slang.git/prelude/slang-torch-prelude.h, branch masterMaking it easier to work with shadershttps://git.yummers.dev/slang.git/atom?h=master2024-11-26T02:57:48+00:00Fix issue with slang-embed & include ordering (#5680)2024-11-26T02:57:48+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2024-11-26T02:57:48+00:00urn:sha1:e6cf93e3e638cb981a9be392a2f48ea06acd4e3f
* Fix issue with slang-embed & include ordering
* Update CMakeLists.txtformat2024-10-29T06:49:26+00:00Ellie Hermaszewskaellieh@nvidia.com2024-10-29T06:49:26+00:00urn:sha1:f65d756bff8d4c5cbc15bd0322a2ae8e6b896a21
* format
* Minor test fixes
* enable checking cpp format in ciAdd `float16` support to slang-torch (#4584)2024-07-10T20:19:06+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2024-07-10T20:19:06+00:00urn:sha1:667e50498a226103278d0997528cc76979b2c4efError out when constructing tensor views from tensors with 0 stride. (#4516)2024-07-01T18:50:04+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2024-07-01T18:50:04+00:00urn:sha1:6e550430cf24a6324ce22f821181f2cab904d543
This avoids a problem with broadcasted tensors. Our tensor-view platform is designed to allow unrestricted access to tensor memory, while broadcasted tensors were designed for 'read-only' use-cases. Trying to write into a broadcasted tensor needs re-allocation, which Slang is not designed to do.
For now, we enforce contiguity on tensors with any 0 strides.
In the future, we will introduce a ConstTensorView object to allow such tensors to be used as an input.
This patch also propagates name-hint information through structs & arrays of tensors, to allow sensible names for the error messages (before this the error messages were temporary inst numbers, which is nearly impossible to debug)Prevent pointer validation for zero-size arrays (#4021)2024-04-24T23:50:50+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2024-04-24T23:50:50+00:00urn:sha1:941961e5dad5afa51095450b8d57380c322900c0More `slangpy` features + polishing (#3233)2023-09-23T16:11:45+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2023-09-23T16:11:45+00:00urn:sha1:ab04bd0dd7dd6a818bbac8c5fef9372c4f597352
* Update user-guide with new slangpy features
* More polishing of new slangpy docs
* Update a1-02-slangpy.md
* Only require contiguity for vector element types
* Added `loadOnce/storeOnce` and subscript operations
* Added docs, `DiffTensorView.dims()` & `DiffTensorView.stride(uint)`
* Add constructors, remove storeOnce/loadOnce test
* Adjusted intrinsic definitionsAdd check for contiguous tensors (#3199)2023-09-09T00:12:25+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2023-09-09T00:12:25+00:00urn:sha1:87bb0b503544f1b8c6ec818e25c695b31cda24b7
Otherwise, this can lead to undetected scenario where the strides are incorrect for non-scalar types (`float2`, `float3`, etc..)
Users must call `tensor = tensor.contiguous()` on the inputs to avoid this error.Remove unsupported torch types + add bool type. (#3197)2023-09-08T21:03:14+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2023-09-08T21:03:14+00:00urn:sha1:26a7cf79526b86a3dff4084d42dde8f1a8c9ac1d
Co-authored-by: Yong He <yonghe@outlook.com>Avoid implicit casts or device transfers. (#2992)2023-07-14T22:12:38+00:00Sai Praveen Bangaru31557731+saipraveenb25@users.noreply.github.com2023-07-14T22:12:38+00:00urn:sha1:1b778811dbc1468ed8d1f6f82117017064de2c96Fix native string emit for CUDA/Cpp backend. (#2980)2023-07-12T21:07:10+00:00Yong Heyonghe@outlook.com2023-07-12T21:07:10+00:00urn:sha1:4ed3aafa20b667329f2f9dea94d7c65dc2e80db4
Co-authored-by: Yong He <yhe@nvidia.com>