Wires up the user-facing section-cut tool on top of the clipping
plumbing landed in the previous commit.
- K toggles the tool.
- LMB while the tool is active:
* On an existing plane's arrow gizmo (screen-space line-segment
hit test, 12 px grab radius) → select + start drag.
* Otherwise on geometry → pickSurfaceAt + addSectionPlaneAt-
Surface, select the new plane.
* Otherwise → deselect.
- LMB drag updates the plane's origin by projecting the cursor
delta onto the screen-space normal axis and converting back to
metres. d is rederived from the new origin each frame.
- Delete removes the selected plane; Esc exits the tool.
- Each plane renders a 2x2 m quad outline plus a yellow arrow
along +n at its origin. Selected plane draws cyan and
thicker.
LMB object-pick is suppressed while the tool is active so plane
creation does not also change selection.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a clip-plane pipeline used by the upcoming section tool:
- Up to 8 SectionPlane{n, d} entries, AND-combined as
fragment-shader discard against world position. Main and pick
fragment shaders both honour the planes, so cut areas are
neither drawn nor selectable.
- Main vertex shader now passes v_world_pos through.
- Pick FBO grows two attachments (RGB32F world position, RGB16F
world normal) and the pick shader writes both alongside the
object id. pickSurfaceAt() does a single readback of all
three. Existing pickObjectAt() still works unchanged for
callers that just want the id.
- addSectionPlaneAtSurface(point, normal) auto-flips the normal
toward the camera so the first click immediately cuts the
camera-facing half.
No UI yet — that's the next commit (gizmo, drag, K shortcut).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
F (no modifier) re-aims the orbit camera at the selected object's
world AABB centroid and dollies camera_distance_ so the bounding
sphere fits the current viewport. Home does the same for the union
of all finalized models. Both preserve yaw/pitch so the user keeps
their orientation; both no-op in FPS mode.
Scene AABB prefers the per-model BVH root when available and falls
back to walking InstanceCpu world AABBs. Object AABB unions every
matching instance. Distance accounts for portrait windows by using
the tighter of the horizontal and vertical FOV constraints.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
A small RGB axis cross is rendered at camera_target_ while the user is
orbiting, panning, or has just zoomed. Visibility toggles on
middle-mouse press/release; the wheel arms a single-shot QTimer that
hides it 750 ms after the last notch.
Drawn in two passes: GL_GREATER at 30% alpha for the occluded portion
(X-ray cue) and GL_LEQUAL at full alpha for the visible portion. Arm
length is computed from camera_distance_, fovy, and viewport height so
the cross stays ~30 px on screen across zoom levels.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Federation (JSON) tracks an ordered list of model sources plus an
optional home-view camera state. Sources are stored relative when
under the federation file's directory, absolute otherwise.
File menu now exposes New / Open / Save / Save As; Add Files moves
to Ctrl+Shift+O. View menu gains Set/Go to Home View. Window title
binds to dirty state via setWindowModified, and the close-window
prompt offers Save/Discard/Cancel.
Per-model transform (4x4 column-major) and visible round-trip
through load/save but are not yet applied at the viewport — the
georeferencing work uses them.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
WASD strafe, Q/E down/up, mouse-look (cursor hidden + recentered),
Shift to sprint, scrollwheel scales speed, click or Esc returns to
orbit. Exiting drops back to the same viewpoint because rotation
re-pins camera_target_ to keep camera_eye_ stationary.
Movement integrates wall-clock dt inside render() and the next frame
self-schedules via requestUpdate() while any key is held. A QTimer
would fight Qt's event loop during long swapBuffers blocks and produce
"camera pauses one frame" stalls; render-driven integration keeps
movement phase-locked to vsync and absorbs slow frames in a single
catch-up step.
IFC_FPS_HITCH_MS=<n> logs frames slower than n ms while in fly mode.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Buffer viewport model mutations until the OpenGL context is initialized so loads that start before first exposure do not silently drop geometry or model state.
Generated with the assistance of an AI coding tool.
Replace i16x2 octahedral normals with i8x2, filling the 2-byte padding
after position and saving 4 bytes per vertex. int8 gives ~1.4 deg
worst-case angular error — invisible for BIM geometry which is
overwhelmingly axis-aligned. 25% VBO reduction; sidecar files shrink
~15% overall (5.4 GB -> 4.6 GB on a 111-model test scene). Bumps
sidecar format to v7.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Benchmarks showed negligible gain (52 vs 51 fps) — the CPU BVH path
already culls efficiently, and the GPU path still read back to CPU for
LOD/winding/HiZ. Removes ~570 lines of dead weight: compute shader,
async readback, one-frame-late consume, per-model AABB SSBOs, and
profiling counters.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add --camera tx,ty,tz,dist,yaw,pitch and --benchmark N CLI args for
reproducible performance measurement. The benchmark orbits the camera
(0.5°/frame yaw) for N frames after a 5-frame warmup, prints
avg/median/p1/p99 frame times, then exits. Press C during interactive
use to print the current camera as a --camera argument.
Fix settle recull to fire after ANY camera motion (not just when
IFC_MIN_PX_MOTION is set), ensuring HiZ artifacts from motion frames
are always cleared when the camera stops.
Document Phase 3G (motion-adaptive culling + HiZ during motion) in
README with benchmark results from 1.06M-instance scene:
- Baseline: 16.3 fps
- IFC_MIN_PX_MOTION=10: 26.5 fps (1.6x)
- IFC_HIZ_MOTION=1: 46.6 fps (2.9x)
- Both combined: 51.0 fps (3.1x)
- + GPU_CULL: 52.0 fps (3.2x, negligible gain)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
During camera motion, use a larger pixel-radius threshold (IFC_MIN_PX_MOTION)
to aggressively cull small objects, dramatically reducing sub_draws and
improving orbit fps (e.g. 29→67 fps on 1M-instance scene). When the camera
stops, automatically re-cull at the base threshold to restore full detail.
Key behaviors:
- IFC_MIN_PX_MOTION=N sets the motion threshold (0 = disabled)
- Settle recull fires on the first still frame after motion
- HiZ pyramid invalidated on settle (stale from sparse motion frame)
- GPU cull results skipped on settle (dispatched at motion threshold)
- requestUpdate() ensures the settle frame actually runs
Also adds IFC_SUBDRAW_DIAG=1 diagnostic for sub-draw composition analysis
and documents Phase 3E/3F experiment results in README.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The HiZ pipeline had two bugs causing false occlusions:
1. The scaling depth blit (glBlitFramebuffer from window-size to HiZ-size)
produced GL_INVALID_VALUE on some drivers. Replace with a fullscreen-
triangle shader that samples the resolved depth and writes gl_FragDepth.
2. The resolve texture used GL_DEPTH_COMPONENT24 but Qt's default FBO uses
D24S8 (depth+stencil). Mismatched formats cause the MSAA resolve blit
to fail. Fix by using GL_DEPTH24_STENCIL8 for the resolve texture.
Additionally, the occlusion test was too aggressive for scenes with
compressed depth ranges (entire scene in 0.99-1.0). Change from
"max over coarse mip texels" to "reject only if ALL fine-mip texels
agree the AABB is behind them", with early-out on first non-occluding
texel and a 64-sample cap.
Also fix IFC_HIZ_MOTION=0 being treated as enabled (checked env var
existence, not value).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Only clear and emit mesh buckets that received survivors in the previous
frame, converting both phases from O(total_meshes) to O(active_meshes).
Adds per-sub-phase timing (bin/clr/class/emit) to the stats line.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace the CPU BVH traversal + frustum + contribution stages with a
GPU compute path (IFC_GPU_CULL=1). A single scene-wide dispatch tests
all instances against frustum planes and screen-space contribution
threshold, compacting survivors into a flat uint32 buffer via atomicAdd.
Uses one-frame-late async readback: frame N dispatches and fences,
frame N+1 polls the fence (non-blocking) and reads the persistent-
mapped result buffer with zero GPU sync cost. CPU still handles HiZ,
LOD selection, winding bucketing, and indirect command generation from
the compact survivor list; draw path is unchanged.
On a 1M-instance / 111-model scene (GTX 1650):
GPU dispatch: 0.70 ms (frustum + contribution, brute-force)
Readback: 0.00 ms (fence already signaled, persistent map)
CPU consume: 5.7–6.7 ms (parallel emit across models)
Cull wall: 5.8–6.9 ms (vs 9.6–15.2 ms CPU-only path)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pack compute shader compacts non-empty indirect commands into
contiguous fwd/rev ranges, eliminating ~690k empty sub-draws that
dominated command-processor overhead. GL 4.6 entrypoint loaded via
getProcAddress with ARB fallback; graceful degradation to uncompacted
MDI when unavailable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two-phase compute-cull dispatch when IFC_GPU_CULL=1:
Phase 1 frustum + contribution + LOD, no HiZ → survivors
Depth render survivors depth-only into half-viewport FBO
Build GPU compute max-reduce depth → R32F mip pyramid
Phase 2 same cull + HiZ test → final survivors
Color render final survivors
The compact shader's new hizOccluded() projects 8 AABB corners to
screen space, picks the mip level where the covered rect fits in ≤2×2
texels, and rejects when the AABB's near-depth exceeds the pyramid's
max depth.
New GPU resources (per-window):
hiz_gpu_fbo_ / hiz_gpu_depth_tex_ — depth-only FBO at half viewport
hiz_gpu_pyramid_tex_ — R32F mipmapped pyramid
hiz_gpu_copy_prog_ — compute: depth → pyramid L0
hiz_gpu_reduce_prog_ — compute: max-reduce L(n-1)→L(n)
hiz_gpu_depth_prog_ — vertex + trivial fragment
On a dense 18-model BIM dataset:
survivors: 140k → 65k (HiZ rejects ~50%)
triangles: 22M → 13M
gpu_cull: 0.06ms → 22.5ms (depth pre-pass CP overhead)
The depth pre-pass suffers the same empty-sub-draws CP overhead as the
color pass (690k commands, most with instanceCount=0). Once MDI
compaction lands, both passes will be fast. For now, net FPS is flat
(savings on color ≈ cost of depth pre-pass).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The compact shader now computes per-instance pixel radius and routes
survivors to LOD1 buckets when the projected sphere falls below the
LOD1 threshold (default 30 px, same as CPU path, tunable via
IFC_LOD1_PX).
Layout expanded from 2 to 4 buckets per mesh:
[0..M) fwd_lod0 [M..2M) fwd_lod1
[2M..3M) rev_lod0 [3M..4M) rev_lod1
Two MDIs per model: CCW for [0..2M), CW for [2M..4M). Per-mesh
has_lod1 flags live in a new gpu_mesh_flags_ssbo (binding 4).
Contribution cull refactored: the compact shader now computes
pixelRadius() once and uses it for both the min_pixel_radius rejection
and LOD routing, matching the CPU path's logic.
Visible-buffer worst case is 2 × total_instances (each LOD bucket
reserves the full fwd/rev capacity per mesh, since LOD selection is
dynamic).
Tri count drops ~60% on the test dataset (53M → 22M) thanks to LOD1
decimated meshes. FPS recovers from 16 to 36 despite 690k sub_draws
(4M layout). MDI compaction remains the final perf fix.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extend the GPU-cull indirect buffer from M to 2M commands: the first M
are the forward (non-reflected, CCW) bucket, the second M are the
reverse (reflected, CW) bucket. The compact shader reads flags bit 0
from the AABB SSBO and routes each survivor to the appropriate bucket
via bucket = reflected ? mesh_id + M : mesh_id.
uploadGpuCullStaticBuffers() now precomputes exact per-mesh fwd/rev
instance counts so each bucket reserves only the slots it needs
(total visible_ssbo size unchanged — sum of fwd + rev = total).
Draw loop issues two MDIs per model under IFC_GPU_CULL: first M
commands CCW, next M commands CW.
Sub-draws doubled (172k → 345k) which further regresses FPS due to
command-processor overhead from zero-instance sub-draws — the same
issue noted in 3a. MDI compaction remains the fix.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Promote the compute cull from a validation shader to the actual draw
driver. With the gate on, the CPU cull fan-out is skipped and MDI
consumes gpu_indirect_buffer / gpu_visible_ssbo directly.
- uploadGpuCullStaticBuffers() pre-fills per-mesh DrawElementsIndirect
commands and a mesh_base prefix sum so the compact shader can scatter
survivors into a fixed per-mesh range. Instance count for each
command is zeroed by a tiny reset dispatch, then the compact shader
atomically writes survivors and increments instanceCount.
- Draw loop branches on the gate: single CCW MDI with all mesh
commands. Fwd/rev winding split, LOD selection, and HiZ are still
CPU-path-only; reflected instances render with wrong winding under
this gate (step 3b).
- Once-per-second readback of each model's indirect buffer populates
the survivor / visible-object / visible-triangle stats so the
[frame] line reflects what the GPU actually drew.
Known regression: sub_draws is the full mesh count per model (~172k on
the test dataset) vs the handful of non-empty commands the CPU path
produces. Command-processor overhead from zero-instance sub-draws is
what drives the FPS drop, not the cull itself (0.05 ms). Compacting
non-empty commands requires glMultiDrawElementsIndirectCount, a GL 4.6
entrypoint not exposed by Qt's QOpenGLFunctions_4_5_Core; deferring to
3a-followup so we don't bolt a getProcAddress loader into the renderer
mid-restructure.
IFC_GPU_CULL is off by default, so this does not affect normal runs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
First Phase 3E milestone: a compute shader that reads the per-instance
world-AABB SSBO added in the last commit, tests each instance against
the 6 frustum planes, and atomicAdds a global counter. No visible list
or indirect-buffer writes yet — the output is just a survivor count,
cross-checked each frame against the CPU cull's numbers in the stats
line (`gpu_cull[Xms in=A surv=B]`) so we can verify the plumbing end-
to-end before we hand the GPU responsibility for the actual render data.
Dispatched from render() after the CPU cull completes, only when
IFC_GPU_CULL=1 and the camera moved (the skipped-cull still-frame path
doesn't re-check either). The readback is synchronous — that's fine
for a validation path; it'll go away once the GPU writes indirect
commands directly.
Expected invariant: gpu_cull.surv >= cpu_cull.visible_objects, since
the GPU path does frustum-only and CPU adds contribution + HiZ cuts on
top. A large mismatch (orders of magnitude, or surv < visible) means
the SSBO upload or shader logic is wrong.
No shader/buffer bindings overlap with the draw path (compute uses
bindings 0/1, restored before drawing; draw programs rebind 0/1/2).
Scaffolding for Phase 3E (GPU compute cull). After finalizeModel /
applyCachedModel, pack each InstanceCpu's world AABB + mesh_id +
reflection bit into a std430-friendly 32 B record and push it to a
per-model aabb_ssbo. No consumer yet — the CPU cull still drives
rendering — but the next commits will point a compute shader at this
buffer and have it produce the visible list + indirect commands
directly on the GPU.
Cost: 32 B per instance, ~18 MB for the 569 k-instance test scene.
One-shot upload at finalize time; streaming-time appends aren't
mirrored (the CPU cull doesn't need the SSBO, and finalizeModel
rebuilds the whole thing in one go).
Two stability bugs:
1. Clicking an object left the scene with wrong shading until the camera
moved. The pick pass re-culls every model with its own parameters
(min_pixel_radius=0, no HiZ) and overwrites each model's visible_ssbo
and indirect buffer. The next render() saw an unchanged camera,
skipped the cull via the have_cached_cull_ shortcut, and drew the
stale pick-pass buffers. Fix: invalidate have_cached_cull_ at the
end of pickObjectAt().
2. Loading two sidecar-cached models made the second model's picked
properties resolve to the first model's elements. Sidecars store raw
object_id / model_id values from the session that wrote them, and
both files start at object_id=1, so element_map_ entries collided.
Fix: on load, rebase every PackedElementInfo and InstanceCpu by
(next_object_id_ - min_id_in_sidecar) and overwrite model_id with
the freshly-assigned handle before the elements hit element_map_.
Also document both in the README — the pick-pass note under 3A
contribution culling, the sidecar rebase under the sidecar format
section.
HiZ from last frame encodes depth from last frame's viewpoint. When
the camera moves, projecting a current-frame AABB through the stored
VP answers 'was this occluded last frame?' rather than 'is it occluded
now?' — a self-reinforcing feedback loop where objects culled in
prior frames never appear in any depth buffer and stay permanently
hidden at certain camera angles.
Fix: require hiz_vp_ == current VP for the HiZ test to apply. HiZ
still helps static views (kicks in one frame after camera stops) but
no longer produces false occlusions during orbit. The correct fix for
orbit coverage is a depth pre-pass feeding fresh HiZ — planned as
part of Phase 3E GPU compute cull.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Split cullAndUploadVisible into cullModelCpu (CPU-only, thread-safe) and
uploadCullResults (GL-only, main thread). render() fans the per-model
culls out via std::async and joins before the serial upload pass.
The cull scratch (vis_fwd/rev_lod0/1, visible_flat, indirect_scratch)
moved onto ModelGpuData so each worker owns its output buffers. Phase
timers and hiz_reject_count_ are atomic since workers fetch_add into
them. A new wall-clock timer around the dispatch block reports the
actual frame-time contribution; the existing clr/trv/emt counters are
now documented as per-thread sums.
Measured on the 18-model / 569k-instance test scene: wall-clock cull
dropped from ~25 ms to ~5 ms while the aggregate CPU work (trv) stayed
~30 ms. Frame time 34 ms -> 19 ms. IFC_CULL_THREADS=0 forces the
single-threaded fallback.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Position now u16x3 normalized against each mesh's local AABB; normal
oct-encoded to i16x2; RGBA8 colour unchanged. Per-mesh dequant basis
lives in a new MeshGpu SSBO at binding 2; both main and pick shaders
mix() against it before applying the instance transform.
Drops VBO and sidecar size by ~43 % (28 -> 16 B/vert), which matters
mostly for warm-load downloads of precomputed sidecars and steady-state
VRAM. LodBuilder dequantizes positions into a scratch buffer before
calling meshopt, since meshoptimizer needs float positions.
Also fixes a streaming-time crash in cullAndUploadVisible: bvh_items
was only populated at finalize, but the linear fallback indexes it
during streaming. Mirror BvhItem appends in uploadInstanceChunk so the
hot path stays valid before the BVH is built.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replaced the 16ms QTimer with QEvent::UpdateRequest delivered via
requestUpdate(), posted from every state mutator (mouse/wheel, model
lifecycle, selection, visibility, resize). A static BIM scene — the
common case for a viewer — now does no work at all between user actions.
FPS is now measured as time spent inside render() rather than wall-clock
gap between frames, so idle gaps don't pollute the 1-second window and
the headline number reflects real render throughput. Headline fps still
caps at vsync; sub-vsync profiling lives in the cull[...] phase timers.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
render() was re-running the full cull every 16 ms timer tick even when
nothing had changed — the camera matrices, scene state, and therefore
visible set were all identical to the previous frame's. The GPU was
still happy to redraw from the cached indirect buffer, but the CPU was
burning 21 ms/frame rebuilding the same visible list.
Detect the no-op case by comparing view/proj against last_cull_view_ /
last_cull_proj_ and checking a scene-dirty flag (have_cached_cull_)
that every mutator on models_gpu_ invalidates — finalizeModel,
applyCachedModel, applyLodExtension, hide/show/remove/reset, and
uploadInstanceChunk. When the check passes we skip both
cullAndUploadVisible and buildHizPyramid (the depth buffer is
bit-identical, so re-reading it produces the same pyramid).
Per-model visible_objects / visible_triangles stats now live on
ModelGpuData so the stats line reports correct numbers on skipped
frames instead of reading from a stale indirect_scratch_.
Measured on a 569k-object overview: still frames go 22 fps → 62 fps;
orbiting goes 23 fps → ~30-50 fps depending on how hard you move the
mouse (the cull only pays its full cost on the ~25 % of frames where
the camera actually moved). The stats line gains a "skipped N/M"
field so you can see the ratio live.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
cullAndUploadVisible was reading each instance's AABB through
m.instances[idx] — a 104-byte InstanceCpu struct — for the frustum /
contribution / HiZ tests. Only 24 of those bytes (the two float[3]
AABBs) are actually used by the tests; the rest (4×4 transform +
header) is pure cache-line waste, and with 569k instances the array
is 59 MB, well past any cache.
bvh_items[idx] already stores a 1:1 compact 28-byte record with the
same AABB, built unconditionally in buildBvhForModel(). Switch the
hot test path to read from it, and only touch InstanceCpu once an
instance has passed all three tests (for mesh_id). Modest ~20 %
drop in cull-traverse time on a 569k-object overview (26 ms → 21 ms).
Also add four cull-phase timers (clr / trv / emt / upl) to the
per-second stats line so future optimisation work has concrete
numbers to chase. Confirmed via these timers that bucket clears,
emit and GPU upload are all <1 ms combined; traversal is where the
remaining CPU cost lives.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
After the main draw, blit the MSAA default-framebuffer depth to a
single-sample 256×128 depth texture, read it back, and build a CPU
max-reduced mip pyramid. Next frame's cullAndUploadVisible projects
each BVH node / instance AABB through the previous frame's VP and
compares the AABB's nearest depth against the pyramid's deepest value
at the matching mip level; strictly-beyond AABBs are rejected.
Conservative direction (aabb_near > hiz_max) — never wrongly rejects a
visible instance, so no flicker. BVH subtree-level test lets a single
8-corner projection reject up to a leaf's worth of instances.
Tuning knobs: IFC_NO_HIZ=1 disables; IFC_HIZ_SIZE overrides base width.
New stats counter hiz_rej shows rejects/frame.
Measured: big win on interior views (GPU-bound), roughly zero net
effect on exterior overviews (CPU-bound on cull traversal, so the
saved GPU work is masked). Tried a 3-deep PBO ring for async readback
and reverted — the extra frame of staleness produced visible flicker
on fast orbit, and the synchronous readback wasn't actually a measured
bottleneck at 256×128.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Decimate each unique mesh once at sidecar-build time and swap to the
reduced index slice per-instance per-frame when projected sphere radius
drops below IFC_LOD1_PX (default 30). Same VBO, same SSBO, just a
different firstIndex/count in the indirect command.
Extends MeshInfo (48→56 B) with lod1_ebo_byte_offset + lod1_index_count
and bumps the sidecar to v5. buildLods() runs inside
onStreamingFinished, appends decimated indices to sd.indices,
applyLodExtension pushes the EBO suffix to the live GPU state, and the
sidecar is written with LOD1 baked in.
simplifySloppy (voxel clustering) is used instead of the default
edge-collapse meshopt_simplify because BIM brep output is per-triangle-
unwelded and non-manifold after welding — simplify returned the input
unchanged for every mesh tested. Sloppy ignores topology. Knobs
(IFC_LOD_SLOPPY, IFC_LOD_ERROR, IFC_LOD_RATIO, IFC_LOD_MIN_SAVINGS,
IFC_LOD_LOCK_BORDER, IFC_LOD_DEBUG) are available for A/B tuning.
Result on the 128M-tri 10-model test scene (GTX 1650, 2px contribution
cull): 20.2 → 43.2 fps, 40M → 14M visible triangles, no change in
object count. LOD build adds 100–600 ms per model on first open,
cached thereafter.
README Phase 3B section is now a full writeup of pipeline, selection,
decimator-choice rationale, env vars, and measured numbers.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Reject frustum-visible objects whose bounding sphere projects below a
pixel-radius threshold. Applied at both BVH-node level (whole subtrees
pruned) and per-instance level; short-circuits when the camera is
inside the AABB so nothing-you're-standing-next-to is ever lost.
Pick pass passes threshold 0 so sub-pixel objects stay clickable.
Threshold defaults to 2 px (radius), overridable via IFC_MIN_PX env
var. Measured on the 128 M-tri test scene (GTX 1650):
0 px (off): 6.7 fps, 128 M tris
2 px: 20.2 fps, 40 M tris (31%)
4 px: 30.3 fps, 15 M tris (12%)
The metric is sphere-based (cheap: one sqrt per test) rather than
AABB-corner projection; loses a little precision on very elongated
bounds but costs ~5x less per test and the BVH-node pre-cull means
the long-tail-of-small-things case is already handled by subtree
pruning before we touch individual instances.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Earlier probes pointed at per-frame glNamedBufferSubData uploads as the
bottleneck (60 fps when those two calls were commented out). That was a
false reading — zeroing the uploads also emptied the indirect buffer, so
MDI drew nothing. "No upload" and "no draw" were indistinguishable.
Two new diagnostic env vars in render() isolate the real costs:
IFC_SKIP_MDI=1 keep cull + upload + binds, skip only the MDI
draws. Gives 62 fps with everything else running,
confirming the non-draw path fits in ~16 ms.
IFC_MAX_SUBDRAWS=N cap each MDI's drawcount. 67k -> 30k sub-draws
saves 0 ms, confirming sub-draw count itself is
not the bottleneck; the long tail of sub-draws
carries ~no triangles.
On a GTX 1650 with 128 M triangles in view, nvidia-smi sits at 95 %
GPU util and FPS scales with triangle work, not sub-draw count. The
card is simply rasterising at ~850 M tri/s. No CPU-side or upload
trick recovers it.
Revised Phase 3 is therefore shedding triangles, not bytes:
3A screen-space contribution culling (next)
3B LOD
3C HiZ occlusion
3D GPU-side compute culling
README Phase 3 section rewritten around the diagnosis, including the
false lead, so future work doesn't re-tread the upload path. The
aborted staging+resident ring-buffer implementation was reverted (the
uncommitted working tree is gone — pure glNamedBufferSubData retained
for the visible + indirect buffers, which we now know is fine).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Enables GL_CULL_FACE by default (user-toggleable in Settings) so
closed solids skip shading their back halves. The catch is that
IFC placements can contain reflections (mat4 with det<0 — mirrored
families, symmetric instances). Naively culling would make every
mirrored instance vanish because the rasterizer sees its screen-space
winding as backwards.
Fix: detect reflections at upload time via determinant sign, bucket
visible instances into forward (det>=0) and reverse (det<0) per mesh
during culling, and issue two glMultiDrawElementsIndirect calls per
model with glFrontFace toggled CCW/CW between them. The indirect
buffer is still one buffer — just split into a forward slice followed
by a reverse slice, with m.indirect_forward_count recording the split.
Vertex shader flips the normal when the transform has negative
determinant, keeping lighting correct on mirrored instances. The
fragment shader keeps the gl_FrontFacing fallback as a safety net
when culling is disabled (e.g. for files with open shells).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two bugs conflated as "weird colors":
1. Two-sided lighting. IFC placements often embed reflection
matrices (mirrored families). Transforming a_normal by
mat3(inst.transform) produces a normal pointing the wrong way
on those instances, and max(n·L, 0) then clamps the surface to
pure ambient — reads as dark / washed out. Use gl_FrontFacing
to flip n in the fragment shader so both winding orientations
shade correctly. The proper fix (ship an inverse-transpose
normal matrix or a det-sign bit per instance) is still owed;
that would unlock re-enabling GL_CULL_FACE for a big fragment-
work win on closed solids.
2. Stats label "inst_draws" was counting indirect sub-draws, not
actual GL draw calls — misleading since MDI collapses N sub-
draws into one glMultiDrawElementsIndirect. Split into
gl_draw_calls (real GL calls, = drawn-model count) and
indirect_sub_draws (packed sub-commands). For a BIM model
with 47k unique meshes at full view this now correctly reads
"1 gl_draws (47092 sub)" rather than suggesting 47k driver
dispatches.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each visible model now issues a single glMultiDrawElementsIndirect
call instead of one glDrawElementsInstancedBaseVertex per mesh. The
CPU BVH cull populates an array of DrawElementsIndirectCommand
records plus the flat visible-instance list, uploads both, and draws
the whole model in one GL call.
Vertex shaders switch from a uniform u_instance_offset to
gl_BaseInstanceARB (ARB_shader_draw_parameters), so per-draw offset
comes from the indirect command's baseInstance field.
Draw-call counts for BIM scenes with hundreds of unique meshes drop
from hundreds-per-frame to one-per-model, cutting driver overhead.
This also sets up the plumbing for the follow-up compute-shader cull
that will populate the indirect buffer entirely on-GPU.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pre-allocate the instance SSBO on model creation (4 MB, grow-on-demand)
and append each arriving InstanceChunk directly to the GPU-side
InstanceGpu array in uploadInstanceChunk. This makes a model drawable
as soon as its first mesh + first instance chunk land, rather than
waiting for finalizeModel.
The visible-list architecture already decouples SSBO order from the
draw path, so appending in insertion order is correct — no sorting
required. finalizeModel collapses to:
- compute per-mesh instance counts (for stats + sidecar round-trip)
- build the per-model BVH over instance world AABBs
Render / pick loops now gate on ssbo_instance_count > 0 rather than
the finalized flag. Stats include in-progress models in totals
(excluding only hidden).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Re-wires the BVH acceleration structure on top of the new instanced
renderer. Per model, build a BVH over per-instance world AABBs at
finalize (and on sidecar apply). Each frame, traverse the BVH against
the camera frustum to produce a visible-instance index list, bucket by
mesh_id, and upload to a per-model SSBO at binding=1. The main and
pick vertex shaders do a double-indirection
`instances[visible[u_offset + gl_InstanceID]]` so draws only touch
instances that passed the frustum test.
Models with fewer than BVH_MIN_OBJECTS instances skip the BVH build
and fall back to a linear per-instance frustum test.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Commit B of the instancing migration. The sidecar on-disk format is
reintroduced at version 4 with MeshInfo + InstanceCpu sections in place
of v3's flat per-object draw-info array.
After streaming finishes, MainWindow asks the viewport for a post-
finalise snapshot (VBO + EBO are read back from the GPU, meshes and
instances come from the CPU-side arrays) and writes it alongside
PackedElementInfo + the string table. On a subsequent load,
readSidecar rehydrates the whole struct and ViewportWindow::
applyCachedModel uploads VBO/EBO/SSBO in a single step, bypassing the
iterator entirely.
Staleness check is still by source file size.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Commit A of the instancing migration (Phase 3a). The streamer now runs
the iterator with use-world-coords=false and dedupes by the geometry's
representation id, emitting a MeshChunk once per unique geometry and an
InstanceChunk per placement. The viewport keeps geometry in local
coordinates (28 B/vertex, down from 32) and applies the per-instance
transform in the vertex shader via an std430 SSBO indexed by
gl_InstanceID + a per-draw uniform offset. After streaming finishes
finalizeModel() stable-sorts instances by mesh_id, assigns each mesh a
contiguous range, and uploads the SSBO; render then issues one
glDrawElementsInstancedBaseVertex per mesh.
BvhAccel is reshaped to operate on a generic BvhItem (world AABB +
model_id) so it can drive instance-level culling, but the path is not
wired in yet -- every instance is drawn every frame in this commit.
Progressive-during-streaming rendering is likewise disabled: a model
appears when its SSBO is uploaded, not incrementally. Sidecar cache
is stubbed (reads miss, writes are no-ops); the v4 on-disk format with
MeshInfo + InstanceGpu sections lands in Commit B.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When a BVH leaf passes the frustum test, emit a single glMultiDrawElements
record covering the leaf's entire index range instead of one per object.
Leaves are contiguous in the EBO after reorderEbo, so the range is just
[first_object.index_offset, sum(index_count)]. Cuts draw calls by ~8x
(BVH_MAX_LEAF_SIZE) and shifts the bottleneck from CPU/driver per-draw
overhead toward GPU vertex throughput.
Per-object features (selection highlight, per-vertex color, object_id
picking) are unchanged — they operate on vertex attributes, not draw
state. Future per-object hide/override will use SSBO lookups sampled
by object_id in the fragment shader.
Slight overdraw from skipping per-object frustum tests within a leaf is
negligible given median-split BVH tightness and spare tri throughput.
Also adds visible_objects_ counter so stats still report true object
counts (not leaf counts), plus leaf_draws/model_draws breakdown in the
per-second frame log.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Per-second frame log reports fps/ms, visible/total object & triangle
ratios, VRAM breakdown (VBO+EBO), model count, and pending uploads.
Upload-complete log includes per-model VBO/EBO MB and scene total VRAM.
Streamer runs an instancing analysis keyed on geom.id(): total shapes,
unique representations, dedup ratio, theoretical VBO/EBO/SSBO sizes if
instanced, potential savings, and top-5 most-duplicated representations.
Used to validate whether GPU instancing is worth the architectural
rewrite for a given dataset.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 2 performance: BVH acceleration with median-split build, per-model
trees, and EBO re-sorting for GPU cache coherence. Raw binary .ifcview
sidecar stores full geometry + BVH for instant subsequent loads (skip
tessellation entirely).
Per-model GPU buffers (VAO/VBO/EBO per model) eliminate cross-model buffer
copies on growth. Sidecar reads happen on a background thread. Bulk GPU
uploads are progressive (48 MB/frame chunks) so the viewport stays
interactive while multi-GB models stream in.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Introduce ModelHandle and per-model GeometryStreamers so multiple IFC
files can be loaded simultaneously. Object IDs are globally unique
(monotonically increasing across models). File picker is now multiselect.
Each model gets a top-level tree node. Property lookup uses the correct
model's ifcopenshell::file. ViewportWindow supports hide/show/remove
per model via model_id filtering in the frustum cull pass.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Show FPS, frame time, visible/total objects, and visible/total
triangles in the status bar. Toggled via Settings > Show Performance
Stats, persisted in app settings.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Track per-object AABB and index range during upload. Each frame,
extract frustum planes from the view-projection matrix and cull
objects whose AABB is entirely outside any plane. Draw only visible
objects via glMultiDrawElements. Document the three-phase rendering
performance strategy in README.md.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Dion Moult