laie

[proposal] Adding RELATIVE_IMM operand type may improve architecture in general

2

~After analyzing the ZydisEncoder's dual encoding modes (`ZydisEncoderEncodeInstructionAbsolute` and `ZydisEncoderEncodeInstruction`), I've identified an opportunity to improve the structural design by explicitly separating relative immediate values from regular immediate operands.~ ~Currently,...

decoded instruction size is bigger than essential

8

[1] The fully decoded instruction byte count is reported 1128 bytes on my build environment [2] This big decoded instruction size makes infeasible to fully decompile some real world EXE...

cannot encode mov EAX, 80000000h when imm is unsigned

11

``` namespace report_case { std::pair testcase_assemble_abs(size_t runtime_address, unsigned char* buffer, size_t length, ZydisMnemonic mnem, const std::vector& operands) { if (5 < operands.size()) throw std::runtime_error("Too many operands"); unsigned char ret_length =...

[Proposal] Features to get maximum possible encode length of given mnemonic(or +given operands)

9

Currently I am preallocating 15 bytes(absolute maximum size) per an insn, for a recompilation. However, most of the instructions are not that large. And the preallocation gets too big compared...