fix(rt): split Sponza's combined image sampler into a heap texture and sampler
examples/Sponza no longer built natively (#164). Its closest-hit declared a descriptor_heap sampler2D, which glslang has rejected since c8253623 ("Reject descriptor_heap combined image samplers"), and main.cpp wrote it through VkResourceDescriptorDataEXT::pCombinedImageSampler, which that union no longer has: the descriptor heap has no combined image samplers. Both went unnoticed because nothing builds the examples. Sponza - closesthit.glsl declares texture2D and sampler heap arrays and joins them where it samples, as ui-shared.glsl and HeapDynamicIndexHit do. The sampler slot is a second spec constant beside the image slot, so the shader assumes nothing about where the host put it, and a constant index keeps the example running on NVIDIA 610.43.02 (see its README). The comment no longer says the image slot is offset by bufferStartElement; only buffer slots are. - main.cpp gives the heap a sampler slot, writes the albedo as a SAMPLED_IMAGE from a view create-info, and writes the sampler into the sampler heap from a VkSamplerCreateInfo with vkWriteSamplerDescriptorsEXT, as UIRenderer::RegisterSampler does. SamplerVulkan's VkSampler had no use left. The sampler keeps its linear filtering and repeat addressing and drops anisotropy, which a hit shader cannot use: it samples at LOD 0, without derivatives. Docs - README.md and examples/Sponza/README.md quote the new declarations, and the Sponza README says why the texture and the sampler are two descriptors. Its texture2DArray materials pattern declared a descriptor_heap sampler2DArray, which glslang rejects as well; it now joins a texture2DArray and a sampler. - examples/SponzaBench/README.md drops its caveat that Sponza does not compile. Tests - ExampleShaders compiles every GLSL shader under examples/ with glslang as crafter-build does for the example (Vulkan rules, SPIR-V 1.4, shaders/ on the include path, and the descriptor-set build where the example registers one), so a glslang that starts rejecting one fails a test rather than an example build nobody runs. It fails on Sponza's old closest-hit and needs glslang only. Closes #164. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
parent
c97a3048fb
commit
6e94cc6209
GPG key ID:
2937E59CDCC1BCFB
7 changed files with 230 additions and 40 deletions
|
|
@ -14,9 +14,16 @@ it via ray tracing on both Vulkan (native) and WebGPU (wasm). Same
|
|||
at the barycentric attribs as UVs — proof-of-binding, not visually
|
||||
accurate. Per-vertex UV interpolation is the next step.
|
||||
|
||||
The closest-hit reads its texture through a **spec constant**
|
||||
(`albedo[albedoSlot]`), not a runtime index. That is deliberate — see
|
||||
below.
|
||||
The closest-hit reads its texture and its sampler through **spec
|
||||
constants** (`textures[albedoSlot]`, `samplers[samplerSlot]`), not a
|
||||
runtime index. That is deliberate — see below.
|
||||
|
||||
On Vulkan the texture and the sampler are separate descriptors: a sampled
|
||||
image in the resource heap and a sampler in the sampler heap, joined where
|
||||
the closest-hit samples them,
|
||||
`texture(sampler2D(textures[albedoSlot], samplers[samplerSlot]), uv)`.
|
||||
The descriptor heap has no combined image samplers, and glslang rejects a
|
||||
`descriptor_heap` `sampler2D` (#164).
|
||||
|
||||
## Native RT limitation: dynamic `descriptor_heap` indexing in hit shaders
|
||||
|
||||
|
|
@ -44,7 +51,8 @@ closest-hit:
|
|||
|
||||
- A closest-hit that reads only `lightHeap[lightSlot]` where `lightSlot`
|
||||
is a **spec constant** survives indefinitely. ✅ (This example's
|
||||
`albedo[albedoSlot]` is exactly this case.)
|
||||
`textures[albedoSlot]` and `samplers[samplerSlot]` are exactly this
|
||||
case.)
|
||||
- Reading `indexHeap[assetIndexStart + gl_InstanceCustomIndexEXT]` /
|
||||
`vertexHeap[...]` — a heap index offset by a **runtime** value —
|
||||
device-losts on the first geometry hit. ❌
|
||||
|
|
@ -116,11 +124,14 @@ indexing the heap:
|
|||
slot and index by **layer**:
|
||||
|
||||
```glsl
|
||||
// ✅ one array texture (spec-constant slot), dynamic layer index
|
||||
// ✅ one array texture (spec-constant slots), dynamic layer index
|
||||
layout(constant_id = 1) const uint16_t materialArraySlot = 0us;
|
||||
layout(descriptor_heap) uniform sampler2DArray materials[];
|
||||
layout(constant_id = 2) const uint16_t samplerSlot = 0us;
|
||||
layout(descriptor_heap) uniform texture2DArray materials[];
|
||||
layout(descriptor_heap) uniform sampler samplers[];
|
||||
uint layer = materialLayer[gl_InstanceCustomIndexEXT]; // from a bound SSBO
|
||||
vec3 albedo = texture(materials[materialArraySlot], vec3(uv, layer)).rgb;
|
||||
vec3 albedo = texture(sampler2DArray(materials[materialArraySlot], samplers[samplerSlot]),
|
||||
vec3(uv, layer)).rgb;
|
||||
```
|
||||
|
||||
This is precisely what the WebGPU path already does — bucketed texture
|
||||
|
|
|
|||
|
|
@ -8,19 +8,24 @@
|
|||
#extension GL_EXT_descriptor_heap : enable
|
||||
#extension GL_EXT_nonuniform_qualifier : enable
|
||||
|
||||
// Specialization constant: descriptor-heap slot of the albedo texture.
|
||||
// Set from descriptorHeap.bufferStartElement + the slot allocated for
|
||||
// the Image2D<RGBA8> on the host side. Sampling uses gl_HitAttributeEXT
|
||||
// barycentrics as UVs — proof-of-binding rather than UV-correct shading.
|
||||
// Per-vertex UV interpolation lands when Mesh on Vulkan exposes the
|
||||
// data-region buffer.
|
||||
// Specialization constants: the albedo's slot in the resource heap and its
|
||||
// sampler's slot in the sampler heap, as the host allocated them (image and
|
||||
// sampler slots index their heaps directly; only buffer slots are offset by
|
||||
// bufferStartElement). The descriptor heap has no combined image samplers,
|
||||
// and glslang rejects a descriptor_heap sampler2D (#164), so the texture
|
||||
// and the sampler are separate descriptors, joined where they are sampled.
|
||||
// Sampling uses gl_HitAttributeEXT barycentrics as UVs — proof-of-binding
|
||||
// rather than UV-correct shading. Per-vertex UV interpolation lands when
|
||||
// Mesh on Vulkan exposes the data-region buffer.
|
||||
layout(constant_id = 0) const uint16_t albedoSlot = 0us;
|
||||
layout(descriptor_heap) uniform sampler2D albedo[];
|
||||
layout(constant_id = 1) const uint16_t samplerSlot = 0us;
|
||||
layout(descriptor_heap) uniform texture2D textures[];
|
||||
layout(descriptor_heap) uniform sampler samplers[];
|
||||
|
||||
hitAttributeEXT vec2 hitAttrs;
|
||||
layout(location = 0) rayPayloadInEXT vec3 hitValue;
|
||||
|
||||
void main() {
|
||||
vec2 bary = vec2(hitAttrs.x, hitAttrs.y);
|
||||
hitValue = texture(albedo[albedoSlot], bary).rgb;
|
||||
hitValue = texture(sampler2D(textures[albedoSlot], samplers[samplerSlot]), bary).rgb;
|
||||
}
|
||||
|
|
|
|||
|
|
@ -79,26 +79,32 @@ int main() {
|
|||
VkCommandBuffer cmd = window.StartInit();
|
||||
|
||||
DescriptorHeapVulkan descriptorHeap;
|
||||
descriptorHeap.Initialize(/*images*/ 2, /*buffers*/ 1, /*samplers*/ 0);
|
||||
descriptorHeap.Initialize(/*images*/ 2, /*buffers*/ 1, /*samplers*/ 1);
|
||||
|
||||
// Two specialization constants: the TLAS slot offset (shared with
|
||||
// VulkanTriangle pattern) and the albedo slot index for closesthit.
|
||||
// Specialization constants: the TLAS slot offset for raygen (shared
|
||||
// with the VulkanTriangle pattern), and the albedo's image and sampler
|
||||
// slots for closesthit.
|
||||
VkSpecializationMapEntry raygenEntry = { .constantID = 0, .offset = 0, .size = sizeof(std::uint16_t) };
|
||||
VkSpecializationInfo raygenSpec = {
|
||||
.mapEntryCount = 1, .pMapEntries = &raygenEntry,
|
||||
.dataSize = sizeof(std::uint16_t), .pData = &descriptorHeap.bufferStartElement,
|
||||
};
|
||||
|
||||
// Allocate the albedo slot first so its index is known when we
|
||||
// Allocate the albedo's slots first so their indices are known when we
|
||||
// compile closesthit.spv.
|
||||
auto imgSlots = descriptorHeap.AllocateImageSlots(2);
|
||||
auto bufSlots = descriptorHeap.AllocateBufferSlots(1);
|
||||
auto imgSlots = descriptorHeap.AllocateImageSlots(2);
|
||||
auto bufSlots = descriptorHeap.AllocateBufferSlots(1);
|
||||
auto samplerSlots = descriptorHeap.AllocateSamplerSlots(1);
|
||||
std::uint16_t albedoHeapSlot = static_cast<std::uint16_t>(imgSlots.firstElement + 1);
|
||||
|
||||
VkSpecializationMapEntry hitEntry = { .constantID = 0, .offset = 0, .size = sizeof(std::uint16_t) };
|
||||
const std::array<std::uint16_t, 2> hitConstants { albedoHeapSlot, samplerSlots.firstElement };
|
||||
const std::array<VkSpecializationMapEntry, 2> hitEntries {{
|
||||
{ .constantID = 0, .offset = 0, .size = sizeof(std::uint16_t) },
|
||||
{ .constantID = 1, .offset = sizeof(std::uint16_t), .size = sizeof(std::uint16_t) },
|
||||
}};
|
||||
VkSpecializationInfo hitSpec = {
|
||||
.mapEntryCount = 1, .pMapEntries = &hitEntry,
|
||||
.dataSize = sizeof(std::uint16_t), .pData = &albedoHeapSlot,
|
||||
.mapEntryCount = static_cast<std::uint32_t>(hitEntries.size()), .pMapEntries = hitEntries.data(),
|
||||
.dataSize = sizeof(hitConstants), .pData = hitConstants.data(),
|
||||
};
|
||||
|
||||
std::array<VulkanShader, 3> shaders {{
|
||||
|
|
@ -140,7 +146,6 @@ int main() {
|
|||
VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT,
|
||||
VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
|
||||
albedo.Update(loadedTex, cmd, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
|
||||
SamplerVulkan<RGBA8> sampler;
|
||||
|
||||
static RenderingElement3D renderer;
|
||||
renderer.instance = {
|
||||
|
|
@ -161,13 +166,26 @@ int main() {
|
|||
window.FinishInit();
|
||||
|
||||
// Write descriptors: TLAS at bufSlots[0], output image at imgSlots[0],
|
||||
// albedo (combined image+sampler) at imgSlots[1]. Per-frame replicated.
|
||||
// albedo (a sampled image) at imgSlots[1], and its sampler at
|
||||
// samplerSlots[0] in the sampler heap. Per-frame replicated.
|
||||
VkDeviceAddressRangeKHR tlasRanges[Window::numFrames];
|
||||
VkImageDescriptorInfoEXT outImgInfos[Window::numFrames];
|
||||
VkDescriptorImageInfo albedoInfo {
|
||||
.sampler = sampler.textureSampler,
|
||||
.imageView = albedo.imageView,
|
||||
.imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
|
||||
// The heap takes an image view's create-info, not a VkImageView.
|
||||
const VkImageViewCreateInfo albedoView {
|
||||
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
|
||||
.image = albedo.image,
|
||||
.viewType = VK_IMAGE_VIEW_TYPE_2D,
|
||||
.format = VK_FORMAT_R8G8B8A8_UNORM,
|
||||
.subresourceRange = {
|
||||
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
|
||||
.levelCount = albedo.mipLevels,
|
||||
.layerCount = 1,
|
||||
},
|
||||
};
|
||||
const VkImageDescriptorInfoEXT albedoInfo {
|
||||
.sType = VK_STRUCTURE_TYPE_IMAGE_DESCRIPTOR_INFO_EXT,
|
||||
.pView = &albedoView,
|
||||
.layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
|
||||
};
|
||||
for (std::uint32_t f = 0; f < Window::numFrames; ++f) {
|
||||
tlasRanges[f] = { .address = RenderingElement3D::tlases[f].address };
|
||||
|
|
@ -205,8 +223,8 @@ int main() {
|
|||
});
|
||||
resources.push_back({
|
||||
.sType = VK_STRUCTURE_TYPE_RESOURCE_DESCRIPTOR_INFO_EXT,
|
||||
.type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
|
||||
.data = { .pCombinedImageSampler = &albedoInfo },
|
||||
.type = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE,
|
||||
.data = { .pImage = &albedoInfo },
|
||||
});
|
||||
destinations.push_back({
|
||||
.address = descriptorHeap.resourceHeap[f].value
|
||||
|
|
@ -221,6 +239,34 @@ int main() {
|
|||
descriptorHeap.resourceHeap[f].FlushDevice();
|
||||
}
|
||||
|
||||
// The sampler heap takes a sampler's create-info, as
|
||||
// UIRenderer::RegisterSampler writes it.
|
||||
const VkSamplerCreateInfo albedoSamplerInfo {
|
||||
.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
|
||||
.magFilter = VK_FILTER_LINEAR,
|
||||
.minFilter = VK_FILTER_LINEAR,
|
||||
.mipmapMode = VK_SAMPLER_MIPMAP_MODE_LINEAR,
|
||||
.addressModeU = VK_SAMPLER_ADDRESS_MODE_REPEAT,
|
||||
.addressModeV = VK_SAMPLER_ADDRESS_MODE_REPEAT,
|
||||
.addressModeW = VK_SAMPLER_ADDRESS_MODE_REPEAT,
|
||||
.maxLod = VK_LOD_CLAMP_NONE,
|
||||
};
|
||||
VkSamplerCreateInfo samplerInfos[Window::numFrames];
|
||||
VkHostAddressRangeEXT samplerDestinations[Window::numFrames];
|
||||
for (std::uint32_t f = 0; f < Window::numFrames; ++f) {
|
||||
samplerInfos[f] = albedoSamplerInfo;
|
||||
samplerDestinations[f] = {
|
||||
.address = descriptorHeap.samplerHeap[f].value
|
||||
+ descriptorHeap.SamplerByteOffset(samplerSlots.firstElement),
|
||||
.size = Device::descriptorHeapProperties.samplerDescriptorSize,
|
||||
};
|
||||
}
|
||||
Device::vkWriteSamplerDescriptorsEXT(Device::device, Window::numFrames,
|
||||
samplerInfos, samplerDestinations);
|
||||
for (std::uint32_t f = 0; f < Window::numFrames; ++f) {
|
||||
descriptorHeap.samplerHeap[f].FlushDevice();
|
||||
}
|
||||
|
||||
window.descriptorHeap = &descriptorHeap;
|
||||
RTPass rtPass(&pipeline);
|
||||
window.passes.push_back(&rtPass);
|
||||
|
|
|
|||
|
|
@ -119,13 +119,6 @@ scene for the dynamic-upload PRs) would be needed.
|
|||
|
||||
## Caveats / notes
|
||||
|
||||
* The stock `examples/Sponza` **native** path does not currently compile
|
||||
against the (unpinned) Vulkan-Headers `main`: the
|
||||
`VkResourceDescriptorDataEXT` union no longer has the
|
||||
`pCombinedImageSampler` member the example uses. SponzaBench sidesteps
|
||||
this by not binding a combined image+sampler. (Pre-existing, unrelated to
|
||||
#155 — worth a follow-up to port Sponza to the split sampled-image +
|
||||
sampler-heap model the engine's own UI renderer already uses.)
|
||||
* Numbers are CPU/GPU specific. Re-run `run-bench.sh` locally for your
|
||||
hardware; the *deltas* between revisions are the point, not the absolute
|
||||
figures.
|
||||
|
|
|
|||
Loading…
Reference in a new issue