Skip to content

Documentation versions

Researching the engine ​

We strongly recommend researching with AI, preferably on strong models such as Sonnet 5.5, GPT-6 Sol or newer. There are no hard requirements; this is a recommendation.

AI research is welcome here. It is valued above manual work because the whole process stays visible: notes, evidence and doubts. So please keep the research history, even when the result looks finished.

The library ​

Download libasphalt8.so. This is Asphalt 8 version 1.6.0 from a Lenovo A7000, ARMv7.

text
SHA-256: ea1026652474b1068ec9c03f6a403b410bb0021e23bbb1b28fb3f26e31335374
Size: 28680208 bytes

Check the hash: addresses from another game version can lead to different code even when the names match.

Ghidra ​

  1. Install Ghidra using its official guide.
  2. Create a project and import the file as ELF. Check that ARM, little-endian and 32 bits are selected.
  3. Run the automatic analysis.
  4. Read code in the Decompiler window and call sites in References.

Engine Atlas uses a Ghidra image base of 0x10000, so an offset is the Ghidra address minus 0x10000. It is not an address in the running game: there the library's load address is added to the offset.

What counts as a result ​

Decompilation suggests, it does not prove. Note what you saw in the code, what you concluded yourself and what you tested in the game. If a type is unknown, say so rather than filling in a similar one.

Connecting AI ​

See Engine Atlas MCP: setup, working with a model and how to tell a good result from a bad one.