int cShaderObject::LoadShaderImpl(eShaderType type, const char* shader_source)
{
const char* shader_name = mObjectName.c_str();
const char* shader_stage;
uint32_t shader_id;
const char* type_define;
switch (type)
{
case VERTEX_SHADER:
shader_id = mVertexShader;
type_define = "#define VERTEX_SHADER 1\n";
shader_stage = "vertex";
break;
case GEOMETRY_SHADER:
shader_stage = "geometry";
shader_id = mGeometryShader;
type_define = "#define GEOMETRY_SHADER 1\n";
break;
case FRAGMENT_SHADER:
shader_stage = "fragment";
shader_id = mFragmentShader;
type_define = "#define FRAGMENT_SHADER 1\n";
break;
default:
FatalError::Die("unknown shader type");
break;
}
cArray<16, const char*> string_pointers;
string_pointers.push_back(mVersionString.c_str());
string_pointers.push_back(type_define);
string_pointers.push_back("#line 2\n");
string_pointers.push_back(shader_source);
glShaderSource(shader_id, string_pointers.size(), &string_pointers[0], NULL);
if(glCheckAllErrors(__FILE__, __LINE__))
{
Log::Error("shader source error for shader '%s' (ID 0x%x).",
shader_name, shader_id );
return 1;
}
if(GL_FALSE == glIsShader(shader_id))
{
Log::Error("somehow this is not a shader object (id = %x).", shader_id );
return 1;
}
glCompileShader(shader_id);
if(glCheckAllErrors(__FILE__, __LINE__))
{
AKJ_THROW("Errors compiling shader "+Twine(shader_id)+
": \""+shader_name+"\".");
}
const char* compile_status_string = "Failed";
GLint compiled_ok = GL_FALSE;
glGetShaderiv(shader_id, GL_COMPILE_STATUS, &compiled_ok);
if(compiled_ok == GL_TRUE)
{
compile_status_string = "Succeeded";
}
Log::Info("%s: Compilation of shader \"%s\" (%s).",
compile_status_string, shader_name, shader_stage);
GLint log_length = 0;
glGetShaderiv(shader_id, GL_INFO_LOG_LENGTH, &log_length);
std::string info_log;
if(log_length > 1 && compiled_ok != GL_TRUE)
{
GLsizei actual_length = 0;
info_log.resize(log_length, ' ');
glGetShaderInfoLog(shader_id, log_length, &actual_length, &info_log[0]);
info_log.resize(actual_length);
AKJ_THROW(Twine("Shader ") + shader_name
+ " Compile failed: \n" + info_log.c_str());
}
if(glCheckAllErrors(__FILE__, __LINE__) || GL_FALSE == compiled_ok)
{
AKJ_THROW("Errors compiling shader "+Twine(shader_id)+
": \""+shader_name+"\".");
}
if(!IsShaderAttached(shader_id))
{
glAttachShader(mShaderProgram, shader_id);
}
if(glCheckAllErrors(__FILE__, __LINE__))
{
Log::Warn("there were errors attaching shader to program for shader %x"
", name: %s.", shader_id, shader_name );
return 1;
}
//everything went better than expected :)
return GL_NO_ERROR;
}
bool IntrinsicCleanerPass::runOnBasicBlock(BasicBlock &b, Module &M) {
bool dirty = false;
bool block_split=false;
#if LLVM_VERSION_CODE <= LLVM_VERSION(3, 1)
unsigned WordSize = TargetData.getPointerSizeInBits() / 8;
#else
unsigned WordSize = DataLayout.getPointerSizeInBits() / 8;
#endif
for (BasicBlock::iterator i = b.begin(), ie = b.end();
(i != ie) && (block_split == false);) {
IntrinsicInst *ii = dyn_cast<IntrinsicInst>(&*i);
// increment now since LowerIntrinsic deletion makes iterator invalid.
++i;
if(ii) {
switch (ii->getIntrinsicID()) {
case Intrinsic::vastart:
case Intrinsic::vaend:
break;
// Lower vacopy so that object resolution etc is handled by
// normal instructions.
//
// FIXME: This is much more target dependent than just the word size,
// however this works for x86-32 and x86-64.
case Intrinsic::vacopy: { // (dst, src) -> *((i8**) dst) = *((i8**) src)
Value *dst = ii->getArgOperand(0);
Value *src = ii->getArgOperand(1);
if (WordSize == 4) {
Type *i8pp = PointerType::getUnqual(PointerType::getUnqual(Type::getInt8Ty(getGlobalContext())));
Value *castedDst = CastInst::CreatePointerCast(dst, i8pp, "vacopy.cast.dst", ii);
Value *castedSrc = CastInst::CreatePointerCast(src, i8pp, "vacopy.cast.src", ii);
Value *load = new LoadInst(castedSrc, "vacopy.read", ii);
new StoreInst(load, castedDst, false, ii);
} else {
assert(WordSize == 8 && "Invalid word size!");
Type *i64p = PointerType::getUnqual(Type::getInt64Ty(getGlobalContext()));
Value *pDst = CastInst::CreatePointerCast(dst, i64p, "vacopy.cast.dst", ii);
Value *pSrc = CastInst::CreatePointerCast(src, i64p, "vacopy.cast.src", ii);
Value *val = new LoadInst(pSrc, std::string(), ii); new StoreInst(val, pDst, ii);
Value *off = ConstantInt::get(Type::getInt64Ty(getGlobalContext()), 1);
pDst = GetElementPtrInst::Create(pDst, off, std::string(), ii);
pSrc = GetElementPtrInst::Create(pSrc, off, std::string(), ii);
val = new LoadInst(pSrc, std::string(), ii); new StoreInst(val, pDst, ii);
pDst = GetElementPtrInst::Create(pDst, off, std::string(), ii);
pSrc = GetElementPtrInst::Create(pSrc, off, std::string(), ii);
val = new LoadInst(pSrc, std::string(), ii); new StoreInst(val, pDst, ii);
}
ii->removeFromParent();
delete ii;
break;
}
case Intrinsic::sadd_with_overflow:
case Intrinsic::ssub_with_overflow:
case Intrinsic::smul_with_overflow:
case Intrinsic::uadd_with_overflow:
case Intrinsic::usub_with_overflow:
case Intrinsic::umul_with_overflow: {
IRBuilder<> builder(ii->getParent(), ii);
Value *op1 = ii->getArgOperand(0);
Value *op2 = ii->getArgOperand(1);
Value *result = 0;
Value *result_ext = 0;
Value *overflow = 0;
unsigned int bw = op1->getType()->getPrimitiveSizeInBits();
unsigned int bw2 = op1->getType()->getPrimitiveSizeInBits()*2;
if ((ii->getIntrinsicID() == Intrinsic::uadd_with_overflow) ||
(ii->getIntrinsicID() == Intrinsic::usub_with_overflow) ||
(ii->getIntrinsicID() == Intrinsic::umul_with_overflow)) {
Value *op1ext =
builder.CreateZExt(op1, IntegerType::get(M.getContext(), bw2));
Value *op2ext =
builder.CreateZExt(op2, IntegerType::get(M.getContext(), bw2));
Value *int_max_s =
ConstantInt::get(op1->getType(), APInt::getMaxValue(bw));
Value *int_max =
builder.CreateZExt(int_max_s, IntegerType::get(M.getContext(), bw2));
if (ii->getIntrinsicID() == Intrinsic::uadd_with_overflow){
result_ext = builder.CreateAdd(op1ext, op2ext);
} else if (ii->getIntrinsicID() == Intrinsic::usub_with_overflow){
result_ext = builder.CreateSub(op1ext, op2ext);
} else if (ii->getIntrinsicID() == Intrinsic::umul_with_overflow){
result_ext = builder.CreateMul(op1ext, op2ext);
}
overflow = builder.CreateICmpUGT(result_ext, int_max);
} else if ((ii->getIntrinsicID() == Intrinsic::sadd_with_overflow) ||
(ii->getIntrinsicID() == Intrinsic::ssub_with_overflow) ||
(ii->getIntrinsicID() == Intrinsic::smul_with_overflow)) {
Value *op1ext =
builder.CreateSExt(op1, IntegerType::get(M.getContext(), bw2));
Value *op2ext =
builder.CreateSExt(op2, IntegerType::get(M.getContext(), bw2));
Value *int_max_s =
ConstantInt::get(op1->getType(), APInt::getSignedMaxValue(bw));
Value *int_min_s =
ConstantInt::get(op1->getType(), APInt::getSignedMinValue(bw));
Value *int_max =
builder.CreateSExt(int_max_s, IntegerType::get(M.getContext(), bw2));
Value *int_min =
builder.CreateSExt(int_min_s, IntegerType::get(M.getContext(), bw2));
if (ii->getIntrinsicID() == Intrinsic::sadd_with_overflow){
result_ext = builder.CreateAdd(op1ext, op2ext);
} else if (ii->getIntrinsicID() == Intrinsic::ssub_with_overflow){
result_ext = builder.CreateSub(op1ext, op2ext);
} else if (ii->getIntrinsicID() == Intrinsic::smul_with_overflow){
result_ext = builder.CreateMul(op1ext, op2ext);
}
overflow = builder.CreateOr(builder.CreateICmpSGT(result_ext, int_max),
builder.CreateICmpSLT(result_ext, int_min));
}
// This trunc cound be replaced by a more general trunc replacement
// that allows to detect also undefined behavior in assignments or
// overflow in operation with integers whose dimension is smaller than
// int's dimension, e.g.
// uint8_t = uint8_t + uint8_t;
// if one desires the wrapping should write
// uint8_t = (uint8_t + uint8_t) & 0xFF;
// before this, must check if it has side effects on other operations
result = builder.CreateTrunc(result_ext, op1->getType());
Value *resultStruct =
builder.CreateInsertValue(UndefValue::get(ii->getType()), result, 0);
resultStruct = builder.CreateInsertValue(resultStruct, overflow, 1);
ii->replaceAllUsesWith(resultStruct);
ii->removeFromParent();
delete ii;
dirty = true;
break;
}
case Intrinsic::dbg_value:
case Intrinsic::dbg_declare:
// Remove these regardless of lower intrinsics flag. This can
// be removed once IntrinsicLowering is fixed to not have bad
// caches.
ii->eraseFromParent();
dirty = true;
break;
case Intrinsic::trap: {
// Intrisic instruction "llvm.trap" found. Directly lower it to
// a call of the abort() function.
Function *F = cast<Function>(
M.getOrInsertFunction(
"abort", Type::getVoidTy(getGlobalContext()), NULL));
F->setDoesNotReturn();
F->setDoesNotThrow();
CallInst::Create(F, Twine(), ii);
new UnreachableInst(getGlobalContext(), ii);
ii->eraseFromParent();
dirty = true;
break;
}
case Intrinsic::objectsize: {
// We don't know the size of an object in general so we replace
// with 0 or -1 depending on the second argument to the intrinsic.
assert(ii->getNumArgOperands() == 2 && "wrong number of arguments");
Value *minArg = ii->getArgOperand(1);
assert(minArg && "Failed to get second argument");
ConstantInt *minArgAsInt = dyn_cast<ConstantInt>(minArg);
assert(minArgAsInt && "Second arg is not a ConstantInt");
assert(minArgAsInt->getBitWidth() == 1 && "Second argument is not an i1");
Value *replacement = NULL;
LLVM_TYPE_Q IntegerType *intType = dyn_cast<IntegerType>(ii->getType());
assert(intType && "intrinsic does not have integer return type");
if (minArgAsInt->isZero()) {
// min=false
replacement = ConstantInt::get(intType, -1, /*isSigned=*/true);
} else {
// min=true
replacement = ConstantInt::get(intType, 0, /*isSigned=*/false);
}
ii->replaceAllUsesWith(replacement);
ii->eraseFromParent();
dirty = true;
break;
}
default:
if (LowerIntrinsics)
IL->LowerIntrinsicCall(ii);
dirty = true;
break;
}
}
}
return dirty;
}