The renderer process now takes each view's chunks on a thread of its own as
they come, and draws the newest frame on its main thread: the client no longer
waits for the previous frame to be drawn (on a VM, 22 of a frame's 25 ms).
The adapter records the client's steps (SceneViewer::step, by replacing its
entry in the exported IFrameListener vtable, after checking it), and the frame
pacer doesn't ask for a frame that wouldn't end before the next step is due: a
late step made the client catch up two at once, which showed as hitches. While
the renderer process covers the view, the client's renderer doesn't begin or
present frames, which could wait for the display.
The engine's stats now include the time in the client renderer's methods.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The adapter starts one renderer process per client, with the first engine
(again if it has exited), and attaches each engine's channel to it through a
control section (protocol version 5: Control, the -attach event). The renderer
process keeps a window, swap chain and scene per channel on one Direct3D
device, takes the views' frames in turns, and closes a view when its engine
does.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
While the renderer process runs, the adapter makes the client's render targets
itself and the renderer process draws frames into them: texture targets are
drawn with like any texture, and a pixmap target's frame is read back through
the channel into its Pixmap (protocol version 4: BeginFrame names its target,
CreateRenderTarget, Channel::readSize and the -read event). Frames into render
targets are drawn even while the window is hidden.
The client uses pixmap targets for snapshots (outfit pictures, photos), checked
to look right. The render target interfaces are declared from IMVU's vtables.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The full view is now the default. While the renderer process covers the whole
view (Channel::covering, set after each frame it presents), the adapter skips
the client's renderer's clears and draws, decided per frame; the client's
renderer still gets state and resources, and draws again as soon as the
renderer process stops covering the view.
The client's window no longer gets WS_CLIPCHILDREN: the client draws when its
window gets WM_PAINT, which a window entirely covered by clipped children never
gets (it fell to about one frame a second). The flip model swap chain shows
above the client's drawing anyway.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
IMVU's projections give depth from -1 to 1; its Direct3D renderer maps it to 0 to 1, and now so does ours. Without that, the nearer half of the depth range was clipped, the whole 2D overlay (name tags, the frame rate, the room's title) with it. The adapter also draws drawTrianglesFromMemory without indices, and sums up what the client draws in its log.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Particles: Direct3D 9's size limit evidently isn't the documented 64 pixels; unclamped (up to 2048) they match the client's renderer. A compare view shows the left half's scene over the right half, both renderers side by side; %TEMP%\t5de-renderer-view.txt switches views and %TEMP%\t5de-renderer-shaders\fixed_function.hlsl replaces the shaders while the client runs. The renderer process runs from a copy, so the program can be replaced meanwhile. The adapter recognizes only the textures and buffers it made (others reach the client renderer unchanged) and logs the others and render targets.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Fogging by depth buffer value, as Direct3D 9's documentation suggests for IMVU's projections, looks nothing like the client; fogging by view depth matched it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Particles (point sprites) are drawn as instanced camera-facing squares sized like Direct3D 9's. Fog is by depth buffer value, as Direct3D 9 does with IMVU's projections. At most one frame is queued for display, frames carry the adapter's timestamp so the renderer process logs the delay to presenting, and the log's lines from several processes no longer mix.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The adapter records everything the client draws (state, transforms, lights, fog, textures and buffers) as a command stream, sent in chunks through the shared memory channel, while IMVU's Direct3D renderer still draws the whole view. The renderer process draws it with Diligent, with shaders reproducing IMVU's Direct3D 9 fixed-function setup, over the right half of the 3D view to compare. It skips to the newest frame when it falls behind, and logs its errors and frame times.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The renderer becomes two programs: the adapter (t5deRenderEngine.dll, 32-bit, in IMVU's process) and a 64-bit renderer process (t5de-renderer.exe) drawing with Diligent Engine, fetched with FetchContent and pinned. For now a spike: the adapter starts the renderer process for each engine and signals each frame through a shared memory channel, and the renderer process draws a box into a child window in the client's 3D view.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
dhkatzandClaude Opus 5.5