void VertexShaderCache::Shutdown()
{
for (int i = 0; i < MAX_SSAA_SHADERS; i++)
{
if (SimpleVertexShader[i])
SimpleVertexShader[i]->Release();
SimpleVertexShader[i] = NULL;
}
if (ClearVertexShader)
ClearVertexShader->Release();
ClearVertexShader = NULL;
Clear();
g_vs_disk_cache.Sync();
g_vs_disk_cache.Close();
}
bool CompileShaders(std::string &errorMsg) {
if (!CompileVertexShader(vscode, &pFramebufferVertexShader, NULL, errorMsg)) {
OutputDebugStringA(errorMsg.c_str());
return false;
}
if (!CompilePixelShader(pscode, &pFramebufferPixelShader, NULL, errorMsg)) {
OutputDebugStringA(errorMsg.c_str());
if (pFramebufferVertexShader) {
pFramebufferVertexShader->Release();
}
return false;
}
pD3Ddevice->CreateVertexDeclaration(VertexElements, &pFramebufferVertexDecl);
pD3Ddevice->SetVertexDeclaration(pFramebufferVertexDecl);
pD3Ddevice->CreateVertexDeclaration(SoftTransVertexElements, &pSoftVertexDecl);
return true;
}
namespace DX9
{
VertexShaderCache::VSCache VertexShaderCache::vshaders;
const VertexShaderCache::VSCacheEntry *VertexShaderCache::last_entry;
#define MAX_SSAA_SHADERS 3
static LPDIRECT3DVERTEXSHADER9 SimpleVertexShader[MAX_SSAA_SHADERS];
static LPDIRECT3DVERTEXSHADER9 ClearVertexShader;
LinearDiskCache<VERTEXSHADERUID, u8> g_vs_disk_cache;
LPDIRECT3DVERTEXSHADER9 VertexShaderCache::GetSimpleVertexShader(int level)
{
return SimpleVertexShader[level % MAX_SSAA_SHADERS];
}
LPDIRECT3DVERTEXSHADER9 VertexShaderCache::GetClearVertexShader()
{
return ClearVertexShader;
}
// this class will load the precompiled shaders into our cache
class VertexShaderCacheInserter : public LinearDiskCacheReader<VERTEXSHADERUID, u8>
{
public:
void Read(const VERTEXSHADERUID &key, const u8 *value, u32 value_size)
{
VertexShaderCache::InsertByteCode(key, value, value_size, false);
}
};
void VertexShaderCache::Init()
{
char* vProg = new char[2048];
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[0] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vColor0 : COLOR0;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float4 inColor0: COLOR0)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vColor0 = inColor0;\n"
"return OUT;\n"
"}\n");
ClearVertexShader = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inInvTexSize : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
// HACK: Scale the texture coordinate range from (0,width) to (0,width-1), otherwise the linear filter won't average our samples correctly
"OUT.vTexCoord = inTEX0 * (float2(1.f,1.f) / inInvTexSize - float2(1.f,1.f)) * inInvTexSize;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[1] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-1.0f,-0.5f, 1.0f,-0.5f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 1.0f, 0.5f,-1.0f, 0.5f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[2] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
Clear();
delete [] vProg;
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sdx9-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
}
void VertexShaderCache::Clear()
{
for (VSCache::iterator iter = vshaders.begin(); iter != vshaders.end(); ++iter)
iter->second.Destroy();
vshaders.clear();
memset(&last_vertex_shader_uid, 0xFF, sizeof(last_vertex_shader_uid));
}
void VertexShaderCache::Shutdown()
{
for (int i = 0; i < MAX_SSAA_SHADERS; i++)
{
if (SimpleVertexShader[i])
SimpleVertexShader[i]->Release();
SimpleVertexShader[i] = NULL;
}
if (ClearVertexShader)
ClearVertexShader->Release();
ClearVertexShader = NULL;
Clear();
g_vs_disk_cache.Sync();
g_vs_disk_cache.Close();
}
bool VertexShaderCache::SetShader(u32 components)
{
VERTEXSHADERUID uid;
GetVertexShaderId(&uid, components);
if (uid == last_vertex_shader_uid && vshaders[uid].frameCount == frameCount)
{
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
return (vshaders[uid].shader != NULL);
}
memcpy(&last_vertex_shader_uid, &uid, sizeof(VERTEXSHADERUID));
VSCache::iterator iter = vshaders.find(uid);
if (iter != vshaders.end())
{
iter->second.frameCount = frameCount;
const VSCacheEntry &entry = iter->second;
last_entry = &entry;
if (entry.shader) D3D::SetVertexShader(entry.shader);
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
return (entry.shader != NULL);
}
const char *code = GenerateVertexShaderCode(components, API_D3D9);
u8 *bytecode;
int bytecodelen;
if (!D3D::CompileVertexShader(code, (int)strlen(code), &bytecode, &bytecodelen))
{
if (g_ActiveConfig.bShowShaderErrors)
{
PanicAlert("Failed to compile Vertex Shader:\n\n%s", code);
}
GFX_DEBUGGER_PAUSE_AT(NEXT_ERROR, true);
return false;
}
g_vs_disk_cache.Append(uid, bytecode, bytecodelen);
g_vs_disk_cache.Sync();
bool result = InsertByteCode(uid, bytecode, bytecodelen, true);
delete [] bytecode;
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
return result;
}
bool VertexShaderCache::InsertByteCode(const VERTEXSHADERUID &uid, const u8 *bytecode, int bytecodelen, bool activate) {
LPDIRECT3DVERTEXSHADER9 shader = D3D::CreateVertexShaderFromByteCode(bytecode, bytecodelen);
// Make an entry in the table
VSCacheEntry entry;
entry.shader = shader;
entry.frameCount = frameCount;
vshaders[uid] = entry;
last_entry = &vshaders[uid];
if (!shader)
return false;
INCSTAT(stats.numVertexShadersCreated);
SETSTAT(stats.numVertexShadersAlive, (int)vshaders.size());
if (activate)
{
D3D::SetVertexShader(shader);
return true;
}
return false;
}
void Renderer::SetVSConstant4f(unsigned int const_number, float f1, float f2, float f3, float f4)
{
const float f[4] = { f1, f2, f3, f4 };
DX9::D3D::dev->SetVertexShaderConstantF(const_number, f, 1);
}
void Renderer::SetVSConstant4fv(unsigned int const_number, const float *f)
{
DX9::D3D::dev->SetVertexShaderConstantF(const_number, f, 1);
}
void Renderer::SetMultiVSConstant3fv(unsigned int const_number, unsigned int count, const float *f)
{
float buf[4*C_VENVCONST_END];
for (unsigned int i = 0; i < count; i++)
{
buf[4*i ] = *f++;
buf[4*i+1] = *f++;
buf[4*i+2] = *f++;
buf[4*i+3] = 0.f;
}
DX9::D3D::dev->SetVertexShaderConstantF(const_number, buf, count);
}
void Renderer::SetMultiVSConstant4fv(unsigned int const_number, unsigned int count, const float *f)
{
DX9::D3D::dev->SetVertexShaderConstantF(const_number, f, count);
}
} // namespace DX9
namespace DX9 {
LPDIRECT3DDEVICE9 pD3Ddevice = NULL;
LPDIRECT3DDEVICE9EX pD3DdeviceEx = NULL;
LPDIRECT3D9 pD3D = NULL;
static const char * vscode =
"struct VS_IN {\n"
" float4 ObjPos : POSITION;\n"
" float2 Uv : TEXCOORD0;\n"
"};"
"struct VS_OUT {\n"
" float4 ProjPos : POSITION;\n"
" float2 Uv : TEXCOORD0;\n"
"};\n"
"VS_OUT main( VS_IN In ) {\n"
" VS_OUT Out;\n"
" Out.ProjPos = In.ObjPos;\n"
" Out.Uv = In.Uv;\n"
" return Out;\n"
"}\n";
//--------------------------------------------------------------------------------------
// Pixel shader
//--------------------------------------------------------------------------------------
static const char * pscode =
"sampler s: register(s0);\n"
"struct PS_IN {\n"
" float2 Uv : TEXCOORD0;\n"
"};\n"
"float4 main( PS_IN In ) : COLOR {\n"
" float4 c = tex2D(s, In.Uv);\n"
" c.a = 1.0f;\n"
" return c;\n"
"}\n";
IDirect3DVertexDeclaration9* pFramebufferVertexDecl = NULL;
static const D3DVERTEXELEMENT9 VertexElements[] = {
{ 0, 0, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0 },
{ 0, 12, D3DDECLTYPE_FLOAT2, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_TEXCOORD, 0 },
D3DDECL_END()
};
IDirect3DVertexDeclaration9* pSoftVertexDecl = NULL;
static const D3DVERTEXELEMENT9 SoftTransVertexElements[] = {
{ 0, 0, D3DDECLTYPE_FLOAT4, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_POSITION, 0 },
{ 0, 16, D3DDECLTYPE_FLOAT3, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_TEXCOORD, 0 },
{ 0, 28, D3DDECLTYPE_UBYTE4N, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_COLOR, 0 },
{ 0, 32, D3DDECLTYPE_UBYTE4N, D3DDECLMETHOD_DEFAULT, D3DDECLUSAGE_COLOR, 1 },
D3DDECL_END()
};
LPDIRECT3DVERTEXSHADER9 pFramebufferVertexShader = NULL; // Vertex Shader
LPDIRECT3DPIXELSHADER9 pFramebufferPixelShader = NULL; // Pixel Shader
bool CompilePixelShader(const char *code, LPDIRECT3DPIXELSHADER9 *pShader, LPD3DXCONSTANTTABLE *pShaderTable, std::string &errorMessage) {
ID3DXBuffer* pShaderCode = NULL;
ID3DXBuffer* pErrorMsg = NULL;
HRESULT hr = -1;
// Compile pixel shader.
hr = dyn_D3DXCompileShader(code,
(UINT)strlen(code),
NULL,
NULL,
"main",
"ps_2_0",
0,
&pShaderCode,
&pErrorMsg,
pShaderTable);
if (pErrorMsg) {
errorMessage = (CHAR *)pErrorMsg->GetBufferPointer();
pErrorMsg->Release();
} else {
errorMessage = "";
}
if (FAILED(hr)) {
if (pShaderCode)
pShaderCode->Release();
return false;
}
// Create pixel shader.
pD3Ddevice->CreatePixelShader( (DWORD*)pShaderCode->GetBufferPointer(),
pShader );
pShaderCode->Release();
return true;
}
bool CompileVertexShader(const char *code, LPDIRECT3DVERTEXSHADER9 *pShader, LPD3DXCONSTANTTABLE *pShaderTable, std::string &errorMessage) {
ID3DXBuffer* pShaderCode = NULL;
ID3DXBuffer* pErrorMsg = NULL;
HRESULT hr = -1;
// Compile pixel shader.
hr = dyn_D3DXCompileShader(code,
(UINT)strlen(code),
NULL,
NULL,
"main",
"vs_2_0",
0,
&pShaderCode,
&pErrorMsg,
pShaderTable);
if (pErrorMsg) {
errorMessage = (CHAR *)pErrorMsg->GetBufferPointer();
pErrorMsg->Release();
} else {
errorMessage = "";
}
if (FAILED(hr)) {
if (pShaderCode)
pShaderCode->Release();
return false;
}
// Create pixel shader.
pD3Ddevice->CreateVertexShader( (DWORD*)pShaderCode->GetBufferPointer(),
pShader );
pShaderCode->Release();
return true;
}
void CompileShaders() {
std::string errorMsg;
HRESULT hr = -1;
if (!CompileVertexShader(vscode, &pFramebufferVertexShader, NULL, errorMsg)) {
OutputDebugStringA(errorMsg.c_str());
DebugBreak();
}
if (!CompilePixelShader(pscode, &pFramebufferPixelShader, NULL, errorMsg)) {
OutputDebugStringA(errorMsg.c_str());
DebugBreak();
}
pD3Ddevice->CreateVertexDeclaration(VertexElements, &pFramebufferVertexDecl);
pD3Ddevice->SetVertexDeclaration(pFramebufferVertexDecl);
pD3Ddevice->CreateVertexDeclaration(SoftTransVertexElements, &pSoftVertexDecl);
}
void DestroyShaders() {
if (pFramebufferVertexShader) {
pFramebufferVertexShader->Release();
}
if (pFramebufferPixelShader) {
pFramebufferPixelShader->Release();
}
if (pFramebufferVertexDecl) {
pFramebufferVertexDecl->Release();
}
if (pSoftVertexDecl) {
pSoftVertexDecl->Release();
}
}
// Only used by Headless! TODO: Remove
void DirectxInit(HWND window) {
pD3D = Direct3DCreate9( D3D_SDK_VERSION );
// Set up the structure used to create the D3DDevice. Most parameters are
// zeroed out. We set Windowed to TRUE, since we want to do D3D in a
// window, and then set the SwapEffect to "discard", which is the most
// efficient method of presenting the back buffer to the display. And
// we request a back buffer format that matches the current desktop display
// format.
D3DPRESENT_PARAMETERS d3dpp;
ZeroMemory(&d3dpp, sizeof(d3dpp));
// TODO?
d3dpp.Windowed = TRUE;
d3dpp.MultiSampleType = D3DMULTISAMPLE_NONE;
d3dpp.MultiSampleQuality = 0;
d3dpp.BackBufferCount = 1;
d3dpp.EnableAutoDepthStencil = TRUE;
d3dpp.AutoDepthStencilFormat = D3DFMT_D24S8;
d3dpp.SwapEffect = D3DSWAPEFFECT_DISCARD;
d3dpp.PresentationInterval = D3DPRESENT_INTERVAL_IMMEDIATE;
//d3dpp.PresentationInterval = (useVsync == true)?D3DPRESENT_INTERVAL_ONE:D3DPRESENT_INTERVAL_IMMEDIATE;
//d3dpp.RingBufferParameters = d3dr;
HRESULT hr = pD3D->CreateDevice( D3DADAPTER_DEFAULT, D3DDEVTYPE_HAL, window,
D3DCREATE_HARDWARE_VERTEXPROCESSING,
&d3dpp, &pD3Ddevice);
if (hr != D3D_OK) {
// TODO
}
#ifdef _XBOX
pD3Ddevice->SetRingBufferParameters( &d3dr );
#endif
CompileShaders();
fbo_init(pD3D);
}
};
namespace DX9
{
VertexShaderCache::VSCache VertexShaderCache::vshaders;
const VertexShaderCache::VSCacheEntry *VertexShaderCache::last_entry;
VertexShaderUid VertexShaderCache::last_uid;
UidChecker<VertexShaderUid,VertexShaderCode> VertexShaderCache::vertex_uid_checker;
#define MAX_SSAA_SHADERS 3
static LPDIRECT3DVERTEXSHADER9 SimpleVertexShader[MAX_SSAA_SHADERS];
static LPDIRECT3DVERTEXSHADER9 ClearVertexShader;
LinearDiskCache<VertexShaderUid, u8> g_vs_disk_cache;
LPDIRECT3DVERTEXSHADER9 VertexShaderCache::GetSimpleVertexShader(int level)
{
return SimpleVertexShader[level % MAX_SSAA_SHADERS];
}
LPDIRECT3DVERTEXSHADER9 VertexShaderCache::GetClearVertexShader()
{
return ClearVertexShader;
}
// this class will load the precompiled shaders into our cache
class VertexShaderCacheInserter : public LinearDiskCacheReader<VertexShaderUid, u8>
{
public:
void Read(const VertexShaderUid &key, const u8 *value, u32 value_size)
{
VertexShaderCache::InsertByteCode(key, value, value_size, false);
}
};
void VertexShaderCache::Init()
{
char* vProg = new char[2048];
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[0] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg,"struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vColor0 : COLOR0;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float4 inColor0: COLOR0)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vColor0 = inColor0;\n"
"return OUT;\n"
"}\n");
ClearVertexShader = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-0.495f,-0.495f, 0.495f,-0.495f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 0.495f, 0.495f,-0.495f, 0.495f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[1] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
sprintf(vProg, "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-0.9f,-0.45f, 0.9f,-0.45f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 0.9f, 0.45f,-0.9f, 0.45f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\n");
SimpleVertexShader[2] = D3D::CompileAndCreateVertexShader(vProg, (int)strlen(vProg));
Clear();
delete [] vProg;
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sdx9-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = NULL;
}
void VertexShaderCache::Clear()
{
for (VSCache::iterator iter = vshaders.begin(); iter != vshaders.end(); ++iter)
iter->second.Destroy();
vshaders.clear();
vertex_uid_checker.Invalidate();
last_entry = NULL;
}
void VertexShaderCache::Shutdown()
{
for (int i = 0; i < MAX_SSAA_SHADERS; i++)
{
if (SimpleVertexShader[i])
SimpleVertexShader[i]->Release();
SimpleVertexShader[i] = NULL;
}
if (ClearVertexShader)
ClearVertexShader->Release();
ClearVertexShader = NULL;
Clear();
g_vs_disk_cache.Sync();
g_vs_disk_cache.Close();
}
bool VertexShaderCache::SetShader(u32 components)
{
VertexShaderUid uid;
GetVertexShaderUid(uid, components, API_D3D9);
if (g_ActiveConfig.bEnableShaderDebugging)
{
VertexShaderCode code;
GenerateVertexShaderCode(code, components, API_D3D9);
vertex_uid_checker.AddToIndexAndCheck(code, uid, "Vertex", "v");
}
if (last_entry)
{
if (uid == last_uid)
{
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
return (last_entry->shader != NULL);
}
}
last_uid = uid;
VSCache::iterator iter = vshaders.find(uid);
if (iter != vshaders.end())
{
const VSCacheEntry &entry = iter->second;
last_entry = &entry;
if (entry.shader) D3D::SetVertexShader(entry.shader);
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
return (entry.shader != NULL);
}
VertexShaderCode code;
GenerateVertexShaderCode(code, components, API_D3D9);
u8 *bytecode;
int bytecodelen;
if (!D3D::CompileVertexShader(code.GetBuffer(), (int)strlen(code.GetBuffer()), &bytecode, &bytecodelen))
{
GFX_DEBUGGER_PAUSE_AT(NEXT_ERROR, true);
return false;
}
g_vs_disk_cache.Append(uid, bytecode, bytecodelen);
bool success = InsertByteCode(uid, bytecode, bytecodelen, true);
if (g_ActiveConfig.bEnableShaderDebugging && success)
{
vshaders[uid].code = code.GetBuffer();
}
delete [] bytecode;
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
return success;
}
bool VertexShaderCache::InsertByteCode(const VertexShaderUid &uid, const u8 *bytecode, int bytecodelen, bool activate) {
LPDIRECT3DVERTEXSHADER9 shader = D3D::CreateVertexShaderFromByteCode(bytecode, bytecodelen);
// Make an entry in the table
VSCacheEntry entry;
entry.shader = shader;
vshaders[uid] = entry;
last_entry = &vshaders[uid];
if (!shader)
return false;
INCSTAT(stats.numVertexShadersCreated);
SETSTAT(stats.numVertexShadersAlive, (int)vshaders.size());
if (activate)
{
D3D::SetVertexShader(shader);
return true;
}
return false;
}
float VSConstantbuffer[4*C_VENVCONST_END];
void Renderer::SetVSConstant4f(unsigned int const_number, float f1, float f2, float f3, float f4)
{
float* VSConstantbuffer_pointer = &VSConstantbuffer[const_number];
VSConstantbuffer_pointer[0] = f1;
VSConstantbuffer_pointer[1] = f2;
VSConstantbuffer_pointer[2] = f3;
VSConstantbuffer_pointer[3] = f4;
DX9::D3D::dev->SetVertexShaderConstantF(const_number, VSConstantbuffer_pointer, 1);
}
void Renderer::SetVSConstant4fv(unsigned int const_number, const float *f)
{
DX9::D3D::dev->SetVertexShaderConstantF(const_number, f, 1);
}
void Renderer::SetMultiVSConstant3fv(unsigned int const_number, unsigned int count, const float *f)
{
float* VSConstantbuffer_pointer = &VSConstantbuffer[const_number];
for (unsigned int i = 0; i < count; i++)
{
*VSConstantbuffer_pointer++ = *f++;
*VSConstantbuffer_pointer++ = *f++;
*VSConstantbuffer_pointer++ = *f++;
*VSConstantbuffer_pointer++ = 0.f;
}
DX9::D3D::dev->SetVertexShaderConstantF(const_number, &VSConstantbuffer[const_number], count);
}
void Renderer::SetMultiVSConstant4fv(unsigned int const_number, unsigned int count, const float *f)
{
DX9::D3D::dev->SetVertexShaderConstantF(const_number, f, count);
}
} // namespace DX9
namespace DX9
{
VertexShaderCache::VSCache VertexShaderCache::vshaders;
const VertexShaderCache::VSCacheEntry *VertexShaderCache::last_entry;
VertexShaderUid VertexShaderCache::last_uid;
VertexShaderUid VertexShaderCache::external_last_uid;
static HLSLAsyncCompiler *Compiler;
static Common::SpinLock<true> vshaderslock;
#define MAX_SSAA_SHADERS 2
static LPDIRECT3DVERTEXSHADER9 SimpleVertexShader[MAX_SSAA_SHADERS];
static LPDIRECT3DVERTEXSHADER9 ClearVertexShader;
LinearDiskCache<VertexShaderUid, u8> g_vs_disk_cache;
LPDIRECT3DVERTEXSHADER9 VertexShaderCache::GetSimpleVertexShader(int level)
{
return SimpleVertexShader[level ? 1 : 0];
}
LPDIRECT3DVERTEXSHADER9 VertexShaderCache::GetClearVertexShader()
{
return ClearVertexShader;
}
// this class will load the precompiled shaders into our cache
class VertexShaderCacheInserter : public LinearDiskCacheReader<VertexShaderUid, u8>
{
public:
void Read(const VertexShaderUid &key, const u8 *value, u32 value_size)
{
VertexShaderCache::InsertByteCode(key, value, value_size);
}
};
void VertexShaderCache::Init()
{
Compiler = &HLSLAsyncCompiler::getInstance();
vshaderslock.unlock();
const char* code = "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float2 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0;\n"
"OUT.vTexCoord1 = inTEX2;\n"
"return OUT;\n"
"}\0";
SimpleVertexShader[0] = D3D::CompileAndCreateVertexShader(code, (int)strlen(code));
code = "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vColor0 : COLOR0;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float4 inColor0: COLOR0)\n"
"{\n"
"VSOUTPUT OUT;\n"
"OUT.vPosition = inPosition;\n"
"OUT.vColor0 = inColor0;\n"
"return OUT;\n"
"}\0";
ClearVertexShader = D3D::CompileAndCreateVertexShader(code, (int)strlen(code));
code = "struct VSOUTPUT\n"
"{\n"
"float4 vPosition : POSITION;\n"
"float4 vTexCoord : TEXCOORD0;\n"
"float vTexCoord1 : TEXCOORD1;\n"
"float4 vTexCoord2 : TEXCOORD2;\n"
"float4 vTexCoord3 : TEXCOORD3;\n"
"};\n"
"VSOUTPUT main(float4 inPosition : POSITION,float2 inTEX0 : TEXCOORD0,float2 inTEX1 : TEXCOORD1,float inTEX2 : TEXCOORD2)\n"
"{\n"
"VSOUTPUT OUT;"
"OUT.vPosition = inPosition;\n"
"OUT.vTexCoord = inTEX0.xyyx;\n"
"OUT.vTexCoord1 = inTEX2.x;\n"
"OUT.vTexCoord2 = inTEX0.xyyx + (float4(-0.375f,-0.125f,-0.375f, 0.125f) * inTEX1.xyyx);\n"
"OUT.vTexCoord3 = inTEX0.xyyx + (float4( 0.375f, 0.125f, 0.375f,-0.125f) * inTEX1.xyyx);\n"
"return OUT;\n"
"}\0";
SimpleVertexShader[1] = D3D::CompileAndCreateVertexShader(code, (int)strlen(code));
Clear();
if (!File::Exists(File::GetUserPath(D_SHADERCACHE_IDX)))
File::CreateDir(File::GetUserPath(D_SHADERCACHE_IDX).c_str());
SETSTAT(stats.numVertexShadersCreated, 0);
SETSTAT(stats.numVertexShadersAlive, 0);
char cache_filename[MAX_PATH];
sprintf(cache_filename, "%sIDX9-%s-vs.cache", File::GetUserPath(D_SHADERCACHE_IDX).c_str(),
SConfig::GetInstance().m_strUniqueID.c_str());
VertexShaderCacheInserter inserter;
vshaderslock.lock();
g_vs_disk_cache.OpenAndRead(cache_filename, inserter);
vshaderslock.unlock();
if (g_Config.bEnableShaderDebugging)
Clear();
last_entry = NULL;
}
void VertexShaderCache::Clear()
{
vshaderslock.lock();
for (VSCache::iterator iter = vshaders.begin(); iter != vshaders.end(); ++iter)
iter->second.Destroy();
vshaders.clear();
vshaderslock.unlock();
last_entry = NULL;
}
void VertexShaderCache::Shutdown()
{
if (Compiler)
{
Compiler->WaitForFinish();
}
for (int i = 0; i < MAX_SSAA_SHADERS; i++)
{
if (SimpleVertexShader[i])
SimpleVertexShader[i]->Release();
SimpleVertexShader[i] = NULL;
}
if (ClearVertexShader)
ClearVertexShader->Release();
ClearVertexShader = NULL;
Clear();
g_vs_disk_cache.Sync();
g_vs_disk_cache.Close();
}
void VertexShaderCache::PrepareShader(u32 components, const XFMemory &xfr, const BPMemory &bpm, bool ongputhread)
{
VertexShaderUid uid;
GetVertexShaderUID(uid, components, xfr, bpm);
if (ongputhread)
{
Compiler->ProcCompilationResults();
#if defined(_DEBUG) || defined(DEBUGFAST)
if (g_ActiveConfig.bEnableShaderDebugging)
{
ShaderCode code;
GenerateVertexShaderCodeD3D9(code, uid.GetUidData());
}
#endif
if (last_entry)
{
if (uid == last_uid)
{
return;
}
}
last_uid = uid;
GFX_DEBUGGER_PAUSE_AT(NEXT_VERTEX_SHADER_CHANGE, true);
}
else
{
if (external_last_uid == uid)
{
return;
}
external_last_uid = uid;
}
vshaderslock.lock();
VSCacheEntry *entry = &vshaders[uid];
vshaderslock.unlock();
if (ongputhread)
{
last_entry = entry;
}
// Compile only when we have a new instance
if (entry->initialized.test_and_set())
{
return;
}
ShaderCompilerWorkUnit *wunit = Compiler->NewUnit(VERTEXSHADERGEN_BUFFERSIZE);
wunit->GenerateCodeHandler = [uid](ShaderCompilerWorkUnit* wunit)
{
ShaderCode code;
code.SetBuffer(wunit->code.data());
GenerateVertexShaderCodeD3D9(code, uid.GetUidData());
wunit->codesize = (u32)code.BufferSize();
};
wunit->entrypoint = "main";
wunit->flags = D3DCOMPILE_SKIP_VALIDATION | D3DCOMPILE_OPTIMIZATION_LEVEL3;
wunit->target = D3D::VertexShaderVersionString();
wunit->ResultHandler = [uid, entry](ShaderCompilerWorkUnit* wunit)
{
if (SUCCEEDED(wunit->cresult))
{
ID3DBlob* shaderBuffer = wunit->shaderbytecode;
const u8* bytecode = (const u8*)shaderBuffer->GetBufferPointer();
u32 bytecodelen = (u32)shaderBuffer->GetBufferSize();
g_vs_disk_cache.Append(uid, bytecode, bytecodelen);
PushByteCode(uid, bytecode, bytecodelen, entry);
#if defined(_DEBUG) || defined(DEBUGFAST)
if (g_ActiveConfig.bEnableShaderDebugging)
{
entry->code = wunit->code.data();
}
#endif
}
else
{
static int num_failures = 0;
std::string filename = StringFromFormat("%sbad_vs_%04i.txt", File::GetUserPath(D_DUMP_IDX).c_str(), num_failures++);
std::ofstream file;
OpenFStream(file, filename, std::ios_base::out);
file << ((const char*)wunit->code.data());
file << ((const char*)wunit->error->GetBufferPointer());
file.close();
PanicAlert("Failed to compile vertex shader!\nThis usually happens when trying to use Dolphin with an outdated GPU or integrated GPU like the Intel GMA series.\n\nIf you're sure this is Dolphin's error anyway, post the contents of %s along with this error message at the forums.\n\nDebug info (%s):\n%s",
filename,
D3D::VertexShaderVersionString(),
(char*)wunit->error->GetBufferPointer());
}
};
Compiler->CompileShaderAsync(wunit);
}
bool VertexShaderCache::TestShader()
{
int count = 0;
while (!last_entry->compiled)
{
Compiler->ProcCompilationResults();
if (g_ActiveConfig.bFullAsyncShaderCompilation)
{
break;
}
Common::cYield(count++);
}
if (last_entry->shader && last_entry->compiled)
{
D3D::SetVertexShader(last_entry->shader);
return true;
}
return false;
}
void VertexShaderCache::PushByteCode(const VertexShaderUid &uid, const u8 *bytecode, int bytecodelen, VertexShaderCache::VSCacheEntry* entry)
{
entry->shader = D3D::CreateVertexShaderFromByteCode(bytecode, bytecodelen);
entry->compiled = true;
if (entry->shader)
{
INCSTAT(stats.numVertexShadersCreated);
SETSTAT(stats.numVertexShadersAlive, (int)vshaders.size());
}
}
void VertexShaderCache::InsertByteCode(const VertexShaderUid &uid, const u8 *bytecode, int bytecodelen)
{
VSCacheEntry *entry = &vshaders[uid];
entry->initialized.test_and_set();
PushByteCode(uid, bytecode, bytecodelen, entry);
}
} // namespace DX9
