Executable Format

The executable format used by ParaVM is called PVC (Parallella Virtual Code). It's a simple binary representation of the PVA (Parallella Virtual Assembly) assembly language.

The format encodes all integers as little endian. Strings are stored as a 32-bit unsigned length followed by the string's characters (without the null terminator).

Header Fields

The file header can be expressed as a pseudo-C structure like this:

struct PVCHeader
{
    uint32_t fourcc;
    uint32_t version;
};

The fourcc field must equal 0x43565000 ("PVC\0"). It identifies the file format.

The version field identifies the version of ParaVM the module was written by/for. A virtual machine or tool is allowed to reject the module if the version is not supported.

Main Structure

The main structure of the format can be expressed as a pseudo-C structure like this:

struct PVCModule
{
    uint32_t function_count;
    struct
    {
        char *function_name;
        uint32_t register_count;
        struct
        {
            char *register_name;
            uint8_t is_argument;
        } registers[register_count];
        uint32_t block_count;
        struct
        {
            char *block_name;
        } block_names[block_count];
        struct
        {
            char *block_name;
            uint8_t has_handler_block;
            if (has_handler_block)
            {
                char *handler_block_name;
            }
            uint8_t has_exception_register;
            if (has_exception_register)
            {
                char *exception_register_name;
            }
            uint32_t instruction_count;
            struct
            {
                uint8_t code;
                uint32_t instruction_register_count;
                struct
                {
                    char *instruction_register_name;
                } instruction_registers[instruction_register_count];
                union
                {
                    char *string;
                    char *block;
                    char *blocks[2];
                } operand;
            } instructions[instruction_count];
        } blocks[block_count];
    } functions[function_count];
};

The purpose of the block name array is so that the program reading the file can create the blocks before reading instructions that refer to them.