-
Notifications
You must be signed in to change notification settings - Fork 2k
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
- Status: Open.#3393 In NVIDIA/cutlass;
[QST] Why was the elect_one programming model for TMA and bulk copy changed in DSL 4.6?
questionQuestionQuestionStatus: Open.#3391 In NVIDIA/cutlass;- Status: Open.#3390 In NVIDIA/cutlass;
[BUG] "cutlass" in CuTe DSL generated kernel names triggers heavy ptxas -O3 spilling on NVVM 13.1 PTX (sm_100a): 168 -> 80 regs, 4400 B stack frame, +31% kernel time
bugSomething isn't workingSomething isn't workingStatus: Open.#3389 In NVIDIA/cutlass;- Status: Open.#3388 In NVIDIA/cutlass;
[BUG] TensorSSA FP32->FP8 cvt doesn't work
bugSomething isn't workingSomething isn't workingStatus: Open.#3382 In NVIDIA/cutlass;[BUG] Filenames too long trip up windows
bugSomething isn't workingSomething isn't workingStatus: Open.#3368 In NVIDIA/cutlass;[FEA] Interactive kernel selection guide for GEMM configurations
feature requestNew feature or requestNew feature or requestStatus: Open.#3367 In NVIDIA/cutlass;[BUG]
CUTE_DSL_KEEP=sassis ignored in cutlass-dsl 4.6.0bugSomething isn't workingSomething isn't workingStatus: Open.#3365 In NVIDIA/cutlass;[FEA] nvidia-cutlass-dsl-libs-base / nvidia-cutlass-dsl-libs-cu13 release for Windows
feature requestNew feature or requestNew feature or requestStatus: Open.#3361 In NVIDIA/cutlass;[BUG] Versioned DLPack fails when not using tvm-ffi
bugSomething isn't workingSomething isn't workingStatus: Open.#3359 In NVIDIA/cutlass;[BUG] cvt int8/uint8 -> bfloat16 emits a sm_90-only instruction on Ampere/Ada instead of the software fallback (arch gate too loose)
bugSomething isn't workingSomething isn't workingStatus: Open.#3354 In NVIDIA/cutlass;