Cleanup GPU init/shutdown a bit

This commit is contained in:
Henrik Rydgård 2012-12-25 15:28:34 +01:00
parent 7c64f9c61e
commit d95dbca9be
16 changed files with 80 additions and 1141 deletions

View File

@ -806,8 +806,6 @@ add_library(GPU OBJECT
GPU/GPUState.h
GPU/Math3D.cpp
GPU/Math3D.h
# GPU/Null/NullDisplayListInterpreter.cpp
# GPU/Null/NullDisplayListInterpreter.h
GPU/Null/NullGpu.cpp
GPU/Null/NullGpu.h
GPU/ge_constants.h)

View File

@ -46,10 +46,7 @@
// Internal drawing library
#include "../Util/PPGeDraw.h"
extern ShaderManager shaderManager;
struct FrameBufferState
{
struct FrameBufferState {
u32 topaddr;
PspDisplayPixelFormat pspFramebufFormat;
int pspFramebufLinesize;
@ -95,8 +92,7 @@ std::vector<WaitVBlankInfo> vblankWaitingThreads;
void hleEnterVblank(u64 userdata, int cyclesLate);
void hleLeaveVblank(u64 userdata, int cyclesLate);
void __DisplayInit()
{
void __DisplayInit() {
gpuStats.reset();
hasSetMode = false;
framebufIsLatched = false;
@ -118,29 +114,24 @@ void __DisplayInit()
InitGfxState();
}
void __DisplayShutdown()
{
void __DisplayShutdown() {
vblankListeners.clear();
vblankWaitingThreads.clear();
ShutdownGfxState();
}
void __DisplayListenVblank(VblankCallback callback)
{
void __DisplayListenVblank(VblankCallback callback) {
vblankListeners.push_back(callback);
}
void __DisplayFireVblank()
{
for (std::vector<VblankCallback>::iterator iter = vblankListeners.begin(), end = vblankListeners.end(); iter != end; ++iter)
{
void __DisplayFireVblank() {
for (std::vector<VblankCallback>::iterator iter = vblankListeners.begin(), end = vblankListeners.end(); iter != end; ++iter) {
VblankCallback cb = *iter;
cb();
}
}
void hleEnterVblank(u64 userdata, int cyclesLate)
{
void hleEnterVblank(u64 userdata, int cyclesLate) {
int vbCount = userdata;
DEBUG_LOG(HLE, "Enter VBlank %i", vbCount);
@ -162,8 +153,7 @@ void hleEnterVblank(u64 userdata, int cyclesLate)
CoreTiming::ScheduleEvent(msToCycles(vblankMs) - cyclesLate, leaveVblankEvent, vbCount+1);
// TODO: Should this be done here or in hleLeaveVblank?
if (framebufIsLatched)
{
if (framebufIsLatched) {
DEBUG_LOG(HLE, "Setting latched framebuffer %08x (prev: %08x)", latchedFramebuf.topaddr, framebuf.topaddr);
framebuf = latchedFramebuf;
framebufIsLatched = false;
@ -181,8 +171,7 @@ void hleEnterVblank(u64 userdata, int cyclesLate)
// Now we can subvert the Ge engine in order to draw custom overlays like stat counters etc.
// Here we will be drawing to the non buffered front surface.
if (g_Config.bShowDebugStats)
{
if (g_Config.bShowDebugStats) {
gpu->UpdateStats();
char stats[512];
sprintf(stats,
@ -205,12 +194,12 @@ void hleEnterVblank(u64 userdata, int cyclesLate)
gpuStats.numFragmentShaders,
gpuStats.numShaders
);
float zoom = 0.7f; /// g_Config.iWindowZoom;
PPGeBegin();
PPGeDrawText(stats, 0, 0, 0, zoom, 0xFFc0c0c0);
PPGeEnd();
gpuStats.resetFrame();
}
@ -240,9 +229,6 @@ void hleEnterVblank(u64 userdata, int cyclesLate)
host->BeginFrame();
gpu->BeginFrame();
shaderManager.DirtyShader();
shaderManager.DirtyUniform(DIRTY_ALL);
// Tell the emu core that it's time to stop emulating
// Win32 doesn't need this.
#ifndef _WIN32
@ -250,9 +236,7 @@ void hleEnterVblank(u64 userdata, int cyclesLate)
#endif
}
void hleLeaveVblank(u64 userdata, int cyclesLate)
{
void hleLeaveVblank(u64 userdata, int cyclesLate) {
isVblank = 0;
DEBUG_LOG(HLE,"Leave VBlank %i", (int)userdata - 1);
vCount++;
@ -260,19 +244,16 @@ void hleLeaveVblank(u64 userdata, int cyclesLate)
CoreTiming::ScheduleEvent(msToCycles(frameMs - vblankMs) - cyclesLate, enterVblankEvent, userdata);
}
void sceDisplayIsVblank()
{
void sceDisplayIsVblank() {
DEBUG_LOG(HLE,"%i=sceDisplayIsVblank()",isVblank);
RETURN(isVblank);
}
u32 sceDisplaySetMode(u32 unknown, u32 xres, u32 yres)
{
u32 sceDisplaySetMode(u32 unknown, u32 xres, u32 yres) {
DEBUG_LOG(HLE,"sceDisplaySetMode(%d,%d,%d)",unknown,xres,yres);
host->BeginFrame();
if (!hasSetMode)
{
if (!hasSetMode) {
gpu->InitClear();
hasSetMode = true;
}
@ -280,9 +261,7 @@ u32 sceDisplaySetMode(u32 unknown, u32 xres, u32 yres)
return 0;
}
u32 sceDisplaySetFramebuf()
{
//host->EndFrame();
u32 sceDisplaySetFramebuf() {
u32 topaddr = PARAM(0);
int linesize = PARAM(1);
int pixelformat = PARAM(2);
@ -290,19 +269,15 @@ u32 sceDisplaySetFramebuf()
FrameBufferState fbstate;
DEBUG_LOG(HLE,"sceDisplaySetFramebuf(topaddr=%08x,linesize=%d,pixelsize=%d,sync=%d)",topaddr,linesize,pixelformat,sync);
if (topaddr == 0)
{
if (topaddr == 0) {
DEBUG_LOG(HLE,"- screen off");
}
else
{
} else {
fbstate.topaddr = topaddr;
fbstate.pspFramebufFormat = (PspDisplayPixelFormat)pixelformat;
fbstate.pspFramebufLinesize = linesize;
}
if (sync == PSP_DISPLAY_SETBUF_IMMEDIATE)
{
if (sync == PSP_DISPLAY_SETBUF_IMMEDIATE) {
// Write immediately to the current framebuffer parameters
if (topaddr != 0)
{
@ -311,9 +286,7 @@ u32 sceDisplaySetFramebuf()
}
else
WARN_LOG(HLE, "%s: PSP_DISPLAY_SETBUF_IMMEDIATE without topaddr?", __FUNCTION__);
}
else if (topaddr != 0)
{
} else if (topaddr != 0) {
// Delay the write until vblank
latchedFramebuf = fbstate;
framebufIsLatched = true;
@ -321,12 +294,11 @@ u32 sceDisplaySetFramebuf()
return 0;
}
u32 sceDisplayGetFramebuf(u32 topaddrPtr, u32 linesizePtr, u32 pixelFormatPtr, int mode)
{
u32 sceDisplayGetFramebuf(u32 topaddrPtr, u32 linesizePtr, u32 pixelFormatPtr, int mode) {
const FrameBufferState &fbState = mode == 1 ? latchedFramebuf : framebuf;
DEBUG_LOG(HLE,"sceDisplayGetFramebuf(*%08x = %08x, *%08x = %08x, *%08x = %08x, %i)",
topaddrPtr, fbState.topaddr, linesizePtr, fbState.pspFramebufLinesize, pixelFormatPtr, fbState.pspFramebufFormat, mode);
if (Memory::IsValidAddress(topaddrPtr))
Memory::Write_U32(fbState.topaddr, topaddrPtr);
if (Memory::IsValidAddress(linesizePtr))
@ -337,52 +309,45 @@ u32 sceDisplayGetFramebuf(u32 topaddrPtr, u32 linesizePtr, u32 pixelFormatPtr, i
return 0;
}
void sceDisplayWaitVblankStart()
{
void sceDisplayWaitVblankStart() {
DEBUG_LOG(HLE,"sceDisplayWaitVblankStart()");
vblankWaitingThreads.push_back(WaitVBlankInfo(__KernelGetCurThread()));
__KernelWaitCurThread(WAITTYPE_VBLANK, 0, 0, 0, false);
}
void sceDisplayWaitVblank()
{
void sceDisplayWaitVblank() {
DEBUG_LOG(HLE,"sceDisplayWaitVblank()");
vblankWaitingThreads.push_back(WaitVBlankInfo(__KernelGetCurThread()));
__KernelWaitCurThread(WAITTYPE_VBLANK, 0, 0, 0, false);
}
void sceDisplayWaitVblankStartMulti()
{
void sceDisplayWaitVblankStartMulti() {
DEBUG_LOG(HLE,"sceDisplayWaitVblankStartMulti()");
vblankWaitingThreads.push_back(WaitVBlankInfo(__KernelGetCurThread()));
__KernelWaitCurThread(WAITTYPE_VBLANK, 0, 0, 0, false);
}
void sceDisplayWaitVblankCB()
{
DEBUG_LOG(HLE,"sceDisplayWaitVblankCB()");
void sceDisplayWaitVblankCB() {
DEBUG_LOG(HLE,"sceDisplayWaitVblankCB()");
vblankWaitingThreads.push_back(WaitVBlankInfo(__KernelGetCurThread()));
__KernelWaitCurThread(WAITTYPE_VBLANK, 0, 0, 0, true);
}
void sceDisplayWaitVblankStartCB()
{
DEBUG_LOG(HLE,"sceDisplayWaitVblankStartCB()");
void sceDisplayWaitVblankStartCB() {
DEBUG_LOG(HLE,"sceDisplayWaitVblankStartCB()");
vblankWaitingThreads.push_back(WaitVBlankInfo(__KernelGetCurThread()));
__KernelWaitCurThread(WAITTYPE_VBLANK, 0, 0, 0, true);
}
void sceDisplayWaitVblankStartMultiCB()
{
DEBUG_LOG(HLE,"sceDisplayWaitVblankStartMultiCB()");
void sceDisplayWaitVblankStartMultiCB() {
DEBUG_LOG(HLE,"sceDisplayWaitVblankStartMultiCB()");
vblankWaitingThreads.push_back(WaitVBlankInfo(__KernelGetCurThread()));
__KernelWaitCurThread(WAITTYPE_VBLANK, 0, 0, 0, true);
}
u32 sceDisplayGetVcount()
{
u32 sceDisplayGetVcount() {
// Too spammy
// DEBUG_LOG(HLE,"%i=sceDisplayGetVcount()", vCount);
// DEBUG_LOG(HLE,"%i=sceDisplayGetVcount()", vCount);
// Puyo Puyo Fever polls this as a substitute for waiting for vblank.
// As a result, the game never gets to reschedule so it doesn't mix audio and things break.
@ -393,29 +358,24 @@ u32 sceDisplayGetVcount()
return vCount;
}
void sceDisplayGetCurrentHcount()
{
void sceDisplayGetCurrentHcount() {
RETURN(hCount++);
}
void sceDisplayGetAccumulatedHcount()
{
void sceDisplayGetAccumulatedHcount() {
// Just do an estimate
u32 accumHCount = CoreTiming::GetTicks() / (222000000 / 60 / 272);
DEBUG_LOG(HLE,"%i=sceDisplayGetAccumulatedHcount()", accumHCount);
DEBUG_LOG(HLE,"%i=sceDisplayGetAccumulatedHcount()", accumHCount);
RETURN(accumHCount);
}
float sceDisplayGetFramePerSec()
{
float sceDisplayGetFramePerSec() {
float fps = 59.9400599f;
DEBUG_LOG(HLE,"%f=sceDisplayGetFramePerSec()", fps);
return fps; // (9MHz * 1)/(525 * 286)
return fps; // (9MHz * 1)/(525 * 286)
}
const HLEFunction sceDisplay[] =
{
const HLEFunction sceDisplay[] = {
{0x0E20F177,WrapU_UUU<sceDisplaySetMode>, "sceDisplaySetMode"},
{0x289D82FE,WrapU_V<sceDisplaySetFramebuf>, "sceDisplaySetFramebuf"},
{0xEEDA2E54,WrapU_UUUI<sceDisplayGetFramebuf>,"sceDisplayGetFrameBuf"},
@ -435,10 +395,8 @@ const HLEFunction sceDisplay[] =
{0xB4F378FA,0,"sceDisplayIsForeground"},
{0x31C4BAA8,0,"sceDisplayGetBrightness"},
{0x4D4E10EC,sceDisplayIsVblank,"sceDisplayIsVblank"},
};
void Register_sceDisplay()
{
void Register_sceDisplay() {
RegisterModule("sceDisplay", ARRAY_SIZE(sceDisplay), sceDisplay);
}

View File

@ -18,6 +18,7 @@
#pragma once
void __DisplayInit();
void __DisplayShutdown();
void Register_sceDisplay();

View File

@ -122,6 +122,7 @@ void __KernelShutdown()
__UtilityShutdown();
__GeShutdown();
__SasShutdown();
__DisplayShutdown();
__AudioShutdown();
__IoShutdown();
__InterruptsShutdown();

View File

@ -103,7 +103,7 @@ int sceUtilityMsgDialogInitStart(u32 structAddr)
int sceUtilityMsgDialogShutdownStart(u32 unknown)
{
DEBUG_LOG(HLE,"FAKE sceUtilityMsgDialogShutdownStart(%i)", unknown);
DEBUG_LOG(HLE,"sceUtilityMsgDialogShutdownStart(%i)", unknown);
return msgDialog.Shutdown();
}
@ -124,25 +124,25 @@ int sceUtilityMsgDialogGetStatus()
int sceUtilityOskInitStart(u32 oskPtr)
{
ERROR_LOG(HLE,"FAKE sceUtilityOskInitStart(%i)", PARAM(0));
DEBUG_LOG(HLE,"sceUtilityOskInitStart(%i)", PARAM(0));
return oskDialog.Init(oskPtr);
}
int sceUtilityOskShutdownStart()
{
ERROR_LOG(HLE,"FAKE sceUtilityOskShutdownStart(%i)", PARAM(0));
DEBUG_LOG(HLE,"sceUtilityOskShutdownStart(%i)", PARAM(0));
return oskDialog.Shutdown();
}
int sceUtilityOskUpdate(unsigned int unknown)
{
ERROR_LOG(HLE,"FAKE sceUtilityOskUpdate(%i)", unknown);
DEBUG_LOG(HLE,"FAKE sceUtilityOskUpdate(%i)", unknown);
return oskDialog.Update();
}
int sceUtilityOskGetStatus()
{
int status = oskDialog.GetStatus();
int status = oskDialog.GetStatus();
// Seems that 4 is the cancelled status for OSK?
if (status == 4)
{
@ -185,13 +185,13 @@ int sceUtilityScreenshotGetStatus()
void sceUtilityGamedataInstallInitStart(u32 unkown)
{
DEBUG_LOG(HLE,"FAKE sceUtilityGamedataInstallInitStart(%i)", unkown);
DEBUG_LOG(HLE,"UNIMPL sceUtilityGamedataInstallInitStart(%i)", unkown);
}
int sceUtilityGamedataInstallGetStatus()
{
u32 retval = 0;//__UtilityGetStatus();
DEBUG_LOG(HLE,"%i=sceUtilityGamedataInstallGetStatus()", retval);
DEBUG_LOG(HLE,"UNIMPL %i=sceUtilityGamedataInstallGetStatus()", retval);
return retval;
}

View File

@ -42,7 +42,6 @@
MetaFileSystem pspFileSystem;
static CoreParameter coreParameter;
extern ShaderManager shaderManager;
bool PSP_Init(const CoreParameter &coreParam, std::string *error_string)
{
@ -76,9 +75,6 @@ bool PSP_Init(const CoreParameter &coreParam, std::string *error_string)
return false;
}
shaderManager.DirtyShader();
shaderManager.DirtyUniform(DIRTY_ALL);
// Setup JIT here.
if (coreParameter.startPaused)
coreState = CORE_STEPPING;
@ -97,7 +93,6 @@ void PSP_Shutdown()
pspFileSystem.UnmountAll();
TextureCache_Clear(true);
shaderManager.ClearCache(true);
CoreTiming::ClearPendingEvents();
CoreTiming::UnregisterAllEvents();

View File

@ -33,8 +33,6 @@
#include "../../Core/HLE/sceKernelThread.h"
#include "../../Core/HLE/sceKernelInterrupt.h"
ShaderManager shaderManager;
extern u32 curTextureWidth;
extern u32 curTextureHeight;
@ -157,11 +155,11 @@ GLES_GPU::GLES_GPU(int renderWidth, int renderHeight)
displayFramebufPtr_(0),
renderWidth_(renderWidth),
renderHeight_(renderHeight),
dlIdGenerator(1),
transformDraw_(&shaderManager) {
dlIdGenerator(1) {
renderWidthFactor_ = (float)renderWidth / 480.0f;
renderHeightFactor_ = (float)renderHeight / 272.0f;
shaderManager_ = &shaderManager;
shaderManager_ = new ShaderManager();
transformDraw_.SetShaderManager(shaderManager_);
TextureCache_Init();
// Sanity check gstate
if ((int *)&gstate.transferstart - (int *)&gstate != 0xEA) {
@ -183,10 +181,18 @@ GLES_GPU::~GLES_GPU() {
delete (*iter);
}
vfbs_.clear();
shaderManager_->ClearCache(true);
delete shaderManager_;
delete [] flushBeforeCommand_;
}
void GLES_GPU::DeviceLost() {
// Simply drop all caches and textures.
// FBO:s appear to survive? Or no?
shaderManager_->ClearCache(false);
TextureCache_Clear(false);
}
void GLES_GPU::InitClear() {
if (!g_Config.bBufferedRendering) {
glClearColor(0,0,0,1);
@ -199,6 +205,11 @@ void GLES_GPU::InitClear() {
void GLES_GPU::BeginFrame() {
TextureCache_Decimate();
shaderManager_->DirtyShader();
// Not sure if this is really needed.
shaderManager_->DirtyUniform(DIRTY_ALL);
// NOTE - this is all wrong. At the beginning of the frame is a TERRIBLE time to draw the fb.
if (g_Config.bDisplayFramebuffer && displayFramebufPtr_) {
INFO_LOG(HLE, "Drawing the framebuffer");

View File

@ -51,6 +51,7 @@ public:
virtual void BeginFrame();
virtual void UpdateStats();
virtual void InvalidateCache(u32 addr, int size);
virtual void DeviceLost(); // Only happens on Android. Drop all textures and shaders.
virtual void Flush();
private:

View File

@ -41,10 +41,10 @@ const GLuint glprim[8] = {
GL_TRIANGLES, // With OpenGL ES we have to expand sprites into triangles, tripling the data instead of doubling. sigh. OpenGL ES, Y U NO SUPPORT GL_QUADS?
};
TransformDrawEngine::TransformDrawEngine(ShaderManager *shaderManager)
TransformDrawEngine::TransformDrawEngine()
: numVerts(0),
lastVType(-1),
shaderManager_(shaderManager) {
shaderManager_(0) {
decoded = new u8[65536 * 48];
decIndex = new u16[65536];
transformed = new TransformedVertex[65536];

View File

@ -27,11 +27,14 @@ struct DecVtxFormat;
// Handles transform, lighting and drawing.
class TransformDrawEngine {
public:
TransformDrawEngine(ShaderManager *shaderManager);
TransformDrawEngine();
~TransformDrawEngine();
void SubmitPrim(void *verts, void *inds, int prim, int vertexCount, float *customUV, int forceIndexType, int *bytesRead);
void DrawBezier(int ucount, int vcount);
void Flush();
void SetShaderManager(ShaderManager *shaderManager) {
shaderManager_ = shaderManager;
}
private:
void SoftwareTransformAndDraw(int prim, u8 *decoded, LinkedShader *program, int vertexCount, void *inds, int indexType, const DecVtxFormat &decVtxFormat, int maxIndex);

View File

@ -52,5 +52,6 @@ public:
// Internal hack to avoid interrupts from "PPGe" drawing (utility UI, etc)
virtual void EnableInterrupts(bool enable) = 0;
virtual void DeviceLost() = 0;
virtual void Flush() = 0;
};

View File

@ -102,14 +102,14 @@ void ReapplyGfxState()
for (int i = GE_CMD_MORPHWEIGHT0; i < GE_CMD_PATCHFACING; i++)
{
gpu->ExecuteOp(gstate.cmdmem[i], 0xFFFFFFFF);
gpu->ExecuteOp(gstate.cmdmem[i], 0xFFFFFFFF);
}
// There are a few here in the middle that we shouldn't execute...
for (int i = GE_CMD_VIEWPORTX1; i < GE_CMD_TRANSFERSTART; i++)
{
gpu->ExecuteOp(gstate.cmdmem[i], 0xFFFFFFFF);
gpu->ExecuteOp(gstate.cmdmem[i], 0xFFFFFFFF);
}
// TODO: there's more...

View File

@ -1,997 +0,0 @@
// 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 "../../Core/MemMap.h"
#include "../../Core/Host.h"
#include "../GPUState.h"
#include "../ge_constants.h"
#include "ShaderManager.h"
#include "DisplayListInterpreter.h"
#include "TransformPipeline.h"
#include "../../Core/HLE/sceKernelThread.h"
#include "../../Core/HLE/sceKernelInterrupt.h"
inline void glEnDis(GLuint cmd, int value)
{
(value ? glEnable : glDisable)(cmd);
}
struct DisplayState
{
u32 pc;
u32 stallAddr;
};
static DisplayState dcontext;
ShaderManager shaderManager;
extern u32 curTextureWidth;
extern u32 curTextureHeight;
int dlIdGenerator = 1;
struct DisplayList
{
int id;
u32 listpc;
u32 stall;
};
std::vector<DisplayList> dlQueue;
static u32 prev;
u32 stack[2];
u32 stackptr = 0;
bool finished;
u8 bezierBuf[16000];
bool ProcessDLQueue()
{
std::vector<DisplayList>::iterator iter = dlQueue.begin();
while (!(iter == dlQueue.end()))
{
DisplayList &l = *iter;
dcontext.pc = l.listpc;
dcontext.stallAddr = l.stall;
// DEBUG_LOG(G3D,"Okay, starting DL execution at %08 - stall = %08x", context.pc, stallAddr);
if (!GPU::InterpretList())
{
l.listpc = dcontext.pc;
l.stall = dcontext.stallAddr;
return false;
}
else
{
//At the end, we can remove it from the queue and continue
dlQueue.erase(iter);
//this invalidated the iterator, let's fix it
iter = dlQueue.begin();
}
}
return true; //no more lists!
}
u32 GPU::EnqueueList(u32 listpc, u32 stall)
{
DisplayList dl;
dl.id = dlIdGenerator++;
dl.listpc = listpc&0xFFFFFFF;
dl.stall = stall&0xFFFFFFF;
dlQueue.push_back(dl);
if (!ProcessDLQueue())
return dl.id;
else
return 0;
}
void GPU::UpdateStall(int listid, u32 newstall)
{
// this needs improvement....
for (std::vector<DisplayList>::iterator iter = dlQueue.begin(); iter != dlQueue.end(); iter++)
{
DisplayList &l = *iter;
if (l.id == listid)
{
l.stall = newstall & 0xFFFFFFF;
}
}
ProcessDLQueue();
}
// Just to get something on the screen, we'll just not subdivide correctly.
void drawBezier(int ucount, int vcount)
{
u16 indices[3 * 3 * 6];
float customUV[32];
int c = 0;
for (int y = 0; y < 3; y++) {
for (int x = 0; x < 3; x++) {
indices[c++] = y * 4 + x;
indices[c++] = y * 4 + x + 1;
indices[c++] = (y + 1) * 4 + x + 1;
indices[c++] = (y + 1) * 4 + x + 1;
indices[c++] = (y + 1) * 4 + x;
indices[c++] = y * 4 + x;
}
}
for (int y = 0; y < 4; y++) {
for (int x = 0; x < 4; x++) {
customUV[(y * 4 + x) * 2 + 0] = (float)x/3.0f;
customUV[(y * 4 + x) * 2 + 1] = (float)y/3.0f;
}
}
LinkedShader *linkedShader = shaderManager.ApplyShader();
TransformAndDrawPrim(Memory::GetPointer(gstate.vertexAddr), &indices[0], GE_PRIM_TRIANGLES, 3 * 3 * 6, linkedShader, customUV, GE_VTYPE_IDX_16BIT);
}
void EnterClearMode(u32 data)
{
bool colMask = (data >> 8) & 1;
bool alphaMask = (data >> 9) & 1;
bool updateZ = (data >> 10) & 1;
glColorMask(colMask, colMask, colMask, alphaMask);
glDepthMask(updateZ); // Update Z or not
// Note that depth test must be enabled for depth writes to go through! So we use GL_ALWAYS
glEnable(GL_DEPTH_TEST);
glDepthFunc(GL_ALWAYS);
glDisable(GL_CULL_FACE); // ??
}
void LeaveClearMode()
{
// We have to reset the following state as per the state of the command registers:
// Back face culling
// Texture map enable (meh)
// Fogging
// Antialiasing
// Alpha test
glDepthMask(1);
glColorMask(1,1,1,1);
glEnDis(GL_DEPTH_TEST, gstate.zTestEnable & 1);
glDepthFunc(GL_LEQUAL); // TODO
// dirtyshader?
}
void SetBlendModePSP(u32 data)
{
// This can't be done exactly as there are several PSP blend modes that are impossible to do on OpenGL ES 2.0, and some even on regular OpenGL for desktop.
// HOWEVER - we should be able to approximate the 2x modes in the shader, although they will clip wrongly.
const GLint aLookup[] = {
GL_DST_COLOR,
GL_ONE_MINUS_DST_COLOR,
GL_SRC_ALPHA,
GL_ONE_MINUS_SRC_ALPHA,
GL_DST_ALPHA,
GL_ONE_MINUS_DST_ALPHA,
GL_SRC_ALPHA, // should be 2x
GL_ONE_MINUS_SRC_ALPHA, // should be 2x
GL_DST_ALPHA, // should be 2x
GL_ONE_MINUS_DST_ALPHA, // should be 2x - and COLOR?
GL_SRC_ALPHA, // should be FIXA
};
const GLint bLookup[] = {
GL_SRC_COLOR,
GL_ONE_MINUS_SRC_COLOR,
GL_SRC_ALPHA,
GL_ONE_MINUS_SRC_ALPHA,
GL_DST_ALPHA,
GL_ONE_MINUS_DST_ALPHA,
GL_SRC_ALPHA, // should be 2x
GL_ONE_MINUS_SRC_ALPHA, // should be 2x
GL_DST_ALPHA, // should be 2x
GL_ONE_MINUS_DST_ALPHA, // should be 2x
GL_SRC_ALPHA, // should be FIXB
};
const GLint eqLookup[] = {
GL_FUNC_ADD,
GL_FUNC_SUBTRACT,
GL_FUNC_REVERSE_SUBTRACT,
#if defined(USING_GLES2)
GL_FUNC_ADD,
GL_FUNC_ADD,
#else
GL_MIN,
GL_MAX,
#endif
GL_FUNC_ADD, // should be abs(diff)
};
int a = data & 0xF;
int b = (data >> 4) & 0xF;
int eq = (data >> 8) & 0x7;
glBlendFunc(aLookup[a], bLookup[b]);
glBlendEquation(eqLookup[eq]);
}
void GPU::ExecuteOp(u32 op, u32 diff)
{
u32 cmd = op >> 24;
u32 data = op & 0xFFFFFF;
// Handle control and drawing commands here directly. The others we delegate.
switch (cmd)
{
case GE_CMD_BASE:
DEBUG_LOG(G3D,"DL BASE: %06x", data);
break;
case GE_CMD_VADDR: /// <<8????
gstate.vertexAddr = (gstate.base<<8)|data;
DEBUG_LOG(G3D,"DL VADDR: %06x", gstate.vertexAddr);
break;
case GE_CMD_IADDR:
gstate.indexAddr = (gstate.base<<8)|data;
DEBUG_LOG(G3D,"DL IADDR: %06x", gstate.indexAddr);
break;
case GE_CMD_PRIM:
{
u32 count = data & 0xFFFF;
u32 type = data >> 16;
static const char* types[7] = {
"POINTS=0,",
"LINES=1,",
"LINE_STRIP=2,",
"TRIANGLES=3,",
"TRIANGLE_STRIP=4,",
"TRIANGLE_FAN=5,",
"RECTANGLES=6,",
};
DEBUG_LOG(G3D, "DL DrawPrim type: %s count: %i vaddr= %08x, iaddr= %08x", type<7 ? types[type] : "INVALID", count, gstate.vertexAddr, gstate.indexAddr);
LinkedShader *linkedShader = shaderManager.ApplyShader();
// TODO: Split this so that we can collect sequences of primitives, can greatly speed things up
// on platforms where draw calls are expensive like mobile and D3D
void *verts = Memory::GetPointer(gstate.vertexAddr);
void *inds = 0;
if ((gstate.vertType & GE_VTYPE_IDX_MASK) != GE_VTYPE_IDX_NONE)
inds = Memory::GetPointer(gstate.indexAddr);
TransformAndDrawPrim(verts, inds, type, count, linkedShader);
}
break;
// The arrow and other rotary items in Puzbob are bezier patches, strangely enough.
case GE_CMD_BEZIER:
{
int bz_ucount = data & 0xFF;
int bz_vcount = (data >> 8) & 0xFF;
drawBezier(bz_ucount, bz_vcount);
DEBUG_LOG(G3D,"DL DRAW BEZIER: %i x %i", bz_ucount, bz_vcount);
}
break;
case GE_CMD_SPLINE:
{
int sp_ucount = data & 0xFF;
int sp_vcount = (data >> 8) & 0xFF;
int sp_utype = (data >> 16) & 0x3;
int sp_vtype = (data >> 18) & 0x3;
//drawSpline(sp_ucount, sp_vcount, sp_utype, sp_vtype);
DEBUG_LOG(G3D,"DL DRAW SPLINE: %i x %i, %i x %i", sp_ucount, sp_vcount, sp_utype, sp_vtype);
}
break;
case GE_CMD_JUMP:
{
u32 target = ((gstate.base << 8) | (op & 0xFFFFFC)) & 0x0FFFFFFF;
DEBUG_LOG(G3D,"DL CMD JUMP - %08x to %08x", dcontext.pc, target);
dcontext.pc = target - 4; // pc will be increased after we return, counteract that
}
break;
case GE_CMD_CALL:
{
u32 retval = dcontext.pc + 4;
stack[stackptr++] = retval;
u32 target = ((gstate.base << 8) | (op & 0xFFFFFC)) & 0xFFFFFFF;
DEBUG_LOG(G3D,"DL CMD CALL - %08x to %08x, ret=%08x", dcontext.pc, target, retval);
dcontext.pc = target - 4; // pc will be increased after we return, counteract that
}
break;
case GE_CMD_RET:
//TODO : debug!
{
u32 target = stack[--stackptr] & 0xFFFFFFF;
DEBUG_LOG(G3D,"DL CMD RET - from %08x to %08x", dcontext.pc, target);
dcontext.pc = target - 4;
}
break;
case GE_CMD_SIGNAL:
ERROR_LOG(G3D, "DL GE_CMD_SIGNAL %08x", data & 0xFFFFFF);
{
// int behaviour = (data >> 16) & 0xFF;
// int signal = data & 0xFFFF;
}
// This should generate a GE Interrupt
// __TriggerInterrupt(PSP_GE_INTR);
// Apparently, these callbacks should be done in a special interrupt way.
//for (size_t i = 0; i < signalCallbacks.size(); i++)
//{
// __KernelNotifyCallback(-1, signalCallbacks[i].first, signal);
//}
break;
case GE_CMD_BJUMP:
// bounding box jump. Let's just not jump, for now.
DEBUG_LOG(G3D,"DL BBOX JUMP - unimplemented");
break;
case GE_CMD_BOUNDINGBOX:
// bounding box test. Let's do nothing.
DEBUG_LOG(G3D,"DL BBOX TEST - unimplemented");
break;
case GE_CMD_ORIGIN:
gstate.offsetAddr = dcontext.pc & 0xFFFFFF;
break;
case GE_CMD_VERTEXTYPE:
DEBUG_LOG(G3D,"DL SetVertexType: %06x", data);
if (diff & GE_VTYPE_THROUGH) {
// Throughmode changed, let's make the proj matrix dirty.
shaderManager.DirtyUniform(DIRTY_PROJMATRIX);
}
if (data & GE_VTYPE_THROUGH) {
glDisable(GL_CULL_FACE);
}
// This sets through-mode or not, as well.
break;
case GE_CMD_OFFSETADDR:
// offsetAddr = data<<8;
break;
case GE_CMD_FINISH:
DEBUG_LOG(G3D,"DL CMD FINISH");
// Trigger the finish callbacks
{
// Apparently, these callbacks should be done in a special interrupt way.
//for (size_t i = 0; i < finishCallbacks.size(); i++)
//{
// __KernelNotifyCallback(-1, finishCallbacks[i].first, 0);
//}
}
break;
case GE_CMD_END:
DEBUG_LOG(G3D,"DL CMD END");
{
switch (prev >> 24)
{
case GE_CMD_FINISH:
finished = true;
break;
default:
DEBUG_LOG(G3D,"Ah, not finished: %06x", prev & 0xFFFFFF);
break;
}
}
// This should generate a Reading Ended interrupt
// __TriggerInterrupt(PSP_GE_INTR);
break;
case GE_CMD_REGION1:
{
int x1 = data & 0x3ff;
int y1 = data >> 10;
//topleft
DEBUG_LOG(G3D,"DL Region TL: %d %d", x1, y1);
}
break;
case GE_CMD_REGION2:
{
int x2 = data & 0x3ff;
int y2 = data >> 10;
DEBUG_LOG(G3D,"DL Region BR: %d %d", x2, y2);
}
break;
case GE_CMD_CLIPENABLE:
DEBUG_LOG(G3D, "DL Clip Enable: %i (ignoring)", data);
//we always clip, this is opengl
break;
case GE_CMD_CULLFACEENABLE:
DEBUG_LOG(G3D, "DL CullFace Enable: %i (ignoring)", data);
glEnDis(GL_CULL_FACE, data&1);
break;
case GE_CMD_TEXTUREMAPENABLE:
DEBUG_LOG(G3D, "DL Texture map enable: %i", data);
glEnDis(GL_TEXTURE_2D, data&1);
break;
case GE_CMD_LIGHTINGENABLE:
DEBUG_LOG(G3D, "DL Lighting enable: %i", data);
data += 1;
//We don't use OpenGL lighting
break;
case GE_CMD_FOGENABLE:
DEBUG_LOG(G3D, "DL Fog Enable: %i", gstate.fogEnable);
break;
case GE_CMD_DITHERENABLE:
DEBUG_LOG(G3D, "DL Dither Enable: %i", gstate.ditherEnable);
break;
case GE_CMD_OFFSETX:
DEBUG_LOG(G3D, "DL Offset X: %i", gstate.offsetx);
break;
case GE_CMD_OFFSETY:
DEBUG_LOG(G3D, "DL Offset Y: %i", gstate.offsety);
break;
case GE_CMD_TEXSCALEU:
gstate.uScale = getFloat24(data);
DEBUG_LOG(G3D, "DL Texture U Scale: %f", gstate.uScale);
break;
case GE_CMD_TEXSCALEV:
gstate.vScale = getFloat24(data);
DEBUG_LOG(G3D, "DL Texture V Scale: %f", gstate.vScale);
break;
case GE_CMD_TEXOFFSETU:
gstate.uOff = getFloat24(data);
DEBUG_LOG(G3D, "DL Texture U Offset: %f", gstate.uOff);
break;
case GE_CMD_TEXOFFSETV:
gstate.vOff = getFloat24(data);
DEBUG_LOG(G3D, "DL Texture V Offset: %f", gstate.vOff);
break;
case GE_CMD_SCISSOR1:
{
int x1 = data & 0x3ff;
int y1 = data >> 10;
DEBUG_LOG(G3D, "DL Scissor TL: %i, %i", x1,y1);
}
break;
case GE_CMD_SCISSOR2:
{
int x2 = data & 0x3ff;
int y2 = data >> 10;
DEBUG_LOG(G3D, "DL Scissor BR: %i, %i", x2, y2);
}
break;
case GE_CMD_MINZ:
DEBUG_LOG(G3D, "DL MinZ: %i", data);
break;
case GE_CMD_MAXZ:
DEBUG_LOG(G3D, "DL MaxZ: %i", data);
break;
case GE_CMD_FRAMEBUFPTR:
{
u32 ptr = op & 0xFFE000;
DEBUG_LOG(G3D, "DL FramebufPtr: %08x", ptr);
}
break;
case GE_CMD_FRAMEBUFWIDTH:
{
u32 w = data & 0xFFFFFF;
DEBUG_LOG(G3D, "DL FramebufWidth: %i", w);
}
break;
case GE_CMD_FRAMEBUFPIXFORMAT:
break;
case GE_CMD_TEXADDR0:
gstate.textureChanged=true;
case GE_CMD_TEXADDR1:
case GE_CMD_TEXADDR2:
case GE_CMD_TEXADDR3:
case GE_CMD_TEXADDR4:
case GE_CMD_TEXADDR5:
case GE_CMD_TEXADDR6:
case GE_CMD_TEXADDR7:
DEBUG_LOG(G3D,"DL Texture address %i: %06x", cmd-GE_CMD_TEXADDR0, data);
break;
case GE_CMD_TEXBUFWIDTH0:
gstate.textureChanged=true;
case GE_CMD_TEXBUFWIDTH1:
case GE_CMD_TEXBUFWIDTH2:
case GE_CMD_TEXBUFWIDTH3:
case GE_CMD_TEXBUFWIDTH4:
case GE_CMD_TEXBUFWIDTH5:
case GE_CMD_TEXBUFWIDTH6:
case GE_CMD_TEXBUFWIDTH7:
DEBUG_LOG(G3D,"DL Texture BUFWIDTHess %i: %06x", cmd-GE_CMD_TEXBUFWIDTH0, data);
break;
case GE_CMD_CLUTADDR:
//DEBUG_LOG(G3D,"CLUT base addr: %06x", data);
break;
case GE_CMD_CLUTADDRUPPER:
DEBUG_LOG(G3D,"DL CLUT addr: %08x", ((gstate.clutaddrupper & 0xFF0000)<<8) | (gstate.clutaddr & 0xFFFFFF));
break;
case GE_CMD_LOADCLUT:
{
u32 clutAttr = ((gstate.clutaddrupper & 0xFF0000)<<8) | (gstate.clutaddr & 0xFFFFFF);
if (clutAttr)
{
u16 *clut = (u16*)Memory::GetPointer(clutAttr);
if (clut) {
int numColors = 16 * (data&0x3F);
memcpy(&gstate.paletteMem[0], clut, numColors * 2);
}
DEBUG_LOG(G3D,"DL Clut load: %i palettes", data);
}
else
{
DEBUG_LOG(G3D,"DL Empty Clut load");
}
// Should hash and invalidate all paletted textures on use
}
break;
// case GE_CMD_TRANSFERSRC:
case GE_CMD_TRANSFERSRCW:
{
u32 xferSrc = gstate.transfersrc | ((data&0xFF0000)<<8);
u32 xferSrcW = gstate.transfersrcw & 1023;
DEBUG_LOG(G3D,"Block Transfer Src: %08x W: %i", xferSrc, xferSrcW);
break;
}
// case GE_CMD_TRANSFERDST:
case GE_CMD_TRANSFERDSTW:
{
u32 xferDst= gstate.transferdst | ((data&0xFF0000)<<8);
u32 xferDstW = gstate.transferdstw & 1023;
DEBUG_LOG(G3D,"Block Transfer Dest: %08x W: %i", xferDst, xferDstW);
break;
}
case GE_CMD_TRANSFERSRCPOS:
{
u32 x = (data & 1023)+1;
u32 y = ((data>>10) & 1023)+1;
DEBUG_LOG(G3D, "DL Block Transfer Src Rect TL: %i, %i", x, y);
break;
}
case GE_CMD_TRANSFERDSTPOS:
{
u32 x = (data & 1023)+1;
u32 y = ((data>>10) & 1023)+1;
DEBUG_LOG(G3D, "DL Block Transfer Dest Rect TL: %i, %i", x, y);
break;
}
case GE_CMD_TRANSFERSIZE:
{
u32 w = (data & 1023)+1;
u32 h = ((data>>10) & 1023)+1;
DEBUG_LOG(G3D, "DL Block Transfer Rect Size: %i x %i", w, h);
break;
}
case GE_CMD_TRANSFERSTART:
{
DEBUG_LOG(G3D, "DL Texture Transfer Start: PixFormat %i", data);
// TODO: Here we should check if the transfer overlaps a framebuffer or any textures,
// and take appropriate action. If not, this should just be a block transfer within
// GPU memory which could be implemented by a copy loop.
break;
}
case GE_CMD_TEXSIZE0:
gstate.textureChanged=true;
gstate.curTextureWidth = 1 << (gstate.texsize[0] & 0xf);
gstate.curTextureHeight = 1 << ((gstate.texsize[0]>>8) & 0xf);
//fall thru - ignoring the mipmap sizes for now
case GE_CMD_TEXSIZE1:
case GE_CMD_TEXSIZE2:
case GE_CMD_TEXSIZE3:
case GE_CMD_TEXSIZE4:
case GE_CMD_TEXSIZE5:
case GE_CMD_TEXSIZE6:
case GE_CMD_TEXSIZE7:
DEBUG_LOG(G3D,"DL Texture Size: %06x", data);
break;
case GE_CMD_ZBUFPTR:
{
u32 ptr = op & 0xFFE000;
DEBUG_LOG(G3D,"Zbuf Ptr: %06x", ptr);
}
break;
case GE_CMD_ZBUFWIDTH:
{
u32 w = data & 0xFFFFFF;
DEBUG_LOG(G3D,"Zbuf Width: %i", w);
}
break;
case GE_CMD_AMBIENTCOLOR:
DEBUG_LOG(G3D,"DL Ambient Color: %06x", data);
break;
case GE_CMD_AMBIENTALPHA:
DEBUG_LOG(G3D,"DL Ambient Alpha: %06x", data);
break;
case GE_CMD_MATERIALAMBIENT:
DEBUG_LOG(G3D,"DL Material Ambient Color: %06x", data);
break;
case GE_CMD_MATERIALDIFFUSE:
DEBUG_LOG(G3D,"DL Material Diffuse Color: %06x", data);
break;
case GE_CMD_MATERIALEMISSIVE:
DEBUG_LOG(G3D,"DL Material Emissive Color: %06x", data);
break;
case GE_CMD_MATERIALSPECULAR:
DEBUG_LOG(G3D,"DL Material Specular Color: %06x", data);
break;
case GE_CMD_MATERIALALPHA:
DEBUG_LOG(G3D,"DL Material Alpha Color: %06x", data);
break;
case GE_CMD_MATERIALSPECULARCOEF:
DEBUG_LOG(G3D,"DL Material specular coef: %f", getFloat24(data));
break;
case GE_CMD_LIGHTTYPE0:
case GE_CMD_LIGHTTYPE1:
case GE_CMD_LIGHTTYPE2:
case GE_CMD_LIGHTTYPE3:
DEBUG_LOG(G3D,"DL Light %i type: %06x", cmd-GE_CMD_LIGHTTYPE0, data);
break;
case GE_CMD_LX0:case GE_CMD_LY0:case GE_CMD_LZ0:
case GE_CMD_LX1:case GE_CMD_LY1:case GE_CMD_LZ1:
case GE_CMD_LX2:case GE_CMD_LY2:case GE_CMD_LZ2:
case GE_CMD_LX3:case GE_CMD_LY3:case GE_CMD_LZ3:
{
int n = cmd - GE_CMD_LX0;
int l = n / 3;
int c = n % 3;
float val = getFloat24(data);
DEBUG_LOG(G3D,"DL Light %i %c pos: %f", l, c+'X', val);
gstate.lightpos[l][c] = val;
}
break;
case GE_CMD_LDX0:case GE_CMD_LDY0:case GE_CMD_LDZ0:
case GE_CMD_LDX1:case GE_CMD_LDY1:case GE_CMD_LDZ1:
case GE_CMD_LDX2:case GE_CMD_LDY2:case GE_CMD_LDZ2:
case GE_CMD_LDX3:case GE_CMD_LDY3:case GE_CMD_LDZ3:
{
int n = cmd - GE_CMD_LDX0;
int l = n / 3;
int c = n % 3;
float val = getFloat24(data);
DEBUG_LOG(G3D,"DL Light %i %c dir: %f", l, c+'X', val);
gstate.lightdir[l][c] = val;
}
break;
case GE_CMD_LKA0:case GE_CMD_LKB0:case GE_CMD_LKC0:
case GE_CMD_LKA1:case GE_CMD_LKB1:case GE_CMD_LKC1:
case GE_CMD_LKA2:case GE_CMD_LKB2:case GE_CMD_LKC2:
case GE_CMD_LKA3:case GE_CMD_LKB3:case GE_CMD_LKC3:
{
int n = cmd - GE_CMD_LKA0;
int l = n / 3;
int c = n % 3;
float val = getFloat24(data);
DEBUG_LOG(G3D,"DL Light %i %c att: %f", l, c+'X', val);
gstate.lightatt[l][c] = val;
}
break;
case GE_CMD_LAC0:case GE_CMD_LAC1:case GE_CMD_LAC2:case GE_CMD_LAC3:
case GE_CMD_LDC0:case GE_CMD_LDC1:case GE_CMD_LDC2:case GE_CMD_LDC3:
case GE_CMD_LSC0:case GE_CMD_LSC1:case GE_CMD_LSC2:case GE_CMD_LSC3:
{
float r = (float)(data>>16)/255.0f;
float g = (float)((data>>8) & 0xff)/255.0f;
float b = (float)(data & 0xff)/255.0f;
int l = (cmd - GE_CMD_LAC0) / 3;
int t = (cmd - GE_CMD_LAC0) % 3;
gstate.lightColor[t][l].r = r;
gstate.lightColor[t][l].g = g;
gstate.lightColor[t][l].b = b;
}
break;
case GE_CMD_VIEWPORTX1:
case GE_CMD_VIEWPORTY1:
case GE_CMD_VIEWPORTZ1:
case GE_CMD_VIEWPORTX2:
case GE_CMD_VIEWPORTY2:
case GE_CMD_VIEWPORTZ2:
DEBUG_LOG(G3D,"DL Viewport param %i: %f", cmd-GE_CMD_VIEWPORTX1, getFloat24(data));
break;
case GE_CMD_LIGHTENABLE0:
case GE_CMD_LIGHTENABLE1:
case GE_CMD_LIGHTENABLE2:
case GE_CMD_LIGHTENABLE3:
DEBUG_LOG(G3D,"DL Light %i enable: %d", cmd-GE_CMD_LIGHTENABLE0, data);
break;
case GE_CMD_CULL:
DEBUG_LOG(G3D,"DL cull: %06x", data);
glCullFace(data ? GL_BACK : GL_FRONT);
break;
case GE_CMD_LMODE:
DEBUG_LOG(G3D,"DL Shade mode: %06x", data);
break;
case GE_CMD_PATCHDIVISION:
gstate.patch_div_s = data & 0xFF;
gstate.patch_div_t = (data >> 8) & 0xFF;
DEBUG_LOG(G3D, "DL Patch subdivision: %i x %i", gstate.patch_div_s, gstate.patch_div_t);
break;
case GE_CMD_MATERIALUPDATE:
DEBUG_LOG(G3D,"DL Material Update: %d", data);
break;
//////////////////////////////////////////////////////////////////
// CLEARING
//////////////////////////////////////////////////////////////////
case GE_CMD_CLEARMODE:
// If it becomes a performance problem, check diff&1
if (data & 1)
EnterClearMode(data);
else
LeaveClearMode();
DEBUG_LOG(G3D,"DL Clear mode: %06x", data);
break;
//////////////////////////////////////////////////////////////////
// ALPHA BLENDING
//////////////////////////////////////////////////////////////////
case GE_CMD_ALPHABLENDENABLE:
DEBUG_LOG(G3D,"DL Alpha blend enable: %d", data);
glEnDis(GL_BLEND, data);
break;
case GE_CMD_BLENDMODE:
DEBUG_LOG(G3D,"DL Blend mode: %06x", data);
SetBlendModePSP(data);
break;
case GE_CMD_BLENDFIXEDA:
DEBUG_LOG(G3D,"DL Blend fix A: %06x", data);
break;
case GE_CMD_BLENDFIXEDB:
DEBUG_LOG(G3D,"DL Blend fix B: %06x", data);
break;
case GE_CMD_ALPHATESTENABLE:
DEBUG_LOG(G3D,"DL Alpha test enable: %d", data);
// This is done in the shader.
break;
case GE_CMD_ALPHATEST:
DEBUG_LOG(G3D,"DL Alpha test settings");
shaderManager.DirtyUniform(DIRTY_ALPHAREF);
break;
case GE_CMD_TEXFUNC:
{
DEBUG_LOG(G3D,"DL TexFunc %i", data&7);
/*
int m=GL_MODULATE;
switch (data & 7)
{
case 0: m=GL_MODULATE; break;
case 1: m=GL_DECAL; break;
case 2: m=GL_BLEND; break;
case 3: m=GL_REPLACE; break;
case 4: m=GL_ADD; break;
}*/
/*
glTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_COMBINE);
glTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_RGB, GL_MODULATE);
glTexEnvi(GL_TEXTURE_ENV, GL_SOURCE0_RGB, GL_CONSTANT);
glTexEnvi(GL_TEXTURE_ENV, GL_OPERAND0_RGB, GL_SRC_COLOR);
glTexEnvi(GL_TEXTURE_ENV, GL_SOURCE1_RGB, GL_TEXTURE);
glTexEnvi(GL_TEXTURE_ENV, GL_OPERAND1_RGB, GL_SRC_COLOR);
glTexEnvi(GL_TEXTURE_ENV, GL_RGB_SCALE, 1);
glTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, m);
glTexEnvi(GL_TEXTURE_ENV, GL_COMBINE_ALPHA, GL_MODULATE);*/
break;
}
case GE_CMD_TEXFILTER:
{
int min = data & 7;
int mag = (data >> 8) & 1;
DEBUG_LOG(G3D,"DL TexFilter min: %i mag: %i", min, mag);
}
break;
//////////////////////////////////////////////////////////////////
// Z/STENCIL TESTING
//////////////////////////////////////////////////////////////////
case GE_CMD_ZTESTENABLE:
glEnDis(GL_DEPTH_TEST, data & 1);
DEBUG_LOG(G3D,"DL Z test enable: %d", data & 1);
break;
case GE_CMD_STENCILTESTENABLE:
DEBUG_LOG(G3D,"DL Stencil test enable: %d", data);
break;
case GE_CMD_ZTEST:
{
static const GLuint ztests[8] =
{
GL_NEVER, GL_ALWAYS, GL_EQUAL, GL_NOTEQUAL,
GL_LESS, GL_LEQUAL, GL_GREATER, GL_GEQUAL
};
//glDepthFunc(ztests[data&7]);
glDepthFunc(GL_LEQUAL);
DEBUG_LOG(G3D,"DL Z test mode: %i", data);
}
break;
case GE_CMD_MORPHWEIGHT0:
case GE_CMD_MORPHWEIGHT1:
case GE_CMD_MORPHWEIGHT2:
case GE_CMD_MORPHWEIGHT3:
case GE_CMD_MORPHWEIGHT4:
case GE_CMD_MORPHWEIGHT5:
case GE_CMD_MORPHWEIGHT6:
case GE_CMD_MORPHWEIGHT7:
{
int index = cmd - GE_CMD_MORPHWEIGHT0;
float weight = getFloat24(data);
DEBUG_LOG(G3D,"DL MorphWeight %i = %f", index, weight);
gstate.morphWeights[index] = weight;
}
break;
case GE_CMD_DITH0:
case GE_CMD_DITH1:
case GE_CMD_DITH2:
case GE_CMD_DITH3:
DEBUG_LOG(G3D,"DL DitherMatrix %i = %06x",cmd-GE_CMD_DITH0,data);
break;
case GE_CMD_WORLDMATRIXNUMBER:
DEBUG_LOG(G3D,"DL World matrix # %i", data);
gstate.worldmtxnum = data&0xF;
break;
case GE_CMD_WORLDMATRIXDATA:
DEBUG_LOG(G3D,"DL World matrix data # %f", getFloat24(data));
gstate.worldMatrix[gstate.worldmtxnum++] = getFloat24(data);
break;
case GE_CMD_VIEWMATRIXNUMBER:
DEBUG_LOG(G3D,"DL VIEW matrix # %i", data);
gstate.viewmtxnum = data&0xF;
break;
case GE_CMD_VIEWMATRIXDATA:
DEBUG_LOG(G3D,"DL VIEW matrix data # %f", getFloat24(data));
gstate.viewMatrix[gstate.viewmtxnum++] = getFloat24(data);
break;
case GE_CMD_PROJMATRIXNUMBER:
DEBUG_LOG(G3D,"DL PROJECTION matrix # %i", data);
gstate.projmtxnum = data&0xF;
break;
case GE_CMD_PROJMATRIXDATA:
DEBUG_LOG(G3D,"DL PROJECTION matrix data # %f", getFloat24(data));
gstate.projMatrix[gstate.projmtxnum++] = getFloat24(data);
shaderManager.DirtyUniform(DIRTY_PROJMATRIX);
break;
case GE_CMD_TGENMATRIXNUMBER:
DEBUG_LOG(G3D,"DL TGEN matrix # %i", data);
gstate.texmtxnum = data&0xF;
break;
case GE_CMD_TGENMATRIXDATA:
DEBUG_LOG(G3D,"DL TGEN matrix data # %f", getFloat24(data));
gstate.tgenMatrix[gstate.texmtxnum++] = getFloat24(data);
break;
case GE_CMD_BONEMATRIXNUMBER:
DEBUG_LOG(G3D,"DL BONE matrix #%i", data);
gstate.boneMatrixNumber = data;
break;
case GE_CMD_BONEMATRIXDATA:
DEBUG_LOG(G3D,"DL BONE matrix data #%i %f", gstate.boneMatrixNumber, getFloat24(data));
gstate.boneMatrix[gstate.boneMatrixNumber++] = getFloat24(data);
break;
default:
DEBUG_LOG(G3D,"DL Unknown: %08x @ %08x", op, dcontext.pc);
break;
//ETC...
}
}
bool GPU::InterpretList()
{
// Reset stackptr for safety
stackptr = 0;
u32 op = 0;
prev = 0;
finished = false;
while (!finished)
{
if (dcontext.pc == dcontext.stallAddr)
return false;
op = Memory::ReadUnchecked_U32(dcontext.pc); //read from memory
u32 cmd = op >> 24;
u32 diff = op ^ gstate.cmdmem[cmd];
gstate.cmdmem[cmd] = op; // crashes if I try to put the whole op there??
ExecuteOp(op, diff);
dcontext.pc += 4;
prev = op;
}
return true;
}

View File

@ -1,31 +0,0 @@
// 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/.
#pragma once
#include "../../Globals.h"
class ShaderManager;
class GPU
{
public:
static u32 EnqueueList(u32 listpc, u32 stall);
static void UpdateStall(int listid, u32 newstall);
static void ExecuteOp(u32 op, u32 diff);
static bool InterpretList();
};

View File

@ -44,6 +44,8 @@ public:
virtual void InvalidateCache(u32 addr, int size);
virtual void Flush() {}
virtual void DeviceLost() {}
private:
bool ProcessDLQueue();
bool interruptsEnabled_;

View File

@ -29,9 +29,8 @@
#include "../../Core/Host.h"
#include "../../Core/System.h"
#include "../../Core/MIPS/MIPS.h"
#include "../../GPU/GLES/TextureCache.h"
#include "../../GPU/GLES/ShaderManager.h"
#include "../../GPU/GPUState.h"
#include "../../GPU/GPUInterface.h"
#include "../../Core/HLE/sceCtrl.h"
#include "GamepadEmu.h"
@ -40,8 +39,6 @@
#include "MenuScreens.h"
#include "EmuScreen.h"
extern ShaderManager shaderManager;
EmuScreen::EmuScreen(const std::string &filename) : invalid_(true)
{
std::string fileToStart = filename;
@ -200,6 +197,5 @@ void EmuScreen::render()
void EmuScreen::deviceLost()
{
TextureCache_Clear(false); // This doesn't seem to help?
shaderManager.ClearCache(false);
gpu->DeviceLost();
}