Merge pull request #6362 from unknownbrackets/gpu-minor

Upload stencil/color on FBO create, clear depth/stencil to 0, fix software transform

GPU/GLES/Framebuffer.cpp

@@ -70,8 +70,7 @@ static const char tex_fs[] =
 	"uniform sampler2D sampler0;\n"
 	"varying vec2 v_texcoord0;\n"
 	"void main() {\n"
-	"  gl_FragColor.rgb = texture2D(sampler0, v_texcoord0).rgb;\n"
-	"  gl_FragColor.a = 1.0;\n"
+	"  gl_FragColor = texture2D(sampler0, v_texcoord0);\n"
 	"}\n";
 
 static const char basic_vs[] =
@@ -176,17 +175,28 @@ void FramebufferManager::ClearBuffer() {
 	glstate.scissorTest.disable();
 	glstate.depthWrite.set(GL_TRUE);
 	glstate.colorMask.set(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
-	glstate.stencilFunc.set(GL_ALWAYS, 0xFF, 0xFF);
-	glClearColor(0.0f, 0.0f, 0.0f, 1.0f);
-	glClearStencil(0xFF);
+	glstate.stencilFunc.set(GL_ALWAYS, 0, 0);
+	glClearColor(0.0f, 0.0f, 0.0f, 0.0f);
+	glClearStencil(0);
 #ifdef USING_GLES2
-	glClearDepthf(1.0f);
+	glClearDepthf(0.0f);
 #else
-	glClearDepth(1.0);
+	glClearDepth(0.0);
 #endif
 	glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT);
 }
 
+void FramebufferManager::ClearDepthBuffer() {
+	glstate.scissorTest.disable();
+	glstate.depthWrite.set(GL_TRUE);
+#ifdef USING_GLES2
+	glClearDepthf(0.0f);
+#else
+	glClearDepth(0.0);
+#endif
+	glClear(GL_DEPTH_BUFFER_BIT);
+}
+
 void FramebufferManager::DisableState() {
 	glstate.blend.disable();
 	glstate.cullFace.disable();
@@ -961,7 +971,6 @@ void FramebufferManager::DoSetRenderFrameBuffer() {
 		}
 
 		INFO_LOG(SCEGE, "Creating FBO for %08x : %i x %i x %i", vfb->fb_address, vfb->width, vfb->height, vfb->format);
-		ClearBuffer();
 
 		textureCache_->NotifyFramebuffer(vfb->fb_address, vfb, NOTIFY_FB_CREATED);
 
@@ -977,6 +986,15 @@ void FramebufferManager::DoSetRenderFrameBuffer() {
 			framebufRangeEnd_ = fb_address_mem + byteSize;
 		}
 
+		if (useBufferedRendering_ && !updateVRAM_) {
+			gpu->PerformMemoryUpload(fb_address_mem, byteSize);
+			gpu->PerformStencilUpload(fb_address_mem, byteSize);
+			// TODO: Is it worth trying to upload the depth buffer?
+			ClearDepthBuffer();
+		} else {
+			ClearBuffer();
+		}
+
 		// Let's check for depth buffer overlap.  Might be interesting.
 		bool sharingReported = false;
 		for (size_t i = 0, end = vfbs_.size(); i < end; ++i) {

GPU/GLES/Framebuffer.h

@@ -251,6 +251,7 @@ private:
 	void EstimateDrawingSize(int &drawing_width, int &drawing_height);
 	static void DisableState();
 	static void ClearBuffer();
+	static void ClearDepthBuffer();
 	static bool MaskedEqual(u32 addr1, u32 addr2);
 
 	void SetColorUpdated(VirtualFramebuffer *dstBuffer) {

GPU/GLES/SoftwareTransform.cpp

@@ -203,34 +203,40 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 			// Scale UV?
 		} else {
 			// We do software T&L for now
-			float out[3], norm[3];
-			float pos[3], nrm[3];
+			float out[3];
+			float pos[3];
 			Vec3f normal(0, 0, 1);
+			Vec3f worldnormal(0, 0, 1);
 			reader.ReadPos(pos);
-			if (reader.hasNormal())
-				reader.ReadNrm(nrm);
 
 			if (!skinningEnabled) {
 				Vec3ByMatrix43(out, pos, gstate.worldMatrix);
 				if (reader.hasNormal()) {
-					Norm3ByMatrix43(norm, nrm, gstate.worldMatrix);
-					normal = Vec3f(norm).Normalized();
+					reader.ReadNrm(normal.AsArray());
+					if (gstate.areNormalsReversed()) {
+						normal = -normal;
+					}
+					Norm3ByMatrix43(worldnormal.AsArray(), normal.AsArray(), gstate.worldMatrix);
+					worldnormal = worldnormal.Normalized();
 				}
 			} else {
 				float weights[8];
 				reader.ReadWeights(weights);
+				if (reader.hasNormal())
+					reader.ReadNrm(normal.AsArray());
+
 				// Skinning
-				Vec3f psum(0,0,0);
-				Vec3f nsum(0,0,0);
+				Vec3f psum(0, 0, 0);
+				Vec3f nsum(0, 0, 0);
 				for (int i = 0; i < vertTypeGetNumBoneWeights(vertType); i++) {
 					if (weights[i] != 0.0f) {
 						Vec3ByMatrix43(out, pos, gstate.boneMatrix+i*12);
 						Vec3f tpos(out);
 						psum += tpos * weights[i];
 						if (reader.hasNormal()) {
-							Norm3ByMatrix43(norm, nrm, gstate.boneMatrix+i*12);
-							Vec3f tnorm(norm);
-							nsum += tnorm * weights[i];
+							Vec3f norm;
+							Norm3ByMatrix43(norm.AsArray(), normal.AsArray(), gstate.boneMatrix+i*12);
+							nsum += norm * weights[i];
 						}
 					}
 				}
@@ -238,8 +244,12 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 				// Yes, we really must multiply by the world matrix too.
 				Vec3ByMatrix43(out, psum.AsArray(), gstate.worldMatrix);
 				if (reader.hasNormal()) {
-					Norm3ByMatrix43(norm, nsum.AsArray(), gstate.worldMatrix);
-					normal = Vec3f(norm).Normalized();
+					normal = nsum;
+					if (gstate.areNormalsReversed()) {
+						normal = -normal;
+					}
+					Norm3ByMatrix43(worldnormal.AsArray(), normal.AsArray(), gstate.worldMatrix);
+					worldnormal = worldnormal.Normalized();
 				}
 			}
 
@@ -254,7 +264,7 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 			if (gstate.isLightingEnabled()) {
 				float litColor0[4];
 				float litColor1[4];
-				lighter.Light(litColor0, litColor1, unlitColor.AsArray(), out, normal);
+				lighter.Light(litColor0, litColor1, unlitColor.AsArray(), out, worldnormal);
 
 				// Don't ignore gstate.lmode - we should send two colors in that case
 				for (int j = 0; j < 4; j++) {
@@ -317,20 +327,16 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 						break;
 
 					case GE_PROJMAP_NORMALIZED_NORMAL: // Use normalized normal as source
-						if (reader.hasNormal()) {
-							source = Vec3f(norm).Normalized();
-						} else {
+						source = normal.Normalized();
+						if (!reader.hasNormal()) {
 							ERROR_LOG_REPORT(G3D, "Normal projection mapping without normal?");
-							source = Vec3f(0.0f, 0.0f, 1.0f);
 						}
 						break;
 
 					case GE_PROJMAP_NORMAL: // Use non-normalized normal as source!
-						if (reader.hasNormal()) {
-							source = Vec3f(norm);
-						} else {
+						source = normal;
+						if (!reader.hasNormal()) {
 							ERROR_LOG_REPORT(G3D, "Normal projection mapping without normal?");
-							source = Vec3f(0.0f, 0.0f, 1.0f);
 						}
 						break;
 					}
@@ -349,8 +355,8 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 					Vec3f lightpos0 = Vec3f(&lighter.lpos[gstate.getUVLS0() * 3]).Normalized();
 					Vec3f lightpos1 = Vec3f(&lighter.lpos[gstate.getUVLS1() * 3]).Normalized();
 
-					uv[0] = (1.0f + Dot(lightpos0, normal))/2.0f;
-					uv[1] = (1.0f + Dot(lightpos1, normal))/2.0f;
+					uv[0] = (1.0f + Dot(lightpos0, worldnormal))/2.0f;
+					uv[1] = (1.0f + Dot(lightpos1, worldnormal))/2.0f;
 					uv[2] = 1.0f;
 				}
 				break;

GPU/GLES/StencilBuffer.cpp

@@ -88,6 +88,7 @@ bool FramebufferManager::NotifyStencilUpload(u32 addr, int size) {
 
 	shaderManager_->DirtyLastShader();
 
+	// TODO: Check if it's all the same value, commonly 0?
 	MakePixelTexture(Memory::GetPointer(addr), dstBuffer->format, dstBuffer->fb_stride, dstBuffer->bufferWidth, dstBuffer->bufferHeight);
 	DisableState();
 	glstate.colorMask.set(GL_FALSE, GL_FALSE, GL_FALSE, GL_TRUE);