Rome 2 CPU Performance Fix

Source code for the Total War: ROME 2 CPU performance mod: a dinput8.dll that the game loads at startup and that patches the engine in memory. No game files are changed.

Download the ready-made DLL:

• Steam Workshop: https://steamcommunity.com/sharedfiles/filedetails/?id=3826423359 (after subscribing, run install.bat in the mod's folder)
• or build it yourself (below), or take it from the latest run under Actions

Created with the help of AI (Claude by Anthropic). Tested in game by JGdaCaesar.

The mod does not appear in the game launcher's mod list: it is a DLL in the game folder, not a .pack mod. It is active whenever dinput8.dll is in the game folder; rome2fix.log shows what it did.

Version 1.1 adds faster loading (a Divide et Impera save: 38 s with 1.0, 20 s with 1.1 on the test PC; see the player README for all measurements).

What each file does

File	Fix
proxy.c	Entry point (DllMain, or a DLL load notification when something loaded dinput8.dll before the game DLL), dinput8 forwarding, rome2fix.log, off switches
tilescan.c	The campaign/battle tile-feedback scan: every frame the main thread counts a screen-sized GPU readback of texture-tile requests pixel by pixel. Runs of identical entries are now counted in one step, with the same counters and the same request list in the same order. Its entry hook is also the per-frame callback for the x87 safety check
x87fix.c	About 1,200 x87 float-to-integer conversions that switch the FPU rounding mode (fstcw/fldcw/fistp) replaced by fisttp / SSE4.1 roundss (from the Attila mod, unchanged: it finds the blocks by pattern)
sortfix.c	Two render-list sorts (by +0x6C/+0x68 and +0x44/+0x40): key extraction plus radix sort, same order as the game's sorts (from the Attila mod, with Rome 2's addresses and keys)
attrcache.c	Loading: a cache of missing folders behind the game's GetFileAttributesW import. Before using a packed file the game checks every search root (Data, each Workshop mod, language folders) for a loose copy, millions of failed checks per load with big mod lists. A failed check records the first missing name below the deepest existing folder; later checks under it get Windows' own answer (ERROR_FILE_NOT_FOUND / ERROR_PATH_NOT_FOUND) without asking. That folder and every folder above it are watched for created/renamed/deleted names and polled on every call; any change clears the cache. Unusual path forms (/, ., .., device names, ...) go straight to Windows
fileio.c	Loading: .pack files opened without FILE_FLAG_NO_BUFFERING (so Windows caches them), and small reads of read-only files in the game's user folder (the unpacked save) served from a 64 KB buffer with the exact file position kept; writes to the same file drop the buffer
strfix.c	Loading: the game's string equality, compare and assign helpers and a pair-table search, rewritten with 16-byte compares and memcpy (same results, byte-for-byte checked sites)
diag.c	Diagnostics in rome2fix.log: mod version, game build, system, settings, other hook DLLs, and crash records (a vectored exception handler that only logs and never handles the exception)
update.c	Self-update from the Steam Workshop copy (item 3826423359): a newer version there replaces the DLL in the game folder for the next start; never downgrades, no network
version.h	Mod version
dinput8.def	The exports a dinput8.dll must have

Every patch first compares the game's bytes at its address with what it expects (with relocated addresses taken into account), and skips itself (and says so in rome2fix.log) if they differ. So on another game version it does nothing.

The developer build adds measuring tools (timing probes, A/B switches); they are not part of this source or of the released DLL.

Building

You need llvm-mingw (the release used: 20261006, UCRT, x86_64 host), with its bin folder on PATH. Then run:

build.bat

The DLL lands in build\dinput8.dll. The build is reproducible: with the same llvm-mingw version you should get exactly the same file as the released one. Compare the SHA-256 (certutil -hashfile build\dinput8.dll SHA256) with the one given for each release. Always build it under the name dinput8.dll: the name is stored inside the file.

Automated builds

Every commit is also built on GitHub's own machines by .github/workflows/build.yml, with the compiler downloaded from its official release (checksum pinned). The run then compares the SHA-256 of that DLL with dinput8.dll.sha256, the hash of the DLL released on the Steam Workshop, and fails if they differ. Open the latest run under Actions: its summary shows both hashes, and the DLL itself can be downloaded there. Builds also carry a signed build attestation (gh attestation verify dinput8.dll -R JSa25/Rome2CPUPerfFix).

The Nexus Mods file is uploaded by .github/workflows/nexus.yml (started by hand): it builds the DLL the same way, stops unless the hash matches the released one, zips it with Rome2CPUPerfFix_README.txt and adds it to the Nexus file as a new version.

Tests

The tests run the game's own code next to the replacements and check the results are identical: the tile scan on 3,000 random buffers (11.6 million entries), every x87 block on 2,740 edge-case and random inputs (3.3 million runs), both sorts on 110 lists of up to 30,000 items, the string helpers on 400,000 random cases. fileio_test compares 300,000 random reads and seeks with plain Windows calls, and attrcache_test compares 450,000 random file checks with plain Windows (attributes and error codes) while files and folders are created, renamed and deleted between checks; with the folder watches switched off the same test finds thousands of differences, so it would catch a stale cache. diag_test checks the crash records on deliberate faults, selfupd_test the update rules on throwaway files, and lateload_test loads the built DLL before the game DLL (as an overlay can) and checks every fix still installs from Windows' load notification. They read your copy of the game's empire.retail.dll (it is not included here):

build.bat test "C:\...\steamapps\common\Total War Rome II\empire.retail.dll"

License

See LICENSE.