Marcos Slomp

@zvookin Would it be possible to use the `halide_mutex` that `scoped_mutex_lock.h` uses? https://github.com/halide/Halide/blob/d03140864ad4b7816b89f450a90f5c0e1effad57/src/runtime/scoped_mutex_lock.h#L12 @abadams mentioned to me that it implements an adaptive mutex (spins just for a bit, then defers...

Yeah, I am also not entirely convinced that spinning even if for just a bit is a good idea. My understanding is that the main benefits of an adaptive mutex...

Yup, I did browse that WebKit blog post, but the microbenchmarks seemed a bit contrived and artificial. > if I remove spinning from our mutex the mullapudi2016_histogram test gets a...

Linus is probably basing his claims on more grounded real world scenarios reported in the wild, when there are other "busy" processes competing for a slice of the scheduler. I...

For what is worth, I found this article on spinlocks from the perspective of real-time audio synthesis: https://timur.audio/using-locks-in-real-time-audio-processing-safely The article describes an implementation of a spinlock with exponential back-off that...

Related: https://github.com/wolfpld/tracy/issues/572

I think it is possible, the only issue would be to ensure the `Input` buffers used by the generator have some data bound to them. That could be done by...

@shoaibkamil I recommend pre-pending the error message with something that reminds the user that this error has happened unexpectedly during the execution of a previous pipeline.

Hi @soufianekhiat, For SM 6 support, I think we'll have to introduce a companion feature flag to Direct3D 12 (say, `d3d12compute-sm6`), and then extend the [current HLSL code generator]( https://github.com/halide/Halide/blob/main/src/CodeGen_D3D12Compute_Dev.cpp)...

> I'm still working on it. Do we need a process to create a branch? [Halide_sk_d3d12_hlsl_6.patch](https://github.com/halide/Halide/files/14996882/Halide_sk_d3d12_hlsl_6.patch) Let me know if you have any comment. [WIP] Just follow standard GitHub practice:...