// Copyright (c) 2012- PPSSPP Project.

// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0 or later versions.

// 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 2.0 for more details.

// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/

// Official git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.

#include "base/logging.h"
#include "profiler/profiler.h"
#include "i18n/i18n.h"

#include "Common/ChunkFile.h"
#include "Common/GraphicsContext.h"

#include "Core/Config.h"
#include "Core/Debugger/Breakpoints.h"
#include "Core/MemMapHelpers.h"
#include "Core/Host.h"
#include "Core/Config.h"
#include "Core/Reporting.h"
#include "Core/System.h"
#include "Core/ELF/ParamSFO.h"

#include "GPU/GPUState.h"
#include "GPU/ge_constants.h"
#include "GPU/GeDisasm.h"
#include "GPU/Common/FramebufferCommon.h"

#include "ext/native/gfx/GLStateCache.h"
#include "GPU/GLES/ShaderManagerGLES.h"
#include "GPU/GLES/GPU_GLES.h"
#include "GPU/GLES/FramebufferManagerGLES.h"
#include "GPU/GLES/DrawEngineGLES.h"
#include "GPU/GLES/TextureCacheGLES.h"

#include "Core/MIPS/MIPS.h"
#include "Core/HLE/sceKernelThread.h"
#include "Core/HLE/sceKernelInterrupt.h"
#include "Core/HLE/sceGe.h"

#ifdef _WIN32
#include "Windows/GPU/WindowsGLContext.h"
#endif

struct GLESCommandTableEntry {
	uint8_t cmd;
	uint8_t flags;
	uint64_t dirty;
	GPU_GLES::CmdFunc func;
};

// This table gets crunched into a faster form by init.
// TODO: Share this table between the backends. Will have to make another indirection for the function pointers though..
static const GLESCommandTableEntry commandTable[] = {
	// Changes that dirty the current texture.
	{ GE_CMD_TEXSIZE0, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTE, DIRTY_UVSCALEOFFSET, &GPU_GLES::Execute_TexSize0 },

	{ GE_CMD_STENCILTEST, FLAG_FLUSHBEFOREONCHANGE, DIRTY_STENCILREPLACEVALUE | DIRTY_BLEND_STATE | DIRTY_DEPTHSTENCIL_STATE },

	// Changing the vertex type requires us to flush.
	{ GE_CMD_VERTEXTYPE, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTEONCHANGE, 0, &GPU_GLES::Execute_VertexType },

	{ GE_CMD_PRIM, FLAG_EXECUTE, 0, &GPU_GLES::Execute_Prim },
	{ GE_CMD_BEZIER, FLAG_FLUSHBEFORE | FLAG_EXECUTE, 0, &GPU_GLES::Execute_Bezier },
	{ GE_CMD_SPLINE, FLAG_FLUSHBEFORE | FLAG_EXECUTE, 0, &GPU_GLES::Execute_Spline },

	// Changes that trigger data copies. Only flushing on change for LOADCLUT must be a bit of a hack...
	{ GE_CMD_LOADCLUT, FLAG_FLUSHBEFOREONCHANGE | FLAG_EXECUTE, 0, &GPU_GLES::Execute_LoadClut },
};

GPU_GLES::CommandInfo GPU_GLES::cmdInfo_[256];

GPU_GLES::GPU_GLES(GraphicsContext *gfxCtx, Draw::DrawContext *draw)
: GPUCommon(gfxCtx, draw) {
	UpdateVsyncInterval(true);
	CheckGPUFeatures();

	shaderManagerGL_ = new ShaderManagerGLES();
	framebufferManagerGL_ = new FramebufferManagerGLES(draw);
	framebufferManager_ = framebufferManagerGL_;
	textureCacheGL_ = new TextureCacheGLES(draw);
	textureCache_ = textureCacheGL_;
	drawEngineCommon_ = &drawEngine_;
	shaderManager_ = shaderManagerGL_;
	drawEngineCommon_ = &drawEngine_;

	drawEngine_.SetShaderManager(shaderManagerGL_);
	drawEngine_.SetTextureCache(textureCacheGL_);
	drawEngine_.SetFramebufferManager(framebufferManagerGL_);
	drawEngine_.SetFragmentTestCache(&fragmentTestCache_);
	framebufferManagerGL_->Init();
	framebufferManagerGL_->SetTextureCache(textureCacheGL_);
	framebufferManagerGL_->SetShaderManager(shaderManagerGL_);
	framebufferManagerGL_->SetDrawEngine(&drawEngine_);
	textureCacheGL_->SetFramebufferManager(framebufferManagerGL_);
	textureCacheGL_->SetDepalShaderCache(&depalShaderCache_);
	textureCacheGL_->SetShaderManager(shaderManagerGL_);
	textureCacheGL_->SetDrawEngine(&drawEngine_);
	fragmentTestCache_.SetTextureCache(textureCacheGL_);

	// Sanity check gstate
	if ((int *)&gstate.transferstart - (int *)&gstate != 0xEA) {
		ERROR_LOG(G3D, "gstate has drifted out of sync!");
	}

	memset(cmdInfo_, 0, sizeof(cmdInfo_));

	// Import both the global and local command tables, and check for dupes
	std::set<u8> dupeCheck;
	for (size_t i = 0; i < commonCommandTableSize; i++) {
		const u8 cmd = commonCommandTable[i].cmd;
		if (dupeCheck.find(cmd) != dupeCheck.end()) {
			ERROR_LOG(G3D, "Command table Dupe: %02x (%i)", (int)cmd, (int)cmd);
		} else {
			dupeCheck.insert(cmd);
		}
		cmdInfo_[cmd].flags |= (uint64_t)commonCommandTable[i].flags | (commonCommandTable[i].dirty << 8);
		cmdInfo_[cmd].func = commonCommandTable[i].func;
		if ((cmdInfo_[cmd].flags & (FLAG_EXECUTE | FLAG_EXECUTEONCHANGE)) && !cmdInfo_[cmd].func) {
			Crash();
		}
	}

	for (size_t i = 0; i < ARRAY_SIZE(commandTable); i++) {
		const u8 cmd = commandTable[i].cmd;
		if (dupeCheck.find(cmd) != dupeCheck.end()) {
			ERROR_LOG(G3D, "Command table Dupe: %02x (%i)", (int)cmd, (int)cmd);
		} else {
			dupeCheck.insert(cmd);
		}
		cmdInfo_[cmd].flags |= (uint64_t)commandTable[i].flags | (commandTable[i].dirty << 8);
		cmdInfo_[cmd].func = commandTable[i].func;
		if ((cmdInfo_[cmd].flags & (FLAG_EXECUTE | FLAG_EXECUTEONCHANGE)) && !cmdInfo_[cmd].func) {
			Crash();
		}
	}
	// Find commands missing from the table.
	for (int i = 0; i < 0xEF; i++) {
		if (dupeCheck.find((u8)i) == dupeCheck.end()) {
			ERROR_LOG(G3D, "Command missing from table: %02x (%i)", i, i);
		}
	}

	// No need to flush before the tex scale/offset commands if we are baking
	// the tex scale/offset into the vertices anyway.

	UpdateCmdInfo();

	BuildReportingInfo();
	// Update again after init to be sure of any silly driver problems.
	UpdateVsyncInterval(true);

	// Some of our defaults are different from hw defaults, let's assert them.
	// We restore each frame anyway, but here is convenient for tests.
	glstate.Restore();
	drawEngine_.RestoreVAO();
	textureCacheGL_->NotifyConfigChanged();

	// Load shader cache.
	std::string discID = g_paramSFO.GetDiscID();
	if (discID.size()) {
		File::CreateFullPath(GetSysDirectory(DIRECTORY_APP_CACHE));
		shaderCachePath_ = GetSysDirectory(DIRECTORY_APP_CACHE) + "/" + discID + ".glshadercache";
		shaderManagerGL_->LoadAndPrecompile(shaderCachePath_);
	}

	if (g_Config.bHardwareTessellation) {
		// Disable hardware tessellation if device is unsupported.
		if (!gstate_c.SupportsAll(GPU_SUPPORTS_INSTANCE_RENDERING | GPU_SUPPORTS_VERTEX_TEXTURE_FETCH | GPU_SUPPORTS_TEXTURE_FLOAT)) {
			// TODO: Check unsupported device name list.(Above gpu features are supported but it has issues with weak gpu, memory, shader compiler etc...)
			g_Config.bHardwareTessellation = false;
			ERROR_LOG(G3D, "Hardware Tessellation is unsupported, falling back to software tessellation");
			I18NCategory *gr = GetI18NCategory("Graphics");
			host->NotifyUserMessage(gr->T("Turn off Hardware Tessellation - unsupported"), 2.5f, 0xFF3030FF);
		}
	}
}

GPU_GLES::~GPU_GLES() {
	framebufferManagerGL_->DestroyAllFBOs();
	shaderManagerGL_->ClearCache(true);
	depalShaderCache_.Clear();
	fragmentTestCache_.Clear();
	if (!shaderCachePath_.empty()) {
		shaderManagerGL_->Save(shaderCachePath_);
	}
	delete shaderManagerGL_;
	shaderManagerGL_ = nullptr;
	delete framebufferManagerGL_;
	delete textureCacheGL_;
#ifdef _WIN32
	gfxCtx_->SwapInterval(0);
#endif
}

// Take the raw GL extension and versioning data and turn into feature flags.
void GPU_GLES::CheckGPUFeatures() {
	u32 features = 0;

	features |= GPU_SUPPORTS_16BIT_FORMATS;

	if (gl_extensions.ARB_blend_func_extended || gl_extensions.EXT_blend_func_extended) {
		if (gl_extensions.gpuVendor == GPU_VENDOR_INTEL || !gl_extensions.VersionGEThan(3, 0, 0)) {
			// Don't use this extension to off on sub 3.0 OpenGL versions as it does not seem reliable
			// Also on Intel, see https://github.com/hrydgard/ppsspp/issues/4867
		} else {
#ifdef __ANDROID__
			// This appears to be broken on nVidia Shield TV.
			if (gl_extensions.gpuVendor != GPU_VENDOR_NVIDIA) {
				features |= GPU_SUPPORTS_DUALSOURCE_BLEND;
			}
#else
			features |= GPU_SUPPORTS_DUALSOURCE_BLEND;
#endif
		}
	}

	if (gl_extensions.IsGLES) {
		if (gl_extensions.GLES3)
			features |= GPU_SUPPORTS_GLSL_ES_300;
	} else {
		if (gl_extensions.VersionGEThan(3, 3, 0))
			features |= GPU_SUPPORTS_GLSL_330;
	}

	if (gl_extensions.EXT_shader_framebuffer_fetch || gl_extensions.NV_shader_framebuffer_fetch || gl_extensions.ARM_shader_framebuffer_fetch) {
		// This has caused problems in the past.  Let's only enable on GLES3.
		if (features & GPU_SUPPORTS_GLSL_ES_300) {
			features |= GPU_SUPPORTS_ANY_FRAMEBUFFER_FETCH;
		}
	}
	
	if (gl_extensions.ARB_framebuffer_object || gl_extensions.EXT_framebuffer_object || gl_extensions.IsGLES) {
		features |= GPU_SUPPORTS_FBO;
	}
	if (gl_extensions.ARB_framebuffer_object || gl_extensions.GLES3) {
		features |= GPU_SUPPORTS_ARB_FRAMEBUFFER_BLIT;
	}
	if (gl_extensions.NV_framebuffer_blit) {
		features |= GPU_SUPPORTS_NV_FRAMEBUFFER_BLIT;
	}
	if (gl_extensions.ARB_vertex_array_object && gl_extensions.IsCoreContext) {
		features |= GPU_SUPPORTS_VAO;
	}

	bool useCPU = false;
	if (!gl_extensions.IsGLES) {
		// Urrgh, we don't even define FB_READFBOMEMORY_CPU on mobile
#ifndef USING_GLES2
		useCPU = g_Config.iRenderingMode == FB_READFBOMEMORY_CPU;
#endif
		// Some cards or drivers seem to always dither when downloading a framebuffer to 16-bit.
		// This causes glitches in games that expect the exact values.
		// It has not been experienced on NVIDIA cards, so those are left using the GPU (which is faster.)
		if (g_Config.iRenderingMode == FB_BUFFERED_MODE) {
			if (gl_extensions.gpuVendor != GPU_VENDOR_NVIDIA || gl_extensions.ver[0] < 3) {
				useCPU = true;
			}
		}
	} else {
		useCPU = true;
	}

	if (useCPU)
		features |= GPU_PREFER_CPU_DOWNLOAD;

	if ((gl_extensions.gpuVendor == GPU_VENDOR_NVIDIA) || (gl_extensions.gpuVendor == GPU_VENDOR_AMD))
		features |= GPU_PREFER_REVERSE_COLOR_ORDER;

	if (gl_extensions.OES_texture_npot)
		features |= GPU_SUPPORTS_OES_TEXTURE_NPOT;

	if (gl_extensions.EXT_unpack_subimage)
		features |= GPU_SUPPORTS_UNPACK_SUBIMAGE;

	if (gl_extensions.EXT_blend_minmax)
		features |= GPU_SUPPORTS_BLEND_MINMAX;

	if (gl_extensions.OES_copy_image || gl_extensions.NV_copy_image || gl_extensions.EXT_copy_image || gl_extensions.ARB_copy_image)
		features |= GPU_SUPPORTS_ANY_COPY_IMAGE;

	if (!gl_extensions.IsGLES)
		features |= GPU_SUPPORTS_LOGIC_OP;

	if (gl_extensions.GLES3 || !gl_extensions.IsGLES)
		features |= GPU_SUPPORTS_TEXTURE_LOD_CONTROL;

	if (gl_extensions.EXT_texture_filter_anisotropic)
		features |= GPU_SUPPORTS_ANISOTROPY;

	bool canUseInstanceID = gl_extensions.EXT_draw_instanced || gl_extensions.ARB_draw_instanced;
	bool canDefInstanceID = gl_extensions.IsGLES || gl_extensions.EXT_gpu_shader4 || gl_extensions.VersionGEThan(3, 1);
	bool instanceRendering = gl_extensions.GLES3 || (canUseInstanceID && canDefInstanceID);
		features |= GPU_SUPPORTS_INSTANCE_RENDERING;

	int maxVertexTextureImageUnits;
	glGetIntegerv(GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS, &maxVertexTextureImageUnits);
	if (maxVertexTextureImageUnits >= 3) // At least 3 for hardware tessellation
		features |= GPU_SUPPORTS_VERTEX_TEXTURE_FETCH;

	if (gl_extensions.ARB_texture_float || gl_extensions.OES_texture_float)
		features |= GPU_SUPPORTS_TEXTURE_FLOAT;

	// If we already have a 16-bit depth buffer, we don't need to round.
	bool prefer24 = draw_->GetDeviceCaps().preferredDepthBufferFormat == Draw::DataFormat::D24_S8;
	if (prefer24) {
		if (!g_Config.bHighQualityDepth && (features & GPU_SUPPORTS_ACCURATE_DEPTH) != 0) {
			features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
		} else if (PSP_CoreParameter().compat.flags().PixelDepthRounding) {
			if (!gl_extensions.IsGLES || gl_extensions.GLES3) {
				// Use fragment rounding on desktop and GLES3, most accurate.
				features |= GPU_ROUND_FRAGMENT_DEPTH_TO_16BIT;
			} else if (prefer24 && (features & GPU_SUPPORTS_ACCURATE_DEPTH) != 0) {
				// Here we can simulate a 16 bit depth buffer by scaling.
				// Note that the depth buffer is fixed point, not floating, so dividing by 256 is pretty good.
				features |= GPU_SCALE_DEPTH_FROM_24BIT_TO_16BIT;
			} else {
				// At least do vertex rounding if nothing else.
				features |= GPU_ROUND_DEPTH_TO_16BIT;
			}
		} else if (PSP_CoreParameter().compat.flags().VertexDepthRounding) {
			features |= GPU_ROUND_DEPTH_TO_16BIT;
		}
	}

	// The Phantasy Star hack :(
	if (PSP_CoreParameter().compat.flags().DepthRangeHack && (features & GPU_SUPPORTS_ACCURATE_DEPTH) == 0) {
		features |= GPU_USE_DEPTH_RANGE_HACK;
	}

	if (PSP_CoreParameter().compat.flags().ClearToRAM) {
		features |= GPU_USE_CLEAR_RAM_HACK;
	}

#ifdef MOBILE_DEVICE
	// Arguably, we should turn off GPU_IS_MOBILE on like modern Tegras, etc.
	features |= GPU_IS_MOBILE;
#endif

	gstate_c.featureFlags = features;
}

// Let's avoid passing nulls into snprintf().
static const char *GetGLStringAlways(GLenum name) {
	const GLubyte *value = glGetString(name);
	if (!value)
		return "?";
	return (const char *)value;
}

// Needs to be called on GPU thread, not reporting thread.
void GPU_GLES::BuildReportingInfo() {
	const char *glVendor = GetGLStringAlways(GL_VENDOR);
	const char *glRenderer = GetGLStringAlways(GL_RENDERER);
	const char *glVersion = GetGLStringAlways(GL_VERSION);
	const char *glSlVersion = GetGLStringAlways(GL_SHADING_LANGUAGE_VERSION);
	const char *glExtensions = nullptr;

	if (gl_extensions.VersionGEThan(3, 0)) {
		glExtensions = g_all_gl_extensions.c_str();
	} else {
		glExtensions = GetGLStringAlways(GL_EXTENSIONS);
	}

	char temp[16384];
	snprintf(temp, sizeof(temp), "%s (%s %s), %s (extensions: %s)", glVersion, glVendor, glRenderer, glSlVersion, glExtensions);
	reportingPrimaryInfo_ = glVendor;
	reportingFullInfo_ = temp;

	Reporting::UpdateConfig();
}

void GPU_GLES::DeviceLost() {
	ILOG("GPU_GLES: DeviceLost");
	// Should only be executed on the GL thread.

	// Simply drop all caches and textures.
	// FBOs appear to survive? Or no?
	// TransformDraw has registered as a GfxResourceHolder.
	shaderManagerGL_->ClearCache(false);
	textureCacheGL_->Clear(false);
	fragmentTestCache_.Clear(false);
	depalShaderCache_.Clear();
	framebufferManagerGL_->DeviceLost();
}

void GPU_GLES::DeviceRestore() {
	ILOG("GPU_GLES: DeviceRestore");

	UpdateCmdInfo();
	UpdateVsyncInterval(true);
}

void GPU_GLES::Reinitialize() {
	GPUCommon::Reinitialize();
	textureCacheGL_->Clear(true);
	depalShaderCache_.Clear();
	framebufferManagerGL_->DestroyAllFBOs();
}

void GPU_GLES::InitClear() {
	bool useNonBufferedRendering = g_Config.iRenderingMode == FB_NON_BUFFERED_MODE;
	if (useNonBufferedRendering) {
		glstate.depthWrite.set(GL_TRUE);
		glstate.colorMask.set(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
		glClearColor(0,0,0,1);
		glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT);
	}
	glstate.viewport.set(0, 0, PSP_CoreParameter().pixelWidth, PSP_CoreParameter().pixelHeight);
}

void GPU_GLES::BeginHostFrame() {
	GPUCommon::BeginHostFrame();
	UpdateCmdInfo();
	if (resized_) {
		CheckGPUFeatures();
		framebufferManager_->Resized();
		drawEngine_.Resized();
		shaderManagerGL_->DirtyShader();
		textureCacheGL_->NotifyConfigChanged();
	}
}

inline void GPU_GLES::UpdateVsyncInterval(bool force) {
#ifdef _WIN32
	int desiredVSyncInterval = g_Config.bVSync ? 1 : 0;
	if (PSP_CoreParameter().unthrottle) {
		desiredVSyncInterval = 0;
	}
	if (PSP_CoreParameter().fpsLimit == 1) {
		// For an alternative speed that is a clean factor of 60, the user probably still wants vsync.
		if (g_Config.iFpsLimit == 0 || (g_Config.iFpsLimit != 15 && g_Config.iFpsLimit != 30 && g_Config.iFpsLimit != 60)) {
			desiredVSyncInterval = 0;
		}
	}

	if (desiredVSyncInterval != lastVsync_ || force) {
		// Disabled EXT_swap_control_tear for now, it never seems to settle at the correct timing
		// so it just keeps tearing. Not what I hoped for...
		//if (gl_extensions.EXT_swap_control_tear) {
		//	// See http://developer.download.nvidia.com/opengl/specs/WGL_EXT_swap_control_tear.txt
		//	glstate.SetVSyncInterval(-desiredVSyncInterval);
		//} else {
		gfxCtx_->SwapInterval(desiredVSyncInterval);
		//}
		lastVsync_ = desiredVSyncInterval;
	}
#endif
}

void GPU_GLES::UpdateCmdInfo() {
	if (g_Config.bSoftwareSkinning) {
		cmdInfo_[GE_CMD_VERTEXTYPE].flags &= ~FLAG_FLUSHBEFOREONCHANGE;
		cmdInfo_[GE_CMD_VERTEXTYPE].func = &GPU_GLES::Execute_VertexTypeSkinning;
	} else {
		cmdInfo_[GE_CMD_VERTEXTYPE].flags |= FLAG_FLUSHBEFOREONCHANGE;
		cmdInfo_[GE_CMD_VERTEXTYPE].func = &GPU_GLES::Execute_VertexType;
	}
}

void GPU_GLES::ReapplyGfxState() {
	drawEngine_.RestoreVAO();
	glstate.Restore();
	GPUCommon::ReapplyGfxState();
}

void GPU_GLES::BeginFrame() {
	UpdateVsyncInterval(resized_);
	resized_ = false;

	textureCacheGL_->StartFrame();
	drawEngine_.DecimateTrackedVertexArrays();
	drawEngine_.DecimateBuffers();
	depalShaderCache_.Decimate();
	fragmentTestCache_.Decimate();

	GPUCommon::BeginFrame();

	// Save the cache from time to time. TODO: How often?
	if (!shaderCachePath_.empty() && (gpuStats.numFlips & 1023) == 0) {
		shaderManagerGL_->Save(shaderCachePath_);
	}

	shaderManagerGL_->DirtyShader();

	// Not sure if this is really needed.
	gstate_c.Dirty(DIRTY_ALL_UNIFORMS);

	framebufferManagerGL_->BeginFrame();
}

void GPU_GLES::SetDisplayFramebuffer(u32 framebuf, u32 stride, GEBufferFormat format) {
	host->GPUNotifyDisplay(framebuf, stride, format);
	framebufferManagerGL_->SetDisplayFramebuffer(framebuf, stride, format);
}

bool GPU_GLES::FramebufferDirty() {
	VirtualFramebuffer *vfb = framebufferManagerGL_->GetDisplayVFB();
	if (vfb) {
		bool dirty = vfb->dirtyAfterDisplay;
		vfb->dirtyAfterDisplay = false;
		return dirty;
	}
	return true;
}

bool GPU_GLES::FramebufferReallyDirty() {
	VirtualFramebuffer *vfb = framebufferManagerGL_->GetDisplayVFB();
	if (vfb) {
		bool dirty = vfb->reallyDirtyAfterDisplay;
		vfb->reallyDirtyAfterDisplay = false;
		return dirty;
	}
	return true;
}

void GPU_GLES::CopyDisplayToOutput() {
	// Flush anything left over.
	framebufferManagerGL_->RebindFramebuffer();
	drawEngine_.Flush();

	shaderManagerGL_->DirtyLastShader();

	glstate.depthWrite.set(GL_TRUE);
	glstate.colorMask.set(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);

	framebufferManagerGL_->CopyDisplayToOutput();
	framebufferManagerGL_->EndFrame();

	// If buffered, discard the depth buffer of the backbuffer. Don't even know if we need one.
#if 0
#ifdef USING_GLES2
	if (gl_extensions.EXT_discard_framebuffer && g_Config.iRenderingMode != 0) {
		GLenum attachments[] = {GL_DEPTH_EXT, GL_STENCIL_EXT};
		glDiscardFramebufferEXT(GL_FRAMEBUFFER, 2, attachments);
	}
#endif
#endif
}

// Maybe should write this in ASM...
void GPU_GLES::FastRunLoop(DisplayList &list) {
	PROFILE_THIS_SCOPE("gpuloop");
	const CommandInfo *cmdInfo = cmdInfo_;
	int dc = downcount;
	for (; dc > 0; --dc) {
		// We know that display list PCs have the upper nibble == 0 - no need to mask the pointer
		const u32 op = *(const u32 *)(Memory::base + list.pc);
		const u32 cmd = op >> 24;
		const CommandInfo &info = cmdInfo[cmd];
		const u32 diff = op ^ gstate.cmdmem[cmd];
		if (diff == 0) {
			if (info.flags & FLAG_EXECUTE) {
				downcount = dc;
				(this->*info.func)(op, diff);
				dc = downcount;
			}
		} else {
			uint64_t flags = info.flags;
			if (flags & FLAG_FLUSHBEFOREONCHANGE) {
				drawEngine_.Flush();
			}
			gstate.cmdmem[cmd] = op;
			if (flags & (FLAG_EXECUTE | FLAG_EXECUTEONCHANGE)) {
				downcount = dc;
				(this->*info.func)(op, diff);
				dc = downcount;
			} else {
				uint64_t dirty = flags >> 8;
				if (dirty)
					gstate_c.Dirty(dirty);
			}
		}
		list.pc += 4;
	}
	downcount = 0;
}

void GPU_GLES::FinishDeferred() {
	// This finishes reading any vertex data that is pending.
	drawEngine_.FinishDeferred();
}

inline void GPU_GLES::CheckFlushOp(int cmd, u32 diff) {
	const u8 cmdFlags = cmdInfo_[cmd].flags;
	if (diff && (cmdFlags & FLAG_FLUSHBEFOREONCHANGE)) {
		if (dumpThisFrame_) {
			NOTICE_LOG(G3D, "================ FLUSH ================");
		}
		drawEngine_.Flush();
	}
}

void GPU_GLES::PreExecuteOp(u32 op, u32 diff) {
	CheckFlushOp(op >> 24, diff);
}

void GPU_GLES::ExecuteOp(u32 op, u32 diff) {
	const u8 cmd = op >> 24;
	const CommandInfo info = cmdInfo_[cmd];
	const u8 cmdFlags = info.flags;
	if ((cmdFlags & FLAG_EXECUTE) || (diff && (cmdFlags & FLAG_EXECUTEONCHANGE))) {
		(this->*info.func)(op, diff);
	} else if (diff) {
		uint64_t dirty = info.flags >> 8;
		if (dirty)
			gstate_c.Dirty(dirty);
	}
}

void GPU_GLES::Execute_Prim(u32 op, u32 diff) {
	// This drives all drawing. All other state we just buffer up, then we apply it only
	// when it's time to draw. As most PSP games set state redundantly ALL THE TIME, this is a huge optimization.

	u32 data = op & 0xFFFFFF;
	u32 count = data & 0xFFFF;
	if (count == 0)
		return;

	// Upper bits are ignored.
	GEPrimitiveType prim = static_cast<GEPrimitiveType>((data >> 16) & 7);
	SetDrawType(DRAW_PRIM, prim);

	// Discard AA lines as we can't do anything that makes sense with these anyway. The SW plugin might, though.

	if (gstate.isAntiAliasEnabled()) {
		// Discard AA lines in DOA
		if (prim == GE_PRIM_LINE_STRIP)
			return;
		// Discard AA lines in Summon Night 5
		if ((prim == GE_PRIM_LINES) && gstate.isSkinningEnabled())
			return;
	}

	// This also makes skipping drawing very effective. This function can change the framebuffer.
	framebufferManagerGL_->SetRenderFrameBuffer(gstate_c.IsDirty(DIRTY_FRAMEBUF), gstate_c.skipDrawReason);
	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB))	{
		drawEngine_.SetupVertexDecoder(gstate.vertType);
		// Rough estimate, not sure what's correct.
		cyclesExecuted += EstimatePerVertexCost() * count;
		return;
	}

	if (!Memory::IsValidAddress(gstate_c.vertexAddr)) {
		ERROR_LOG_REPORT(G3D, "Bad vertex address %08x!", gstate_c.vertexAddr);
		return;
	}

	void *verts = Memory::GetPointerUnchecked(gstate_c.vertexAddr);
	void *inds = 0;
	if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
		if (!Memory::IsValidAddress(gstate_c.indexAddr)) {
			ERROR_LOG_REPORT(G3D, "Bad index address %08x!", gstate_c.indexAddr);
			return;
		}
		inds = Memory::GetPointerUnchecked(gstate_c.indexAddr);
	}

#ifndef MOBILE_DEVICE
	if (prim > GE_PRIM_RECTANGLES) {
		ERROR_LOG_REPORT_ONCE(reportPrim, G3D, "Unexpected prim type: %d", prim);
	}
#endif

	if (gstate_c.dirty & DIRTY_VERTEXSHADER_STATE) {
		vertexCost_ = EstimatePerVertexCost();
	}
	gpuStats.vertexGPUCycles += vertexCost_ * count;
	cyclesExecuted += vertexCost_* count;

	int bytesRead = 0;
	UpdateUVScaleOffset();
	drawEngine_.SubmitPrim(verts, inds, prim, count, gstate.vertType, &bytesRead);

	// After drawing, we advance the vertexAddr (when non indexed) or indexAddr (when indexed).
	// Some games rely on this, they don't bother reloading VADDR and IADDR.
	// The VADDR/IADDR registers are NOT updated.
	AdvanceVerts(gstate.vertType, count, bytesRead);
}

void GPU_GLES::Execute_VertexType(u32 op, u32 diff) {
	if (diff)
		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE);
	if (diff & (GE_VTYPE_TC_MASK | GE_VTYPE_THROUGH_MASK)) {
		gstate_c.Dirty(DIRTY_UVSCALEOFFSET);
		if (diff & GE_VTYPE_THROUGH_MASK)
			gstate_c.Dirty(DIRTY_RASTER_STATE | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_FRAGMENTSHADER_STATE);
	}
}

void GPU_GLES::Execute_VertexTypeSkinning(u32 op, u32 diff) {
	// Don't flush when weight count changes, unless morph is enabled.
	if ((diff & ~GE_VTYPE_WEIGHTCOUNT_MASK) || (op & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
		// Restore and flush
		gstate.vertType ^= diff;
		Flush();
		gstate.vertType ^= diff;
		if (diff & (GE_VTYPE_TC_MASK | GE_VTYPE_THROUGH_MASK))
			gstate_c.Dirty(DIRTY_UVSCALEOFFSET);
		// In this case, we may be doing weights and morphs.
		// Update any bone matrix uniforms so it uses them correctly.
		if ((op & GE_VTYPE_MORPHCOUNT_MASK) != 0) {
			gstate_c.Dirty(gstate_c.deferredVertTypeDirty);
			gstate_c.deferredVertTypeDirty = 0;
		}
		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE);
	}
	if (diff & GE_VTYPE_THROUGH_MASK)
		gstate_c.Dirty(DIRTY_RASTER_STATE | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_FRAGMENTSHADER_STATE);
}

void GPU_GLES::Execute_Bezier(u32 op, u32 diff) {
	Flush();

	// We don't dirty on normal changes anymore as we prescale, but it's needed for splines/bezier.
	gstate_c.Dirty(DIRTY_UVSCALEOFFSET);

	// This also make skipping drawing very effective.
	framebufferManagerGL_->SetRenderFrameBuffer(gstate_c.IsDirty(DIRTY_FRAMEBUF), gstate_c.skipDrawReason);
	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB))	{
		// TODO: Should this eat some cycles?  Probably yes.  Not sure if important.
		return;
	}

	if (!Memory::IsValidAddress(gstate_c.vertexAddr)) {
		ERROR_LOG_REPORT(G3D, "Bad vertex address %08x!", gstate_c.vertexAddr);
		return;
	}

	void *control_points = Memory::GetPointerUnchecked(gstate_c.vertexAddr);
	void *indices = NULL;
	if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
		if (!Memory::IsValidAddress(gstate_c.indexAddr)) {
			ERROR_LOG_REPORT(G3D, "Bad index address %08x!", gstate_c.indexAddr);
			return;
		}
		indices = Memory::GetPointerUnchecked(gstate_c.indexAddr);
	}

	if (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) {
		DEBUG_LOG_REPORT(G3D, "Bezier + morph: %i", (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) >> GE_VTYPE_MORPHCOUNT_SHIFT);
	}
	if (vertTypeIsSkinningEnabled(gstate.vertType)) {
		DEBUG_LOG_REPORT(G3D, "Bezier + skinning: %i", vertTypeGetNumBoneWeights(gstate.vertType));
	}

	GEPatchPrimType patchPrim = gstate.getPatchPrimitiveType();
	SetDrawType(DRAW_BEZIER, PatchPrimToPrim(patchPrim));

	int bz_ucount = op & 0xFF;
	int bz_vcount = (op >> 8) & 0xFF;
	bool computeNormals = gstate.isLightingEnabled();
	bool patchFacing = gstate.patchfacing & 1;

	if (g_Config.bHardwareTessellation && g_Config.bHardwareTransform && !g_Config.bSoftwareRendering) {
		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE);
		gstate_c.bezier = true;
		if (gstate_c.spline_count_u != bz_ucount) {
			gstate_c.Dirty(DIRTY_BEZIERSPLINE);
			gstate_c.spline_count_u = bz_ucount;
		}
	}

	int bytesRead = 0;
	UpdateUVScaleOffset();
	drawEngine_.SubmitBezier(control_points, indices, gstate.getPatchDivisionU(), gstate.getPatchDivisionV(), bz_ucount, bz_vcount, patchPrim, computeNormals, patchFacing, gstate.vertType, &bytesRead);

	if (gstate_c.bezier)
		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE);
	gstate_c.bezier = false;

	// After drawing, we advance pointers - see SubmitPrim which does the same.
	int count = bz_ucount * bz_vcount;
	AdvanceVerts(gstate.vertType, count, bytesRead);
}

void GPU_GLES::Execute_Spline(u32 op, u32 diff) {
	Flush();

	// We don't dirty on normal changes anymore as we prescale, but it's needed for splines/bezier.
	gstate_c.Dirty(DIRTY_UVSCALEOFFSET);

	// This also make skipping drawing very effective.
	framebufferManagerGL_->SetRenderFrameBuffer(gstate_c.IsDirty(DIRTY_FRAMEBUF), gstate_c.skipDrawReason);
	if (gstate_c.skipDrawReason & (SKIPDRAW_SKIPFRAME | SKIPDRAW_NON_DISPLAYED_FB))	{
		// TODO: Should this eat some cycles?  Probably yes.  Not sure if important.
		return;
	}

	if (!Memory::IsValidAddress(gstate_c.vertexAddr)) {
		ERROR_LOG_REPORT(G3D, "Bad vertex address %08x!", gstate_c.vertexAddr);
		return;
	}

	void *control_points = Memory::GetPointerUnchecked(gstate_c.vertexAddr);
	void *indices = NULL;
	if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE) {
		if (!Memory::IsValidAddress(gstate_c.indexAddr)) {
			ERROR_LOG_REPORT(G3D, "Bad index address %08x!", gstate_c.indexAddr);
			return;
		}
		indices = Memory::GetPointerUnchecked(gstate_c.indexAddr);
	}

	if (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) {
		DEBUG_LOG_REPORT(G3D, "Spline + morph: %i", (gstate.vertType & GE_VTYPE_MORPHCOUNT_MASK) >> GE_VTYPE_MORPHCOUNT_SHIFT);
	}
	if (vertTypeIsSkinningEnabled(gstate.vertType)) {
		DEBUG_LOG_REPORT(G3D, "Spline + skinning: %i", vertTypeGetNumBoneWeights(gstate.vertType));
	}

	int sp_ucount = op & 0xFF;
	int sp_vcount = (op >> 8) & 0xFF;
	int sp_utype = (op >> 16) & 0x3;
	int sp_vtype = (op >> 18) & 0x3;
	GEPatchPrimType patchPrim = gstate.getPatchPrimitiveType();
	SetDrawType(DRAW_SPLINE, PatchPrimToPrim(patchPrim));
	bool computeNormals = gstate.isLightingEnabled();
	bool patchFacing = gstate.patchfacing & 1;
	u32 vertType = gstate.vertType;

	if (g_Config.bHardwareTessellation && g_Config.bHardwareTransform && !g_Config.bSoftwareRendering) {
		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE);
		gstate_c.spline = true;
		bool countsChanged = gstate_c.spline_count_u != sp_ucount || gstate_c.spline_count_v != sp_vcount;
		bool typesChanged = gstate_c.spline_type_u != sp_utype || gstate_c.spline_type_v != sp_vtype;
		if (countsChanged || typesChanged) {
			gstate_c.Dirty(DIRTY_BEZIERSPLINE);
			gstate_c.spline_count_u = sp_ucount;
			gstate_c.spline_count_v = sp_vcount;
			gstate_c.spline_type_u = sp_utype;
			gstate_c.spline_type_v = sp_vtype;
		}
	}

	int bytesRead = 0;
	UpdateUVScaleOffset();
	drawEngine_.SubmitSpline(control_points, indices, gstate.getPatchDivisionU(), gstate.getPatchDivisionV(), sp_ucount, sp_vcount, sp_utype, sp_vtype, patchPrim, computeNormals, patchFacing, vertType, &bytesRead);

	if (gstate_c.spline)
		gstate_c.Dirty(DIRTY_VERTEXSHADER_STATE);
	gstate_c.spline = false;

	// After drawing, we advance pointers - see SubmitPrim which does the same.
	int count = sp_ucount * sp_vcount;
	AdvanceVerts(gstate.vertType, count, bytesRead);
}

void GPU_GLES::Execute_TexSize0(u32 op, u32 diff) {
	// Render to texture may have overridden the width/height.
	// Don't reset it unless the size is different / the texture has changed.
	if (diff || gstate_c.IsDirty(DIRTY_TEXTURE_IMAGE | DIRTY_TEXTURE_PARAMS)) {
		gstate_c.curTextureWidth = gstate.getTextureWidth(0);
		gstate_c.curTextureHeight = gstate.getTextureHeight(0);
		gstate_c.Dirty(DIRTY_UVSCALEOFFSET | DIRTY_TEXTURE_PARAMS);
	}
}

void GPU_GLES::Execute_LoadClut(u32 op, u32 diff) {
	gstate_c.Dirty(DIRTY_TEXTURE_PARAMS);
	textureCacheGL_->LoadClut(gstate.getClutAddress(), gstate.getClutLoadBytes());
}

void GPU_GLES::GetStats(char *buffer, size_t bufsize) {
	float vertexAverageCycles = gpuStats.numVertsSubmitted > 0 ? (float)gpuStats.vertexGPUCycles / (float)gpuStats.numVertsSubmitted : 0.0f;
	snprintf(buffer, bufsize - 1,
		"DL processing time: %0.2f ms\n"
		"Draw calls: %i, flushes %i\n"
		"Cached Draw calls: %i\n"
		"Num Tracked Vertex Arrays: %i\n"
		"GPU cycles executed: %d (%f per vertex)\n"
		"Commands per call level: %i %i %i %i\n"
		"Vertices submitted: %i\n"
		"Cached, Uncached Vertices Drawn: %i, %i\n"
		"FBOs active: %i\n"
		"Textures active: %i, decoded: %i  invalidated: %i\n"
		"Vertex, Fragment, Programs loaded: %i, %i, %i\n",
		gpuStats.msProcessingDisplayLists * 1000.0f,
		gpuStats.numDrawCalls,
		gpuStats.numFlushes,
		gpuStats.numCachedDrawCalls,
		gpuStats.numTrackedVertexArrays,
		gpuStats.vertexGPUCycles + gpuStats.otherGPUCycles,
		vertexAverageCycles,
		gpuStats.gpuCommandsAtCallLevel[0], gpuStats.gpuCommandsAtCallLevel[1], gpuStats.gpuCommandsAtCallLevel[2], gpuStats.gpuCommandsAtCallLevel[3],
		gpuStats.numVertsSubmitted,
		gpuStats.numCachedVertsDrawn,
		gpuStats.numUncachedVertsDrawn,
		(int)framebufferManagerGL_->NumVFBs(),
		(int)textureCacheGL_->NumLoadedTextures(),
		gpuStats.numTexturesDecoded,
		gpuStats.numTextureInvalidations,
		shaderManagerGL_->GetNumVertexShaders(),
		shaderManagerGL_->GetNumFragmentShaders(),
		shaderManagerGL_->GetNumPrograms());
}

void GPU_GLES::ClearCacheNextFrame() {
	textureCacheGL_->ClearNextFrame();
}

void GPU_GLES::ClearShaderCache() {
	shaderManagerGL_->ClearCache(true);
}

void GPU_GLES::CleanupBeforeUI() {
	// Clear any enabled vertex arrays.
	shaderManagerGL_->DirtyLastShader();
	glstate.arrayBuffer.bind(0);
	glstate.elementArrayBuffer.bind(0);
}

void GPU_GLES::DoState(PointerWrap &p) {
	GPUCommon::DoState(p);

	// TODO: Some of these things may not be necessary.
	// None of these are necessary when saving.
	// In Freeze-Frame mode, we don't want to do any of this.
	if (p.mode == p.MODE_READ && !PSP_CoreParameter().frozen) {
		textureCacheGL_->Clear(true);
		depalShaderCache_.Clear();
		drawEngine_.ClearTrackedVertexArrays();

		gstate_c.Dirty(DIRTY_TEXTURE_IMAGE);
		framebufferManagerGL_->DestroyAllFBOs();
		shaderManagerGL_->ClearCache(true);
	}
}

std::vector<std::string> GPU_GLES::DebugGetShaderIDs(DebugShaderType type) {
	switch (type) {
	case SHADER_TYPE_VERTEXLOADER:
		return drawEngine_.DebugGetVertexLoaderIDs();
	case SHADER_TYPE_DEPAL:
		return depalShaderCache_.DebugGetShaderIDs(type);
	default:
		return shaderManagerGL_->DebugGetShaderIDs(type);
	}
}

std::string GPU_GLES::DebugGetShaderString(std::string id, DebugShaderType type, DebugShaderStringType stringType) {
	switch (type) {
	case SHADER_TYPE_VERTEXLOADER:
		return drawEngine_.DebugGetVertexLoaderString(id, stringType);
	case SHADER_TYPE_DEPAL:
		return depalShaderCache_.DebugGetShaderString(id, type, stringType);
	default:
		return shaderManagerGL_->DebugGetShaderString(id, type, stringType);
	}
}