-
### What happened?
When compiling CLIP Text Encoder 1 ([stable_diffusion_xl_base_1_0_64_fp16_clip_1.mlir](https://storage.googleapis.com/shark-public/ean/sdxl-turbine/stable_diffusion_xl_base_1_0_64_…
-
Blocks vkQuake3 on gfx-portability. We are likely blocked on specialization support in a larger sense after this.
kvark updated
4 months ago
-
Consider following HLSL
```HLSL
struct PixelInput
{
float4 pos : SV_Position0;
float2 texCoord : TEXCOORD0;
};
Texture2D Tex0;
SamplerState SS0;
Texture2D Tex1;
SamplerState SS1;
fl…
-
**Description**
We are seeing a crash in our shader debug code which uses template functions to output strings to a `RWByteAddressBuffer`.
Only SPIR-V compilation crashes. DXIL compilation works …
-
WGSL doesn't currently support `mat / scalar` though it does support `mat * scalar`. Apparently this is because MSL and SPIR-V don't support it. GLSL, HLSL do. Is it trivial to transform the expressio…
-
Upstream have started to add SPIRV target that register spirv64.
```
extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeSPIRVTargetInfo() {
RegisterTarget X(getTheSPIRV32Target(), "spirv32…
-
I am trying to implement the RWTexture2DMS feature from SM 6.7 using DXC packages from the Agility SDK.
This is a dummy shader with minimal code.
Surprisingly, the shader compiles fine for DXIL, htt…
-
Currently, SPIR-V backend does not support `EvaluateAttribute*` intrinsics.
According to the discussion in https://github.com/KhronosGroup/glslang/issues/2584, `OpInterpolate*` needs a pointer oper…
-
The following shader generates invalid SPIR-V:
```hlsl
RWTexture2D textures[];
cbuffer CB {
uint index;
}
RWTexture2D GetTexture() {
return textures[index];
}
[numthreads(1, 1, 1)]
…
-
There is a plan to enable AOT Inductor for Intel GPU in PyTorch 2.6. While working on the design, PyTorch Team realized that the Triton kernel is now saved as SPIR-V(IR), while CUDA is cubin(device c…