8c4be30·Author: Alexandro Sanchez·Closed by: dennisaa·2017-10-02
Description
This patch implements the switch detection passes for ARM/AArch64/MIPS and fixes a couple of issues with the ARM/MIPS disassembler (e.g. mislabeled instructions, delay slot issues).
Pending:
Support for little-endian MIPS/PPC: Shouldn't be to hard, just detecting the endianness from the ELF's identification header and switch between read_be and read_le accordingly.
Support for MIPS64* (PIC): The instruction sequences to load the jump table base address involve some memory accesses.
Fixing issues with the MIPS delay slot: Although I modifies the disassembler to support it, manual testing reveals some issues (padding nop's being detected as reachable for some weird reason). I need to debug it.
PS: I've extended the scripts and configuration files from the virtual machine from the “An In-Depth Analysis of Disassembly on Full-Scale x86/x64 Binaries” paper to compile the SPEC CPU2006 benchmarks for: arm, aarch64, mips, mipsel, mips64, mips64el, powerpc, powerpc64, powerpc64le, and computing the F-scores as done in the Nucleus paper. I'll provide the detailed results ASAP.
This patch implements the switch detection passes for ARM/AArch64/MIPS and fixes a couple of issues with the ARM/MIPS disassembler (e.g. mislabeled instructions, delay slot issues).
Pending:
Support for little-endian MIPS/PPC: Shouldn't be to hard, just detecting the endianness from the ELF's identification header and switch between
read_beandread_leaccordingly.Support for MIPS64* (PIC): The instruction sequences to load the jump table base address involve some memory accesses.
Fixing issues with the MIPS delay slot: Although I modifies the disassembler to support it, manual testing reveals some issues (padding
nop's being detected as reachable for some weird reason). I need to debug it.PS: I've extended the scripts and configuration files from the virtual machine from the “An In-Depth Analysis of Disassembly on Full-Scale x86/x64 Binaries” paper to compile the SPEC CPU2006 benchmarks for: arm, aarch64, mips, mipsel, mips64, mips64el, powerpc, powerpc64, powerpc64le, and computing the F-scores as done in the Nucleus paper. I'll provide the detailed results ASAP.