Shared/Local-Memory Load Memory
Read a value from on-chip, workgroup/CTA-scoped shared (local) memory.
Vendor-Neutral Definition
d = *addr, where addr resolves in the on-chip shared/local-data-share memory scoped to the executing workgroup.
⚠️ Cross-vendor mappings describe semantic relationships and are not guaranteed one-to-one compiler
translations.
equivalent with restrictions
Both read from on-chip, workgroup-scoped memory (PTX shared state space / AMDGPU Local Data Share).
| Property | NVIDIA PTX | AMDGPU |
|---|---|---|
| Instruction | ld | ds_read_b32 |
| ISA Layer | Virtual | Native |
| Data Types | b16, b32, b64, b8, f16, f32, f64, s16, s32, s64, s8, u16, u32, u64, u8 | b32 |
| Version / Target Introduced | PTX ISA 1.0 | gfx942/gfx1100 |
Important Differences
- LDS banking, size, and allocation granularity differ per GFX target and are not identical to CUDA shared-memory sizing on any given NVIDIA SM.
Source Evidence
- Parallel Thread Execution ISA ↗ — NVIDIA Corporation
- User Guide for AMDGPU Backend ↗ — LLVM Project
Verification method: documentation