/*
* All or portions of this file Copyright (c) Amazon.com, Inc. or its affiliates or
* its licensors.
*
* For complete copyright and license terms please see the LICENSE at the root of this
* distribution (the "License"). All use of this software is governed by the License,
* or, if provided, by the license below or the license accompanying this file. Do not
* remove or modify any license notices. This file is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
*
*/
// Original file Copyright Crytek GMBH or its affiliates, used under license.
#include "StdAfx.h"
#include "DriverD3D.h"
#include "I3DEngine.h"
#include "CryCrc32.h"
#include "UnalignedBlit.h"
#if defined(AZ_RESTRICTED_PLATFORM)
#undef AZ_RESTRICTED_SECTION
#define D3DHWSHADERCOMPILING_CPP_SECTION_1 1
#define D3DHWSHADERCOMPILING_CPP_SECTION_2 2
#define D3DHWSHADERCOMPILING_CPP_SECTION_3 3
#define D3DHWSHADERCOMPILING_CPP_SECTION_4 4
#define D3DHWSHADERCOMPILING_CPP_SECTION_5 5
#define D3DHWSHADERCOMPILING_CPP_SECTION_6 6
#endif
#if defined(AZ_RESTRICTED_PLATFORM)
#define AZ_RESTRICTED_SECTION D3DHWSHADERCOMPILING_CPP_SECTION_1
#if defined(AZ_PLATFORM_XENIA)
#include "Xenia/D3DHWShaderCompiling_cpp_xenia.inl"
#elif defined(AZ_PLATFORM_PROVO)
#include "Provo/D3DHWShaderCompiling_cpp_provo.inl"
#elif defined(AZ_PLATFORM_SALEM)
#include "Salem/D3DHWShaderCompiling_cpp_salem.inl"
#endif
#endif
#if defined(AZ_RESTRICTED_SECTION_IMPLEMENTED)
#undef AZ_RESTRICTED_SECTION_IMPLEMENTED
#elif !defined(OPENGL)
#include <D3D11Shader.h>
#include <D3DCompiler.h>
#endif
#ifdef CRY_USE_METAL
#import <Foundation/Foundation.h>
#endif
#include "../Common/Shaders/RemoteCompiler.h"
#include "../Common/RenderCapabilities.h"
#include <AzFramework/IO/FileOperations.h>
static CryCriticalSection g_cAILock;
static SShaderAsyncInfo g_PendingList;
static SShaderAsyncInfo g_PendingListT;
#ifdef SHADER_ASYNC_COMPILATION
static SShaderAsyncInfo g_BuildList;
#endif
SShaderAsyncInfo& SShaderAsyncInfo::PendingList() { return g_PendingList; }
SShaderAsyncInfo& SShaderAsyncInfo::PendingListT() { return g_PendingListT; }
#ifdef SHADER_ASYNC_COMPILATION
SShaderAsyncInfo& CAsyncShaderTask::BuildList() { return g_BuildList; }
#endif
CryEvent SShaderAsyncInfo::s_RequestEv;
int CHWShader_D3D::s_nDevicePSDataSize;
int CHWShader_D3D::s_nDeviceVSDataSize;
class CSpinLock
{
public:
CSpinLock()
{
#if defined (WIN32) || defined(LINUX) || defined(APPLE) || defined(IMPLEMENT_SPIN_LOCK_WITH_INTERLOCKED_COMPARE_EXCHANGE)
while (CryInterlockedCompareExchange(&s_locked, 1L, 0L) == 1L)
{
Sleep(0);
}
#endif
}
~CSpinLock()
{
#if defined (WIN32)
InterlockedExchange(&s_locked, 0L);
#elif defined(LINUX) || defined(APPLE) || defined(IMPLEMENT_SPIN_LOCK_WITH_INTERLOCKED_COMPARE_EXCHANGE)
while (CryInterlockedCompareExchange(&s_locked, 0L, 1L) == 0L)
{
;
}
#endif
}
private:
static volatile LONG s_locked;
};
volatile LONG CSpinLock::s_locked = 0L;
volatile int SShaderAsyncInfo::s_nPendingAsyncShaders = 0;
//------------------------------------------------------------------------------
// This is the single method that is used to map a sampler and its HW slot.
// It is called when parsing a shader / loading from cache.
//------------------------------------------------------------------------------
bool CHWShader_D3D::mfAddFXSampler(SHWSInstance* pInst, SShaderFXParams& FXParams, SFXSampler* pr, const char* ParamName, SCGBind* pBind, CShader* ef, EHWShaderClass eSHClass)
{
assert(pBind);
if (!pBind)
{
return false;
}
std::vector<SCGSampler>* pParams;
pParams = &pInst->m_Samplers;
uint32 nOffs = pParams->size();
bool bRes = gRenDev->m_cEF.mfParseFXSampler(FXParams, pr, ParamName, ef, pBind->m_RegisterCount, pParams, eSHClass);
if (pParams->size() > nOffs)
{
for (uint32 i = 0; i < pParams->size() - nOffs; i++)
{
SCGSampler& p = (*pParams)[nOffs + i];
p.m_RegisterOffset = pBind->m_RegisterOffset + i;
p.m_BindingSlot = pBind->m_BindingSlot;
p.m_Name = pBind->m_Name;
}
}
// Parameter without semantic
return bRes;
}
//------------------------------------------------------------------------------
// The main texture binding method.
// This is the single method that is used to map a texture and its HW slot.
// It is called when going over a shader / loading from cache.
//------------------------------------------------------------------------------
bool CHWShader_D3D::mfAddFXTexture(
SHWSInstance* pInst, SShaderFXParams& FXParams, SFXTexture* pr, const char* ParamName,
SCGBind* pBind, CShader* ef, EHWShaderClass eSHClass)
{
assert(pBind);
if (!pBind)
{
return false;
}
std::vector<SCGTexture>* pParams;
pParams = &pInst->m_Textures;
uint32 nOffs = pParams->size();
bool bRes = gRenDev->m_cEF.mfParseFXTexture(FXParams, pr, ParamName, ef, pBind->m_RegisterCount, pParams, eSHClass);
if (pParams->size() > nOffs)
{ // If the texture was added
for (uint32 i = 0; i < pParams->size() - nOffs; i++)
{
SCGTexture& p = (*pParams)[nOffs + i];
p.m_RegisterOffset = pBind->m_RegisterOffset + i; // Offset by one and set it for this texture
p.m_BindingSlot = pBind->m_BindingSlot; // Set the passed binding slot
p.m_Name = pBind->m_Name;
}
}
// Parameter without semantic
return bRes;
}
//------------------------------------------------------------------------------
// This is the single method that is used to map constant parameter and its HW slot.
// It is called when parsing a shader / loading from cache.
//------------------------------------------------------------------------------
void CHWShader_D3D::mfAddFXParameter(
SHWSInstance* pInst, SParamsGroup& OutParams, SShaderFXParams& FXParams,
SFXParam* pr, const char* ParamName, SCGBind* pBind, CShader* ef, bool bInstParam, EHWShaderClass eSHClass)
{
SCGParam CGpr;
assert(pBind);
if (!pBind)
{
return;
}
int nComps = 0;
int nParams = pBind->m_RegisterCount;
if (!pr->m_Semantic.empty())
{
nComps = pr->m_ComponentCount;
}
else
{
for (int i = 0; i < pr->m_ComponentCount; i++)
{
CryFixedStringT<128> cur;
pr->GetParamComp(i, cur);
if (cur.empty())
{
break;
}
nComps++;
}
}
// Process parameters only with semantics
if (nComps && nParams)
{
std::vector<SCGParam>* pParams;
if (bInstParam)
{
pParams = &OutParams.Params_Inst;
}
else
{
pParams = &OutParams.Params[0];
}
uint32 nOffs = pParams->size();
bool bRes = gRenDev->m_cEF.mfParseFXParameter(FXParams, pr, ParamName, ef, bInstParam, pBind->m_RegisterCount, pParams, eSHClass, false);
AZ_Assert(bRes, "Error: CHWShader_D3D::mfAddFXParameter - bRes is false");
if (pParams->size() > nOffs)
{
for (uint32 i = 0; i < pParams->size() - nOffs; i++)
{
//assert(pBind->m_nComponents == 1);
SCGParam& p = (*pParams)[nOffs + i];
p.m_RegisterOffset = pBind->m_RegisterOffset + i;
p.m_BindingSlot = pBind->m_BindingSlot;
}
}
}
// Parameter without semantic
}
struct SAliasSampler
{
STexSamplerRT Sampler;
string NameTex;
};
bool CHWShader_D3D::mfAddFXParameter(SHWSInstance* pInst, SParamsGroup& OutParams, SShaderFXParams& FXParams, const char* param, SCGBind* bn, bool bInstParam, EHWShaderClass eSHClass, CShader* pFXShader)
{
if (bn->m_RegisterOffset & SHADER_BIND_TEXTURE)
{
SFXTexture* pr = gRenDev->m_cEF.mfGetFXTexture(FXParams.m_FXTextures, param);
if (pr)
{
if (bn->m_RegisterCount < 0)
{
bn->m_RegisterCount = pr->m_nArray;
}
bool bRes = mfAddFXTexture(pInst, FXParams, pr, param, bn, pFXShader, eSHClass);
return true;
}
}
else
if (bn->m_RegisterOffset & SHADER_BIND_SAMPLER)
{
SFXSampler* pr = gRenDev->m_cEF.mfGetFXSampler(FXParams.m_FXSamplers, param);
if (pr)
{
if (bn->m_RegisterCount < 0)
{
bn->m_RegisterCount = pr->m_nArray;
}
bool bRes = mfAddFXSampler(pInst, FXParams, pr, param, bn, pFXShader, eSHClass);
return true;
}
}
else
{
SFXParam* pr = gRenDev->m_cEF.mfGetFXParameter(FXParams.m_FXParams, param);
if (pr)
{
if (bn->m_RegisterCount < 0)
{
bn->m_RegisterCount = pr->m_RegisterCount;
}
mfAddFXParameter(pInst, OutParams, FXParams, pr, param, bn, pFXShader, bInstParam, eSHClass);
return true;
}
}
return false;
}
//==================================================================================================================
int CGBindCallback(const VOID* arg1, const VOID* arg2)
{
SCGBind* pi1 = (SCGBind*)arg1;
SCGBind* pi2 = (SCGBind*)arg2;
if (pi1->m_RegisterOffset < pi2->m_RegisterOffset)
{
return -1;
}
if (pi1->m_RegisterOffset > pi2->m_RegisterOffset)
{
return 1;
}
return 0;
}
const char* ReflectedConstantBufferNames[eConstantBufferShaderSlot_ReflectedCount] = {"PER_BATCH", "PER_INSTANCE", "PER_MATERIAL"};
void CHWShader_D3D::mfCreateBinds(SHWSInstance* pInst, void* pConstantTable, byte* pShader, int nSize)
{
uint32 i;
ID3D11ShaderReflection* pShaderReflection = (ID3D11ShaderReflection*)pConstantTable;
D3D11_SHADER_DESC Desc;
pShaderReflection->GetDesc(&Desc);
ID3D11ShaderReflectionConstantBuffer* pCB = NULL;
for (uint32 n = 0; n < Desc.ConstantBuffers; n++)
{
pCB = pShaderReflection->GetConstantBufferByIndex(n);
D3D11_SHADER_BUFFER_DESC SBDesc;
pCB->GetDesc(&SBDesc);
if (SBDesc.Type == D3D11_CT_RESOURCE_BIND_INFO)
{
continue;
}
int nCB;
if (!strcmp("$Globals", SBDesc.Name))
{
nCB = eConstantBufferShaderSlot_PerBatch;
}
else
{
for (nCB = 0; nCB < eConstantBufferShaderSlot_ReflectedCount; nCB++)
{
if (!strcmp(ReflectedConstantBufferNames[nCB], SBDesc.Name))
{
break;
}
}
}
if (nCB >= eConstantBufferShaderSlot_ReflectedCount) // Allow having custom cbuffers in shaders
{
continue;
}
for (i = 0; i < SBDesc.Variables; i++)
{
uint32 nCount = 1;
ID3D11ShaderReflectionVariable* pCV = pCB->GetVariableByIndex(i);
ID3D11ShaderReflectionType* pVT = pCV->GetType();
D3D11_SHADER_VARIABLE_DESC CDesc;
D3D11_SHADER_TYPE_DESC CTDesc;
pVT->GetDesc(&CTDesc);
pCV->GetDesc(&CDesc);
if (!(CDesc.uFlags & D3D10_SVF_USED))
{
continue;
}
if (CTDesc.Class == D3D10_SVC_VECTOR || CTDesc.Class == D3D10_SVC_SCALAR || CTDesc.Class == D3D10_SVC_MATRIX_COLUMNS || CTDesc.Class == D3D10_SVC_MATRIX_ROWS)
{
SCGBind cgp;
assert(!(CDesc.StartOffset & 0xf));
//assert(!(CDesc.Size & 0xf));
int nReg = CDesc.StartOffset >> 4;
cgp.m_RegisterOffset = nReg; //<<2;
cgp.m_BindingSlot = nCB;
cgp.m_RegisterCount = (CDesc.Size + 15) >> 4;
cgp.m_Name = CDesc.Name;
cgp.m_Flags = CParserBin::GetCRC32(CDesc.Name);
pInst->m_pBindVars.push_back(cgp);
}
else
{
assert(false);
}
}
}
D3D11_SHADER_INPUT_BIND_DESC IBDesc;
for (i = 0; i < Desc.BoundResources; i++)
{
ZeroStruct(IBDesc);
pShaderReflection->GetResourceBindingDesc(i, &IBDesc);
SCGBind cgp;
if (IBDesc.Type == D3D10_SIT_TEXTURE)
{
cgp.m_RegisterOffset = IBDesc.BindPoint | SHADER_BIND_TEXTURE;
}
else
if (IBDesc.Type == D3D10_SIT_SAMPLER)
{
cgp.m_RegisterOffset = IBDesc.BindPoint | SHADER_BIND_SAMPLER;
}
else
{
continue;
}
if (IBDesc.Dimension == D3D10_SRV_DIMENSION_BUFFER)
{
continue;
}
cgp.m_BindingSlot = IBDesc.BindPoint;
cgp.m_RegisterCount = IBDesc.BindCount;
cgp.m_Name = IBDesc.Name;
cgp.m_Flags = CParserBin::GetCRC32(IBDesc.Name);
pInst->m_pBindVars.push_back(cgp);
}
}
void CHWShader_D3D::mfGatherFXParameters(SHWSInstance* pInst, std::vector<SCGBind>* InstBindVars, CHWShader_D3D* pSH, int nFlags, CShader* pFXShader)
{
// LOADING_TIME_PROFILE_SECTION(iSystem);
std::vector<SCGBind>& parameters = pInst->m_pBindVars;
uint32 i, j;
SAliasSampler samps[MAX_TMU];
int nMaxSampler = -1;
int nParam = 0;
SParamsGroup Group;
SShaderFXParams& FXParams = gRenDev->m_cEF.m_Bin.mfGetFXParams(pInst->m_bFallback ? CShaderMan::s_ShaderFallback : pFXShader);
if (parameters.size())
{
std::list<uint32> skipped;
for (i = 0; i < parameters.size(); i++)
{
SCGBind* bn = ¶meters[i];
const char* param = bn->m_Name.c_str();
if (!strncmp(param, "_g_", 3))
{
continue;
}
bool bRes = mfAddFXParameter(pInst, Group, FXParams, param, bn, false, pSH->m_eSHClass, pFXShader);
if (!bRes && (bn->m_RegisterOffset & SHADER_BIND_TEXTURE)) // try to find old samplers (Without semantics)
{
skipped.push_back(i);
}
}
bool setSamplers[256] = { false };
bool setTextures[256] = { false };
for (i = 0; i < pInst->m_Samplers.size(); i++)
{
setSamplers[pInst->m_Samplers[i].m_RegisterOffset & 0xff] = true;
}
for (i = 0; i < pInst->m_Textures.size(); i++)
{
setTextures[pInst->m_Textures[i].m_RegisterOffset & 0xff] = true;
}
while (skipped.size() > 0)
{
i = skipped.front();
skipped.pop_front();
SCGBind *bn = ¶meters[i];
const char *param = bn->m_Name.c_str();
{
for (j = 0; j < (uint32)FXParams.m_FXSamplersOld.size(); j++)
{
STexSamplerFX* sm = &FXParams.m_FXSamplersOld[j];
if (!azstricmp(sm->m_szName.c_str(), param))
{
int nSampler = bn->m_RegisterOffset & 0x7f;
if (nSampler < MAX_TMU)
{
nMaxSampler = max(nSampler, nMaxSampler);
samps[nSampler].Sampler = STexSamplerRT(*sm);
samps[nSampler].NameTex = sm->m_szTexture;
samps[nSampler].Sampler.m_nSamplerSlot = (int8)(bn->m_BindingSlot & 0xff);
samps[nSampler].Sampler.m_nTextureSlot = nSampler;
for (int k = 0; k < parameters.size(); k++)
{
if (parameters[k].m_RegisterOffset & SHADER_BIND_TEXTURE)
{
if (!azstricmp(parameters[k].m_Name.c_str(), param))
{
samps[nSampler].Sampler.m_nTextureSlot = (int8)parameters[k].m_BindingSlot;
}
}
}
for (int k = 0; k < parameters.size(); k++)
{
if (parameters[k].m_RegisterOffset & SHADER_BIND_SAMPLER)
{
if (!azstricmp(parameters[k].m_Name.c_str(), param))
{
samps[nSampler].Sampler.m_nSamplerSlot = (int8)parameters[k].m_BindingSlot;
}
}
}
// Texture slot occupied, search an alternative
if (setSamplers[samps[nSampler].Sampler.m_nSamplerSlot])
{
uint32 f = 0; while (setSamplers[f]) ++f;
samps[nSampler].Sampler.m_nSamplerSlot = f;
assert(f < 16);
}
// Sampler slot occupied, search an alternative
if (setTextures[samps[nSampler].Sampler.m_nTextureSlot])
{
uint32 f = 0; while (setTextures[f]) ++f;
samps[nSampler].Sampler.m_nTextureSlot = f;
assert(f < 256);
}
setTextures[samps[nSampler].Sampler.m_nTextureSlot] = true;
setSamplers[samps[nSampler].Sampler.m_nSamplerSlot] = true;
break;
}
}
}
if (j == FXParams.m_FXSamplersOld.size())
{
for (j = 0; j < (uint32)FXParams.m_FXSamplersOld.size(); j++)
{
STexSamplerFX* sm = &FXParams.m_FXSamplersOld[j];
const char* src = sm->m_szName.c_str();
char name[128];
int n = 0;
while (src[n])
{
if (src[n] <= 0x20 || src[n] == '[')
{
break;
}
name[n] = src[n];
n++;
}
name[n] = 0;
if (!azstricmp(name, param))
{
int nSampler = bn->m_RegisterOffset & 0x7f;
if (nSampler < MAX_TMU)
{
samps[nSampler].Sampler = STexSamplerRT(*sm);
samps[nSampler].NameTex = sm->m_szTexture;
samps[nSampler].Sampler.m_nSamplerSlot = (int8)(bn->m_BindingSlot & 0xff);
samps[nSampler].Sampler.m_nTextureSlot = nSampler;
for (int k = 0; k < parameters.size(); k++)
{
if (parameters[k].m_RegisterOffset & SHADER_BIND_TEXTURE)
{
if (!azstricmp(parameters[k].m_Name.c_str(), param))
{
samps[nSampler].Sampler.m_nTextureSlot = (int8)parameters[k].m_BindingSlot;
}
}
}
for (int k = 0; k < parameters.size(); k++)
{
if (parameters[k].m_RegisterOffset & SHADER_BIND_SAMPLER)
{
if (!azstricmp(parameters[k].m_Name.c_str(), param))
{
samps[nSampler].Sampler.m_nSamplerSlot = (int8)parameters[k].m_BindingSlot;
}
}
}
// Texture slot occupied, search an alternative
if (setSamplers[samps[nSampler].Sampler.m_nSamplerSlot])
{
uint32 f = 0; while (setSamplers[f]) ++f;
samps[nSampler].Sampler.m_nSamplerSlot = f;
assert(f < 16);
}
// Sampler slot occupied, search an alternative
if (setTextures[samps[nSampler].Sampler.m_nTextureSlot])
{
uint32 f = 0; while (setTextures[f]) ++f;
samps[nSampler].Sampler.m_nTextureSlot = f;
assert(f < 256);
}
setTextures[samps[nSampler].Sampler.m_nTextureSlot] = true;
setSamplers[samps[nSampler].Sampler.m_nSamplerSlot] = true;
for (int nS = 0; nS < bn->m_RegisterCount; nS++)
{
nMaxSampler = max(nSampler + nS, nMaxSampler);
samps[nSampler + nS].Sampler = samps[nSampler].Sampler;
samps[nSampler + nS].NameTex = sm->m_szTexture;
}
break;
}
}
}
}
}
}
}
if (nFlags != 1)
{
for (i = 0; (int)i <= nMaxSampler; i++)
{
STexSamplerRT& smp = samps[i].Sampler;
smp.m_pTex = gRenDev->m_cEF.mfParseFXTechnique_LoadShaderTexture(&smp, samps[i].NameTex.c_str(), NULL, NULL, i, eCO_NOSET, eCO_NOSET, DEF_TEXARG0, DEF_TEXARG0);
if (!smp.m_pTex)
{
continue;
}
if (smp.m_bGlobal)
{
mfAddGlobalSampler(smp);
}
else
{
pInst->m_pSamplers.push_back(smp);
}
}
}
else
{
assert(pInst->m_pAsync);
if (pInst->m_pAsync && nMaxSampler >= 0)
{
pInst->m_pAsync->m_bPendedSamplers = true;
}
}
pInst->m_nMaxVecs[0] = pInst->m_nMaxVecs[1] = 0;
if (parameters.size())
{
for (i = 0; i < parameters.size(); i++)
{
SCGBind* pB = ¶meters[i];
if (pB->m_RegisterOffset & (SHADER_BIND_SAMPLER | SHADER_BIND_TEXTURE))
{
continue;
}
if (pB->m_BindingSlot < 0 || pB->m_BindingSlot > 2)
{
continue;
}
for (j = 0; j < Group.Params[0].size(); j++)
{
SCGParam* pr = &Group.Params[0][j];
if (pr->m_RegisterOffset == pB->m_RegisterOffset && pr->m_Name == pB->m_Name)
{
break;
}
}
if (j != Group.Params[0].size())
{
continue;
}
if (pB->m_BindingSlot < 3)
{
pInst->m_nMaxVecs[pB->m_BindingSlot] = max(pB->m_RegisterOffset + pB->m_RegisterCount, pInst->m_nMaxVecs[pB->m_BindingSlot]);
}
}
}
if (Group.Params[0].size())
{
for (i = 0; i < Group.Params[0].size(); i++)
{
SCGParam* pr = &Group.Params[0][i];
if (pr->m_Flags & PF_MATERIAL)
{
pInst->m_bHasPMParams = true;
}
}
gRenDev->m_cEF.mfCheckObjectDependParams(Group.Params[0], Group.Params[1], pSH->m_eSHClass, pFXShader);
}
for (i = 0; i < 2; i++)
{
if (Group.Params[i].size())
{
for (j = 0; j < Group.Params[i].size(); j++)
{
const SCGParam* pr = &Group.Params[i][j];
pInst->m_nMaxVecs[i] = max(pr->m_RegisterOffset + pr->m_RegisterCount, pInst->m_nMaxVecs[i]);
}
}
}
int nMax = AzRHI::GetConstantRegisterCountMax(pSH->m_eSHClass);
assert(pInst->m_nMaxVecs[0] < nMax);
assert(pInst->m_nMaxVecs[1] < nMax);
if ((pInst->m_Ident.m_RTMask & g_HWSR_MaskBit[HWSR_INSTANCING_ATTR]) && pSH->m_eSHClass == eHWSC_Vertex)
{
int nNumInst = 0;
if (InstBindVars)
{
for (i = 0; i < (uint32)InstBindVars->size(); i++)
{
SCGBind& b = (*InstBindVars)[i];
int nID = b.m_RegisterOffset;
if (!nNumInst)
{
pInst->m_nInstMatrixID = nID;
}
SCGBind bn;
bn.m_RegisterCount = b.m_RegisterCount;
bn.m_RegisterOffset = nID;
bool bRes = mfAddFXParameter(pInst, Group, FXParams, b.m_Name.c_str(), &bn, true, pSH->m_eSHClass, pFXShader);
nNumInst++;
}
}
//assert(cgi->m_nNumInstAttributes == nNumInst);
pInst->m_nNumInstAttributes = nNumInst;
if (Group.Params_Inst.size())
{
qsort(&Group.Params_Inst[0], Group.Params_Inst.size(), sizeof(SCGParam), CGBindCallback);
pInst->m_nParams_Inst = CGParamManager::GetParametersGroup(Group, 2);
}
}
if (Group.Params[0].size() > 0)
{
qsort(&Group.Params[0][0], Group.Params[0].size(), sizeof(SCGParam), CGBindCallback);
pInst->m_nParams[0] = CGParamManager::GetParametersGroup(Group, 0);
}
if (Group.Params[1].size() > 0)
{
qsort(&Group.Params[1][0], Group.Params[1].size(), sizeof(SCGParam), CGBindCallback);
pInst->m_nParams[1] = CGParamManager::GetParametersGroup(Group, 1);
}
}
// Vertex shader specific
void CHWShader_D3D::mfUpdateFXVertexFormat(SHWSInstance* pInst, CShader* pSH)
{
// Update global FX shader's vertex format / flags
if (pSH)
{
AZ::Vertex::Format vertexFormat = pSH->m_vertexFormat;
bool bCurrent = false;
for (uint32 i = 0; i < pSH->m_HWTechniques.Num(); i++)
{
SShaderTechnique* hw = pSH->m_HWTechniques[i];
for (uint32 j = 0; j < hw->m_Passes.Num(); j++)
{
SShaderPass* pass = &hw->m_Passes[j];
if (pass->m_VShader)
{
if (pass->m_VShader == this)
{
bCurrent = true;
}
bool bUseLM = false;
bool bUseTangs = false;
bool bUseHWSkin = false;
bool bUseVertexVelocity = false;
AZ::Vertex::Format currentVertexFormat = pass->m_VShader->mfVertexFormat(bUseTangs, bUseLM, bUseHWSkin, bUseVertexVelocity);
if (currentVertexFormat > vertexFormat)
{
vertexFormat = currentVertexFormat;
}
if (bUseTangs)
{
pass->m_PassFlags |= VSM_TANGENTS;
}
if (bUseHWSkin)
{
pass->m_PassFlags |= VSM_HWSKIN;
}
if (bUseVertexVelocity)
{
pass->m_PassFlags |= VSM_VERTEX_VELOCITY;
}
}
}
}
//assert (bCurrent);
pSH->m_vertexFormat = vertexFormat;
}
}
void CHWShader_D3D::mfPostVertexFormat(SHWSInstance* pInst, CHWShader_D3D* pHWSH, bool bCol, byte bNormal, bool bTC0, bool bTC1, bool bPSize, bool bTangent[2], bool bBitangent[2], bool bHWSkin, bool bSH[2], bool bVelocity, bool bMorph)
{
if (bBitangent[0])
{
pInst->m_VStreamMask_Decl |= 1 << VSF_TANGENTS;
}
else
if (bTangent[0])
{
pInst->m_VStreamMask_Decl |= 1 << VSF_QTANGENTS;
}
if (bBitangent[1])
{
pInst->m_VStreamMask_Stream |= 1 << VSF_TANGENTS;
}
else
if (bTangent[1])
{
pInst->m_VStreamMask_Stream |= 1 << VSF_QTANGENTS;
}
if (pInst->m_VStreamMask_Decl & (1 << VSF_TANGENTS))
{
bNormal = false;
}
if (bHWSkin)
{
pInst->m_VStreamMask_Decl |= VSM_HWSKIN;
pInst->m_VStreamMask_Stream |= VSM_HWSKIN;
}
if (bVelocity)
{
pInst->m_VStreamMask_Decl |= VSM_VERTEX_VELOCITY;
pInst->m_VStreamMask_Stream |= VSM_VERTEX_VELOCITY;
}
if (bMorph)
{
pInst->m_VStreamMask_Decl |= VSM_MORPHBUDDY;
pInst->m_VStreamMask_Stream |= VSM_MORPHBUDDY;
}
pInst->m_vertexFormat = AZ::Vertex::VertFormatForComponents(bCol, bTC0, bTC1, bPSize, bNormal != 0);
}
AZ::Vertex::Format CHWShader_D3D::mfVertexFormat(bool& bUseTangents, bool& bUseLM, bool& bUseHWSkin, bool& bUseVertexVelocity)
{
uint32 i;
assert (m_eSHClass == eHWSC_Vertex);
AZ::Vertex::Format vertexFormat(eVF_Unknown);
int nStream = 0;
for (i = 0; i < m_Insts.size(); i++)
{
SHWSInstance* pInst = m_Insts[i];
if (pInst->m_vertexFormat > vertexFormat)
{
vertexFormat = pInst->m_vertexFormat;
}
nStream |= pInst->m_VStreamMask_Stream;
}
bUseTangents = (nStream & VSM_TANGENTS) != 0;
bUseLM = false;
bUseHWSkin = (nStream & VSM_HWSKIN) != 0;
bUseVertexVelocity = (nStream & VSM_VERTEX_VELOCITY) != 0;
return vertexFormat;
}
AZ::Vertex::Format CHWShader_D3D::mfVertexFormat(SHWSInstance* pInst, CHWShader_D3D* pSH, LPD3D10BLOB pShader)
{
assert (pSH->m_eSHClass == eHWSC_Vertex);
byte bNormal = false;
bool bTangent[2] = {false, false};
bool bBitangent[2] = {false, false};
bool bHWSkin = false;
bool bVelocity = false;
bool bMorph = false;
bool bBoneSpace = false;
bool bPSize = false;
bool bSH[2] = {false, false};
bool bTC0 = false;
bool bTC1[2] = {false, false};
bool bCol = false;
bool bSecCol = false;
bool bPos = false;
size_t nSize = pShader->GetBufferSize();
void* pData = pShader->GetBufferPointer();
void* pShaderReflBuf;
HRESULT hr = D3DReflect(pData, nSize, IID_ID3D11ShaderReflection, &pShaderReflBuf);
assert (SUCCEEDED(hr));
ID3D11ShaderReflection* pShaderReflection = (ID3D11ShaderReflection*)pShaderReflBuf;
if (!SUCCEEDED(hr))
{
return AZ::Vertex::Format(eVF_Unknown);
}
D3D11_SHADER_DESC Desc;
pShaderReflection->GetDesc(&Desc);
if (!Desc.InputParameters)
{
return AZ::Vertex::Format(eVF_Unknown);
}
D3D11_SIGNATURE_PARAMETER_DESC IDesc;
for (uint32 i = 0; i < Desc.InputParameters; i++)
{
pShaderReflection->GetInputParameterDesc(i, &IDesc);
//if (!IDesc.ReadWriteMask)
// continue;
if (!IDesc.SemanticName)
{
continue;
}
int nIndex;
if (!_strnicmp(IDesc.SemanticName, "POSITION", 8)
#if defined(AZ_RESTRICTED_PLATFORM)
#define AZ_RESTRICTED_SECTION D3DHWSHADERCOMPILING_CPP_SECTION_2
#if defined(AZ_PLATFORM_XENIA)
#include "Xenia/D3DHWShaderCompiling_cpp_xenia.inl"
#elif defined(AZ_PLATFORM_PROVO)
#include "Provo/D3DHWShaderCompiling_cpp_provo.inl"
#elif defined(AZ_PLATFORM_SALEM)
#include "Salem/D3DHWShaderCompiling_cpp_salem.inl"
#endif
#endif
#if defined(AZ_RESTRICTED_SECTION_IMPLEMENTED)
#undef AZ_RESTRICTED_SECTION_IMPLEMENTED
#else
|| !_strnicmp(IDesc.SemanticName, "SV_POSITION", 11)
#endif
)
{
nIndex = IDesc.SemanticIndex;
if (nIndex == 0)
{
bPos = true;
}
else
if (nIndex == 3)
{
bVelocity = true;
}
else
if (nIndex == 4)
{
bHWSkin = true;
}
else
if (nIndex == 8)
{
bMorph = true;
}
else
{
assert(false);
}
}
else
if (!_strnicmp(IDesc.SemanticName, "NORMAL", 6))
{
bNormal = true;
}
else
if (!_strnicmp(IDesc.SemanticName, "TEXCOORD", 8))
{
nIndex = IDesc.SemanticIndex;
if (nIndex == 0)
{
bTC0 = true;
}
else
if (!(pInst->m_Ident.m_RTMask & g_HWSR_MaskBit[HWSR_INSTANCING_ATTR]))
{
if (nIndex == 1)
{
bTC1[0] = true;
if (IDesc.ReadWriteMask)
{
bTC1[1] = true;
}
}
else
if (nIndex == 8)
{
bMorph = true;
}
}
}
else
if (!_strnicmp(IDesc.SemanticName, "COLOR", 5))
{
nIndex = IDesc.SemanticIndex;
if (nIndex == 0)
{
bCol = true;
}
else
if (nIndex == 1)
{
bSecCol = true;
}
else
{
if (nIndex == 2 || nIndex == 3)
{
bSH[0] = true;
if (IDesc.ReadWriteMask)
{
bSH[1] = true;
}
}
else
{
assert(false);
}
}
}
else
if (!azstricmp(IDesc.SemanticName, "TANGENT"))
{
bTangent[0] = true;
if (IDesc.ReadWriteMask)
{
bTangent[1] = true;
}
}
else
if (!azstricmp(IDesc.SemanticName, "BITANGENT") || !azstricmp(IDesc.SemanticName, "BINORMAL"))
{
bBitangent[0] = true;
if (IDesc.ReadWriteMask)
{
bBitangent[1] = true;
}
}
else
if (!_strnicmp(IDesc.SemanticName, "PSIZE", 5))
{
bPSize = true;
}
else
if (!_strnicmp(IDesc.SemanticName, "BLENDWEIGHT", 11) || !_strnicmp(IDesc.SemanticName, "BLENDINDICES", 12))
{
nIndex = IDesc.SemanticIndex;
if (nIndex == 0)
{
bHWSkin = true;
}
else
if (nIndex == 1)
{
bMorph = true;
}
else
{
assert(0);
}
}
else
#if defined(AZ_RESTRICTED_PLATFORM)
#define AZ_RESTRICTED_SECTION D3DHWSHADERCOMPILING_CPP_SECTION_3
#if defined(AZ_PLATFORM_XENIA)
#include "Xenia/D3DHWShaderCompiling_cpp_xenia.inl"
#elif defined(AZ_PLATFORM_PROVO)
#include "Provo/D3DHWShaderCompiling_cpp_provo.inl"
#elif defined(AZ_PLATFORM_SALEM)
#include "Salem/D3DHWShaderCompiling_cpp_salem.inl"
#endif
#endif
#if defined(AZ_RESTRICTED_SECTION_IMPLEMENTED)
#undef AZ_RESTRICTED_SECTION_IMPLEMENTED
#else
if (!_strnicmp(IDesc.SemanticName, "SV_", 3))
#endif
{
// SV_ are valid semantics
}
else
{
assert(0);
}
}
mfPostVertexFormat(pInst, pSH, bCol, bNormal, bTC0, bTC1[0], bPSize, bTangent, bBitangent, bHWSkin, bSH, bVelocity, bMorph);
SAFE_RELEASE(pShaderReflection);
return pInst->m_vertexFormat;
}
void CHWShader_D3D::mfSetDefaultRT(uint64& nAndMask, uint64& nOrMask)
{
uint32 i, j;
SShaderGen* pGen = gRenDev->m_cEF.m_pGlobalExt;
uint32 nBitsPlatform = 0;
if (CParserBin::m_nPlatform == SF_ORBIS)
{
nBitsPlatform |= SHGD_HW_ORBIS;
}
else
if (CParserBin::m_nPlatform == SF_DURANGO)
{
nBitsPlatform |= SHGD_HW_DURANGO;
}
else
if (CParserBin::m_nPlatform == SF_D3D11)
{
nBitsPlatform |= SHGD_HW_DX11;
}
else
if (CParserBin::m_nPlatform == SF_GL4)
{
nBitsPlatform |= SHGD_HW_GL4;
}
else
if (CParserBin::m_nPlatform == SF_GLES3)
{
nBitsPlatform |= SHGD_HW_GLES3;
}
// Confetti Nicholas Baldwin: adding metal shader language support
else
if (CParserBin::m_nPlatform == SF_METAL)
{
nBitsPlatform |= SHGD_HW_METAL;
}
// Make a mask of flags affected by this type of shader
uint32 nType = m_dwShaderType;
if (nType)
{
for (i = 0; i < pGen->m_BitMask.size(); i++)
{
SShaderGenBit* pBit = pGen->m_BitMask[i];
if (!pBit->m_Mask)
{
continue;
}
if (nBitsPlatform & pBit->m_nDependencyReset)
{
nAndMask &= ~pBit->m_Mask;
continue;
}
for (j = 0; j < pBit->m_PrecacheNames.size(); j++)
{
if (pBit->m_PrecacheNames[j] == nType)
{
if (nBitsPlatform & pBit->m_nDependencySet)
{
nOrMask |= pBit->m_Mask;
}
break;
}
}
}
}
}
//==================================================================================================================
static bool sGetMask(char* str, SShaderGen* pGen, uint64& nMask)
{
uint32 i;
for (i = 0; i < pGen->m_BitMask.Num(); i++)
{
SShaderGenBit* pBit = pGen->m_BitMask[i];
if (!strcmp(str, pBit->m_ParamName.c_str()))
{
nMask |= pBit->m_Mask;
return true;
}
}
return false;
}
void CHWShader::mfValidateTokenData(CResFile* pRes)
{
#ifdef _DEBUG
if (pRes == 0)
{
return;
}
bool bTokenValid = true;
ResDir* Dir = pRes->mfGetDirectory();
for (unsigned int i = 0; i < Dir->size(); i++)
{
SDirEntry* pDE = &(*Dir)[i];
if (pDE->flags & RF_RES_$TOKENS)
{
uint32 nSize = pRes->mfFileRead(pDE);
byte* pData = (byte*)pRes->mfFileGetBuf(pDE);
if (!pData)
{
bTokenValid = false;
break;
}
uint32 nL = *(uint32*)pData;
if (CParserBin::m_bEndians)
{
SwapEndian(nL, eBigEndian);
}
if (nL * sizeof(uint32) > nSize)
{
bTokenValid = false;
break;
}
int nTableSize = nSize - (4 + nL * sizeof(uint32));
if (nTableSize < 0)
{
bTokenValid = false;
break;
}
pRes->mfCloseEntry(pDE);
}
}
if (!bTokenValid)
{
CryFatalError("Invalid token data in shader cache file");
}
#endif
}
bool CHWShader_D3D::mfStoreCacheTokenMap(FXShaderToken*& Table, TArray<uint32>*& pSHData, const char* szName)
{
TArray<byte> Data;
FXShaderTokenItor itor;
uint32 nSize = pSHData->size();
if (CParserBin::m_bEndians)
{
uint32 nSizeEnd = nSize;
SwapEndian(nSizeEnd, eBigEndian);
Data.Copy((byte*)&nSizeEnd, sizeof(uint32));
for (uint32 i = 0; i < nSize; i++)
{
uint32 nToken = (*pSHData)[i];
SwapEndian(nToken, eBigEndian);
Data.Copy((byte*)&nToken, sizeof(uint32));
}
}
else
{
Data.Copy((byte*)&nSize, sizeof(uint32));
Data.Copy((byte*)&(*pSHData)[0], nSize * sizeof(uint32));
}
for (itor = Table->begin(); itor != Table->end(); itor++)
{
STokenD T = *itor;
if (CParserBin::m_bEndians)
{
SwapEndian(T.Token, eBigEndian);
}
Data.Copy((byte*)&T.Token, sizeof(DWORD));
Data.Copy((byte*)T.SToken.c_str(), T.SToken.size() + 1);
}
if (!Data.size())
{
return false;
}
SDirEntry de;
de.Name = szName;
de.flags = RF_RES_$TOKENS;
de.size = Data.size();
m_pGlobalCache->m_pRes[CACHE_USER]->mfFileAdd(&de);
SDirEntryOpen* pOE = m_pGlobalCache->m_pRes[CACHE_USER]->mfOpenEntry(&de);
pOE->pData = &Data[0];
m_pGlobalCache->m_pRes[CACHE_USER]->mfFlush();
m_pGlobalCache->m_pRes[CACHE_USER]->mfCloseEntry(&de);
return true;
}
void CHWShader_D3D::mfGetTokenMap(CResFile* pRes, SDirEntry* pDE, FXShaderToken*& Table, TArray<uint32>*& pSHData)
{
uint32 i;
int nSize = pRes->mfFileRead(pDE);
byte* pData = (byte*)pRes->mfFileGetBuf(pDE);
if (!pData)
{
Table = NULL;
return;
}
Table = new FXShaderToken;
pSHData = new TArray<uint32>;
uint32 nL = *(uint32*)pData;
if (CParserBin::m_bEndians)
{
SwapEndian(nL, eBigEndian);
}
pSHData->resize(nL);
if (CParserBin::m_bEndians)
{
uint32* pTokens = (uint32*)&pData[4];
for (i = 0; i < nL; i++)
{
uint32 nToken = pTokens[i];
SwapEndian(nToken, eBigEndian);
(*pSHData)[i] = nToken;
}
}
else
{
memcpy(&(*pSHData)[0], &pData[4], nL * sizeof(uint32));
}
pData += 4 + nL * sizeof(uint32);
nSize -= 4 + nL * sizeof(uint32);
int nOffs = 0;
while (nOffs < nSize)
{
char* pStr = (char*)&pData[nOffs + sizeof(DWORD)];
DWORD nToken;
LoadUnaligned(pData + nOffs, nToken);
if (CParserBin::m_bEndians)
{
SwapEndian(nToken, eBigEndian);
}
int nLen = strlen(pStr) + 1;
STokenD TD;
TD.Token = nToken;
TD.SToken = pStr;
Table->push_back(TD);
nOffs += sizeof(DWORD) + nLen;
}
}
bool CHWShader_D3D::mfGetCacheTokenMap(FXShaderToken*& Table, TArray<uint32>*& pSHData, uint64 nMaskGenFX)
{
if (!m_pGlobalCache || !m_pGlobalCache->isValid())
{
if (m_pGlobalCache)
{
m_pGlobalCache->Release(false);
}
m_pGlobalCache = mfInitCache(NULL, this, true, m_CRC32, true, CRenderer::CV_r_shadersasyncactivation != 0);
}
if (!m_pGlobalCache)
{
assert(false);
return false;
}
char strName[256];
sprintf_s(strName, "$MAP_%llx_%llx", nMaskGenFX, m_maskGenStatic);
if (Table)
{
if (m_pGlobalCache->m_pRes[CACHE_READONLY] && m_pGlobalCache->m_pRes[CACHE_READONLY]->mfFileExist(strName))
{
return true;
}
if (!m_pGlobalCache->m_pRes[CACHE_USER])
{
m_pGlobalCache->Release(false);
m_pGlobalCache = mfInitCache(NULL, this, true, m_CRC32, false);
}
if (!m_pGlobalCache || !m_pGlobalCache->m_pRes[CACHE_USER])
{
assert(false);
return false;
}
if (!m_pGlobalCache->m_pRes[CACHE_USER]->mfFileExist(strName))
{
if (!CRenderer::CV_r_shadersAllowCompilation)
{
return false;
}
//CryLogAlways("Storing MAP entry '%s' in shader cache file '%s'", strName, m_pGlobalCache->m_Name.c_str());
return mfStoreCacheTokenMap(Table, pSHData, strName);
}
return true;
}
SDirEntry* pDE = NULL;
CResFile* pRes = NULL;
for (int i = 0; i < 2; i++)
{
pRes = m_pGlobalCache->m_pRes[i];
if (!pRes)
{
continue;
}
pDE = pRes->mfGetEntry(strName);
if (pDE)
{
break;
}
}
if (!pDE || !pRes)
{
Warning("Couldn't find tokens MAP entry '%s' in shader cache file '%s'", strName, m_pGlobalCache->m_Name.c_str());
ASSERT_IN_SHADER(0);
return false;
}
mfGetTokenMap(pRes, pDE, Table, pSHData);
pRes->mfFileClose(pDE);
return true;
}
//==============================================================================================================================================================
bool CHWShader_D3D::mfGenerateScript(CShader* pSH, SHWSInstance*& pInst, std::vector<SCGBind>& InstBindVars, uint32 nFlags, FXShaderToken* Table, TArray<uint32>* pSHData, TArray<char>& sNewScr)
{
char* cgs = NULL;
bool bTempMap = (Table == NULL);
assert((Table && pSHData) || (!Table && !pSHData));
assert (m_pGlobalCache);
if (CParserBin::m_bEditable && !Table) // Fast path for offline shaders builder
{
Table = &m_TokenTable;
pSHData = &m_TokenData;
bTempMap = false;
}
else
{
if (m_pGlobalCache)
{
mfGetCacheTokenMap(Table, pSHData, m_nMaskGenShader);
}
if (CParserBin::m_bEditable)
{
if (bTempMap)
{
SAFE_DELETE(Table);
SAFE_DELETE(pSHData);
}
Table = &m_TokenTable;
pSHData = &m_TokenData;
bTempMap = false;
}
}
assert (Table && pSHData);
if (!Table || !pSHData)
{
return false;
}
ShaderTokensVec NewTokens;
uint32 eT = eT_unknown;
switch (pInst->m_eClass)
{
case eHWSC_Vertex:
eT = eT__VS;
break;
case eHWSC_Pixel:
eT = eT__PS;
break;
case eHWSC_Geometry:
eT = eT__GS;
break;
case eHWSC_Hull:
eT = eT__HS;
break;
case eHWSC_Compute:
eT = eT__CS;
break;
case eHWSC_Domain:
eT = eT__DS;
break;
default:
assert(0);
}
if (eT != eT_unknown)
{
CParserBin::AddDefineToken(eT, NewTokens);
}
// Include runtime mask definitions in the script
SShaderGen* shg = gRenDev->m_cEF.m_pGlobalExt;
if (shg && pInst->m_Ident.m_RTMask)
{
for (uint32 i = 0; i < shg->m_BitMask.Num(); i++)
{
SShaderGenBit* bit = shg->m_BitMask[i];
if (!(bit->m_Mask & pInst->m_Ident.m_RTMask))
{
continue;
}
CParserBin::AddDefineToken(bit->m_dwToken, NewTokens);
}
}
// Include light mask definitions in the script
if (m_Flags & HWSG_SUPPORTS_MULTILIGHTS)
{
int nLights = pInst->m_Ident.m_LightMask & 0xf;
if (nLights)
{
CParserBin::AddDefineToken(eT__LT_LIGHTS, NewTokens);
}
CParserBin::AddDefineToken(eT__LT_NUM, nLights + eT_0, NewTokens);
bool bHasProj = false;
for (int i = 0; i < 4; i++)
{
int nLightType = (pInst->m_Ident.m_LightMask >> (SLMF_LTYPE_SHIFT + i * SLMF_LTYPE_BITS)) & SLMF_TYPE_MASK;
if (nLightType == SLMF_PROJECTED)
{
bHasProj = true;
}
CParserBin::AddDefineToken(eT__LT_0_TYPE + i, nLightType + eT_0, NewTokens);
}
if (bHasProj)
{
CParserBin::AddDefineToken(eT__LT_HASPROJ, eT_1, NewTokens);
}
}
else
if (m_Flags & HWSG_SUPPORTS_LIGHTING)
{
CParserBin::AddDefineToken(eT__LT_LIGHTS, NewTokens);
int nLightType = (pInst->m_Ident.m_LightMask >> SLMF_LTYPE_SHIFT) & SLMF_TYPE_MASK;
if (nLightType == SLMF_PROJECTED)
{
CParserBin::AddDefineToken(eT__LT_HASPROJ, eT_1, NewTokens);
}
}
// Include modificator mask definitions in the script
if ((m_Flags & HWSG_SUPPORTS_MODIF) && pInst->m_Ident.m_MDMask)
{
const uint32 tcProjMask = HWMD_TEXCOORD_PROJ;
const uint32 tcMatrixMask = HWMD_TEXCOORD_MATRIX;
if (pInst->m_Ident.m_MDMask & tcProjMask)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_PROJ, NewTokens);
}
if (pInst->m_Ident.m_MDMask & tcMatrixMask)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_MATRIX, NewTokens);
}
if (pInst->m_Ident.m_MDMask & HWMD_TEXCOORD_GEN_OBJECT_LINEAR_DIFFUSE)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_GEN_OBJECT_LINEAR_DIFFUSE, NewTokens);
}
if (pInst->m_Ident.m_MDMask & HWMD_TEXCOORD_GEN_OBJECT_LINEAR_EMITTANCE)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_GEN_OBJECT_LINEAR_EMITTANCE, NewTokens);
}
if (pInst->m_Ident.m_MDMask & HWMD_TEXCOORD_GEN_OBJECT_LINEAR_EMITTANCE_MULT)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_GEN_OBJECT_LINEAR_EMITTANCE_MULT, NewTokens);
}
if (pInst->m_Ident.m_MDMask & HWMD_TEXCOORD_GEN_OBJECT_LINEAR_DETAIL)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_GEN_OBJECT_LINEAR_DETAIL, NewTokens);
}
if (pInst->m_Ident.m_MDMask & HWMD_TEXCOORD_GEN_OBJECT_LINEAR_CUSTOM)
{
CParserBin::AddDefineToken(eT__TT_TEXCOORD_GEN_OBJECT_LINEAR_CUSTOM, NewTokens);
}
}
// Include vertex modificator mask definitions in the script
if ((m_Flags & HWSG_SUPPORTS_VMODIF) && pInst->m_Ident.m_MDVMask)
{
int nMDV = pInst->m_Ident.m_MDVMask & 0x0fffffff;
int nType = nMDV & 0xf;
if (nType)
{
CParserBin::AddDefineToken(eT__VT_TYPE, eT_0 + nType, NewTokens);
}
if ((nMDV & MDV_BENDING) || nType == eDT_Bending)
{
CParserBin::AddDefineToken(eT__VT_BEND, eT_1, NewTokens);
if (!(nMDV & 0xf))
{
nType = eDT_Bending;
CParserBin::AddDefineToken(eT__VT_TYPE, eT_0 + nType, NewTokens);
}
}
if (nMDV & MDV_DEPTH_OFFSET)
{
CParserBin::AddDefineToken(eT__VT_DEPTH_OFFSET, eT_1, NewTokens);
}
if (nMDV & MDV_WIND)
{
CParserBin::AddDefineToken(eT__VT_WIND, eT_1, NewTokens);
}
if (nMDV & MDV_DET_BENDING)
{
CParserBin::AddDefineToken(eT__VT_DET_BEND, eT_1, NewTokens);
}
if (nMDV & MDV_DET_BENDING_GRASS)
{
CParserBin::AddDefineToken(eT__VT_GRASS, eT_1, NewTokens);
}
if (nMDV & ~0xf)
{
CParserBin::AddDefineToken(eT__VT_TYPE_MODIF, eT_1, NewTokens);
}
}
if (m_Flags & HWSG_FP_EMULATION)
{
//starting at LSB. 8 bits for colorop, 8 bits for alphaop, 3 bits for color arg1, 3 bits for color arg2, 1 bit for srgbwrite, 1 bit unused, 3 bits for
// alpha arg1, 3 bits for alpha arg2, 2 bits unused.
CParserBin::AddDefineToken(eT__FT0_COP, eT_0 + (pInst->m_Ident.m_LightMask & 0xff), NewTokens);
CParserBin::AddDefineToken(eT__FT0_AOP, eT_0 + ((pInst->m_Ident.m_LightMask & 0xff00) >> 8), NewTokens);
byte CO_0 = ((pInst->m_Ident.m_LightMask & 0xff0000) >> 16) & 7;
CParserBin::AddDefineToken(eT__FT0_CARG1, eT_0 + CO_0, NewTokens);
byte CO_1 = ((pInst->m_Ident.m_LightMask & 0xff0000) >> 19) & 7;
CParserBin::AddDefineToken(eT__FT0_CARG2, eT_0 + CO_1, NewTokens);
byte AO_0 = ((pInst->m_Ident.m_LightMask & 0xff000000) >> 24) & 7;
CParserBin::AddDefineToken(eT__FT0_AARG1, eT_0 + AO_0, NewTokens);
byte AO_1 = ((pInst->m_Ident.m_LightMask & 0xff000000) >> 27) & 7;
CParserBin::AddDefineToken(eT__FT0_AARG2, eT_0 + AO_1, NewTokens);
if (CO_0 == eCA_Specular || CO_1 == eCA_Specular || AO_0 == eCA_Specular || AO_1 == eCA_Specular)
{
CParserBin::AddDefineToken(eT__FT_SPECULAR, NewTokens);
}
if (CO_0 == eCA_Diffuse || CO_1 == eCA_Diffuse || AO_0 == eCA_Diffuse || AO_1 == eCA_Diffuse)
{
CParserBin::AddDefineToken(eT__FT_DIFFUSE, NewTokens);
}
if (CO_0 == eCA_Texture || CO_1 == eCA_Texture || AO_0 == eCA_Texture || AO_1 == eCA_Texture)
{
CParserBin::AddDefineToken(eT__FT_TEXTURE, NewTokens);
}
if (CO_0 == eCA_Texture1 || (CO_1 == eCA_Texture1) || AO_0 == eCA_Texture1 || AO_1 == eCA_Texture1
|| CO_0 == eCA_Previous || (CO_1 == eCA_Previous) || AO_0 == eCA_Previous || AO_1 == eCA_Previous)
{
CParserBin::AddDefineToken(eT__FT_TEXTURE1, NewTokens);
}
if (CO_0 == eCA_Normal || CO_1 == eCA_Normal || AO_0 == eCA_Normal || AO_1 == eCA_Normal)
{
CParserBin::AddDefineToken(eT__FT_NORMAL, NewTokens);
}
if (CO_0 == eCA_Constant || (CO_1 == eCA_Constant) || AO_0 == eCA_Constant || AO_1 == eCA_Constant
|| CO_0 == eCA_Previous || (CO_1 == eCA_Previous) || AO_0 == eCA_Previous || AO_1 == eCA_Previous)
{
CParserBin::AddDefineToken(eT__FT_PSIZE, NewTokens);
}
byte toSRGB = ((pInst->m_Ident.m_LightMask & 0x400000) >> 22) & 1;
if (toSRGB)
{
CParserBin::AddDefineToken(eT__FT_SRGBWRITE, NewTokens);
}
if (nFlags & HWSF_STOREDATA)
{
int nStreams = pInst->m_Ident.m_LightMask & 0xff;
if (nStreams & (1 << VSF_QTANGENTS))
{
CParserBin::AddDefineToken(eT__FT_QTANGENT_STREAM, NewTokens);
}
if (nStreams & (1 << VSF_TANGENTS))
{
CParserBin::AddDefineToken(eT__FT_TANGENT_STREAM, NewTokens);
}
if (nStreams & VSM_HWSKIN)
{
CParserBin::AddDefineToken(eT__FT_SKIN_STREAM, NewTokens);
}
#if ENABLE_NORMALSTREAM_SUPPORT
if (CParserBin::m_nPlatform == SF_D3D11 || CParserBin::m_nPlatform == SF_DURANGO || CParserBin::m_nPlatform == SF_ORBIS || CParserBin::m_nPlatform == SF_GL4 || CParserBin::m_nPlatform == SF_GLES3)
{
if (nStreams & VSM_NORMALS)
{
CParserBin::AddDefineToken(eT__FT_NORMAL, NewTokens);
}
}
#endif
if (nStreams & VSM_VERTEX_VELOCITY)
{
CParserBin::AddDefineToken(eT__FT_VERTEX_VELOCITY_STREAM, NewTokens);
}
}
}
int nT = NewTokens.size();
NewTokens.resize(nT + pSHData->size());
memcpy(&NewTokens[nT], &(*pSHData)[0], pSHData->size() * sizeof(uint32));
CParserBin Parser(NULL, pSH);
Parser.Preprocess(1, NewTokens, Table);
CorrectScriptEnums(Parser, pInst, InstBindVars, Table);
RemoveUnaffectedParameters_D3D10(Parser, pInst, InstBindVars);
ConvertBinScriptToASCII(Parser, pInst, InstBindVars, Table, sNewScr);
if (bTempMap)
{
SAFE_DELETE(Table);
SAFE_DELETE(pSHData);
}
return sNewScr.Num() && sNewScr[0];
}
void CHWShader_D3D::RemoveUnaffectedParameters_D3D10(CParserBin& Parser, SHWSInstance* pInst, std::vector<SCGBind>& InstBindVars)
{
int nPos = Parser.FindToken(0, Parser.m_Tokens.size() - 1, eT_cbuffer);
while (nPos >= 0)
{
uint32 nName = Parser.m_Tokens[nPos + 1];
if (nName == eT_PER_BATCH || nName == eT_PER_INSTANCE)
{
int nPosEnd = Parser.FindToken(nPos + 3, Parser.m_Tokens.size() - 1, eT_br_cv_2);
assert(nPosEnd >= 0);
int nPosN = Parser.FindToken(nPos + 1, Parser.m_Tokens.size() - 1, eT_br_cv_1);
assert(nPosN >= 0);
nPosN++;
while (nPosN < nPosEnd)
{
uint32 nT = Parser.m_Tokens[nPosN + 1];
int nPosCode = Parser.FindToken(nPosEnd + 1, Parser.m_Tokens.size() - 1, nT);
if (nPosCode < 0)
{
assert(nPosN > 0 && nPosN < (int)Parser.m_Tokens.size());
if (InstBindVars.size())
{
size_t i = 0;
CCryNameR nm(Parser.GetString(nT));
for (; i < InstBindVars.size(); i++)
{
SCGBind& b = InstBindVars[i];
if (b.m_Name == nm)
{
break;
}
}
if (i == InstBindVars.size())
{
Parser.m_Tokens[nPosN] = eT_comment;
}
}
else
{
Parser.m_Tokens[nPosN] = eT_comment;
}
}
nPosN = Parser.FindToken(nPosN + 2, nPosEnd, eT_semicolumn);
assert(nPosN >= 0);
nPosN++;
}
nPos = Parser.FindToken(nPosEnd + 1, Parser.m_Tokens.size() - 1, eT_cbuffer);
}
else
{
nPos = Parser.FindToken(nPos + 2, Parser.m_Tokens.size() - 1, eT_cbuffer);
}
}
}
struct SStructData
{
uint32 m_nName;
uint32 m_nTCs;
int m_nPos;
};
void CHWShader_D3D::CorrectScriptEnums(CParserBin& Parser, SHWSInstance* pInst, std::vector<SCGBind>& InstBindVars, FXShaderToken* Table)
{
// correct enumeration of TEXCOORD# interpolators after preprocessing
int nCur = 0;
int nSize = Parser.m_Tokens.size();
uint32* pTokens = &Parser.m_Tokens[0];
int nInstParam = 0;
const uint32 Toks[] = {eT_TEXCOORDN, eT_TEXCOORDN_centroid, eT_unknown};
std::vector<SStructData> SData;
uint32 i;
while (true)
{
nCur = Parser.FindToken(nCur, nSize - 1, eT_struct);
if (nCur < 0)
{
break;
}
int nLastStr = Parser.FindToken(nCur, nSize - 1, eT_br_cv_2);
assert(nLastStr >= 0);
if (nLastStr < 0)
{
break;
}
bool bNested = false;
for (i = 0; i < SData.size(); i++)
{
SStructData& Data = SData[i];
Data.m_nPos = Parser.FindToken(nCur, nLastStr, Data.m_nName);
if (Data.m_nPos > 0)
{
bNested = true;
}
}
uint32 nName = pTokens[nCur + 1];
int n = 0;
while (nCur < nLastStr)
{
int nTN = Parser.FindToken(nCur, nLastStr, Toks);
if (nTN < 0)
{
nCur = nLastStr + 1;
break;
}
int nNested = 0;
if (bNested)
{
for (i = 0; i < SData.size(); i++)
{
SStructData& Data = SData[i];
if (Data.m_nPos > 0 && nTN > Data.m_nPos)
{
nNested += Data.m_nTCs;
}
}
}
assert(pTokens[nTN - 1] == eT_colon);
int nArrSize = 1;
uint32 nTokName;
if (pTokens[nTN - 2] == eT_br_sq_2)
{
nArrSize = pTokens[nTN - 3] - eT_0;
if ((unsigned int) nArrSize > 15)
{
const char* szArrSize = Parser.GetString(pTokens[nTN - 3], *Table);
nArrSize = szArrSize ? atoi(szArrSize) : 0;
}
assert(pTokens[nTN - 4] == eT_br_sq_1);
//const char *szName = Parser.GetString(pTokens[nTN-5], *Table);
nTokName = pTokens[nTN - 5];
}
else
{
uint32 nType = pTokens[nTN - 3];
assert(nType == eT_float || nType == eT_float2 || nType == eT_float3 || nType == eT_float4 || nType == eT_float4x4 || nType == eT_float3x4 || nType == eT_float2x4 || nType == eT_float3x3 ||
nType == eT_half || nType == eT_half2 || nType == eT_half3 || nType == eT_half4 || nType == eT_half4x4 || nType == eT_half3x4 || nType == eT_half2x4 || nType == eT_half3x3);
if (nType == eT_float4x4 || nType == eT_half4x4)
{
nArrSize = 4;
}
else
if (nType == eT_float3x4 || nType == eT_float3x3 || nType == eT_half3x4 || nType == eT_half3x3)
{
nArrSize = 3;
}
else
if (nType == eT_float2x4 || nType == eT_half2x4)
{
nArrSize = 2;
}
nTokName = pTokens[nTN - 2];
}
assert(nArrSize > 0 && nArrSize < 16);
EToken eT = (pTokens[nTN] == eT_TEXCOORDN) ? eT_TEXCOORD0 : eT_TEXCOORD0_centroid;
pTokens[nTN] = n + nNested + eT;
n += nArrSize;
nCur = nTN + 1;
if (pInst->m_Ident.m_RTMask & g_HWSR_MaskBit[HWSR_INSTANCING_ATTR])
{
const char* szName = Parser.GetString(nTokName, *Table);
if (!_strnicmp(szName, "Inst", 4))
{
char newName[256];
int nm = 0;
while (szName[4 + nm] > 0x20 && szName[4 + nm] != '[')
{
newName[nm] = szName[4 + nm];
nm++;
}
newName[nm++] = 0;
SCGBind bn;
bn.m_RegisterOffset = nInstParam;
bn.m_RegisterCount = nArrSize;
bn.m_Name = newName;
InstBindVars.push_back(bn);
nInstParam += nArrSize;
}
}
}
SStructData SD;
SD.m_nName = nName;
SD.m_nPos = -1;
SD.m_nTCs = n;
SData.push_back(SD);
}
if (InstBindVars.size())
{
qsort(&InstBindVars[0], InstBindVars.size(), sizeof(SCGBind), CGBindCallback);
}
pInst->m_nNumInstAttributes = nInstParam;
}
static int sFetchInst(uint32& nCur, uint32* pTokens, uint32 nT, std::vector<uint32>& Parameter)
{
while (true)
{
uint32 nTok = pTokens[nCur];
if (nTok != eT_br_rnd_1 && nTok != eT_br_rnd_2 && nTok != eT_comma)
{
break;
}
nCur++;
}
int nC = 0;
Parameter.push_back(pTokens[nCur]);
nCur++;
while (pTokens[nCur] == eT_dot)
{
nC = 2;
Parameter.push_back(pTokens[nCur]);
Parameter.push_back(pTokens[nCur + 1]);
nCur += 2;
}
return nC;
}
namespace HWShaderD3D {
static void sCR(TArray<char>& Text, int nLevel)
{
Text.AddElem('\n');
for (int i = 0; i < nLevel; i++)
{
Text.AddElem(' ');
Text.AddElem(' ');
}
}
}
bool CHWShader_D3D::ConvertBinScriptToASCII(CParserBin& Parser, SHWSInstance* pInst, std::vector<SCGBind>& InstBindVars, FXShaderToken* Table, TArray<char>& Text)
{
using namespace HWShaderD3D;
uint32 i;
bool bRes = true;
uint32* pTokens = &Parser.m_Tokens[0];
uint32 nT = Parser.m_Tokens.size();
const char* szPrev = " ";
int nLevel = 0;
for (i = 0; i < nT; i++)
{
uint32 nToken = pTokens[i];
if (nToken == 0)
{
Text.Copy("\n", 1);
continue;
}
if (nToken == eT_skip)
{
i++;
continue;
}
if (nToken == eT_skip_1)
{
while (i < nT)
{
nToken = pTokens[i];
if (nToken == eT_skip_2)
{
break;
}
i++;
}
assert(i < nT);
continue;
}
if (nToken == eT_fetchinst)
{
char str[512];
i++;
std::vector<uint32> ParamDst, ParamSrc;
TArray<char> sParamDstFull, sParamDstName, sParamSrc;
int nDst = sFetchInst(i, &Parser.m_Tokens[0], Parser.m_Tokens.size(), ParamDst);
assert(Parser.m_Tokens[i] == eT_eq);
if (Parser.m_Tokens[i] != eT_eq)
{ // Should never happen
int n = CParserBin::FindToken(i, Parser.m_Tokens.size() - 1, &Parser.m_Tokens[0], eT_semicolumn);
if (n > 0)
{
i = n + 1;
}
continue;
}
i++;
int nSrc = sFetchInst(i, &Parser.m_Tokens[0], Parser.m_Tokens.size(), ParamSrc);
CParserBin::ConvertToAscii(&ParamDst[0], ParamDst.size(), *Table, sParamDstFull);
CParserBin::ConvertToAscii(&ParamDst[nDst], 1, *Table, sParamDstName);
CParserBin::ConvertToAscii(&ParamSrc[nSrc], 1, *Table, sParamSrc);
assert(strncmp(&sParamSrc[0], "Inst", 4) == 0);
{
sParamSrc.Free();
CParserBin::ConvertToAscii(&ParamSrc[0], ParamSrc.size(), *Table, sParamSrc);
sprintf_s(str, "%s = %s;\n", &sParamDstFull[0], &sParamSrc[0]);
Text.Copy(str, strlen(str));
}
while (Parser.m_Tokens[i] != eT_semicolumn)
{
i++;
}
continue;
}
const char* szStr = CParserBin::GetString(nToken, *Table, false);
assert(szStr);
if (!szStr || !szStr[0])
{
assert(0);
bRes = CParserBin::CorrectScript(pTokens, i, nT, Text);
}
else
{
#if defined(AZ_RESTRICTED_PLATFORM)
#define AZ_RESTRICTED_SECTION D3DHWSHADERCOMPILING_CPP_SECTION_4
#if defined(AZ_PLATFORM_XENIA)
#include "Xenia/D3DHWShaderCompiling_cpp_xenia.inl"
#elif defined(AZ_PLATFORM_PROVO)
#include "Provo/D3DHWShaderCompiling_cpp_provo.inl"
#elif defined(AZ_PLATFORM_SALEM)
#include "Salem/D3DHWShaderCompiling_cpp_salem.inl"
#endif
#endif
#if defined(AZ_RESTRICTED_SECTION_IMPLEMENTED)
#undef AZ_RESTRICTED_SECTION_IMPLEMENTED
#elif defined (_DEBUG)
int n = 0;
while (szStr[n])
{
char c = szStr[n++];
bool bASC = isascii(c);
assert(bASC);
}
#endif
if (nToken == eT_semicolumn || nToken == eT_br_cv_1)
{
if (nToken == eT_br_cv_1)
{
sCR(Text, nLevel);
nLevel++;
}
Text.Copy(szStr, strlen(szStr));
if (nToken == eT_semicolumn)
{
if (i + 1 < nT && pTokens[i + 1] == eT_br_cv_2)
{
sCR(Text, nLevel - 1);
}
else
{
sCR(Text, nLevel);
}
}
else
if (i + 1 < nT)
{
if (pTokens[i + 1] < eT_br_rnd_1 || pTokens[i + 1] >= eT_float)
{
sCR(Text, nLevel);
}
}
}
else
{
if (i + 1 < nT)
{
if (Text.Num())
{
char cPrev = Text[Text.Num() - 1];
if (!SkipChar((uint8)cPrev) && !SkipChar((uint8)szStr[0]))
{
Text.AddElem(' ');
}
}
}
Text.Copy(szStr, strlen(szStr));
if (nToken == eT_br_cv_2)
{
nLevel--;
if (i + 1 < nT && pTokens[i + 1] != eT_semicolumn)
{
sCR(Text, nLevel);
}
}
}
}
}
Text.AddElem(0);
return bRes;
}
void CHWShader_D3D::mfGetSrcFileName(char* srcName, int nSize)
{
if (!srcName || nSize <= 0)
{
return;
}
if (!m_NameSourceFX.empty())
{
cry_strcpy(srcName, nSize, m_NameSourceFX.c_str());
return;
}
cry_strcpy(srcName, nSize, gRenDev->m_cEF.m_HWPath);
if (m_eSHClass == eHWSC_Vertex)
{
cry_strcat(srcName, nSize, "Declarations/CGVShaders/");
}
else if (m_eSHClass == eHWSC_Pixel)
{
cry_strcat(srcName, nSize, "Declarations/CGPShaders/");
}
else
{
cry_strcat(srcName, nSize, "Declarations/CGGShaders/");
}
cry_strcat(srcName, nSize, GetName());
cry_strcat(srcName, nSize, ".crycg");
}
void CHWShader_D3D::mfGenName(SHWSInstance* pInst, char* dstname, int nSize, byte bType)
{
if (bType)
{
CHWShader::mfGenName(pInst->m_Ident.m_GLMask, pInst->m_Ident.m_RTMask, pInst->m_Ident.m_LightMask, pInst->m_Ident.m_MDMask, pInst->m_Ident.m_MDVMask, pInst->m_Ident.m_pipelineState.opaque, pInst->m_Ident.m_STMask, pInst->m_eClass, dstname, nSize, bType);
}
else
{
CHWShader::mfGenName(0, 0, 0, 0, 0, 0, 0, eHWSC_Num, dstname, nSize, bType);
}
}
void CHWShader_D3D::mfGetDstFileName(SHWSInstance* pInst, CHWShader_D3D* pSH, char* dstname, int nSize, byte bType)
{
cry_strcpy(dstname, nSize, gRenDev->m_cEF.m_ShadersCache.c_str());
if (pSH->m_eSHClass == eHWSC_Vertex)
{
if (bType == 1 || bType == 4)
{
cry_strcat(dstname, nSize, "CGVShaders/Debug/");
}
else if (bType == 0)
{
cry_strcat(dstname, nSize, "CGVShaders/");
}
else if (bType == 2 || bType == 3)
{
cry_strcat(dstname, nSize, "CGVShaders/Pending/");
}
}
else if (pSH->m_eSHClass == eHWSC_Pixel)
{
if (bType == 1 || bType == 4)
{
cry_strcat(dstname, nSize, "CGPShaders/Debug/");
}
else if (bType == 0)
{
cry_strcat(dstname, nSize, "CGPShaders/");
}
else if (bType == 2 || bType == 3)
{
cry_strcat(dstname, nSize, "CGPShaders/Pending/");
}
}
else if (GEOMETRYSHADER_SUPPORT && pSH->m_eSHClass == eHWSC_Geometry)
{
if (bType == 1 || bType == 4)
{
cry_strcat(dstname, nSize, "CGGShaders/Debug/");
}
else if (bType == 0)
{
cry_strcat(dstname, nSize, "CGGShaders/");
}
else if (bType == 2 || bType == 3)
{
cry_strcat(dstname, nSize, "CGGShaders/Pending/");
}
}
else if (GEOMETRYSHADER_SUPPORT && pSH->m_eSHClass == eHWSC_Hull)
{
if (bType == 1 || bType == 4)
{
cry_strcat(dstname, nSize, "CGHShaders/Debug/");
}
else if (bType == 0)
{
cry_strcat(dstname, nSize, "CGHShaders/");
}
else if (bType == 2 || bType == 3)
{
cry_strcat(dstname, nSize, "CGHShaders/Pending/");
}
}
else if (GEOMETRYSHADER_SUPPORT && pSH->m_eSHClass == eHWSC_Domain)
{
if (bType == 1 || bType == 4)
{
cry_strcat(dstname, nSize, "CGDShaders/Debug/");
}
else if (bType == 0)
{
cry_strcat(dstname, nSize, "CGDShaders/");
}
else if (bType == 2 || bType == 3)
{
cry_strcat(dstname, nSize, "CGDShaders/Pending/");
}
}
else if (GEOMETRYSHADER_SUPPORT && pSH->m_eSHClass == eHWSC_Compute)
{
if (bType == 1 || bType == 4)
{
cry_strcat(dstname, nSize, "CGCShaders/Debug/");
}
else if (bType == 0)
{
cry_strcat(dstname, nSize, "CGCShaders/");
}
else if (bType == 2 || bType == 3)
{
cry_strcat(dstname, nSize, "CGCShaders/Pending/");
}
}
cry_strcat(dstname, nSize, pSH->GetName());
if (bType == 2)
{
cry_strcat(dstname, nSize, "_out");
}
if (bType == 0)
{
char* s = strchr(dstname, '(');
if (s)
{
s[0] = 0;
}
}
char szGenName[256];
mfGenName(pInst, szGenName, 256, bType);
cry_strcat(dstname, nSize, szGenName);
}
//========================================================================================================
// Binary cache support
SShaderCache::~SShaderCache()
{
if (m_pStreamInfo)
{
CResFile* pRes = m_pStreamInfo->m_pRes;
bool bWarn = false;
if (pRes)
{
assert(pRes == m_pRes[0] || pRes == m_pRes[1]);
assert(!pRes->mfIsDirStreaming());
//bWarn = pRes->mfIsDirStreaming();
}
/*if (m_pStreamInfo->m_EntriesQueue.size())
bWarn = true;
if (bWarn)
Warning("Warning: SShader`Cache::~SShaderCache(): '%s' Streaming tasks is still in progress!: %d", m_Name.c_str(), m_pStreamInfo->m_EntriesQueue.size());*/
m_pStreamInfo->AbortJobs();
}
CHWShader::m_ShaderCache.erase(m_Name);
SAFE_DELETE(m_pRes[CACHE_USER]);
SAFE_DELETE(m_pRes[CACHE_READONLY]);
SAFE_RELEASE(m_pStreamInfo);
}
void SShaderCache::Cleanup()
{
if (m_pRes[0])
{
m_pRes[0]->mfDeactivate(true);
}
if (m_pRes[1])
{
m_pRes[1]->mfDeactivate(true);
}
}
bool SShaderCache::isValid()
{
return ((m_pRes[CACHE_READONLY] || m_pRes[CACHE_USER]) && CParserBin::m_nPlatform == m_nPlatform);
}
int SShaderCache::Size()
{
int nSize = sizeof(SShaderCache);
if (m_pRes[0])
{
nSize += m_pRes[0]->Size();
}
if (m_pRes[1])
{
nSize += m_pRes[1]->Size();
}
return nSize;
}
int SShaderDevCache::Size()
{
int nSize = sizeof(SShaderDevCache);
nSize += m_DeviceShaders.size() * sizeof(SD3DShader);
return nSize;
}
void SShaderCache::GetMemoryUsage(ICrySizer* pSizer) const
{
pSizer->AddObject(this, sizeof(*this));
pSizer->AddObject(m_pRes[0]);
pSizer->AddObject(m_pRes[1]);
}
void SShaderDevCache::GetMemoryUsage(ICrySizer* pSizer) const
{
pSizer->AddObject(this, sizeof(*this));
pSizer->AddObject(m_DeviceShaders);
}
SShaderDevCache* CHWShader::mfInitDevCache(const char* name, CHWShader* pSH)
{
CCryNameR cryShaderName(name);
return new SShaderDevCache(cryShaderName);
}
#include <LZSS.H>
SShaderCacheHeaderItem* CHWShader_D3D::mfGetCompressedItem(uint32 nFlags, int32& nSize)
{
SHWSInstance* pInst = m_pCurInst;
char name[128];
{
azstrcpy(name, AZ_ARRAY_SIZE(name), GetName());
char* s = strchr(name, '(');
if (s)
{
s[0] = 0;
}
}
CCryNameTSCRC Name = name;
FXCompressedShadersItor it = CHWShader::m_CompressedShaders.find(Name);
if (it == CHWShader::m_CompressedShaders.end())
{
return NULL;
}
SHWActivatedShader* pAS = it->second;
assert(pAS);
if (!pAS)
{
return NULL;
}
mfGenName(pInst, name, 128, 1);
Name = name;
FXCompressedShaderRemapItor itR = pAS->m_Remap.find(Name);
if (itR == pAS->m_Remap.end())
{
return NULL;
}
int nDevID = itR->second;
FXCompressedShaderItor itS = pAS->m_CompressedShaders.find(nDevID);
if (itS == pAS->m_CompressedShaders.end())
{
return NULL;
}
SCompressedData& CD = itS->second;
assert(CD.m_pCompressedShader);
if (!CD.m_pCompressedShader)
{
return NULL;
}
byte* pData = new byte[CD.m_nSizeDecompressedShader];
if (!pData)
{
return NULL;
}
pInst->m_DeviceObjectID = nDevID;
Decodem(CD.m_pCompressedShader, pData, CD.m_nSizeCompressedShader);
SShaderCacheHeaderItem* pIt = (SShaderCacheHeaderItem*)pData;
if (CParserBin::m_bEndians)
{
SwapEndian(*pIt, eBigEndian);
}
nSize = CD.m_nSizeDecompressedShader;
return pIt;
}
SShaderCacheHeaderItem* CHWShader_D3D::mfGetCacheItem(uint32& nFlags, int32& nSize)
{
LOADING_TIME_PROFILE_SECTION(gEnv->pSystem);
SHWSInstance* pInst = m_pCurInst;
byte* pData = NULL;
nSize = 0;
if (!m_pGlobalCache || !m_pGlobalCache->isValid())
{
return NULL;
}
CResFile* rf = NULL;
SDirEntry* de = NULL;
int i;
bool bAsync = false;
int n = CRenderer::CV_r_shadersAllowCompilation == 0 ? 1 : 2;
for (i = 0; i < n; i++)
{
rf = m_pGlobalCache->m_pRes[i];
if (!rf)
{
continue;
}
char name[128];
mfGenName(pInst, name, 128, 1);
de = rf->mfGetEntry(name, &bAsync);
if (de || bAsync)
{
break;
}
}
if (de)
{
if (CRenderer::CV_r_shadersdebug == 3 || CRenderer::CV_r_shadersdebug == 4)
{
iLog->Log("---Cache: LoadedFromGlobal %s': 0x%x", rf->mfGetFileName(), de->Name.get());
}
pInst->m_nCache = i;
SShaderCacheHeaderItem* pIt = NULL;
nSize = rf->mfFileRead(de);
pInst->m_bAsyncActivating = (nSize == -1);
pData = (byte*)rf->mfFileGetBuf(de);
if (pData && nSize > 0)
{
byte* pD = new byte[nSize];
memcpy(pD, pData, nSize);
pIt = (SShaderCacheHeaderItem*)pD;
if (CParserBin::m_bEndians)
{
SwapEndian(*pIt, eBigEndian);
}
pInst->m_DeviceObjectID = de->Name.get();
rf->mfFileClose(de);
}
if (i == CACHE_USER)
{
nFlags |= HWSG_CACHE_USER;
}
return pIt;
}
else
{
pInst->m_bAsyncActivating = bAsync;
return NULL;
}
}
bool CHWShader_D3D::mfAddCacheItem(SShaderCache* pCache, SShaderCacheHeaderItem* pItem, const byte* pData, int nLen, bool bFlush, CCryNameTSCRC Name)
{
if (!pCache)
{
return false;
}
if (!pCache->m_pRes[CACHE_USER])
{
return false;
}
if (CRenderer::CV_r_shadersdebug == 3 || CRenderer::CV_r_shadersdebug == 4)
{
iLog->Log("---Cache: StoredToGlobal %s': 0x%x", pCache->m_pRes[CACHE_USER]->mfGetFileName(), Name.get());
}
pItem->m_CRC32 = CCrc32::Compute(pData, nLen);
//CryLog("Size: %d: CRC: %x", nLen, pItem->m_CRC32);
byte* pNew = new byte[sizeof(SShaderCacheHeaderItem) + nLen];
SDirEntry de;
de.offset = 0;
if (CParserBin::m_bEndians)
{
SShaderCacheHeaderItem IT = *pItem;
SwapEndian(IT, eBigEndian);
memcpy(pNew, &IT, sizeof(SShaderCacheHeaderItem));
}
else
{
memcpy(pNew, pItem, sizeof(SShaderCacheHeaderItem));
}
memcpy(&pNew[sizeof(SShaderCacheHeaderItem)], pData, nLen);
de.Name = Name;
de.flags = RF_COMPRESS | RF_TEMPDATA;
de.size = nLen + sizeof(SShaderCacheHeaderItem);
pCache->m_pRes[CACHE_USER]->mfFileAdd(&de);
SDirEntryOpen* pOE = pCache->m_pRes[CACHE_USER]->mfOpenEntry(&de);
pOE->pData = pNew;
if (bFlush)
{
pCache->m_pRes[CACHE_USER]->mfFlush();
}
return true;
}
SEmptyCombination::Combinations SEmptyCombination::s_Combinations;
bool CHWShader_D3D::mfAddEmptyCombination(CShader* pSH, uint64 nRT, uint64 nGL, uint32 nLT)
{
CD3D9Renderer* rd = gcpRendD3D;
SRenderPipeline& RESTRICT_REFERENCE rRP = rd->m_RP;
SEmptyCombination Comb;
Comb.nGLNew = m_nMaskGenShader;
Comb.nRTNew = rRP.m_FlagsShader_RT & m_nMaskAnd_RT | m_nMaskOr_RT;
Comb.nLTNew = rRP.m_FlagsShader_LT;
Comb.nGLOrg = nGL;
Comb.nRTOrg = nRT & m_nMaskAnd_RT | m_nMaskOr_RT;
Comb.nLTOrg = nLT;
Comb.nMD = rRP.m_FlagsShader_MD;
Comb.nMDV = rRP.m_FlagsShader_MDV;
Comb.nST = m_maskGenStatic;
if (m_eSHClass == eHWSC_Pixel)
{
Comb.nMD &= ~HWMD_TEXCOORD_FLAG_MASK;
Comb.nMDV = 0;
}
Comb.pShader = this;
if (Comb.nRTNew != Comb.nRTOrg || Comb.nGLNew != Comb.nGLOrg || Comb.nLTNew != Comb.nLTOrg)
{
SEmptyCombination::s_Combinations.push_back(Comb);
}
m_nMaskGenShader = nGL;
return true;
}
bool CHWShader_D3D::mfStoreEmptyCombination(SEmptyCombination& Comb)
{
if (!m_pGlobalCache || !m_pGlobalCache->m_pRes[CACHE_USER])
{
return false;
}
CResFile* rf = m_pGlobalCache->m_pRes[CACHE_USER];
char nameOrg[128];
char nameNew[128];
SShaderCombIdent Ident;
Ident.m_GLMask = Comb.nGLNew;
Ident.m_RTMask = Comb.nRTNew;
Ident.m_LightMask = Comb.nLTNew;
Ident.m_MDMask = Comb.nMD;
Ident.m_MDVMask = Comb.nMDV;
Ident.m_STMask = Comb.nST;
SHWSInstance* pInstNew = mfGetInstance(gRenDev->m_RP.m_pShader, Ident, 0);
mfGenName(pInstNew, nameNew, 128, 1);
SDirEntry* deNew = rf->mfGetEntry(nameNew);
//assert(deNew);
if (!deNew)
{
return false;
}
Ident.m_GLMask = Comb.nGLOrg;
Ident.m_RTMask = Comb.nRTOrg;
Ident.m_LightMask = Comb.nLTOrg;
Ident.m_MDMask = Comb.nMD;
Ident.m_MDVMask = Comb.nMDV;
Ident.m_STMask = Comb.nST;
SHWSInstance* pInstOrg = mfGetInstance(gRenDev->m_RP.m_pShader, Ident, 0);
mfGenName(pInstOrg, nameOrg, 128, 1);
SDirEntry* deOrg = rf->mfGetEntry(nameOrg);
if (deOrg)
{
if (deOrg->offset != deNew->offset)
{
deOrg->offset = -abs(deNew->offset);
deOrg->flags |= RF_NOTSAVED;
}
return true;
}
SDirEntry de;
de.Name = CCryNameTSCRC(nameOrg);
de.flags = deNew->flags;
de.size = deNew->size;
de.offset = -abs(deNew->offset);
rf->mfFileAdd(&de);
return true;
}
bool CHWShader_D3D::mfFlushCacheFile()
{
uint32 i;
for (i = 0; i < m_Insts.size(); i++)
{
SHWSInstance* pInst = m_Insts[i];
if (pInst->m_Handle.m_bStatus == 2) // Fake
{
pInst->m_Handle.SetShader(NULL);
}
}
return m_pGlobalCache && m_pGlobalCache->m_pRes[CACHE_USER] && m_pGlobalCache->m_pRes[CACHE_USER]->mfFlush();
}
struct SData
{
CCryNameTSCRC Name;
uint32 nSizeDecomp;
uint32 nSizeComp;
//uint32 CRC;
uint16 flags;
int nOffset;
byte* pData;
byte bProcessed;
bool operator <(const SData& o) const
{
return Name < o.Name;
}
};
#if !defined(CONSOLE)
// Remove shader duplicates
bool CHWShader::mfOptimiseCacheFile(SShaderCache* pCache, bool bForce, SOptimiseStats* pStats)
{
CResFile* pRes = pCache->m_pRes[CACHE_USER];
pRes->mfFlush();
ResDir* Dir = pRes->mfGetDirectory();
uint32 i, j;
#ifdef _DEBUG
/*for (i=0; i<Dir->size(); i++)
{
SDirEntry *pDE = &(*Dir)[i];
for (j=i+1; j<Dir->size(); j++)
{
SDirEntry *pDE1 = &(*Dir)[j];
assert(pDE->Name != pDE1->Name);
}
}
*/
mfValidateTokenData(pRes);
#endif
std::vector<SData> Data;
bool bNeedOptimise = true;
if (pStats)
{
pStats->nEntries += Dir->size();
}
for (i = 0; i < Dir->size(); i++)
{
SDirEntry* pDE = &(*Dir)[i];
if (pDE->flags & RF_RES_$)
{
if (pDE->Name == CShaderMan::s_cNameHEAD)
{
continue;
}
SData d;
d.nSizeComp = d.nSizeDecomp = 0;
d.pData = pRes->mfFileReadCompressed(pDE, d.nSizeDecomp, d.nSizeComp);
assert(d.pData && d.nSizeComp && d.nSizeDecomp);
if (!d.pData || !d.nSizeComp || !d.nSizeDecomp)
{
continue;
}
if (pStats)
{
pStats->nTokenDataSize += d.nSizeDecomp;
}
d.bProcessed = 3;
d.Name = pDE->Name;
//d.CRC = 0;
d.nOffset = 0;
d.flags = (short)pDE->flags;
Data.push_back(d);
continue;
}
SData d;
d.flags = pDE->flags;
d.nSizeComp = d.nSizeDecomp = 0;
d.pData = pRes->mfFileReadCompressed(pDE, d.nSizeDecomp, d.nSizeComp);
assert(d.pData && d.nSizeComp && d.nSizeDecomp);
if (!d.pData || !d.nSizeComp || !d.nSizeDecomp)
{
continue;
}
d.nOffset = pDE->offset;
d.bProcessed = 0;
d.Name = pDE->Name;
//d.CRC = 0;
Data.push_back(d);
pRes->mfCloseEntry(pDE);
}
//FILE *fp = NULL;
int nDevID = 0x10000000;
int nOutFiles = Data.size();
if (bNeedOptimise)
{
for (i = 0; i < Data.size(); i++)
{
/*if (fp)
{
gEnv->pCryPak->FClose(fp);
fp = NULL;
}*/
if (Data[i].bProcessed)
{
continue;
}
byte* pD = Data[i].pData;
Data[i].bProcessed = 1;
Data[i].nOffset = nDevID++;
int nSizeComp = Data[i].nSizeComp;
int nSizeDecomp = Data[i].nSizeDecomp;
for (j = i + 1; j < Data.size(); j++)
{
if (Data[j].bProcessed)
{
continue;
}
byte* pD1 = Data[j].pData;
if (nSizeComp != Data[j].nSizeComp || nSizeDecomp != Data[j].nSizeDecomp)
{
continue;
}
if (!memcmp(pD, pD1, nSizeComp))
{
/*if (!fp && CRenderer::CV_r_shaderscacheoptimiselog)
{
char name[256];
sprintf_s(name, "Optimise/%s/%s.cache", pRes->mfGetFileName(), Data[i].Name.c_str());
fp = gEnv->pCryPak->FOpen(name, "w");
}*/
Data[j].nOffset = Data[i].nOffset;
Data[j].bProcessed = 2;
nOutFiles--;
//if (fp)
// gEnv->pCryPak->FPrintf(fp, "%s\n", Data[j].Name.c_str());
}
}
}
}
/*if (fp)
{
gEnv->pCryPak->FClose(fp);
fp = NULL;
}*/
if (nOutFiles != Data.size() || CRenderer::CV_r_shaderscachedeterministic)
{
if (nOutFiles == Data.size())
{
iLog->Log(" Forcing optimise for deterministic order...");
}
iLog->Log(" Optimising shaders resource '%s' (%" PRISIZE_T " items)...", pCache->m_Name.c_str(), Data.size() - 1);
pRes->mfClose();
pRes->mfOpen(RA_CREATE | (CParserBin::m_bEndians ? RA_ENDIANS : 0), &gRenDev->m_cEF.m_ResLookupDataMan[CACHE_USER]);
float fVersion = FX_CACHE_VER;
uint32 nMinor = (int)(((float)fVersion - (float)(int)fVersion) * 10.1f);
uint32 nMajor = (int)fVersion;
SResFileLookupData* pLookupCache = pCache->m_pRes[CACHE_USER]->GetLookupData(false, 0, 0);
CRY_ASSERT(pLookupCache != NULL);
if (pLookupCache == NULL || pLookupCache->m_CacheMajorVer != nMajor || pLookupCache->m_CacheMinorVer != nMinor)
{
CRY_ASSERT_MESSAGE(pLookupCache == NULL, "Losing ShaderIdents by recreating lookupdata cache");
pLookupCache = pRes->GetLookupData(true, 0, FX_CACHE_VER);
}
pRes->mfFlush();
if (CRenderer::CV_r_shaderscachedeterministic)
{
std::sort(Data.begin(), Data.end());
}
for (i = 0; i < Data.size(); i++)
{
SData* pD = &Data[i];
SDirEntry de;
de.Name = pD->Name;
de.flags = pD->flags;
if (pD->bProcessed == 1)
{
de.offset = pD->nOffset;
de.flags |= RF_COMPRESS | RF_COMPRESSED;
if (pStats)
{
pStats->nSizeUncompressed += pD->nSizeDecomp;
pStats->nSizeCompressed += pD->nSizeComp;
pStats->nUniqueEntries++;
}
assert(pD->pData);
if (pD->pData)
{
de.size = pD->nSizeComp + 4;
SDirEntryOpen* pOE = pRes->mfOpenEntry(&de);
byte* pData = new byte[de.size];
int nSize = pD->nSizeDecomp;
*(int*)pData = nSize;
memcpy(&pData[4], pD->pData, pD->nSizeComp);
de.flags |= RF_TEMPDATA;
pOE->pData = pData;
SAFE_DELETE_ARRAY(pD->pData);
}
}
else
if (pD->bProcessed != 3)
{
de.size = pD->nSizeComp + 4;
de.flags |= RF_COMPRESS;
de.offset = -pD->nOffset;
SAFE_DELETE_ARRAY(pD->pData);
}
else
{
SDirEntryOpen* pOE = pRes->mfOpenEntry(&de);
pOE->pData = pD->pData;
de.size = pD->nSizeDecomp;
}
pRes->mfFileAdd(&de);
}
}
if (nOutFiles != Data.size())
{
iLog->Log(" -- Removed %" PRISIZE_T " duplicated shaders", Data.size() - nOutFiles);
}
Data.clear();
int nSizeDir = pRes->mfFlush(true);
//int nSizeCompr = pRes->mfFlush();
#ifdef _DEBUG
mfValidateTokenData(pRes);
#endif
if (pStats)
{
pStats->nDirDataSize += nSizeDir;
}
for (i = 0; i < Data.size(); i++)
{
SData* pD = &Data[i];
SAFE_DELETE_ARRAY(pD->pData);
}
if (pStats)
{
CryLog(" -- Shader cache '%s' stats: Entries: %d, Unique Entries: %d, Size: %.3f Mb, Compressed Size: %.3f Mb, Token data size: %3f Mb, Directory Size: %.3f Mb", pCache->m_Name.c_str(), pStats->nEntries, pStats->nUniqueEntries, pStats->nSizeUncompressed / 1024.0f / 1024.0f, pStats->nSizeCompressed / 1024.0f / 1024.0f, pStats->nTokenDataSize / 1024.0f / 1024.0f, pStats->nDirDataSize / 1024.0f / 1024.0f);
}
return true;
}
#endif
int __cdecl sSort(const VOID* arg1, const VOID* arg2)
{
SDirEntry** pi1 = (SDirEntry**)arg1;
SDirEntry** pi2 = (SDirEntry**)arg2;
SDirEntry* ti1 = *pi1;
SDirEntry* ti2 = *pi2;
if (ti1->Name < ti2->Name)
{
return -1;
}
if (ti1->Name == ti2->Name)
{
return 0;
}
return 1;
}
bool CHWShader::_OpenCacheFile(float fVersion, SShaderCache* pCache, CHWShader* pSH, bool bCheckValid, uint32 CRC32, int nCache, CResFile* pRF, bool bReadOnly)
{
assert(nCache == CACHE_USER || nCache == CACHE_READONLY);
bool bValid = true;
CHWShader_D3D* pSHHW = (CHWShader_D3D*)pSH;
int nRes = pRF->mfOpen(RA_READ | (CParserBin::m_bEndians ? RA_ENDIANS : 0), &gRenDev->m_cEF.m_ResLookupDataMan[nCache], (nCache == CACHE_READONLY && pCache->m_pStreamInfo) ? pCache->m_pStreamInfo : NULL);
if (nRes == 0)
{
pRF->mfClose();
bValid = false;
}
else
if (nRes > 0)
{
if (bValid)
{
SResFileLookupData* pLookup = pRF->GetLookupData(false, 0, 0);
if (!pLookup)
{
bValid = false;
}
else
if (bCheckValid)
{
if (fVersion && (pLookup->m_CacheMajorVer != (int)fVersion || pLookup->m_CacheMinorVer != (int)(((float)fVersion - (float)(int)fVersion) * 10.1f)))
{
bValid = false;
}
if (!bValid && (CRenderer::CV_r_shadersdebug == 2 || nCache == CACHE_READONLY))
{
LogWarningEngineOnly("WARNING: Shader cache '%s' version mismatch (Cache: %d.%d, Expected: %.1f)", pRF->mfGetFileName(), pLookup->m_CacheMajorVer, pLookup->m_CacheMinorVer, fVersion);
}
if (pSH)
{
if (bValid && pLookup->m_CRC32 != pSHHW->m_CRC32)
{
bValid = false;
if (CRenderer::CV_r_shadersdebug == 2 && (CRenderer::CV_r_shadersdebug == 2 || nCache == CACHE_READONLY))
{
LogWarningEngineOnly("WARNING: Shader cache '%s' CRC mismatch", pRF->mfGetFileName());
}
}
}
}
}
if (nCache == CACHE_USER)
{
pRF->mfClose();
if (bValid)
{
int nAcc = CRenderer::CV_r_shadersAllowCompilation != 0 ? (RA_READ | RA_WRITE) : RA_READ;
if (!pRF->mfOpen(nAcc | (CParserBin::m_bEndians ? RA_ENDIANS : 0), &gRenDev->m_cEF.m_ResLookupDataMan[nCache]))
{
pRF->mfClose();
bValid = false;
}
}
}
}
if (!bValid && bCheckValid)
{
if (nCache == CACHE_USER && !bReadOnly)
{
if (!pRF->mfOpen(RA_CREATE | (CParserBin::m_bEndians ? RA_ENDIANS : 0), &gRenDev->m_cEF.m_ResLookupDataMan[nCache]))
{
return false;
}
SResFileLookupData* pLookup = pRF->GetLookupData(true, CRC32, FX_CACHE_VER);
if (pSHHW)
{
pRF->mfFlush();
}
pCache->m_bNeedPrecache = true;
bValid = true;
}
else
{
SAFE_DELETE(pRF);
}
}
pCache->m_pRes[nCache] = pRF;
pCache->m_bReadOnly[nCache] = bReadOnly;
#ifdef _DEBUG
mfValidateTokenData(pRF);
#endif
return bValid;
}
bool CHWShader::mfOpenCacheFile(const char* szName, float fVersion, SShaderCache* pCache, CHWShader* pSH, bool bCheckValid, uint32 CRC32, bool bReadOnly)
{
bool bValidRO = false;
bool bValidUser = true;
// don't load the readonly cache, when shaderediting is true
if (!CRenderer::CV_r_shadersediting && !pCache->m_pRes[CACHE_READONLY])
{
CResFile* rfRO = new CResFile(szName);
bool bRO = bReadOnly;
if (!CRenderer::CV_r_shadersAllowCompilation)
{
bRO = true;
}
bValidRO = _OpenCacheFile(fVersion, pCache, pSH, bCheckValid, CRC32, CACHE_READONLY, rfRO, bRO);
}
if (!CRenderer::CV_r_shadersAllowCompilation)
{
assert(bReadOnly);
}
if ((!bReadOnly || gRenDev->IsShaderCacheGenMode()) && !pCache->m_pRes[CACHE_USER])
{
stack_string szUser = stack_string(gRenDev->m_cEF.m_szCachePath.c_str()) + stack_string(szName);
CResFile* rfUser = new CResFile(szUser.c_str());
bValidUser = _OpenCacheFile(fVersion, pCache, pSH, bCheckValid, CRC32, CACHE_USER, rfUser, bReadOnly);
}
return (bValidRO || bValidUser);
}
SShaderCache* CHWShader::mfInitCache(const char* name, CHWShader* pSH, bool bCheckValid, uint32 CRC32, bool bReadOnly, bool bAsync)
{
// LOADING_TIME_PROFILE_SECTION(iSystem);
CHWShader_D3D* pSHHW = (CHWShader_D3D*)pSH;
char nameCache[256];
if (!CRenderer::CV_r_shadersAllowCompilation)
{
bCheckValid = false;
}
if (CRenderer::CV_r_shadersediting)
{
bReadOnly = false;
}
if (!name)
{
char namedst[256];
pSHHW->mfGetDstFileName(pSHHW->m_pCurInst, pSHHW, namedst, 256, 0);
fpStripExtension(namedst, nameCache);
fpAddExtension(nameCache, ".fxcb");
name = nameCache;
}
SShaderCache* pCache = NULL;
FXShaderCacheItor it = m_ShaderCache.find(CCryNameR(name));
if (it != m_ShaderCache.end())
{
pCache = it->second;
pCache->AddRef();
if (pSHHW)
{
if (bCheckValid)
{
int nCache[2] = {-1, -1};
if (!CRenderer::CV_r_shadersAllowCompilation)
{
nCache[0] = CACHE_READONLY;
}
else
{
nCache[0] = CACHE_USER;
nCache[1] = CACHE_READONLY;
}
bool bValid;
for (int i = 0; i < 2; i++)
{
if (nCache[i] < 0 || !pCache->m_pRes[i])
{
continue;
}
CResFile* pRF = pCache->m_pRes[i];
SResFileLookupData* pLookup = pRF->GetLookupData(false, 0, FX_CACHE_VER);
bValid = (pLookup && pLookup->m_CRC32 == CRC32);
if (!bValid)
{
SAFE_DELETE(pCache->m_pRes[i]);
}
}
bValid = true;
if (!CRenderer::CV_r_shadersAllowCompilation && !pCache->m_pRes[CACHE_READONLY])
{
bValid = false;
}
else
{
if (bReadOnly && (!pCache->m_pRes[CACHE_READONLY] || !pCache->m_pRes[CACHE_USER]))
{
bValid = false;
}
if (!bReadOnly && !pCache->m_pRes[CACHE_USER])
{
bValid = false;
}
}
if (!bValid)
{
mfOpenCacheFile(name, FX_CACHE_VER, pCache, pSH, bCheckValid, CRC32, bReadOnly);
}
}
}
}
else
{
pCache = new SShaderCache;
if (bAsync)
{
pCache->m_pStreamInfo = new SResStreamInfo(pCache);
}
pCache->m_nPlatform = CParserBin::m_nPlatform;
pCache->m_Name = name;
mfOpenCacheFile(name, FX_CACHE_VER, pCache, pSH, bCheckValid, CRC32, bReadOnly);
m_ShaderCache.insert(FXShaderCacheItor::value_type(CCryNameR(name), pCache));
}
return pCache;
}
byte* CHWShader_D3D::mfBindsToCache(SHWSInstance* pInst, std::vector<SCGBind>* Binds, int nParams, byte* pP)
{
int i;
for (i = 0; i < nParams; i++)
{
SCGBind* cgb = &(*Binds)[i];
SShaderCacheHeaderItemVar* pVar = (SShaderCacheHeaderItemVar*)pP;
pVar->m_nCount = cgb->m_RegisterCount;
pVar->m_Reg = cgb->m_RegisterOffset;
if (CParserBin::m_bEndians)
{
SwapEndian(pVar->m_nCount, eBigEndian);
SwapEndian(pVar->m_Reg, eBigEndian);
}
int len = strlen(cgb->m_Name.c_str()) + 1;
memcpy(pVar->m_Name, cgb->m_Name.c_str(), len);
pP += offsetof(SShaderCacheHeaderItemVar, m_Name) + strlen(pVar->m_Name) + 1;
}
return pP;
}
byte* CHWShader_D3D::mfBindsFromCache(std::vector<SCGBind>*& Binds, int nParams, byte* pP)
{
int i;
for (i = 0; i < nParams; i++)
{
if (!Binds)
{
Binds = new std::vector<SCGBind>;
}
SCGBind cgb;
SShaderCacheHeaderItemVar* pVar = (SShaderCacheHeaderItemVar*)pP;
short nParameters = pVar->m_nCount;
if (CParserBin::m_bEndians)
{
SwapEndian(nParameters, eBigEndian);
}
cgb.m_RegisterCount = nParameters;
cgb.m_Name = pVar->m_Name;
int dwBind = pVar->m_Reg;
if (CParserBin::m_bEndians)
{
SwapEndian(dwBind, eBigEndian);
}
cgb.m_RegisterOffset = dwBind;
Binds->push_back(cgb);
pP += offsetof(SShaderCacheHeaderItemVar, m_Name) + strlen(pVar->m_Name) + 1;
}
return pP;
}
byte* CHWShader::mfIgnoreBindsFromCache(int nParams, byte* pP)
{
int i;
for (i = 0; i < nParams; i++)
{
SShaderCacheHeaderItemVar* pVar = (SShaderCacheHeaderItemVar*)pP;
pP += offsetof(SShaderCacheHeaderItemVar, m_Name) + strlen(pVar->m_Name) + 1;
}
return pP;
}
bool CHWShader_D3D::mfUploadHW(SHWSInstance* pInst, byte* pBuf, uint32 nSize, CShader* pSH, uint32 nFlags)
{
PROFILE_FRAME(Shader_mfUploadHW);
const char* sHwShaderName = _HELP("Vertex Shader");
if (m_eSHClass == eHWSC_Pixel)
{
sHwShaderName = _HELP("Pixel Shader");
}
HRESULT hr = S_OK;
if (!pInst->m_Handle.m_pShader)
{
pInst->m_Handle.SetShader(new SD3DShader);
}
if ((m_eSHClass == eHWSC_Vertex) && (!(nFlags & HWSF_PRECACHE) || gRenDev->m_cEF.m_bActivatePhase) && !pInst->m_bFallback)
{
mfUpdateFXVertexFormat(pInst, pSH);
}
pInst->m_nDataSize = nSize;
if (m_eSHClass == eHWSC_Pixel)
{
s_nDevicePSDataSize += nSize;
}
else
{
s_nDeviceVSDataSize += nSize;
}
if (m_eSHClass == eHWSC_Pixel)
{
hr = gcpRendD3D->GetDevice().CreatePixelShader(alias_cast<DWORD*>(pBuf), nSize, NULL, alias_cast<ID3D11PixelShader**>(&pInst->m_Handle.m_pShader->m_pHandle));
}
else
if (m_eSHClass == eHWSC_Vertex)
{
hr = gcpRendD3D->GetDevice().CreateVertexShader(alias_cast<DWORD*>(pBuf), nSize, NULL, alias_cast<ID3D11VertexShader**>(&pInst->m_Handle.m_pShader->m_pHandle));
}
else
if (m_eSHClass == eHWSC_Geometry)
{
hr = gcpRendD3D->GetDevice().CreateGeometryShader(alias_cast<DWORD*>(pBuf), nSize, NULL, alias_cast<ID3D11GeometryShader**>(&pInst->m_Handle.m_pShader->m_pHandle));
}
else
if (m_eSHClass == eHWSC_Hull)
{
hr = gcpRendD3D->GetDevice().CreateHullShader(alias_cast<DWORD*>(pBuf), nSize, NULL, alias_cast<ID3D11HullShader**>(&pInst->m_Handle.m_pShader->m_pHandle));
}
else
if (m_eSHClass == eHWSC_Compute)
{
hr = gcpRendD3D->GetDevice().CreateComputeShader(alias_cast<DWORD*>(pBuf), nSize, NULL, alias_cast<ID3D11ComputeShader**>(&pInst->m_Handle.m_pShader->m_pHandle));
}
else
if (m_eSHClass == eHWSC_Domain)
{
hr = gcpRendD3D->GetDevice().CreateDomainShader(alias_cast<DWORD*>(pBuf), nSize, NULL, alias_cast<ID3D11DomainShader**>(&pInst->m_Handle.m_pShader->m_pHandle));
}
else
{
assert(0);
}
#if defined(AZ_RESTRICTED_PLATFORM)
#define AZ_RESTRICTED_SECTION D3DHWSHADERCOMPILING_CPP_SECTION_5
#if defined(AZ_PLATFORM_XENIA)
#include "Xenia/D3DHWShaderCompiling_cpp_xenia.inl"
#elif defined(AZ_PLATFORM_PROVO)
#include "Provo/D3DHWShaderCompiling_cpp_provo.inl"
#elif defined(AZ_PLATFORM_SALEM)
#include "Salem/D3DHWShaderCompiling_cpp_salem.inl"
#endif
#endif
// Assign name to Shader for enhanced debugging
#if !defined(RELEASE) && defined(WIN64)
if (pInst->m_Handle.m_pShader->m_pHandle)
{
char name[1024];
azsprintf(name, "%s_%s(LT%x)@(RT%llx)(MD%x)(MDV%x)(GL%llx)(PSS%llx)(ST%llx)", pSH->GetName(), m_EntryFunc.c_str(), pInst->m_Ident.m_LightMask, pInst->m_Ident.m_RTMask, pInst->m_Ident.m_MDMask, pInst->m_Ident.m_MDVMask, pInst->m_Ident.m_GLMask, pInst->m_Ident.m_pipelineState.opaque, pInst->m_Ident.m_STMask);
((ID3D11DeviceChild*)pInst->m_Handle.m_pShader->m_pHandle)->SetPrivateData(WKPDID_D3DDebugObjectName, strlen(name), name);
}
#endif
return (hr == S_OK);
}
bool CHWShader_D3D::mfUploadHW(LPD3D10BLOB pShader, SHWSInstance* pInst, CShader* pSH, uint32 nFlags)
{
bool bResult = true;
if (m_eSHClass == eHWSC_Vertex && !pInst->m_bFallback)
{
mfUpdateFXVertexFormat(pInst, pSH);
}
if (pShader && !(m_Flags & HWSG_PRECACHEPHASE))
{
DWORD* pCode = (DWORD*)pShader->GetBufferPointer();
if (gcpRendD3D->m_cEF.m_nCombinationsProcess >= 0 && !gcpRendD3D->m_cEF.m_bActivatePhase)
{
pInst->m_Handle.SetFake();
}
else
{
bResult = mfUploadHW(pInst, (byte*)pCode, (uint32)pShader->GetBufferSize(), pSH, nFlags);
if (m_eSHClass == eHWSC_Vertex)
{
size_t nSize = pShader->GetBufferSize();
pInst->m_pShaderData = new byte[nSize];
pInst->m_nDataSize = nSize;
memcpy(pInst->m_pShaderData, pCode, nSize);
pInst->m_uniqueNameCRC = AZ::Crc32(GetName());
}
}
if (!bResult)
{
if (m_eSHClass == eHWSC_Vertex)
{
Warning("CHWShader_D3D::mfUploadHW: Could not create vertex shader '%s'(0x%" PRIx64 ")\n", GetName(), pInst->m_Ident.m_GLMask);
}
else
if (m_eSHClass == eHWSC_Pixel)
{
Warning("CHWShader_D3D::mfUploadHW: Could not create pixel shader '%s'(0x%" PRIx64 ")\n", GetName(), pInst->m_Ident.m_GLMask);
}
else
if (m_eSHClass == eHWSC_Geometry)
{
Warning("CHWShader_D3D::mfUploadHW: Could not create geometry shader '%s'(0x%" PRIx64 ")\n", GetName(), pInst->m_Ident.m_GLMask);
}
else
if (m_eSHClass == eHWSC_Domain)
{
Warning("CHWShader_D3D::mfUploadHW: Could not create domain shader '%s'(0x%" PRIx64 ")\n", GetName(), pInst->m_Ident.m_GLMask);
}
else
if (m_eSHClass == eHWSC_Hull)
{
Warning("CHWShader_D3D::mfUploadHW: Could not create hull shader '%s'(0x%" PRIx64 ")\n", GetName(), pInst->m_Ident.m_GLMask);
}
else
if (m_eSHClass == eHWSC_Compute)
{
Warning("CHWShader_D3D::mfUploadHW: Could not create compute shader '%s'(0x%" PRIx64 ")\n", GetName(), pInst->m_Ident.m_GLMask);
}
}
}
return bResult;
}
bool CHWShader_D3D::mfActivateCacheItem(CShader* pSH, SShaderCacheHeaderItem* pItem, uint32 nSize, uint32 nFlags)
{
SHWSInstance* pInst = m_pCurInst;
byte* pData = (byte*)pItem;
pData += sizeof(SShaderCacheHeaderItem);
byte* pBuf = pData;
std::vector<SCGBind>* pInstBinds = NULL;
pInst->Release(m_pDevCache, false);
pBuf = mfBindsFromCache(pInstBinds, pItem->m_nInstBinds, pBuf);
nSize -= (uint32)(pBuf - (byte*)pItem);
pInst->m_eClass = (EHWShaderClass)pItem->m_Class;
#if !defined(_RELEASE)
if (pItem->m_nVertexFormat >= eVF_Max)
{
AZ_Warning("Graphics", false, "Existing vertex format with enum %d not legit (must be less than %d). Is the shader cache out of date? Defaulting to eVF_P3S_C4B_T2S.", pItem->m_nVertexFormat, eVF_Max);
pItem->m_nVertexFormat = eVF_P3S_C4B_T2S;
}
#endif
pInst->m_vertexFormat = AZ::Vertex::Format(gcpRendD3D->m_RP.m_vertexFormats[pItem->m_nVertexFormat]);
pInst->m_nInstructions = pItem->m_nInstructions;
pInst->m_VStreamMask_Decl = pItem->m_StreamMask_Decl;
pInst->m_VStreamMask_Stream = pItem->m_StreamMask_Stream;
bool bResult = true;
SD3DShader* pHandle = NULL;
SShaderDevCache* pCache = m_pDevCache;
if (!(nFlags & HWSG_CACHE_USER))
{
if (pCache)
{
FXDeviceShaderItor it = pCache->m_DeviceShaders.find(pInst->m_DeviceObjectID);
if (it != pCache->m_DeviceShaders.end())
{
pHandle = it->second;
}
}
}
HRESULT hr = S_OK;
if (pHandle)
{
pInst->m_Handle.SetShader(pHandle);
pInst->m_Handle.AddRef();
#if D3DHWSHADERCOMPILING_CPP_TRAIT_VERTEX_FORMAT
if (m_eSHClass == eHWSC_Vertex)
{
ID3D10Blob* pS = NULL;
D3D10CreateBlob(nSize, (LPD3D10BLOB*)&pS);
DWORD* pBuffer = (DWORD*)pS->GetBufferPointer();
memcpy(pBuffer, pBuf, nSize);
mfVertexFormat(pInst, this, pS);
SAFE_RELEASE(pS);
}
#endif
if ((m_eSHClass == eHWSC_Vertex) && (!(nFlags & HWSF_PRECACHE) || gRenDev->m_cEF.m_bActivatePhase) && !pInst->m_bFallback)
{
mfUpdateFXVertexFormat(pInst, pSH);
}
}
else
{
if (gcpRendD3D->m_cEF.m_nCombinationsProcess > 0 && !gcpRendD3D->m_cEF.m_bActivatePhase)
{
pInst->m_Handle.SetFake();
}
else
{
#if D3DHWSHADERCOMPILING_CPP_TRAIT_VERTEX_FORMAT
if (m_eSHClass == eHWSC_Vertex)
{
ID3D10Blob* pS = NULL;
D3D10CreateBlob(nSize, (LPD3D10BLOB*)&pS);
DWORD* pBuffer = (DWORD*)pS->GetBufferPointer();
memcpy(pBuffer, pBuf, nSize);
mfVertexFormat(pInst, this, pS);
SAFE_RELEASE(pS);
}
#endif
bResult = mfUploadHW(pInst, pBuf, nSize, pSH, nFlags);
}
if (!bResult)
{
SAFE_DELETE(pInstBinds);
assert(!"Shader creation error");
iLog->Log("WARNING: cannot create shader '%s' (FX: %s)", m_EntryFunc.c_str(), GetName());
return true;
}
pCache->m_DeviceShaders.insert(FXDeviceShaderItor::value_type(pInst->m_DeviceObjectID, pInst->m_Handle.m_pShader));
}
void* pConstantTable = NULL;
void* pShaderReflBuf = NULL;
hr = D3DReflect(pBuf, nSize, IID_ID3D11ShaderReflection, &pShaderReflBuf);
ID3D11ShaderReflection* pShaderReflection = (ID3D11ShaderReflection*)pShaderReflBuf;
if (SUCCEEDED(hr))
{
pConstantTable = (void*)pShaderReflection;
}
if (m_eSHClass == eHWSC_Vertex || gRenDev->IsEditorMode())
{
pInst->m_pShaderData = new byte[nSize];
pInst->m_nDataSize = nSize;
memcpy(pInst->m_pShaderData, pBuf, nSize);
pInst->m_uniqueNameCRC = AZ::Crc32(GetName());
}
assert(hr == S_OK);
bResult &= (hr == S_OK);
if (pConstantTable)
{
mfCreateBinds(pInst, pConstantTable, pBuf, nSize);
}
mfGatherFXParameters(pInst, pInstBinds, this, 0, pSH);
SAFE_DELETE(pInstBinds);
SAFE_RELEASE(pShaderReflection);
return bResult;
}
bool CHWShader_D3D::mfCreateCacheItem(SHWSInstance* pInst, std::vector<SCGBind>& InstBinds, byte* pData, int nLen, CHWShader_D3D* pSH, bool bShaderThread)
{
if (!pSH->m_pGlobalCache || !pSH->m_pGlobalCache->m_pRes[CACHE_USER])
{
if (pSH->m_pGlobalCache)
{
pSH->m_pGlobalCache->Release(false);
}
pSH->m_pGlobalCache = mfInitCache(NULL, pSH, true, pSH->m_CRC32, false, false);
}
assert(pSH->m_pGlobalCache);
if (!pSH->m_pGlobalCache || !pSH->m_pGlobalCache->m_pRes[CACHE_USER])
{
return false;
}
byte* byteData = NULL;
DWORD* dwordData = NULL;
SShaderCacheHeaderItem h;
h.m_nInstBinds = InstBinds.size();
h.m_nInstructions = pInst->m_nInstructions;
h.m_nVertexFormat = pInst->m_vertexFormat.GetEnum();
h.m_Class = pData ? pInst->m_eClass : 255;
h.m_StreamMask_Decl = pInst->m_VStreamMask_Decl;
h.m_StreamMask_Stream = (byte)pInst->m_VStreamMask_Stream;
int nNewSize = (h.m_nInstBinds) * sizeof(SShaderCacheHeaderItemVar) + nLen;
byte* pNewData = new byte [nNewSize];
byte* pP = pNewData;
pP = mfBindsToCache(pInst, &InstBinds, h.m_nInstBinds, pP);
PREFAST_ASSUME(pData);
memcpy(pP, pData, nLen);
delete[] byteData;
delete[] dwordData;
pP += nLen;
char name[256];
mfGenName(pInst, name, 256, 1);
CCryNameTSCRC nm = CCryNameTSCRC(name);
bool bRes = mfAddCacheItem(pSH->m_pGlobalCache, &h, pNewData, (int)(pP - pNewData), false, nm);
SAFE_DELETE_ARRAY(pNewData);
if (gRenDev->m_cEF.m_bActivatePhase || (!(pSH->m_Flags & HWSG_PRECACHEPHASE) && gRenDev->m_cEF.m_nCombinationsProcess <= 0))
{
if (!gRenDev->m_cEF.m_bActivatePhase)
{
if (bShaderThread && false)
{
if (pInst->m_pAsync)
{
pInst->m_pAsync->m_bPendedFlush = true;
}
}
else
{
pSH->mfFlushCacheFile();
}
}
azstrcpy(name, AZ_ARRAY_SIZE(name), pSH->GetName());
char* s = strchr(name, '(');
if (s)
{
s[0] = 0;
}
if (!bShaderThread || true)
{
byte bStore = 1;
if (pSH->m_Flags & HWSG_FP_EMULATION)
{
bStore = 2;
}
SShaderCombIdent Ident = pInst->m_Ident;
;
Ident.m_GLMask = pSH->m_nMaskGenFX;
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, name, 0, NULL, bStore);
}
}
pInst->m_nCache = CACHE_USER;
return bRes;
}
//============================================================================
void CHWShader_D3D::mfSaveCGFile(const char* scr, const char* path)
{
if (CRenderer::CV_r_shadersdebug < 1)
{
return;
}
char name[1024];
if (path && path[0])
{
sprintf_s(name, "%s/%s(LT%x)/(RT%llx)(MD%x)(MDV%x)(GL%llx)(PSS%llx)(ST%llx).cg", path, GetName(), m_pCurInst->m_Ident.m_LightMask, m_pCurInst->m_Ident.m_RTMask, m_pCurInst->m_Ident.m_MDMask, m_pCurInst->m_Ident.m_MDVMask, m_pCurInst->m_Ident.m_GLMask, m_pCurInst->m_Ident.m_pipelineState.opaque, m_pCurInst->m_Ident.m_STMask);
}
else
{
sprintf_s(name, "@cache@/shaders/fxerror/%s(GL%llx)/(LT%x)(RT%llx)/(MD%x)(MDV%x)(PSS%llx)(ST%llx).cg", GetName(), m_pCurInst->m_Ident.m_GLMask, m_pCurInst->m_Ident.m_LightMask, m_pCurInst->m_Ident.m_RTMask, m_pCurInst->m_Ident.m_MDMask, m_pCurInst->m_Ident.m_MDVMask, m_pCurInst->m_Ident.m_pipelineState.opaque, m_pCurInst->m_Ident.m_STMask);
}
AZ::IO::HandleType fileHandle;
fileHandle = gEnv->pCryPak->FOpen(name, "w");
if (fileHandle != AZ::IO::InvalidHandle)
{
size_t len = strlen(scr);
gEnv->pCryPak->FWrite(scr, len, fileHandle);
gEnv->pCryPak->FClose (fileHandle);
}
}
void CHWShader_D3D::mfOutputCompilerError(string& strErr, const char* szSrc)
{
if (CRenderer::CV_r_shadersdebug)
{
AZ::IO::HandleType fileHandle = fxopen("$$err", "w");
if (fileHandle != AZ::IO::InvalidHandle)
{
AZ::IO::FPutS(szSrc, fileHandle);
gEnv->pFileIO->Close(fileHandle);
}
}
string strE = strErr;
size_t newlinePos = strE.find('\n');
while (newlinePos != string::npos)
{
iLog->LogError("%s", strE.substr(0, newlinePos).c_str());
strE = strE.substr(newlinePos + 1);
newlinePos = strE.find('\n');
}
if (strE.size())
{
iLog->LogError("%s", strE.c_str());
}
}
SShaderAsyncInfo::~SShaderAsyncInfo()
{
{
AUTO_LOCK(g_cAILock);
Unlink();
}
if (m_pFXShader)
{
assert(m_pFXShader->GetID() >= 0 && m_pFXShader->GetID() < MAX_REND_SHADERS);
}
SAFE_RELEASE(m_pFXShader);
SAFE_RELEASE(m_pShader);
}
// Flush pended or processed shaders (main thread task)
void SShaderAsyncInfo::FlushPendingShaders()
{
//assert (gRenDev->m_pRT->IsRenderThread());
if (!SShaderAsyncInfo::PendingList().m_Next)
{
SShaderAsyncInfo::PendingList().m_Next = &SShaderAsyncInfo::PendingList();
SShaderAsyncInfo::PendingList().m_Prev = &SShaderAsyncInfo::PendingList();
SShaderAsyncInfo::PendingListT().m_Next = &SShaderAsyncInfo::PendingListT();
SShaderAsyncInfo::PendingListT().m_Prev = &SShaderAsyncInfo::PendingListT();
}
SShaderAsyncInfo* pAI, * pAINext;
{
AUTO_LOCK(g_cAILock);
for (pAI = PendingListT().m_Next; pAI != &PendingListT(); pAI = pAINext)
{
pAINext = pAI->m_Next;
pAI->Unlink();
pAI->Link(&PendingList());
}
}
for (pAI = PendingList().m_Next; pAI != &PendingList(); pAI = pAINext)
{
pAINext = pAI->m_Next;
CHWShader_D3D* pSH = pAI->m_pShader;
if (pSH)
{
CHWShader_D3D::SHWSInstance* pInst = pSH->mfGetInstance(pAI->m_pFXShader, pAI->m_nHashInst, pSH->m_nMaskGenShader);
if (pInst->m_pAsync != pAI)
{
CryFatalError("Shader instance async info doesn't match queued async info.");
}
pSH->mfAsyncCompileReady(pInst);
}
}
}
void CShader::mfFlushPendedShaders()
{
SShaderAsyncInfo::FlushPendingShaders();
}
void CHWShader::mfFlushPendedShadersWait(int nMaxAllowed)
{
if (nMaxAllowed > 0 && SShaderAsyncInfo::s_nPendingAsyncShaders < nMaxAllowed)
{
return;
}
if (CRenderer::CV_r_shadersasynccompiling > 0)
{
iLog->Log("Flushing pended shaders...");
Start:
while (true)
{
if (SShaderAsyncInfo::s_nPendingAsyncShaders <= 0)
{
break;
}
int n = (int)iTimer->GetAsyncCurTime();
if (!(n % 2))
{
iLog->Update();
}
if (!(n % 8))
{
SShaderAsyncInfo::FlushPendingShaders();
}
else
{
Sleep(1);
}
}
// Compile FXC shaders or next iteration of internal shaders
SShaderAsyncInfo::FlushPendingShaders();
if (SShaderAsyncInfo::s_nPendingAsyncShaders)
{
goto Start;
}
iLog->Log("Finished flushing pended shaders...");
}
}
int CHWShader_D3D::mfAsyncCompileReady(SHWSInstance* pInst)
{
//SHWSInstance *pInst = m_pCurInst;
//assert(pInst->m_pAsync);
if (!pInst->m_pAsync)
{
return 0;
}
gRenDev->m_cEF.m_ShaderCacheStats.m_nNumShaderAsyncCompiles = SShaderAsyncInfo::s_nPendingAsyncShaders;
SShaderAsyncInfo* pAsync = pInst->m_pAsync;
int nFrame = gRenDev->GetFrameID(false);
if (pAsync->m_nFrame == nFrame)
{
if (pAsync->m_fMinDistance > gRenDev->m_RP.m_fMinDistance)
{
pAsync->m_fMinDistance = gRenDev->m_RP.m_fMinDistance;
}
}
else
{
pAsync->m_fMinDistance = gRenDev->m_RP.m_fMinDistance;
pAsync->m_nFrame = nFrame;
}
std::vector<SCGBind> InstBindVars;
LPD3D10BLOB pShader = NULL;
void* pConstantTable = NULL;
LPD3D10BLOB pErrorMsgs = NULL;
string strErr;
char nmDst[256], nameSrc[256];
bool bResult = true;
int nRefCount;
SShaderTechnique* pTech = gRenDev->m_RP.m_pCurTechnique;
CShader* pSH = pAsync->m_pFXShader;
{
if (pAsync->m_bPending)
{
return 0;
}
mfGetDstFileName(pInst, this, nmDst, 256, 3);
gEnv->pCryPak->AdjustFileName(nmDst, nameSrc, AZ_ARRAY_SIZE(nameSrc), 0);
if (pAsync->m_pFXShader && pAsync->m_pFXShader->m_HWTechniques.Num())
{
pTech = pAsync->m_pFXShader->m_HWTechniques[0];
}
if ((pAsync->m_pErrors && !pAsync->m_Errors.empty()) || !pAsync->m_pDevShader)
{
if (CRenderer::CV_r_logShaders)
{
gcpRendD3D->LogShv(SRendItem::m_RecurseLevel[gRenDev->m_RP.m_nProcessThreadID], "Async %d: **Failed to compile 0x%x '%s' shader\n", gRenDev->GetFrameID(false), pInst, nameSrc);
}
string Errors = pAsync->m_Errors;
string Text = pAsync->m_Text;
CShader* pFXShader = pAsync->m_pFXShader;
nRefCount = pFXShader ? pFXShader->GetRefCounter() : 0;
nRefCount = min(nRefCount, pAsync->m_pShader ? pAsync->m_pShader->GetRefCounter() : 0);
if (nRefCount <= 1) // Just exit if shader was deleted
{
pInst->m_pAsync = NULL;
SAFE_DELETE (pAsync);
return -1;
}
mfOutputCompilerError(Errors, Text.c_str());
Warning("Couldn't compile HW shader '%s'", GetName());
mfSaveCGFile(Text.c_str(), NULL);
bResult = false;
}
else if (CRenderer::CV_r_logShaders)
{
gcpRendD3D->LogShv(SRendItem::m_RecurseLevel[gRenDev->m_RP.m_nProcessThreadID], "Async %d: Finished compiling 0x%x '%s' shader\n", gRenDev->GetFrameID(false), pInst, nameSrc);
}
pShader = pAsync->m_pDevShader;
pErrorMsgs = pAsync->m_pErrors;
pConstantTable = pAsync->m_pConstants;
strErr = pAsync->m_Errors;
InstBindVars = decltype(InstBindVars)(pAsync->m_InstBindVars.begin(), pAsync->m_InstBindVars.end());
if (pAsync->m_bPendedEnv)
{
bResult &= CHWShader_D3D::mfCreateShaderEnv(pAsync->m_nThread, pInst, pAsync->m_pDevShader, pAsync->m_pConstants, pAsync->m_pErrors, InstBindVars, this, false, pAsync->m_pFXShader, pAsync->m_nCombination);
assert(bResult == true);
}
// Load samplers
if (pAsync->m_bPendedSamplers)
{
mfGatherFXParameters(pInst, &InstBindVars, this, 2, pAsync->m_pFXShader);
}
if (pAsync->m_bPendedFlush)
{
mfFlushCacheFile();
azstrcpy(nmDst, AZ_ARRAY_SIZE(nmDst), GetName());
char* s = strchr(nmDst, '(');
if (s)
{
s[0] = 0;
}
SShaderCombIdent Ident = pInst->m_Ident;
Ident.m_GLMask = m_nMaskGenFX;
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, nmDst, 0);
}
nRefCount = pAsync->m_pFXShader ? pAsync->m_pFXShader->GetRefCounter() : 0;
nRefCount = min(nRefCount, pAsync->m_pShader ? pAsync->m_pShader->GetRefCounter() : 0);
if (nRefCount <= 1) // Just exit if shader was deleted
{
pInst->m_pAsync = NULL;
SAFE_DELETE(pAsync);
return -1;
}
SAFE_DELETE (pInst->m_pAsync);
}
if (pErrorMsgs && !strErr.empty())
{
return -1;
}
bResult &= mfUploadHW(pShader, pInst, pSH, 0);
SAFE_RELEASE(pShader);
if (bResult)
{
if (pTech)
{
mfUpdatePreprocessFlags(pTech);
}
return 1;
}
return -1;
}
bool CHWShader_D3D::mfRequestAsync(CShader* pSH, SHWSInstance* pInst, std::vector<SCGBind>& InstBindVars, const char* prog_text, const char* szProfile, const char* szEntry)
{
#ifdef SHADER_ASYNC_COMPILATION
char nameSrc[256], nmDst[256];
mfGetDstFileName(pInst, this, nmDst, 256, 3);
gEnv->pCryPak->AdjustFileName(nmDst, nameSrc, AZ_ARRAY_SIZE(nameSrc), 0);
if (!SShaderAsyncInfo::PendingList().m_Next)
{
SShaderAsyncInfo::PendingList().m_Next = &SShaderAsyncInfo::PendingList();
SShaderAsyncInfo::PendingList().m_Prev = &SShaderAsyncInfo::PendingList();
SShaderAsyncInfo::PendingListT().m_Next = &SShaderAsyncInfo::PendingListT();
SShaderAsyncInfo::PendingListT().m_Prev = &SShaderAsyncInfo::PendingListT();
}
if (!m_pGlobalCache || !m_pGlobalCache->m_pRes[CACHE_USER])
{
if (m_pGlobalCache)
{
m_pGlobalCache->Release(false);
}
m_pGlobalCache = mfInitCache(NULL, this, true, m_CRC32, false, false);
}
pInst->m_pAsync = new SShaderAsyncInfo;
pInst->m_pAsync->m_fMinDistance = gRenDev->m_RP.m_fMinDistance;
pInst->m_pAsync->m_nFrame = gRenDev->GetFrameID(false);
pInst->m_pAsync->m_InstBindVars.assign(InstBindVars.data(), InstBindVars.data() + InstBindVars.size());
pInst->m_pAsync->m_pShader = this;
pInst->m_pAsync->m_pShader->AddRef();
pInst->m_pAsync->m_pFXShader = pSH;
pInst->m_pAsync->m_pFXShader->AddRef();
pInst->m_pAsync->m_nCombination = gRenDev->m_cEF.m_nCombinationsProcess;
assert(!azstricmp(m_NameSourceFX.c_str(), pInst->m_pAsync->m_pFXShader->m_NameFile.c_str()));
InstContainer* pInstCont = &m_Insts;
if (m_bUseLookUpTable)
{
pInst->m_pAsync->m_nHashInst = pInst->m_nContIndex;
}
else
{
pInst->m_pAsync->m_nHashInst = pInst->m_Ident.m_nHash;
}
pInst->m_pAsync->m_RTMask = pInst->m_Ident.m_RTMask;
pInst->m_pAsync->m_LightMask = pInst->m_Ident.m_LightMask;
pInst->m_pAsync->m_MDMask = pInst->m_Ident.m_MDMask;
pInst->m_pAsync->m_MDVMask = pInst->m_Ident.m_MDVMask;
pInst->m_pAsync->m_pipelineState.opaque = pInst->m_Ident.m_pipelineState.opaque;
pInst->m_pAsync->m_eClass = pInst->m_eClass;
pInst->m_pAsync->m_Text = prog_text;
pInst->m_pAsync->m_Name = szEntry;
pInst->m_pAsync->m_Profile = szProfile;
// Generate request line text to store on the shaderlist for next shader cache gen
{
char szShaderGenName[512];
cry_strcpy(szShaderGenName, GetName());
char* s = strchr(szShaderGenName, '(');
if (s)
{
s[0] = 0;
}
string RequestLine;
SShaderCombIdent Ident = pInst->m_Ident;
Ident.m_GLMask = m_nMaskGenFX;
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, szShaderGenName, 0, &RequestLine, false);
pInst->m_pAsync->m_RequestLine = RequestLine;
}
CAsyncShaderTask::InsertPendingShader(pInst->m_pAsync);
if (CRenderer::CV_r_logShaders)
{
gcpRendD3D->LogShv(SRendItem::m_RecurseLevel[gRenDev->m_RP.m_nProcessThreadID], "Async %d: Requested compiling 0x%x '%s' shader\n", gRenDev->GetFrameID(false), pInst, nameSrc);
}
#endif
return false;
}
void CHWShader_D3D::mfSubmitRequestLine(SHWSInstance* pInst, string* pRequestLine)
{
// Generate request line text.
char szShaderGenName[512];
cry_strcpy(szShaderGenName, GetName());
char* s = strchr(szShaderGenName, '(');
if (s)
{
s[0] = 0;
}
string RequestLine;
SShaderCombIdent Ident = pInst->m_Ident;
Ident.m_GLMask = m_nMaskGenFX;
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, szShaderGenName, 0, &RequestLine, false);
if (pRequestLine)
{
*pRequestLine = RequestLine;
}
if (!CRenderer::CV_r_shaderssubmitrequestline || !CRenderer::CV_r_shadersremotecompiler || pInst->m_bHasSendRequest)
{
return;
}
// make sure we only send the request once
pInst->m_bHasSendRequest = true;
#ifdef SHADER_ASYNC_COMPILATION
if (CRenderer::CV_r_shadersasynccompiling && !(m_Flags & HWSG_SYNC))
{
if (!SShaderAsyncInfo::PendingList().m_Next)
{
SShaderAsyncInfo::PendingList().m_Next = &SShaderAsyncInfo::PendingList();
SShaderAsyncInfo::PendingList().m_Prev = &SShaderAsyncInfo::PendingList();
SShaderAsyncInfo::PendingListT().m_Next = &SShaderAsyncInfo::PendingListT();
SShaderAsyncInfo::PendingListT().m_Prev = &SShaderAsyncInfo::PendingListT();
}
SShaderAsyncInfo* pAsync = new SShaderAsyncInfo;
if (pAsync)
{
pAsync->m_RequestLine = RequestLine;
pAsync->m_Text = "";
pAsync->m_bDeleteAfterRequest = true;
CAsyncShaderTask::InsertPendingShader(pAsync);
}
}
else
#endif
{
NRemoteCompiler::CShaderSrv::Instance().RequestLine(GetShaderListFilename().c_str(), RequestLine.c_str());
}
}
bool CHWShader_D3D::mfCompileHLSL_Int(CShader* pSH, char* prog_text, LPD3D10BLOB* ppShader, void** ppConstantTable, LPD3D10BLOB* ppErrorMsgs, string& strErr, std::vector<SCGBind>& InstBindVars)
{
HRESULT hr = S_OK;
SHWSInstance* pInst = m_pCurInst;
const char* szProfile = mfProfileString(pInst->m_eClass);
const char* pFunCCryName = m_EntryFunc.c_str();
bool bRes = true;
if (CRenderer::CV_r_shadersdebug == 2)
{
mfSaveCGFile(prog_text, "TestCG");
}
if (CRenderer::CV_r_shadersasynccompiling && !(m_Flags & HWSG_SYNC))
{
return mfRequestAsync(pSH, pInst, InstBindVars, prog_text, szProfile, pFunCCryName);
}
else
if (CRenderer::CV_r_shadersremotecompiler)
{
AZStd::string lsCompilerFlags = NRemoteCompiler::CShaderSrv::Instance().GetShaderCompilerFlags(pInst->m_eClass, pInst->m_Ident.m_pipelineState, pInst->m_Ident.m_MDVMask);
string RequestLine;
mfSubmitRequestLine(pInst, &RequestLine);
std::vector<uint8> Data;
if (NRemoteCompiler::ESOK != NRemoteCompiler::CShaderSrv::Instance().Compile(Data, szProfile, prog_text, pFunCCryName, lsCompilerFlags.c_str(), RequestLine.c_str()))
{
string sErrorText;
sErrorText.reserve(Data.size());
for (uint32 i = 0; i < Data.size(); i++)
{
sErrorText += Data[i];
}
strErr = sErrorText;
return false;
}
D3D10CreateBlob(Data.size(), (LPD3D10BLOB*)ppShader);
LPD3D10BLOB pShader = (LPD3D10BLOB) * ppShader;
DWORD* pBuf = (DWORD*) pShader->GetBufferPointer();
memcpy(pBuf, &Data[0], Data.size());
*ppShader = (LPD3D10BLOB)pShader;
pBuf = (DWORD*)pShader->GetBufferPointer();
size_t nSize = pShader->GetBufferSize();
bool bReflect = true;
#if !defined(CONSOLE)
if (CParserBin::PlatformIsConsole())
{
bReflect = false;
}
#endif
if (bReflect)
{
void* pShaderReflBuf;
hr = D3DReflect(pBuf, nSize, IID_ID3D11ShaderReflection, &pShaderReflBuf);
if (SUCCEEDED(hr))
{
ID3D11ShaderReflection* pShaderReflection = (ID3D11ShaderReflection*)pShaderReflBuf;
*ppConstantTable = (void*)pShaderReflection;
}
else
{
assert(0);
}
}
return hr == S_OK;
}
#if defined(WIN32) || defined(WIN64)
else
{
static bool s_logOnce_WrongPlatform = false;
# if !defined(OPENGL)
# if !defined(_RELEASE)
if (!s_logOnce_WrongPlatform && !(CParserBin::m_nPlatform == SF_D3D11 || CParserBin::m_nPlatform == SF_DURANGO))
{
s_logOnce_WrongPlatform = true;
iLog->LogError("Trying to build non DX11 shader via internal compiler which is not supported. Please use remote compiler instead!");
}
# endif
uint32 nFlags = D3D10_SHADER_PACK_MATRIX_ROW_MAJOR | D3D10_SHADER_ENABLE_BACKWARDS_COMPATIBILITY;
if (CRenderer::CV_r_shadersdebug == 3) // Generate debug information and do not optimize shader for ease of debugging
{
nFlags |= D3D10_SHADER_DEBUG | D3D10_SHADER_SKIP_OPTIMIZATION;
}
else if(CRenderer::CV_r_shadersdebug == 4) // Generate debug information, but still optimize shader
{
nFlags |= D3D10_SHADER_DEBUG;
}
hr = D3DCompile(prog_text, strlen(prog_text), GetName(), NULL, NULL, pFunCCryName, szProfile, nFlags, 0, (ID3DBlob**) ppShader, (ID3DBlob**) ppErrorMsgs);
if (FAILED(hr) || !*ppShader)
{
if (*ppErrorMsgs)
{
const char* err = (const char*)ppErrorMsgs[0]->GetBufferPointer();
strErr += err;
}
else
{
strErr += "D3DXCompileShader failed";
}
bRes = false;
}
else
{
void* pShaderReflBuf;
UINT* pData = (UINT*)ppShader[0]->GetBufferPointer();
UINT nSize = (uint32)ppShader[0]->GetBufferSize();
hr = D3DReflect(pData, nSize, IID_ID3D11ShaderReflection, &pShaderReflBuf);
if (SUCCEEDED(hr))
{
ID3D11ShaderReflection* pShaderReflection = (ID3D11ShaderReflection*)pShaderReflBuf;
*ppConstantTable = (void*)pShaderReflection;
}
else
{
assert(0);
}
}
return bRes;
# endif
}
#endif // #if defined(WIN32) || defined(WIN64)
return false;
}
LPD3D10BLOB CHWShader_D3D::mfCompileHLSL(CShader* pSH, char* prog_text, void** ppConstantTable, LPD3D10BLOB* ppErrorMsgs, uint32 nFlags, std::vector<SCGBind>& InstBindVars)
{
// LOADING_TIME_PROFILE_SECTION(iSystem);
// Test adding source text to context
SHWSInstance* pInst = m_pCurInst;
string strErr;
LPD3D10BLOB pCode = NULL;
HRESULT hr = S_OK;
if (!prog_text)
{
assert(0);
return NULL;
}
if (!CRenderer::CV_r_shadersAllowCompilation)
{
return NULL;
}
bool bResult = mfCompileHLSL_Int(pSH, prog_text, &pCode, ppConstantTable, ppErrorMsgs, strErr, InstBindVars);
if (!pCode)
{
if (CRenderer::CV_r_shadersasynccompiling)
{
return NULL;
}
if (!pCode)
{
{
mfOutputCompilerError(strErr, prog_text);
Warning("Couldn't compile HW shader '%s'", GetName());
mfSaveCGFile(prog_text, NULL);
}
}
}
return pCode;
}
void CHWShader_D3D::mfPrepareShaderDebugInfo(SHWSInstance* pInst, CHWShader_D3D* pSH, const char* szAsm, std::vector<SCGBind>& InstBindVars, void* pConstantTable)
{
if (szAsm)
{
char* szInst = strstr((char*)szAsm, "pproximately ");
if (szInst)
{
pInst->m_nInstructions = atoi(&szInst[13]);
}
}
// Confetti Nicholas Baldwin: adding metal shader language support
if (CParserBin::m_nPlatform == SF_D3D11 || CParserBin::m_nPlatform == SF_DURANGO || CParserBin::m_nPlatform == SF_GL4 || CParserBin::m_nPlatform == SF_GLES3 || CParserBin::m_nPlatform == SF_METAL)
{
ID3D11ShaderReflection* pShaderReflection = (ID3D11ShaderReflection*) pConstantTable;
if (pShaderReflection)
{
D3D11_SHADER_DESC Desc;
pShaderReflection->GetDesc(&Desc);
pInst->m_nInstructions = Desc.InstructionCount;
pInst->m_nTempRegs = Desc.TempRegisterCount;
}
}
if (CRenderer::CV_r_shadersdebug)
{
char nmdst[256];
mfGetDstFileName(pInst, pSH, nmdst, 256, 4);
AZStd::string szName = AZStd::string::format("%s%s.fxca", gRenDev->m_cEF.m_szCachePath.c_str(), nmdst);
AZ::IO::HandleType statusdstFileHandle = gEnv->pCryPak->FOpen(szName.c_str(), "wb");
if (statusdstFileHandle != AZ::IO::InvalidHandle)
{
gEnv->pCryPak->FPrintf(statusdstFileHandle, "\n// %s %s\n\n", "%STARTSHADER", mfProfileString(pInst->m_eClass));
if (pSH->m_eSHClass == eHWSC_Vertex)
{
for (uint32 i = 0; i < (uint32)InstBindVars.size(); i++)
{
SCGBind* pBind = &InstBindVars[i];
gEnv->pCryPak->FPrintf(statusdstFileHandle, "// %s %s %d %d\n", "%%", pBind->m_Name.c_str(), pBind->m_RegisterCount, pBind->m_RegisterOffset);
}
}
gEnv->pCryPak->FPrintf(statusdstFileHandle, "%s", szAsm);
gEnv->pCryPak->FPrintf(statusdstFileHandle, "\n// %s\n", "%ENDSHADER");
gEnv->pCryPak->FClose(statusdstFileHandle);
}
pInst->m_Handle.m_pShader = NULL;
}
}
void CHWShader_D3D::mfPrintCompileInfo(SHWSInstance* pInst)
{
int nConsts = 0;
int nParams = pInst->m_pBindVars.size();
for (int i = 0; i < nParams; i++)
{
SCGBind* pB = &pInst->m_pBindVars[i];
nConsts += pB->m_RegisterCount;
}
char szGenName[512];
cry_strcpy(szGenName, GetName());
char* s = strchr(szGenName, '(');
if (s)
{
s[0] = 0;
}
if (CRenderer::CV_r_shadersdebug == 2)
{
string pName;
SShaderCombIdent Ident(m_nMaskGenFX, pInst->m_Ident);
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, szGenName, 0, &pName, false);
CryLog(" Compile %s (%d instructions, %d tempregs, %d/%d constants) ... ", pName.c_str(), pInst->m_nInstructions, pInst->m_nTempRegs, nParams, nConsts);
int nSize = strlen(szGenName);
mfGenName(pInst, &szGenName[nSize], 512 - nSize, 1);
CryLog(" --- Cache entry: %s", szGenName);
}
else
{
int nSize = strlen(szGenName);
mfGenName(pInst, &szGenName[nSize], 512 - nSize, 1);
CryLog(" Compile %s (%d instructions, %d tempregs, %d/%d constants) ... ", szGenName, pInst->m_nInstructions, pInst->m_nTempRegs, nParams, nConsts);
}
if (gRenDev->m_cEF.m_bActivated && CRenderer::CV_r_shadersdebug > 0)
{
CryLog(" Shader %s (%llx)(%x)(%x)(%x)(%llx)(%llx)(%s) wasn't compiled before preactivating phase",
GetName(), pInst->m_Ident.m_RTMask, pInst->m_Ident.m_LightMask, pInst->m_Ident.m_MDMask, pInst->m_Ident.m_MDVMask, pInst->m_Ident.m_pipelineState.opaque, pInst->m_Ident.m_STMask, mfProfileString(pInst->m_eClass));
}
}
bool CHWShader_D3D::mfCreateShaderEnv(int nThread, SHWSInstance* pInst, LPD3D10BLOB pShader, void* pConstantTable, LPD3D10BLOB pErrorMsgs, std::vector<SCGBind>& InstBindVars, CHWShader_D3D* pSH, bool bShaderThread, CShader* pFXShader, int nCombination, const char* src)
{
// Create asm (.fxca) cache file
assert(pInst);
if (!pInst)
{
return false;
}
CSpinLock lock;
if (pInst->m_pBindVars.size())
{
return true;
}
if (pShader && (nCombination < 0))
{
#if !defined(OPENGL)
ID3D10Blob* pAsm = NULL;
ID3D10Blob* pSrc = (ID3D10Blob*)pShader;
UINT* pBuf = (UINT*)pSrc->GetBufferPointer();
D3DDisassemble(pBuf, pSrc->GetBufferSize(), 0, NULL, &pAsm);
if (pAsm)
{
char* szAsm = (char*)pAsm->GetBufferPointer();
mfPrepareShaderDebugInfo(pInst, pSH, szAsm, InstBindVars, pConstantTable);
}
SAFE_RELEASE(pAsm);
#endif
}
//assert(!pInst->m_pBindVars);
if (pShader)
{
bool bVF = pSH->m_eSHClass == eHWSC_Vertex;
#if !defined(CONSOLE)
if (CParserBin::PlatformIsConsole())
{
bVF = false;
}
#endif
#if !defined(OPENGL)
if (CParserBin::m_nPlatform & (SF_GL4 | SF_GLES3))
{
bVF = false;
}
#endif
#if defined(CRY_USE_METAL)
if (!(CParserBin::m_nPlatform & (SF_METAL)))
{
bVF = false;
}
#endif
if (bVF)
{
mfVertexFormat(pInst, pSH, pShader);
}
if (pConstantTable)
{
mfCreateBinds(pInst, pConstantTable, (byte*)pShader->GetBufferPointer(), (uint32)pShader->GetBufferSize());
}
}
if (!(pSH->m_Flags & HWSG_PRECACHEPHASE))
{
int nConsts = 0;
int nParams = pInst->m_pBindVars.size();
for (int i = 0; i < nParams; i++)
{
SCGBind* pB = &pInst->m_pBindVars[i];
nConsts += pB->m_RegisterCount;
}
if (gRenDev->m_cEF.m_nCombinationsProcess >= 0)
{
//assert(!bShaderThread);
//if (!(gRenDev->m_cEF.m_nCombination & 0xff))
if (!CParserBin::m_nPlatform)
{
CryLog("%d: Compile %s %s (%d out of %d) - (%d/%d constants) ... ", nThread,
mfProfileString(pInst->m_eClass), pSH->GetName(), nCombination, gRenDev->m_cEF.m_nCombinationsProcessOverall,
nParams, nConsts);
}
else
{
CryLog("%d: Compile %s %s (%d out of %d) ... ", nThread,
mfProfileString(pInst->m_eClass), pSH->GetName(), nCombination, gRenDev->m_cEF.m_nCombinationsProcessOverall);
}
}
else
{
pSH->mfPrintCompileInfo(pInst);
}
}
mfGatherFXParameters(pInst, &InstBindVars, pSH, bShaderThread ? 1 : 0, pFXShader);
if (pShader)
{
mfCreateCacheItem(pInst, InstBindVars, (byte*)pShader->GetBufferPointer(), (uint32)pShader->GetBufferSize(), pSH, bShaderThread);
}
else if (CRenderer::CV_r_shadersCacheUnavailableShaders)
{
// If the shader is unavailable, cache an invalid item to avoid requesting its compilation again in future executions
mfCreateCacheItem(pInst, InstBindVars, NULL, 0, pSH, bShaderThread);
}
#if !defined (NULL_RENDERER)
ID3D11ShaderReflection* pRFL = (ID3D11ShaderReflection*)pConstantTable;
ID3D10Blob* pER = (ID3D10Blob*)pErrorMsgs;
SAFE_RELEASE(pRFL);
SAFE_RELEASE(pER);
#endif
return true;
}
// Compile pixel/vertex shader for the current instance properties
bool CHWShader_D3D::mfActivate(CShader* pSH, uint32 nFlags, FXShaderToken* Table, TArray<uint32>* pSHData, bool bCompressedOnly)
{
PROFILE_FRAME(Shader_HWShaderActivate);
AZ_TRACE_METHOD();
bool bResult = true;
SHWSInstance* pInst = m_pCurInst;
mfLogShaderRequest(pInst);
if (mfIsValid(pInst, true) == ED3DShError_NotCompiled)
{
//if (!(m_Flags & HWSG_PRECACHEPHASE) && !(nFlags & HWSF_NEXT))
// mfSetHWStartProfile(nFlags);
bool bCreate = false;
// We need a different source and desination for fpStripExtension
// since a call to strcpy with the same src and dst results in
// undefined behaviour
char nameCacheUnstripped[256];
char nameCache[256];
float t0 = gEnv->pTimer->GetAsyncCurTime();
mfGetDstFileName(pInst, this, nameCacheUnstripped, 256, 0);
fpStripExtension(nameCacheUnstripped, nameCache);
fpAddExtension(nameCache, ".fxcb");
if (!m_pDevCache)
{
m_pDevCache = mfInitDevCache(nameCache, this);
}
int32 nSize;
SShaderCacheHeaderItem* pCacheItem = mfGetCompressedItem(nFlags, nSize);
if (pCacheItem)
{
pInst->m_bCompressed = true;
}
else if (bCompressedOnly)
{
// don't activate if shader isn't found in compressed shader data
return false;
}
else
{
// if shader compiling is enabled, make sure the user folder shader caches are also available
bool bReadOnly = CRenderer::CV_r_shadersAllowCompilation == 0;
if (!m_pGlobalCache || m_pGlobalCache->m_nPlatform != CParserBin::m_nPlatform ||
(!bReadOnly && !m_pGlobalCache->m_pRes[CACHE_USER]))
{
SAFE_RELEASE(m_pGlobalCache);
bool bAsync = CRenderer::CV_r_shadersasyncactivation != 0;
if (nFlags & HWSF_PRECACHE)
{
bAsync = false;
}
m_pGlobalCache = mfInitCache(nameCache, this, true, m_CRC32, bReadOnly, bAsync);
}
if (gRenDev->m_cEF.m_nCombinationsProcess >= 0 && !gRenDev->m_cEF.m_bActivatePhase)
{
mfGetDstFileName(pInst, this, nameCache, 256, 0);
fpStripExtension(nameCache, nameCache);
fpAddExtension(nameCache, ".fxcb");
FXShaderCacheNamesItor it = m_ShaderCacheList.find(nameCache);
if (it == m_ShaderCacheList.end())
{
m_ShaderCacheList.insert(FXShaderCacheNamesItor::value_type(nameCache, m_CRC32));
}
}
pCacheItem = mfGetCacheItem(nFlags, nSize);
}
if (pCacheItem && pCacheItem->m_Class != 255)
{
if (Table && CRenderer::CV_r_shadersAllowCompilation)
{
mfGetCacheTokenMap(Table, pSHData, m_nMaskGenShader);
}
if (((m_Flags & HWSG_PRECACHEPHASE) || gRenDev->m_cEF.m_nCombinationsProcess >= 0) && !gRenDev->m_cEF.m_bActivatePhase)
{
byte* pData = (byte*)pCacheItem;
SAFE_DELETE_ARRAY(pData);
return true;
}
bool bRes = false;
bRes = mfActivateCacheItem(pSH, pCacheItem, nSize, nFlags);
byte* pData = (byte*)pCacheItem;
SAFE_DELETE_ARRAY(pData);
pCacheItem = NULL;
if (CRenderer::CV_r_shaderspreactivate == 2 && !gRenDev->m_cEF.m_bActivatePhase)
{
t0 = gEnv->pTimer->GetAsyncCurTime() - t0;
iLog->Log("Warning: Shader activation (%.3f ms): %s(%llx)(%x)(%x)(%x)(%llx)(%llx)(%s)...",
t0 * 1000.0f, GetName(), pInst->m_Ident.m_RTMask, pInst->m_Ident.m_LightMask, pInst->m_Ident.m_MDMask, pInst->m_Ident.m_MDVMask,
pInst->m_Ident.m_pipelineState.opaque, pInst->m_Ident.m_STMask, mfProfileString(pInst->m_eClass));
char name[256];
azstrcpy(name, AZ_ARRAY_SIZE(name), GetName());
char* s = strchr(name, '(');
if (s)
{
s[0] = 0;
}
string pName;
SShaderCombIdent Ident(m_nMaskGenFX, pInst->m_Ident);
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, name, 0, &pName, false);
iLog->Log("...Shader list entry: %s (%llx)", pName.c_str(), m_nMaskGenFX);
}
if (bRes)
{
return (pInst->m_Handle.m_pShader != NULL);
}
pCacheItem = NULL;
}
else
if (pCacheItem && pCacheItem->m_Class == 255)
{
byte* pData = (byte*)pCacheItem;
SAFE_DELETE_ARRAY(pData);
pCacheItem = NULL;
return false;
}
else
if (gRenDev->m_cEF.m_bActivatePhase)
{
if (CRenderer::CV_r_shadersdebug > 0)
{
iLog->Log("Warning: Shader %s(%llx)(%x)(%x)(%x)(%llx)(%llx)(%s) wasn't compiled before preactivating phase",
GetName(), pInst->m_Ident.m_RTMask, pInst->m_Ident.m_LightMask, pInst->m_Ident.m_MDMask, pInst->m_Ident.m_MDVMask, pInst->m_Ident.m_pipelineState.opaque, pInst->m_Ident.m_STMask, mfProfileString(pInst->m_eClass));
}
byte* pData = (byte*)pCacheItem;
SAFE_DELETE_ARRAY(pData);
pCacheItem = NULL;
return false;
}
if (pCacheItem)
{
byte* pData = (byte*)pCacheItem;
SAFE_DELETE_ARRAY(pData);
pCacheItem = NULL;
}
//assert(!m_TokenData.empty());
TArray<char> newScr;
if (nFlags & HWSF_PRECACHE)
{
gRenDev->m_cEF.m_nCombinationsCompiled++;
}
float fTime0 = iTimer->GetAsyncCurTime();
LPD3D10BLOB pShader = NULL;
void* pConstantTable = NULL;
LPD3D10BLOB pErrorMsgs = NULL;
std::vector<SCGBind> InstBindVars;
m_Flags |= HWSG_WASGENERATED;
bool bScriptSuccess = false;
if (CRenderer::CV_r_shadersAllowCompilation)
{
// MemReplay shows that 16kb should be enough memory to hold the script without having to reallocate
newScr.reserve(16 * 1024);
bScriptSuccess = mfGenerateScript(pSH, pInst, InstBindVars, nFlags, Table, pSHData, newScr);
ASSERT_IN_SHADER(bScriptSuccess);
}
if (!pInst->m_bAsyncActivating && !bCompressedOnly)
{
// report miss in global cache to log and/or callback
mfLogShaderCacheMiss(pInst);
// still sumit the request line when no compiling to be sure that the shadercompiler will
// compile it in the next build
if (!CRenderer::CV_r_shadersAllowCompilation)
{
mfSubmitRequestLine(pInst);
}
}
if (!bScriptSuccess)
{
if (!pInst->m_bAsyncActivating)
{
Warning("Warning: Shader %s(%llx)(%x)(%x)(%x)(%llx)(%llx)(%s) is not existing in the cache\n",
GetName(), pInst->m_Ident.m_RTMask, pInst->m_Ident.m_LightMask, pInst->m_Ident.m_MDMask, pInst->m_Ident.m_MDVMask, pInst->m_Ident.m_pipelineState.opaque, pInst->m_Ident.m_STMask, mfProfileString(pInst->m_eClass));
}
return false;
}
{
PROFILE_FRAME(Shader_CompileHLSL);
pShader = mfCompileHLSL(pSH, newScr.Data(), &pConstantTable, &pErrorMsgs, nFlags, InstBindVars);
}
gRenDev->m_cEF.m_ShaderCacheStats.m_nNumShaderAsyncCompiles = SShaderAsyncInfo::s_nPendingAsyncShaders;
if (!pShader)
{
if (!CRenderer::CV_r_shadersAllowCompilation || pInst->IsAsyncCompiling())
{
return false;
}
}
bResult = mfCreateShaderEnv(0, pInst, pShader, pConstantTable, pErrorMsgs, InstBindVars, this, false, pSH, gRenDev->m_cEF.m_nCombinationsProcess, newScr.Data());
bResult &= mfUploadHW(pShader, pInst, pSH, nFlags);
SAFE_RELEASE(pShader);
fTime0 = iTimer->GetAsyncCurTime() - fTime0;
//iLog->LogToConsole(" Time activate: %.3f", fTime0);
}
else
if (pSHData)
{
mfGetCacheTokenMap(Table, pSHData, m_nMaskGenShader);
}
bool bSuccess = (mfIsValid(pInst, true) == ED3DShError_Ok);
return bSuccess;
}
//////////////////////////////////////////////////////////////////////////
#ifdef SHADER_ASYNC_COMPILATION
#pragma warning(disable: 4355) // warning C4355: 'this' : used in base member initializer list
CAsyncShaderTask::CAsyncShaderTask()
: m_thread(this)
{
}
void CAsyncShaderTask::InsertPendingShader(SShaderAsyncInfo* pAsync)
{
AUTO_LOCK(g_cAILock);
pAsync->Link(&BuildList());
CryInterlockedIncrement(&SShaderAsyncInfo::s_nPendingAsyncShaders);
}
void CAsyncShaderTask::FlushPendingShaders()
{
SShaderAsyncInfo* pAI, * pAI2, * pAINext;
assert(m_flush_list.m_Prev == &m_flush_list && m_flush_list.m_Next == &m_flush_list); // the flush list must be empty - cleared by the shader compile thread
if (BuildList().m_Prev == &BuildList() && BuildList().m_Next == &BuildList())
{
return; // the build list is empty, might need to do some assert here
}
{
AUTO_LOCK(g_cAILock);
int n = 0;
for (pAI = BuildList().m_Prev; pAI != &BuildList(); pAI = pAINext)
{
pAINext = pAI->m_Prev;
pAI->Unlink();
pAI->Link(&m_flush_list);
}
}
// Sorting by distance
if (gRenDev->m_cEF.m_nCombinationsProcess < 0)
{
for (pAI = m_flush_list.m_Next; pAI != &m_flush_list; pAI = pAI->m_Next)
{
assert(pAI);
PREFAST_ASSUME(pAI);
pAINext = NULL;
int nFrame = pAI->m_nFrame;
float fDist = pAI->m_fMinDistance;
for (pAI2 = pAI->m_Next; pAI2 != &m_flush_list; pAI2 = pAI2->m_Next)
{
if (pAI2->m_nFrame < nFrame)
{
continue;
}
if (pAI2->m_nFrame > nFrame || pAI2->m_fMinDistance < fDist)
{
pAINext = pAI2;
nFrame = pAI2->m_nFrame;
fDist = pAI2->m_fMinDistance;
}
}
if (pAINext)
{
assert(pAI != pAINext);
SShaderAsyncInfo* pAIP0 = pAI->m_Prev;
SShaderAsyncInfo* pAIP1 = pAINext->m_Prev == pAI ? pAINext : pAINext->m_Prev;
pAI->m_Next->m_Prev = pAI->m_Prev;
pAI->m_Prev->m_Next = pAI->m_Next;
pAI->m_Next = pAIP1->m_Next;
pAIP1->m_Next->m_Prev = pAI;
pAIP1->m_Next = pAI;
pAI->m_Prev = pAIP1;
pAI = pAINext;
pAI->m_Next->m_Prev = pAI->m_Prev;
pAI->m_Prev->m_Next = pAI->m_Next;
pAI->m_Next = pAIP0->m_Next;
pAIP0->m_Next->m_Prev = pAI;
pAIP0->m_Next = pAI;
pAI->m_Prev = pAIP0;
}
}
}
for (pAI = m_flush_list.m_Next; pAI != &m_flush_list; pAI = pAINext)
{
pAINext = pAI->m_Next;
assert(pAI->m_bPending);
SubmitAsyncRequestLine(pAI);
if (pAI->m_Text.length() > 0)
{
CompileAsyncShader(pAI);
}
CryInterlockedDecrement(&SShaderAsyncInfo::s_nPendingAsyncShaders);
{
AUTO_LOCK(g_cAILock);
pAI->Unlink();
pAI->m_bPending = 0;
pAI->Link(&SShaderAsyncInfo::PendingListT());
}
if (pAI->m_bDeleteAfterRequest)
{
SAFE_DELETE(pAI);
}
}
}
bool CAsyncShaderTask::PostCompile(SShaderAsyncInfo* pAsync)
{
bool bResult = true;
/*if (pAsync->m_nCombination < 0 && false)
{
CHWShader_D3D *pSH = pAsync->m_pShader;
CHWShader_D3D::SHWSInstance *pInst = pSH->mfGetInstance(pAsync->m_nOwner, pSH->m_nMaskGenShader);
bResult = CHWShader_D3D::mfCreateShaderEnv(m_nThread, pInst, pAsync->m_pDevShader, pAsync->m_pConstants, pAsync->m_pErrors, pAsync->m_InstBindVars, pSH, true, pAsync->m_pFXShader, pAsync->m_nCombination);
assert(bResult == true);
}
else*/
{
pAsync->m_nThread = m_nThread;
pAsync->m_bPendedEnv = true;
}
return bResult;
}
void CAsyncShaderTask::SubmitAsyncRequestLine(SShaderAsyncInfo* pAsync)
{
if (CRenderer::CV_r_shadersremotecompiler)
{
if (!pAsync->m_shaderList.empty())
{
NRemoteCompiler::CShaderSrv::Instance().RequestLine(pAsync->m_shaderList.c_str(), pAsync->m_RequestLine.c_str());
}
else
{
NRemoteCompiler::CShaderSrv::Instance().RequestLine(GetShaderListFilename().c_str(), pAsync->m_RequestLine.c_str());
}
}
}
bool CAsyncShaderTask::CompileAsyncShader(SShaderAsyncInfo* pAsync)
{
#if !(defined(MOBILE) || defined(CONSOLE))
const bool outputShaderSourceFiles = CRenderer::CV_r_OutputShaderSourceFiles != 0;
#else
const bool outputShaderSourceFiles = false;
#endif
string shaderSourceOutputFolder;
if ( outputShaderSourceFiles )
{
CryFixedStringT<1024> hlslPath;
// Create a directory for this shader type, strip the .fxcb extension from the folder name
shaderSourceOutputFolder.Format("@cache@/%s",pAsync->m_pShader->m_pDevCache->m_Name.c_str());
PathUtil::RemoveExtension(shaderSourceOutputFolder);
gEnv->pFileIO->CreatePath(shaderSourceOutputFolder);
// Output both HLSL here, and GLSL/Metal below since those require the remote shader compiler to execute first
hlslPath.Format("%s/[0x%08x].hlsl",shaderSourceOutputFolder.c_str(),pAsync->m_nHashInst);
{
AZ::IO::HandleType hlslHandle;
if ( gEnv->pFileIO->Open(hlslPath.c_str(),AZ::IO::OpenMode::ModeWrite,hlslHandle) )
{
gEnv->pFileIO->Write(hlslHandle, pAsync->m_Text.c_str(), pAsync->m_Text.size());
gEnv->pFileIO->Flush(hlslHandle);
gEnv->pFileIO->Close(hlslHandle);
}
}
}
bool bResult = true;
if (CRenderer::CV_r_shadersremotecompiler)
{
AZStd::string lsCompilerFlags = NRemoteCompiler::CShaderSrv::Instance().GetShaderCompilerFlags(pAsync->m_eClass, pAsync->m_pipelineState, pAsync->m_MDVMask);
std::vector<uint8> Data;
if (NRemoteCompiler::ESOK != NRemoteCompiler::CShaderSrv::Instance().Compile(Data, pAsync->m_Profile, pAsync->m_Text.c_str(), pAsync->m_Name.c_str(), lsCompilerFlags.c_str(), pAsync->m_RequestLine.c_str()))
{
#if !defined(NULL_RENDERER)
D3D10CreateBlob(sizeof("D3DXCompileShader failed"), (LPD3D10BLOB*)&pAsync->m_pErrors);
DWORD* pBuf = (DWORD*) pAsync->m_pErrors->GetBufferPointer();
memcpy(pBuf, "D3DXCompileShader failed", sizeof("D3DXCompileShader failed"));
#endif
string sErrorText;
if (!Data.empty())
{
sErrorText.reserve(Data.size());
for (uint32 i = 0; i < Data.size(); i++)
{
sErrorText += Data[i];
}
}
else
{
sErrorText.assign("Unknown Error");
}
pAsync->m_Errors += sErrorText;
return false;
}
HRESULT hr = S_OK;
D3D10CreateBlob(Data.size(), (LPD3D10BLOB*) & pAsync->m_pDevShader);
LPD3D10BLOB pShader = (LPD3D10BLOB)*&pAsync->m_pDevShader;
DWORD* pBuf = (DWORD*)pShader->GetBufferPointer();
memcpy(pBuf, &Data[0], Data.size());
pAsync->m_pDevShader = (LPD3D10BLOB)pShader;
pBuf = (DWORD*)pShader->GetBufferPointer();
size_t nSize = pShader->GetBufferSize();
if ( outputShaderSourceFiles )
{
#if defined(OPENGL)
CryFixedStringT<1024> glslPath;
glslPath.Format("%s/[0x%08x].glsl",shaderSourceOutputFolder.c_str(),pAsync->m_nHashInst);
{
AZ::IO::HandleType glslHandle;
if ( gEnv->pFileIO->Open(glslPath.c_str(),AZ::IO::OpenMode::ModeWrite,glslHandle) )
{
gEnv->pFileIO->Write(glslHandle,static_cast<const void*>(pBuf),nSize);
gEnv->pFileIO->Flush(glslHandle);
gEnv->pFileIO->Close(glslHandle);
}
}
#endif
#if defined(CRY_USE_METAL)
CryFixedStringT<1024> metallibPath;
metallibPath.Format("%s/[0x%08x].metallib",shaderSourceOutputFolder.c_str(),pAsync->m_nHashInst);
{
AZ::IO::HandleType metalHandle;
if ( gEnv->pFileIO->Open(metallibPath.c_str(),AZ::IO::OpenMode::ModeWrite,metalHandle) )
{
gEnv->pFileIO->Write(metalHandle,static_cast<const void*>(pBuf),nSize);
gEnv->pFileIO->Flush(metalHandle);
gEnv->pFileIO->Close(metalHandle);
}
}
#endif
}
bool bReflect = true;
#if !defined(CONSOLE)
if (CParserBin::PlatformIsConsole())
{
bReflect = false;
}
#endif
#if !defined(OPENGL)
if (CParserBin::m_nPlatform & (SF_GL4 | SF_GLES3))
{
bReflect = false;
}
#endif
#if defined(CRY_USE_METAL)
if (!(CParserBin::m_nPlatform & (SF_METAL)))
{
bReflect = false;
}
#endif
if (bReflect)
{
ID3D11ShaderReflection* pShaderReflection;
hr = D3DReflect(pBuf, nSize, IID_ID3D11ShaderReflection, (void**)&pShaderReflection);
if (SUCCEEDED(hr))
{
pAsync->m_pConstants = (void*)pShaderReflection;
}
}
if (SUCCEEDED(hr))
{
bResult = PostCompile(pAsync);
}
else
{
pAsync->m_pDevShader = 0;
assert(0);
}
}
#if defined(WIN32) && !defined(OPENGL)
else
{
static bool s_logOnce_WrongPlatform = false;
# if !defined(_RELEASE)
if (!s_logOnce_WrongPlatform && !(CParserBin::m_nPlatform == SF_D3D11 || CParserBin::m_nPlatform == SF_DURANGO))
{
s_logOnce_WrongPlatform = true;
iLog->LogError("Trying to build non DX11 shader via internal compiler which is not supported. Please use remote compiler instead!");
}
# endif
uint32 nFlags = D3D10_SHADER_PACK_MATRIX_ROW_MAJOR | D3D10_SHADER_ENABLE_BACKWARDS_COMPATIBILITY;
if (CRenderer::CV_r_shadersdebug == 3) // Generate debug information and do not optimize shader for ease of debugging
{
nFlags |= D3D10_SHADER_DEBUG | D3D10_SHADER_SKIP_OPTIMIZATION;
}
else if(CRenderer::CV_r_shadersdebug == 4) // Generate debug information, but still optimize shader
{
nFlags |= D3D10_SHADER_DEBUG;
}
const char* Name = pAsync->m_pShader ? pAsync->m_pShader->GetName() : "Unknown";
HRESULT hr = S_OK;
hr = D3DCompile(pAsync->m_Text.c_str(), pAsync->m_Text.size(), Name, NULL, NULL, pAsync->m_Name.c_str(), pAsync->m_Profile.c_str(), nFlags, 0, (ID3DBlob * *) & pAsync->m_pDevShader, (ID3DBlob**) &pAsync->m_pErrors);
if (FAILED(hr) || !pAsync->m_pDevShader)
{
if (pAsync->m_pErrors)
{
const char* err = (const char*)pAsync->m_pErrors->GetBufferPointer();
pAsync->m_Errors += err;
}
else
{
pAsync->m_Errors += "D3DXCompileShader failed";
}
bResult = false;
}
else
{
ID3D11ShaderReflection* pShaderReflection;
UINT* pData = (UINT*)pAsync->m_pDevShader->GetBufferPointer();
size_t nSize = pAsync->m_pDevShader->GetBufferSize();
hr = D3DReflect(pData, nSize, IID_ID3D11ShaderReflection, (void**)&pShaderReflection);
if (SUCCEEDED(hr))
{
pAsync->m_pConstants = (void*)pShaderReflection;
bResult = PostCompile(pAsync);
}
else
{
iLog->LogWarning("ERROR: Shader Reflection Failed!");
assert(0);
}
}
}
#endif // #if defined(WIN32) || defined(WIN64)
return bResult;
}
void CAsyncShaderTask::CShaderThread::Run()
{
CryThreadSetName(-1, SHADER_THREAD_NAME);
#if defined(AZ_RESTRICTED_PLATFORM)
#define AZ_RESTRICTED_SECTION D3DHWSHADERCOMPILING_CPP_SECTION_6
#if defined(AZ_PLATFORM_XENIA)
#include "Xenia/D3DHWShaderCompiling_cpp_xenia.inl"
#elif defined(AZ_PLATFORM_PROVO)
#include "Provo/D3DHWShaderCompiling_cpp_provo.inl"
#elif defined(AZ_PLATFORM_SALEM)
#include "Salem/D3DHWShaderCompiling_cpp_salem.inl"
#endif
#endif
while (!m_quit)
{
m_task->FlushPendingShaders();
if (!CRenderer::CV_r_shadersasynccompiling)
{
Sleep(250);
}
else
{
Sleep(25);
}
}
}
#endif
//===============================================================================================
// Export/Import
#ifdef SHADERS_SERIALIZING
bool STexSamplerFX::Export(SShaderSerializeContext& SC)
{
bool bRes = true;
SSTexSamplerFX TS;
TS.m_nRTIdx = -1;
TS.m_nsName = SC.AddString(m_szName.c_str());
TS.m_nsNameTexture = SC.AddString(m_szTexture.c_str());
TS.m_eTexType = m_eTexType;
TS.m_nSamplerSlot = m_nSlotId;
TS.m_nTexFlags = m_nTexFlags;
if (m_nTexState > 0)
{
TS.m_bTexState = 1;
STexState* pTS = &CTexture::s_TexStates[m_nTexState];
memcpy(&TS.ST, &CTexture::s_TexStates[m_nTexState], sizeof(TS.ST));
TS.ST.m_pDeviceState = NULL;
}
if (m_pTarget)
{
TS.m_nRTIdx = SC.FXTexRTs.Num();
SHRenderTarget* pRT = m_pTarget;
SSHRenderTarget RT;
RT.m_eOrder = pRT->m_eOrder;
RT.m_nProcessFlags = pRT->m_nProcessFlags;
RT.m_nsTargetName = SC.AddString(pRT->m_TargetName.c_str());
RT.m_nWidth = pRT->m_nWidth;
RT.m_nHeight = pRT->m_nHeight;
RT.m_eTF = pRT->m_eTF;
RT.m_nIDInPool = pRT->m_nIDInPool;
RT.m_eUpdateType = pRT->m_eUpdateType;
RT.m_bTempDepth = pRT->m_bTempDepth;
RT.m_ClearColor = pRT->m_ClearColor;
RT.m_fClearDepth = pRT->m_fClearDepth;
RT.m_nFlags = pRT->m_nFlags;
RT.m_nFilterFlags = pRT->m_nFilterFlags;
SC.FXTexRTs.push_back(RT);
}
//crude workaround for TArray push_back() bug a=b operator. Does not init a, breaks texState dtor for a
SSTexSamplerFX* pNewSampler = SC.FXTexSamplers.AddIndex(1);
memcpy(pNewSampler, &TS, sizeof(SSTexSamplerFX));
return bRes;
}
bool STexSamplerFX::Import(SShaderSerializeContext& SC, SSTexSamplerFX* pTS)
{
bool bRes = true;
m_szName = sString(pTS->m_nsName, SC.Strings);
m_szTexture = sString(pTS->m_nsNameTexture, SC.Strings);
m_eTexType = pTS->m_eTexType;
m_nSlotId = pTS->m_nSamplerSlot;
m_nTexFlags = pTS->m_nTexFlags;
if (pTS->m_bTexState)
{
m_nTexState = CTexture::GetTexState(pTS->ST);
}
if (pTS->m_nRTIdx != -1)
{
SSHRenderTarget* pRT = &SC.FXTexRTs[pTS->m_nRTIdx];
SHRenderTarget* pDst = new SHRenderTarget;
pDst->m_eOrder = pRT->m_eOrder;
pDst->m_nProcessFlags = pRT->m_nProcessFlags;
pDst->m_TargetName = sString(pRT->m_nsTargetName, SC.Strings);
pDst->m_nWidth = pRT->m_nWidth;
pDst->m_nHeight = pRT->m_nHeight;
pDst->m_eTF = pRT->m_eTF;
pDst->m_nIDInPool = pRT->m_nIDInPool;
pDst->m_eUpdateType = pRT->m_eUpdateType;
pDst->m_bTempDepth = pRT->m_bTempDepth != 0;
pDst->m_ClearColor = pRT->m_ClearColor;
pDst->m_fClearDepth = pRT->m_fClearDepth;
pDst->m_nFlags = pRT->m_nFlags;
pDst->m_nFilterFlags = pRT->m_nFilterFlags;
m_pTarget = pDst;
}
PostLoad();
return bRes;
}
bool SFXParam::Export(SShaderSerializeContext& SC)
{
bool bRes = true;
SSFXParam PR;
PR.m_nsName = SC.AddString(m_Name.c_str());
PR.m_nsAnnotations = SC.AddString(m_Annotations.c_str());
PR.m_nsSemantic = SC.AddString(m_Semantic.c_str());
PR.m_nsValues = SC.AddString(m_Values.c_str());
PR.m_eType = m_eType;
PR.m_nCB = m_BindingSlot;
PR.m_nComps = m_ComponentCount;
PR.m_nFlags = m_nFlags;
PR.m_nParameters = m_RegisterCount;
for ( int i=0; i<eHWSC_Num; ++i )
{
PR.m_nRegister[i] = m_Register[i];
}
SC.FXParams.push_back(PR);
return bRes;
}
bool SFXParam::Import(SShaderSerializeContext& SC, SSFXParam* pPR)
{
bool bRes = true;
m_Name = sString(pPR->m_nsName, SC.Strings);
m_Annotations = sString(pPR->m_nsAnnotations, SC.Strings);
m_Semantic = sString(pPR->m_nsSemantic, SC.Strings);
m_Values = sString(pPR->m_nsValues, SC.Strings);
m_eType = pPR->m_eType;
m_BindingSlot = pPR->m_nCB;
m_ComponentCount = pPR->m_nComps;
m_nFlags = pPR->m_nFlags;
m_RegisterCount = pPR->m_nParameters;
for ( int i=0; i<eHWSC_Num; ++i )
{
m_Register[i] = pPR->m_nRegister[i];
}
return bRes;
}
bool SFXSampler::Export(SShaderSerializeContext& SC)
{
bool bRes = true;
SSFXSampler PR;
PR.m_nsName = SC.AddString(m_Name.c_str());
PR.m_nsAnnotations = SC.AddString(m_Annotations.c_str());
PR.m_nsSemantic = SC.AddString(m_Semantic.c_str());
PR.m_nsValues = SC.AddString(m_Values.c_str());
PR.m_eType = m_eType;
PR.m_nArray = m_nArray;
PR.m_nFlags = m_nFlags;
for ( int i=0; i<eHWSC_Num; ++i )
{
PR.m_nRegister[i] = m_Register[i];
}
SC.FXSamplers.push_back(PR);
return bRes;
}
bool SFXSampler::Import(SShaderSerializeContext& SC, SSFXSampler* pPR)
{
bool bRes = true;
m_Name = sString(pPR->m_nsName, SC.Strings);
m_Annotations = sString(pPR->m_nsAnnotations, SC.Strings);
m_Semantic = sString(pPR->m_nsSemantic, SC.Strings);
m_Values = sString(pPR->m_nsValues, SC.Strings);
m_eType = pPR->m_eType;
m_nArray = pPR->m_nArray;
m_nFlags = pPR->m_nFlags;
for ( int i=0; i<eHWSC_Num; ++i )
{
m_Register[i] = pPR->m_nRegister[i];
}
return bRes;
}
bool SFXTexture::Export(SShaderSerializeContext& SC)
{
bool bRes = true;
SSFXTexture PR;
PR.m_nsName = SC.AddString(m_Name.c_str());
PR.m_nsAnnotations = SC.AddString(m_Annotations.c_str());
PR.m_nsSemantic = SC.AddString(m_Semantic.c_str());
PR.m_nsValues = SC.AddString(m_Values.c_str());
PR.m_nsNameTexture = SC.AddString(m_szTexture.c_str());
PR.m_bSRGBLookup = m_bSRGBLookup;
PR.m_eType = m_eType;
PR.m_nArray = m_nArray;
PR.m_nFlags = m_nFlags;
for ( int i=0; i<eHWSC_Num; ++i )
{
PR.m_nRegister[i] = m_Register[i];
}
SC.FXTextures.push_back(PR);
return bRes;
}
bool SFXTexture::Import(SShaderSerializeContext& SC, SSFXTexture* pPR)
{
bool bRes = true;
m_Name = sString(pPR->m_nsName, SC.Strings);
m_Annotations = sString(pPR->m_nsAnnotations, SC.Strings);
m_Semantic = sString(pPR->m_nsSemantic, SC.Strings);
m_Values = sString(pPR->m_nsValues, SC.Strings);
m_szTexture = sString(pPR->m_nsNameTexture, SC.Strings);
m_bSRGBLookup = pPR->m_bSRGBLookup;
m_eType = pPR->m_eType;
m_nArray = pPR->m_nArray;
m_nFlags = pPR->m_nFlags;
for ( int i=0; i<eHWSC_Num; ++i )
{
m_Register[i] = pPR->m_nRegister[i];
}
return bRes;
}
bool CHWShader_D3D::ExportSamplers(SCHWShader& SHW, SShaderSerializeContext& SC)
{
bool bRes = true;
return bRes;
}
bool CHWShader::ImportSamplers(SShaderSerializeContext& SC, SCHWShader* pSHW, byte*& pData, std::vector<STexSamplerRT>& Samplers)
{
bool bRes = true;
return bRes;
}
bool CHWShader_D3D::ExportParams(SCHWShader& SHW, SShaderSerializeContext& SC)
{
bool bRes = true;
SHW.m_nParams = 0;
return bRes;
}
bool CHWShader::ImportParams(SShaderSerializeContext& SC, SCHWShader* pSHW, byte*& pData, std::vector<SFXParam>& Params)
{
bool bRes = true;
return bRes;
}
bool CHWShader_D3D::Export(SShaderSerializeContext& SC)
{
bool bRes = true;
SCHWShader SHW;
char str[256];
azstrcpy(str, AZ_ARRAY_SIZE(str), GetName());
char* c = strchr(str, '(');
if (c)
{
c[0] = 0;
}
SHW.m_nsName = SC.AddString(str);
SHW.m_nsNameSourceFX = SC.AddString(m_NameSourceFX.c_str());
SHW.m_nsEntryFunc = SC.AddString(m_EntryFunc.c_str());
SHW.m_eSHClass = m_eSHClass;
SHW.m_dwShaderType = m_dwShaderType;
SHW.m_nMaskGenFX = m_nMaskGenFX;
SHW.m_nMaskGenShader = m_nMaskGenShader;
SHW.m_nMaskOr_RT = m_nMaskOr_RT;
SHW.m_nMaskAnd_RT = m_nMaskAnd_RT;
SHW.m_Flags = m_Flags;
FXShaderToken* pMap = NULL;
TArray<uint32>* pData = &m_TokenData;
bool bDeleteTokenData = false;
// No longer export any token data, slow and bloated!
const bool bOutputTokens = false;
if (bOutputTokens) // && !pData->size())
{
//always evaluate?
if (1)
{
bDeleteTokenData = true;
mfGetCacheTokenMap(pMap, pData, m_nMaskGenShader /*is this correct? m_nMaskGenFX*/);
}
else
{
assert(0);
}
}
if (bOutputTokens && !pMap)
{
assert(0);
return false;
}
SHW.m_nTokens = bOutputTokens ? pData->size() : 0;
SHW.m_nTableEntries = bOutputTokens ? pMap->size() : 0;
SCHWShader SHWTemp = SHW;
SHW.Export(SC.Data);
uint32 nOffs = 0;
if (bOutputTokens)
{
sAddDataArray_POD(SC.Data, *pData, nOffs);
FXShaderTokenItor itor;
for (itor = pMap->begin(); itor != pMap->end(); itor++)
{
//String pool method
sAddData(SC.Data, itor->Token);
uint32 tokenStrIdx = SC.AddString(itor->SToken.c_str());
sAddData(SC.Data, tokenStrIdx);
}
}
bRes &= ExportSamplers(SHW, SC);
bRes &= ExportParams(SHW, SC);
if (bRes)
{
if (memcmp(&SHW, &SHWTemp, sizeof(SCHWShader)))
{
CryFatalError("Export failed");
}
}
if (bDeleteTokenData)
{
SAFE_DELETE(pData);
SAFE_DELETE(pMap);
}
return bRes;
}
CHWShader* CHWShader::Import(SShaderSerializeContext& SC, int nOffs, uint32 CRC32, CShader* pSH)
{
if (nOffs < 0)
{
return NULL;
}
CHWShader* pHWSH = NULL;
SCHWShader shaderHW;
shaderHW.Import(&SC.Data[nOffs]);
SCHWShader* pSHW = &shaderHW;
const char* szName = sString(pSHW->m_nsName, SC.Strings);
const char* szNameSource = sString(pSHW->m_nsNameSourceFX, SC.Strings);
const char* szNameEntry = sString(pSHW->m_nsEntryFunc, SC.Strings);
TArray<uint32> SHData;
SHData.resize(pSHW->m_nTokens);
FXShaderToken Table, * pTable = NULL;
Table.reserve(pSHW->m_nTableEntries);
nOffs = 0;
//Copy string pool, TODO separate string pool for tokens!
//TArray<char> tokenStringPool = SC.Strings;
// Token data is no longer in export data - See CHWShader_D3D::Export where bOutputTokens == false and disables this path
if (0) //CRenderer::CV_r_shadersAllowCompilation)
{
byte* pData = &SC.Data[nOffs + sizeof(SCHWShader)];
memcpy(&SHData[0], pData, pSHW->m_nTokens * sizeof(uint32));
pData += pSHW->m_nTokens * sizeof(uint32);
pTable = &Table;
for (uint32 i = 0; i < pSHW->m_nTableEntries; i++)
{
// string pool method
DWORD nToken = *(DWORD*)&pData[nOffs];
nOffs += sizeof(DWORD);
uint32 nTokenStrIdx = *(DWORD*)&pData[nOffs];
nOffs += sizeof(uint32);
if (CParserBin::m_bEndians)
{
SwapEndian(nToken, eBigEndian);
SwapEndian(nTokenStrIdx, eBigEndian);
}
STokenD TD;
TD.SToken = sString(nTokenStrIdx, SC.Strings);
TD.Token = nToken;
Table.push_back(TD);
}
pData += nOffs;
std::vector<STexSamplerRT> Samplers;
ImportSamplers(SC, pSHW, pData, Samplers);
std::vector<SFXParam> Params;
ImportParams(SC, pSHW, pData, Params);
}
bool bPrecache = (SC.SSR.m_Flags & EF_PRECACHESHADER) != 0;
//static CHWShader *mfForName(const char *name, const char *nameSource, uint32 CRC32, const char *szEntryFunc, EHWShaderClass eClass, TArray<uint32>& SHData, FXShaderToken *pTable, uint32 dwType, CShader *pFX, uint64 nMaskGen=0, uint64 nMaskGenFX=0);
pHWSH = CHWShader::mfForName(szName, szNameSource, CRC32, szNameEntry, pSHW->m_eSHClass, SHData, pTable, pSHW->m_dwShaderType, pSH, pSHW->m_nMaskGenShader, pSHW->m_nMaskGenFX);
pHWSH->m_eSHClass = shaderHW.m_eSHClass;
pHWSH->m_dwShaderType = shaderHW.m_dwShaderType;
pHWSH->m_nMaskGenFX = shaderHW.m_nMaskGenFX;
pHWSH->m_nMaskGenShader = shaderHW.m_nMaskGenShader;
pHWSH->m_nMaskOr_RT = shaderHW.m_nMaskOr_RT;
pHWSH->m_nMaskAnd_RT = shaderHW.m_nMaskAnd_RT;
pHWSH->m_Flags = shaderHW.m_Flags;
return pHWSH;
}
#else
bool CHWShader_D3D::Export(SShaderSerializeContext& SC)
{
return false;
}
#endif
const char* CHWShader_D3D::mfGetActivatedCombinations(bool bForLevel)
{
TArray <char> Combinations;
char* pPtr = NULL;
uint32 i;
for (i = 0; i < m_Insts.size(); i++)
{
SHWSInstance* pInst = m_Insts[i];
char name[256];
azstrcpy(name, AZ_ARRAY_SIZE(name), GetName());
char* s = strchr(name, '(');
if (s)
{
s[0] = 0;
}
string str;
SShaderCombIdent Ident(m_nMaskGenFX, pInst->m_Ident);
gRenDev->m_cEF.mfInsertNewCombination(Ident, pInst->m_eClass, name, 0, &str, false);
assert (str.size());
if (str.size())
{
assert(str[0] == '<' && str[2] == '>');
string s1;
if (str[0] == '<' && str[2] == '>')
{
s1.Format("<%d>%s", pInst->m_nUsed, &str[3]);
}
else
{
s1 = str;
}
Combinations.Copy(s1.c_str(), s1.size());
Combinations.Copy("\n", 1);
}
}
if (!Combinations.Num())
{
return NULL;
}
pPtr = new char [Combinations.Num() + 1];
memcpy(pPtr, &Combinations[0], Combinations.Num());
pPtr[Combinations.Num()] = 0;
return pPtr;
}
const char* CHWShader::GetCurrentShaderCombinations(bool bForLevel)
{
TArray <char> Combinations;
char* pPtr = NULL;
CCryNameTSCRC Name;
SResourceContainer* pRL;
Name = CHWShader::mfGetClassName(eHWSC_Vertex);
pRL = CBaseResource::GetResourcesForClass(Name);
int nVS = 0;
int nPS = 0;
if (pRL)
{
ResourcesMapItor itor;
for (itor = pRL->m_RMap.begin(); itor != pRL->m_RMap.end(); itor++)
{
CHWShader* vs = (CHWShader*)itor->second;
if (!vs)
{
continue;
}
const char* szCombs = vs->mfGetActivatedCombinations(bForLevel);
if (!szCombs)
{
continue;
}
Combinations.Copy(szCombs, strlen(szCombs));
delete [] szCombs;
nVS++;
}
}
Name = CHWShader::mfGetClassName(eHWSC_Pixel);
pRL = CBaseResource::GetResourcesForClass(Name);
int n = 0;
if (pRL)
{
ResourcesMapItor itor;
for (itor = pRL->m_RMap.begin(); itor != pRL->m_RMap.end(); itor++)
{
CHWShader* ps = (CHWShader*)itor->second;
if (!ps)
{
continue;
}
const char* szCombs = ps->mfGetActivatedCombinations(bForLevel);
if (!szCombs)
{
continue;
}
Combinations.Copy(szCombs, strlen(szCombs));
delete [] szCombs;
nPS++;
}
}
if (!Combinations.Num())
{
return NULL;
}
pPtr = new char [Combinations.Num() + 1];
memcpy(pPtr, &Combinations[0], Combinations.Num());
pPtr[Combinations.Num()] = 0;
return pPtr;
}
bool CHWShader::PreactivateShaders()
{
bool bRes = true;
if (CRenderer::CV_r_shaderspreactivate)
{
gRenDev->m_pRT->RC_PreactivateShaders();
}
return bRes;
}
void CHWShader::RT_PreactivateShaders()
{
gRenDev->m_cEF.mfPreloadBinaryShaders();
}