How to dynamically allocate and access an array using LLVM API?

Question

I'd like to generate LLVM IR for a C++ code like this:

T *arr = (T*) malloc(sizeof(T) * 100);
arr[0] = somevalue;

Here's my code:

llvm::Type* element_type = /... some T type .../;
llvm::Type* int32type = llvm::Type::getInt32Ty(context);

auto element_size = llvm::ConstantInt::get(int32type, data_layout.getTypeAllocSize(element_type));
auto array_size = llvm::ConstantInt::get(int32type, 100);

// malloc:
auto malloc_inst = llvm::CallInst::CreateMalloc(
    ir_builder.GetInsertBlock(), 
    element_type->getPointerTo(), // T*
    element_type,                 // T
    element_size,                 // sizeof(T)
    array_size,                   // 100
    nullptr, 
    "")
ir_builder.Insert(malloc_inst);

// cast to T*:
auto arr = ir_builder.CreatePointerCast(malloc_inst, element_type->getPointerTo());

// arr[0] = somevalue:
llvm::Value *value = /... some value of type T .../
auto element_ptr = ir_builder.CreateInBoundsGEP(
    arr, { 
        llvm::ConstantInt::get(llvm::Type::getInt32Ty(context), 0),
        llvm::ConstantInt::get(llvm::Type::getInt32Ty(context), 0)
    });
ir_builder.CreateStore(value, element_ptr)

However, when running generated IR I'm getting:

Assertion failed: (Ty && "Invalid GetElementPtrInst indices for type!"), function checkGEPType, file /usr/local/opt/llvm/include/llvm/IR/Instructions.h, line 847

I tried different combinations of types passed to the GEP instruction, but it's obvious I'm missing something basic. Can someone point me to a working example for this?

Michael Spector · Accepted Answer · 2019-06-11 07:18:14Z

1

There were two issues:

First, to access element array there's no need to pass two indices into GEP instruction.
Second, 5th argument of CreateMalloc is casted to a pointer for some reason here. After doing the multiplication of element_size * array_size by myself everything worked.

Also, explicit pointer cast (from i8* to T*) was not needed as it happens automatically.

Here's the final working code:

llvm::Type* element_type = /... some T type .../;
llvm::Type* int32type = llvm::Type::getInt32Ty(context);

auto element_size = llvm::ConstantInt::get(int32type, data_layout.getTypeAllocSize(element_type));
auto array_size = llvm::ConstantInt::get(int32type, 100);
auto alloc_size = llvm::ConstantExpr::getMul(element_size, array_size);

// malloc:
auto arr = llvm::CallInst::CreateMalloc(
    ir_builder.GetInsertBlock(), 
    element_type->getPointerTo(), // T*
    element_type,                 // T
    alloc_size,                   // sizeof(T) * 100
    nullptr,
    nullptr, 
    "")
ir_builder.Insert(arr);

// arr[0] = somevalue:
llvm::Value *value = /... some value of type T .../
auto element_ptr = ir_builder.CreateInBoundsGEP(
    arr, llvm::ConstantInt::get(llvm::Type::getInt32Ty(context), 0));
ir_builder.CreateStore(value, element_ptr)

answered Jun 11, 2019 at 7:18

Michael Spector

37k7 gold badges67 silver badges90 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Wave and Matter Over a year ago

I tried to use the information from this answer for my own array allocation/access code. I thought about posting my own question but will try it here first: I get two issues: - verifyFunction complains about "Call parameter type does not match function signature! i32 48" at the malloc call. - verifyFunction complains about "Stored value type does not match pointer operand type! store double 5.000000e+00, double** %acc_tmp, align 8", which is correct (one "*" to much), but i don't know how to fix it.

Collectives™ on Stack Overflow

How to dynamically allocate and access an array using LLVM API?

1 Answer 1

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related