BridgeApi - RemixApi wrapper for x86 games

[xoxor4d]

This PR adds a intermediary api called BridgeApi to the bridge that allows x86 games to use parts of the x64 remix-runtime api.

Overview:

• the client implements BridgeApi and sends the data of received api calls to the server client/bridge_api.cpp
• the server assembles the necessary RemixApi structs and calls the appropriate RemixApi functions server/main.cpp

Behavior on init:

• server/module_processing.cpp grab the device pointer used to initialize the RemixApi in CreateDevice / CreateDeviceEx
• server/main.cpp initialize the RemixApi in InitializeD3D if client option getEnableBridgeApi() is true by calling remixapi_lib_loadRemixDllAndInitialize with the optained device pointer

Usage:

• Using BridgeApi requires the user to be internal (dll/asi injection or by other means - requires game process access)
• Client option client.exposeRemixApi true
• Include bridge_c.h and a predefined macro:

#define BRIDGE_API_IMPORT
#include "bridge_c.h"

• Initialize the api:

// make sure that the bridge had enough time to create a device before you call the following
const auto status = bridgeapi_initialize(&bridge);
if (status == BRIDGEAPI_ERROR_CODE_SUCCESS)
{
  if (bridge.initialized)
  {
    bridge.RegisterDevice();
  }
}

• If no game render logic is hooked (triggered once per frame) or if the dll/asi has to be game agnostic, a callback can be registered that triggers when device->EndScene is called:

void on_endscene()
{
  bridge.DrawLightInstance(my_light_handle);
}

// somewhere after the api was initialized
bridge.RegisterEndSceneCallback(on_endscene);

Notes:

• I've set client option client.exposeRemixApi to true by default (now false by default)

• I'm using a mutex to get the device here. (I don't have much experience working with threads -> not sure if correct)

• I've directly included remix_c.h from dxvk-remix (should be made a dependency that's pulled?)

• There is no RemixApi version check nor does bridge_c.h expose any version

• Sending const char* or wchar_t* via rpc can result in the string no longer having the null terminator in the correct location. This can lead to strings with "junk data" attached to them. The returned length/size is correct tho. This also happens for Bridge_DebugMessage. This leads to special handling in every api case in server/main.cpp that`s using strings:

  • Api_DebugPrint using std::string(data, length) to construct a new string with the proper length
  • Api_CreateOpaqueMaterial using NVPULL_PATH -> std::wstring(data, length) to construct a new wstring with the proper length

---

Example usage of almost every feature:

• iw3xo-dev/blob/feature/remix-api/src/components/modules/rtx/rtx_gui.cpp (search for Developer)
• iw3xo-dev/blob/feature/remix-api/src/components/modules/rtx/rtx_api.cpp

[nv-nfreybler]

Overall looks great, thanks for doing this! There is certainly a promising future for exposing RemixApi like this.

One overall comment: Please rename everything from BridgeApi to RemixApi and similar. We see this more so as exposing the RemixApi directly to the games/mods, as opposed to wrapping it through a layer, and we don't want to conflate the RemixApi with the bridge's actual API, which is basically just the d3d9 API.

[xoxor4d]

@nv-nfreybler Firstly, thanks for taking the time to review this! I've made the required changes and squashed everything into one commit. I've remove almost all traces of bridgeApi as requested with the exception to the bridge_c header file as I'm not sure if the contents will stay in a separate file or if they'll be included within the remix_c header?

[nv-nfreybler]

Thanks for making all these changes! A couple things to note:

1. Part of the reason for my delay in getting back to you is because we've been implementing a versioning handshake.
2. When we're ready, the next step of the process will involve taking in your changes so we can do some internal testing, and make further additions. Please be patient while we deliberate on this PR, in addition to our internal set of tasks.

I've remove almost all traces of bridgeApi as requested with the exception to the bridge_c header file as I'm not sure if the contents will stay in a separate file or if they'll be included within the remix_c header?

Thanks for doing that, I know it can be a long and tedious task to rename things. I appreciate your calling it out in particular, as we are actively discussing the bridge_c header file in particular. A decision will likely be reached in our own internal merge commit.

[nv-nfreybler]

Just wanted to post an update here that we're currently actively internally reviewing and working on this PR, for the better part of the last couple weeks, to be honest. We appreciate your patience, and want to assure you that it is not forgotten.

[xoxor4d]

Thank you for the update, really appreciate it! I hope this hasn't caused too much trouble internally 🙂

[nv-nfreybler]

Please see commit 2dd0a0e.
Binaries

We overhauled quite a bit from the original PR, mostly to eliminate the code duplication between the original bridge_c.h and remix_c.h. We also wanted 1:1 parity with the standard Remix API, so we went with a singular CreateLight etc as opposed to extension-specific APIs. Additionally we found some memory leaks, and needed to refactor accordingly.

Thank you again for the PR. If you consider this sufficient, please go ahead and close this, or else we can.