I hit this as trying to instrument this pipeline (from a a vkd3d title using Descriptor Buffers on Intel Mesa) with GPU-AV running on caused the driver to crash and wanted to create a repro, but hitting
$ fossilize-synth --frag dump_frag.spv --output repro.foz
Fossilize ERROR: Overlap in descriptor type for binding (2, 0) (was 6, now 7).
I pulled out the shader and seems valid https://godbolt.org/z/cadKMvben
I see this is around the whole HLSL having CBV vs SSBO stuff
layout(set = 2, binding = 0, std430) restrict readonly buffer SSBO {
uint _m0[];
} _23[];
layout(set = 2, binding = 0, std430) restrict readonly buffer _26_29 {
uvec4 _m0[];
} _29[];
layout(set = 2, binding = 0, std140) uniform BindlessCBV {
vec4 _m0[4096];
} _36[];
not sure who is at fault here, but if this is valid SPIR-V, I guess not sure if this is a limitation of fossilize-synth... but at the same time, not sure what the VK_EXT_descriptor_buffer API side looks like
I hit this as trying to instrument this pipeline (from a a vkd3d title using Descriptor Buffers on Intel Mesa) with GPU-AV running on caused the driver to crash and wanted to create a repro, but hitting
I pulled out the shader and seems valid https://godbolt.org/z/cadKMvben
I see this is around the whole HLSL having CBV vs SSBO stuff
not sure who is at fault here, but if this is valid SPIR-V, I guess not sure if this is a limitation of
fossilize-synth... but at the same time, not sure what theVK_EXT_descriptor_bufferAPI side looks like