Repository Issue Activity (beta)

#1310 cooperative_groups: block_reduce/block_scan mishandle partial final warps

#1308 Migrate LLVM inline asm semantics to upstream Pliron attributes

#1302 cargo oxide doctor's clang check fails after following its own 'apt install clang-21' remedy

#811 warp_reduce runs a 10-instruction shuffle tree where Ampere's redux.sync does it in one: 4.74x measured, but no way to gate it on the target

#1303 Target feature scan splits LLVM IR on ';', so conjunctive predicates match across the whole module (integer redux + any f32 reads as ReduxF32)

#1300 llc is the pipeline anchor, but nothing checks it against the LLVM that wrote the IR

#337 Tracking: SM89 (Ada Lovelace) device intrinsics — FP8 types, conversions, and MMA

#1298 intrinsics-gen: runtime_validation = "executed" is unreachable for compact admissions (30 guards, no Runtime-stage producer)

#1296 sparse_mma resolver conflates type-kind with metadata mode, blocking admission of imported intrinsics

#1287 core::sync::atomic::AtomicPtr<T> fails to compile in CUDA kernels

#1280 Full debug: support runtime-indexed statement debug locations

#1274 llvm: use upstream GlobalOp constant semantics

#1272 llvm: replace local volatile load/store shim with upstream Pliron support

#1269 CoalescedThreads::ballot returns physical lane bits for sparse groups

#1266 cargo oxide build defaults to sm_80 while cargo oxide run auto-detects the local GPU

#1160 rustlantis: validate projection chains when constructing places

#1217 smoketest: code-shape gates are not applicable to full-debug builds (11 examples)

#1248 Nothing catches a prose reference to a source file that no longer exists

#1162 Rustlantis: support Static and StaticMut operands in custom MIR

#1223 Support grid-constant by-value kernel parameters for TMA descriptors

#1292 Merged PTX bundles collide on .file indices, so lineinfo/debug builds cannot load through the merged loader

#1290 ltoir_cache: the no-symlink open uses x86-64 flag values, inert on aarch64

#1277 `load_all_ptx_bundles_merged` collides on `__shared_mem_N`

#1284 `load()` reads the process executable, so a `#[cuda_module]` compiled into a shared object cannot find its own artifact

#1281 feat(wgmma): add m64n64k32 FP8 E4M3 full-drain path