdFormat {
 	BLEND_MODE, // 4 bits srcfactor, 4 bits dstfactor, 3 bits equation.
 	DITHER_ROW, // 4 s.3.0 fixed point dither offsets.
 	LOGIC_OP, // 4 bits logic operation.
-	ALPHA_PRIM, // 8 bits alpha, 3 bits primitive type.
+	ALPHA_PRIM, // 8 bits alpha, 3 bits primitive type, 1 bit antialias, 6 bit clip?, 1 bit shading, 1 bit cullenable, 1 bit cullface, 1 bit tex enable, 1 bit fog, 1 bit dither.
 };
 
 struct GECmdInfo {

GPU/GPUCommon.cpp

@@ -1530,6 +1530,7 @@ void GPUCommon::Execute_End(u32 op, u32 diff) {
 			break;
 
 		default:
+			FlushImm();
 			currentList->subIntrToken = prev & 0xFFFF;
 			UpdateState(GPUSTATE_DONE);
 			// Since we marked done, we have to restore the context now before the next list runs.
@@ -1651,6 +1652,7 @@ void GPUCommon::Execute_Prim(u32 op, u32 diff) {
 	u32 count = data & 0xFFFF;
 	if (count == 0)
 		return;
+	FlushImm();
 
 	// Upper bits are ignored.
 	GEPrimitiveType prim = static_cast<GEPrimitiveType>((data >> 16) & 7);
@@ -2379,40 +2381,60 @@ void GPUCommon::Execute_ImmVertexAlphaPrim(u32 op, u32 diff) {
 		return;
 	}
 
+	int prim = (op >> 8) & 0x7;
+	if (prim != GE_PRIM_KEEP_PREVIOUS) {
+		// Flush before changing the prim type.  Only continue can be used to continue a prim.
+		FlushImm();
+	}
+
 	TransformedVertex &v = immBuffer_[immCount_++];
 
-	// Formula deduced from ThrillVille's clear.
-	int offsetX = gstate.getOffsetX16();
-	int offsetY = gstate.getOffsetY16();
-	v.x = ((gstate.imm_vscx & 0xFFFFFF) - offsetX) / 16.0f;
-	v.y = ((gstate.imm_vscy & 0xFFFFFF) - offsetY) / 16.0f;
+	// ThrillVille does a clear with this, additional parameters found via tests.
+	// The current vtype affects how the coordinate is processed.
+	if (gstate.isModeThrough()) {
+		v.x = ((int)(gstate.imm_vscx & 0xFFFF) - 0x8000) / 16.0f;
+		v.y = ((int)(gstate.imm_vscy & 0xFFFF) - 0x8000) / 16.0f;
+	} else {
+		int offsetX = gstate.getOffsetX16();
+		int offsetY = gstate.getOffsetY16();
+		v.x = ((int)(gstate.imm_vscx & 0xFFFF) - offsetX) / 16.0f;
+		v.y = ((int)(gstate.imm_vscy & 0xFFFF) - offsetY) / 16.0f;
+	}
 	v.z = gstate.imm_vscz & 0xFFFF;
 	v.pos_w = 1.0f;
 	v.u = getFloat24(gstate.imm_vtcs);
 	v.v = getFloat24(gstate.imm_vtct);
 	v.uv_w = getFloat24(gstate.imm_vtcq);
 	v.color0_32 = (gstate.imm_cv & 0xFFFFFF) | (gstate.imm_ap << 24);
-	v.fog = 0.0f; // we have no information about the scale here
+	// TODO: When !gstate.isModeThrough(), direct fog coefficient (0 = entirely fog), ignore fog flag (also GE_IMM_FOG.)
+	v.fog = (gstate.imm_fc & 0xFF) / 255.0f;
+	// TODO: Apply if gstate.isUsingSecondaryColor() && !gstate.isModeThrough(), ignore lighting flag.
 	v.color1_32 = gstate.imm_scv & 0xFFFFFF;
-	int prim = (op >> 8) & 0x7;
 	if (prim != GE_PRIM_KEEP_PREVIOUS) {
 		immPrim_ = (GEPrimitiveType)prim;
-	} else if (prim == GE_PRIM_KEEP_PREVIOUS && immCount_ == 2) {
+		// Flags seem to only be respected from the first prim.
+		immFlags_ = op & 0x00FFF800;
+	} else if (prim == GE_PRIM_KEEP_PREVIOUS && immPrim_ != GE_PRIM_INVALID) {
+		static constexpr int flushPrimCount[] = { 1, 2, 0, 3, 0, 0, 2, 0 };
 		// Instead of finding a proper point to flush, we just emit a full rectangle every time one
 		// is finished.
-		FlushImm();
-		// Need to reset immCount_ here. If we do it in FlushImm it could get skipped by gstate_c.skipDrawReason.
-		immCount_ = 0;
+		if (immCount_ == flushPrimCount[immPrim_ & 7])
+			FlushImm();
 	} else {
 		ERROR_LOG_REPORT_ONCE(imm_draw_prim, G3D, "Immediate draw: Unexpected primitive %d at count %d", prim, immCount_);
 	}
 }
 
 void GPUCommon::FlushImm() {
+	if (immCount_ == 0 || immPrim_ == GE_PRIM_INVALID)
+		return;
+
 	SetDrawType(DRAW_PRIM, immPrim_);
-	framebufferManager_->SetRenderFrameBuffer(gstate_c.IsDirty(DIRTY_FRAMEBUF), gstate_c.skipDrawReason);
+	if (framebufferManager_)
+		framebufferManager_->SetRenderFrameBuffer(gstate_c.IsDirty(DIRTY_FRAMEBUF), gstate_c.skipDrawReason);
 	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB)) {
 		// No idea how many cycles to skip, heh.
+		immCount_ = 0;
 		return;
 	}
 	UpdateUVScaleOffset();
@@ -2422,23 +2444,69 @@ void GPUCommon::FlushImm() {
 	// through vertices.
 	// Since the only known use is Thrillville and it only uses it to clear, we just use color and pos.
 	struct ImmVertex {
+		float uv[2];
 		uint32_t color;
 		float xyz[3];
 	};
 	ImmVertex temp[MAX_IMMBUFFER_SIZE];
+	uint32_t color1Used = 0;
 	for (int i = 0; i < immCount_; i++) {
+		// Since we're sending through, scale back up to w/h.
+		temp[i].uv[0] = immBuffer_[i].u * gstate.getTextureWidth(0);
+		temp[i].uv[1] = immBuffer_[i].v * gstate.getTextureHeight(0);
 		temp[i].color = immBuffer_[i].color0_32;
 		temp[i].xyz[0] = immBuffer_[i].pos[0];
 		temp[i].xyz[1] = immBuffer_[i].pos[1];
 		temp[i].xyz[2] = immBuffer_[i].pos[2];
+		color1Used |= immBuffer_[i].color1_32;
+	}
+	int vtype = GE_VTYPE_TC_FLOAT | GE_VTYPE_POS_FLOAT | GE_VTYPE_COL_8888 | GE_VTYPE_THROUGH;
+
+	// TODO: Handle fog and secondary color somehow?
+
+	bool antialias = (immFlags_ & GE_IMM_ANTIALIAS) != 0;
+	bool prevAntialias = gstate.isAntiAliasEnabled();
+	bool shading = (immFlags_ & GE_IMM_SHADING) != 0;
+	bool prevShading = gstate.getShadeMode() == GE_SHADE_GOURAUD;
+	bool cullEnable = (immFlags_ & GE_IMM_CULLENABLE) != 0;
+	bool prevCullEnable = gstate.isCullEnabled();
+	int cullMode = (immFlags_ & GE_IMM_CULLFACE) != 0 ? 1 : 0;
+	bool texturing = (immFlags_ & GE_IMM_TEXTURE) != 0;
+	bool prevTexturing = gstate.isTextureMapEnabled();
+	bool dither = (immFlags_ & GE_IMM_DITHER) != 0;
+	bool prevDither = gstate.isDitherEnabled();
+
+	if ((immFlags_ & GE_IMM_CLIPMASK) != 0) {
+		WARN_LOG_REPORT_ONCE(geimmclipvalue, G3D, "Imm vertex used clip value, flags=%06x", immFlags_);
+	} else if ((immFlags_ & GE_IMM_FOG) != 0) {
+		WARN_LOG_REPORT_ONCE(geimmfog, G3D, "Imm vertex used fog, flags=%06x", immFlags_);
+	} else if (color1Used != 0 && gstate.isUsingSecondaryColor()) {
+		WARN_LOG_REPORT_ONCE(geimmcolor1, G3D, "Imm vertex used secondary color, flags=%06x", immFlags_);
+	}
+
+	if (texturing != prevTexturing || cullEnable != prevCullEnable || dither != prevDither || prevShading != shading) {
+		DispatchFlush();
+		gstate.antiAliasEnable = (GE_CMD_ANTIALIASENABLE << 24) | (int)antialias;
+		gstate.shademodel = (GE_CMD_SHADEMODE << 24) | (int)shading;
+		gstate.cullfaceEnable = (GE_CMD_CULLFACEENABLE << 24) | (int)cullEnable;
+		gstate.textureMapEnable = (GE_CMD_TEXTUREMAPENABLE << 24) | (int)texturing;
+		gstate.ditherEnable = (GE_CMD_DITHERENABLE << 24) | (int)dither;
+		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE | DIRTY_FRAGMENTSHADER_STATE | DIRTY_RASTER_STATE);
 	}
-	int vtype = GE_VTYPE_POS_FLOAT | GE_VTYPE_COL_8888 | GE_VTYPE_THROUGH;
 
 	int bytesRead;
 	uint32_t vertTypeID = GetVertTypeID(vtype, 0);
-	drawEngineCommon_->DispatchSubmitImm(temp, nullptr, immPrim_, immCount_, vertTypeID, gstate.getCullMode(), &bytesRead);
-	// TOOD: In the future, make a special path for these.
+	drawEngineCommon_->DispatchSubmitImm(temp, nullptr, immPrim_, immCount_, vertTypeID, cullMode, &bytesRead);
+	// TODO: In the future, make a special path for these.
 	// drawEngineCommon_->DispatchSubmitImm(immBuffer_, immCount_);
+	immCount_ = 0;
+
+	gstate.antiAliasEnable = (GE_CMD_ANTIALIASENABLE << 24) | (int)prevAntialias;
+	gstate.shademodel = (GE_CMD_SHADEMODE << 24) | (int)prevShading;
+	gstate.cullfaceEnable = (GE_CMD_CULLFACEENABLE << 24) | (int)prevCullEnable;
+	gstate.textureMapEnable = (GE_CMD_TEXTUREMAPENABLE << 24) | (int)prevTexturing;
+	gstate.ditherEnable = (GE_CMD_DITHERENABLE << 24) | (int)prevDither;
+	gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE | DIRTY_FRAGMENTSHADER_STATE | DIRTY_RASTER_STATE);
 }
 
 void GPUCommon::ExecuteOp(u32 op, u32 diff) {

GPU/GPUCommon.h

@@ -286,6 +286,7 @@ class GPUCommon : public GPUInterface, public GPUDebugInterface {
 	void UpdatePC(u32 currentPC, u32 newPC);
 	void UpdateState(GPURunState state);
 	void FastLoadBoneMatrix(u32 target);
+	void FlushImm();
 
 	// TODO: Unify this.
 	virtual void FinishDeferred() {}
@@ -352,13 +353,13 @@ class GPUCommon : public GPUInterface, public GPUDebugInterface {
 
 	TransformedVertex immBuffer_[MAX_IMMBUFFER_SIZE];
 	int immCount_ = 0;
-	GEPrimitiveType immPrim_;
+	GEPrimitiveType immPrim_ = GE_PRIM_INVALID;
+	uint32_t immFlags_ = 0;
 
 	std::string reportingPrimaryInfo_;
 	std::string reportingFullInfo_;
 
 private:
-	void FlushImm();
 	void CheckDepthUsage(VirtualFramebuffer *vfb);
 	void DoBlockTransfer(u32 skipDrawReason);
 	void DoExecuteCall(u32 target);

GPU/GeDisasm.cpp

@@ -1354,11 +1354,36 @@ void GeDisassembleOp(u32 pc, u32 op, u32 prev, char *buffer, int bufsize) {
 		break;
 
 	case GE_CMD_VAP:
-		snprintf(buffer, bufsize, "Vertex draw: alpha=%02x, prim=%s, other=%06x", data & 0xFF, primTypes[(data >> 8) & 7], data & ~0x0007FF);
+		{
+			bool antialias = (data & GE_IMM_ANTIALIAS) != 0;
+			int clip = (data & GE_IMM_CLIPMASK) >> 12;
+			bool shading = (data & GE_IMM_SHADING) != 0;
+			bool cullEnable = (data & GE_IMM_CULLENABLE) != 0;
+			int cullMode = (data & GE_IMM_CULLFACE) != 0 ? 1 : 0;
+			bool texturing = (data & GE_IMM_TEXTURE) != 0;
+			bool dither = (data & GE_IMM_DITHER) != 0;
+			char *p = buffer;
+			p += snprintf(p, bufsize - (p - buffer), "Vertex draw: alpha=%02x, prim=%s", data & 0xFF, primTypes[(data >> 8) & 7]);
+			if (antialias)
+				p += snprintf(p, bufsize - (p - buffer), ", antialias");
+			if (clip != 0)
+				p += snprintf(p, bufsize - (p - buffer), ", clip=%02x", clip);
+			if (shading)
+				p += snprintf(p, bufsize - (p - buffer), ", shading");
+			if (cullEnable)
+				p += snprintf(p, bufsize - (p - buffer), ", cull=%s", cullMode == 1 ? "back (CCW)" : "front (CW)");
+			if (texturing)
+				p += snprintf(p, bufsize - (p - buffer), ", texturing");
+			if (dither)
+				p += snprintf(p, bufsize - (p - buffer), ", dither");
+		}
 		break;
 
 	case GE_CMD_VFC:
-		snprintf(buffer, bufsize, "Vertex fog: %06x", data);
+		if (data & ~0xFF)
+			snprintf(buffer, bufsize, "Vertex fog: %02x / %f (extra %04x)", data & 0xFF, (data & 0xFF) / 255.0f, data >> 8);
+		else
+			snprintf(buffer, bufsize, "Vertex fog: %02x / %f", data & 0xFF, (data & 0xFF) / 255.0f);
 		break;
 
 	case GE_CMD_VSCV:

GPU/Software/BinManager.cpp

@@ -161,13 +161,13 @@ BinManager::~BinManager() {
 	}
 }
 
-void BinManager::UpdateState() {
+void BinManager::UpdateState(bool throughMode) {
 	PROFILE_THIS_SCOPE("bin_state");
 	if (HasDirty(SoftDirty::PIXEL_ALL | SoftDirty::SAMPLER_ALL | SoftDirty::RAST_ALL)) {
 		if (states_.Full())
 			Flush("states");
 		stateIndex_ = (int)states_.Push(RasterizerState());
-		ComputeRasterizerState(&states_[stateIndex_]);
+		ComputeRasterizerState(&states_[stateIndex_], throughMode);
 		states_[stateIndex_].samplerID.cached.clut = cluts_[clutIndex_].readable;
 
 		ClearDirty(SoftDirty::PIXEL_ALL | SoftDirty::SAMPLER_ALL | SoftDirty::RAST_ALL);

GPU/Software/BinManager.h

@@ -181,7 +181,7 @@ class BinManager {
 	BinManager();
 	~BinManager();
 
-	void UpdateState();
+	void UpdateState(bool throughMode);
 	void UpdateClut(const void *src);
 
 	const Rasterizer::RasterizerState &State() {

GPU/Software/Clipper.cpp

@@ -132,7 +132,7 @@ static inline bool CheckOutsideZ(ClipCoords p, int &pos, int &neg) {
 }
 
 void ProcessRect(const VertexData &v0, const VertexData &v1, BinManager &binner) {
-	if (!gstate.isModeThrough()) {
+	if (!binner.State().throughMode) {
 		// We may discard the entire rect based on depth values.
 		int outsidePos = 0, outsideNeg = 0;
 		CheckOutsideZ(v0.clippos, outsidePos, outsideNeg);
@@ -181,7 +181,7 @@ void ProcessPoint(const VertexData &v0, BinManager &binner) {
 }
 
 void ProcessLine(const VertexData &v0, const VertexData &v1, BinManager &binner) {
-	if (gstate.isModeThrough()) {
+	if (binner.State().throughMode) {
 		// Actually, should clip this one too so we don't need to do bounds checks in the rasterizer.
 		binner.AddLine(v0, v1);
 		return;
@@ -221,7 +221,7 @@ void ProcessLine(const VertexData &v0, const VertexData &v1, BinManager &binner)
 
 void ProcessTriangle(const VertexData &v0, const VertexData &v1, const VertexData &v2, const VertexData &provoking, BinManager &binner) {
 	int mask = 0;
-	if (!gstate.isModeThrough()) {
+	if (!binner.State().throughMode) {
 		mask |= CalcClipMask(v0.clippos);
 		mask |= CalcClipMask(v1.clippos);
 		mask |= CalcClipMask(v2.clippos);

GPU/Software/FuncId.cpp

@@ -48,11 +48,11 @@ static inline PixelBlendFactor OptimizeAlphaFactor(uint32_t color) {
 	return PixelBlendFactor::FIX;
 }
 
-void ComputePixelFuncID(PixelFuncID *id) {
+void ComputePixelFuncID(PixelFuncID *id, bool throughMode) {
 	id->fullKey = 0;
 
 	// TODO: Could this be minz > 0x0000 || maxz < 0xFFFF?  Maybe unsafe, depending on verts...
-	id->applyDepthRange = !gstate.isModeThrough();
+	id->applyDepthRange = !throughMode;
 	// Dither happens even in clear mode.
 	id->dithering = gstate.isDitherEnabled();
 	id->fbFormat = gstate.FrameBufFormat();
@@ -162,7 +162,7 @@ void ComputePixelFuncID(PixelFuncID *id) {
 		}
 
 		id->applyLogicOp = gstate.isLogicOpEnabled() && gstate.getLogicOp() != GE_LOGIC_COPY;
-		id->applyFog = gstate.isFogEnabled() && !gstate.isModeThrough();
+		id->applyFog = gstate.isFogEnabled() && !throughMode;
 	}
 
 	// Cache some values for later convenience.

GPU/Software/FuncId.h

@@ -243,7 +243,7 @@ struct hash<SamplerID> {
 
 };
 
-void ComputePixelFuncID(PixelFuncID *id);
+void ComputePixelFuncID(PixelFuncID *id, bool throughMode);
 std::string DescribePixelFuncID(const PixelFuncID &id);
 
 void ComputeSamplerID(SamplerID *id);

GPU/Software/Rasterizer.cpp

@@ -101,8 +101,8 @@ static inline Vec4<float> Interpolate(const float &c0, const float &c1, const fl
 	return Interpolate(c0, c1, c2, w0.Cast<float>(), w1.Cast<float>(), w2.Cast<float>(), wsum_recip);
 }
 
-void ComputeRasterizerState(RasterizerState *state) {
-	ComputePixelFuncID(&state->pixelID);
+void ComputeRasterizerState(RasterizerState *state, bool throughMode) {
+	ComputePixelFuncID(&state->pixelID, throughMode);
 	state->drawPixel = Rasterizer::GetSingleFunc(state->pixelID);
 
 	state->enableTextures = gstate.isTextureMapEnabled() && !state->pixelID.clearMode;
@@ -140,7 +140,7 @@ void ComputeRasterizerState(RasterizerState *state) {
 	}
 
 	state->shadeGouraud = gstate.getShadeMode() == GE_SHADE_GOURAUD;
-	state->throughMode = gstate.isModeThrough();
+	state->throughMode = throughMode;
 	state->antialiasLines = gstate.isAntiAliasEnabled();
 
 	state->screenOffsetX = gstate.getOffsetX16();

GPU/Software/Rasterizer.h

@@ -67,7 +67,7 @@ struct RasterizerState {
 	}
 };
 
-void ComputeRasterizerState(RasterizerState *state);
+void ComputeRasterizerState(RasterizerState *state, bool throughMode);
 
 // Draws a triangle if its vertices are specified in counter-clockwise order
 void DrawTriangle(const VertexData &v0, const VertexData &v1, const VertexData &v2, const BinCoords &range, const RasterizerState &state);