-
A module specifying version 1.6 implies`VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT`, which makes `gl_SubgroupSize`/`WaveGetLaneCount` behave on implementations with several subgro…
-
I have a function that loads spv file:
```cpp
std::vector loadSpv(std::string path) {
std::ifstream file(path, std::ios::ate | std::ios::binary);
if (!file.is_open()) {
throw …
-
Using 0 as a constant value for a groupshared array causes a `fatal error: generated SPIR-V is invalid: OpTypeArray Length '11[%uint_0]' default value must be at least 1: found 0`
Not exactly a bi…
-
Following my AR from OpenCL Tooling TSG. The translator claims SPIR-V 1.4 support, yet there are few features missing
- [x] `OpCopyLogical` instruction
- [x] `OpPtrEqual `and `OpPtrNotEqual` instr…
-
**Description**
The HLSL code that compiles successfully in DXIL output mode results in a compilation error in SPIRV output mode.
Maybe related to #6389 ?
**Steps to Reproduce**
`dxc -T cs_6_0…
-
When compiling this shader:
```
StructuredBuffer g_mask;
float4 main(uint4 input0 : ATTR0) : SV_Target {
uint4 mask = g_mask[0];
uint4 res1 = WaveMultiPrefixProduct(input0, mask);
…
-
Blocks vkQuake3 on gfx-portability. We are likely blocked on specialization support in a larger sense after this.
kvark updated
2 months ago
-
(llm) E:\chatcode>set SYCL_CACHE_PERSISTENT=1
(llm) E:\chatcode>set BIGDL_LLM_XMX_DISABLED=1
(llm) E:\chatcode>**python chatglm3_infer_gpu.py**
D:\Users\admin\anaconda3\envs\llm\lib\site-packag…
-
By design, it's possible to generate arbitrary and therefore invalid SPIR-V with inline SPIR-V. In this case, validation errors are much less likely to be the result of compiler bugs, so we should rem…
-
**Environment:**
- OS: windows 10
- GPU and driver version: nvidia 556.12
- SDK or header version if building from repo: 1.3.283.0
- Options enabled (synchronization, best practices, etc.): sy…