GPU instancing: streamer, viewport, shaders rewritten

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>
This commit is contained in:
Dion Moult
2026-04-12 19:53:06 +10:00
parent 8c8ef5c32b
commit 07a5c59359
11 changed files with 836 additions and 1432 deletions
+21 -18
View File
@@ -17,22 +17,28 @@
* *
********************************************************************************/
// NOTE: Sidecar format v3 is being rewritten to v4 (instanced geometry layout).
// During the instancing rewrite (Commit A) the cache is a no-op: reads always
// miss and writes always succeed without producing a file. Commit B will
// re-introduce the on-disk format with MeshInfo + InstanceGpu sections.
#ifndef SIDECARCACHE_H
#define SIDECARCACHE_H
#include "BvhAccel.h"
#include "InstancedGeometry.h"
#include <cstdint>
#include <optional>
#include <string>
#include <vector>
#include <memory>
static constexpr uint32_t SIDECAR_MAGIC = 0x49465657; // "IFVW"
static constexpr uint32_t SIDECAR_VERSION = 3;
static constexpr uint32_t SIDECAR_VERSION = 4;
static constexpr uint32_t SIDECAR_ENDIAN = 0x01020304;
// Fixed-size element record for the sidecar. Strings are stored as
// (offset, length) pairs into a separate string table.
// Fixed-size element record. Strings are stored as (offset, length) pairs
// into a separate string table.
struct PackedElementInfo {
uint32_t object_id;
uint32_t model_id;
@@ -46,30 +52,27 @@ struct PackedElementInfo {
uint32_t type_length;
};
// Everything the viewer needs to display a model without tessellating.
// Everything needed to display an already-tessellated model without
// re-running the iterator. v4 schema: instanced geometry.
struct SidecarData {
// GPU geometry (ready to upload as-is)
std::vector<float> vertices; // interleaved, 8 floats per vertex
std::vector<uint32_t> indices; // global (already remapped)
// Per-model GPU geometry (local coords). 28 bytes/vertex.
std::vector<float> vertices;
std::vector<uint32_t> indices;
// Per-object metadata
std::vector<ObjectDrawInfo> draw_info;
// Mesh dictionary and per-instance data.
std::vector<MeshInfo> meshes; // indexed by local_mesh_id
std::vector<InstanceCpu> instances; // sorted by mesh_id
// Element tree metadata
// Element tree metadata.
std::vector<PackedElementInfo> elements;
std::string string_table; // concatenated UTF-8
// BVH acceleration
std::shared_ptr<BvhSet> bvh_set;
std::string string_table;
};
// Write a full sidecar next to the IFC file.
// Returns true on success.
// v4 writer/reader are stubbed for Commit A — no disk I/O happens.
bool writeSidecar(const std::string& ifc_path,
const SidecarData& data,
uint64_t ifc_file_size);
// Read a sidecar. Returns nullopt on any failure (missing, stale, corrupt).
std::optional<SidecarData> readSidecar(const std::string& ifc_path,
uint64_t ifc_file_size);