[llvm] [AMDGPU] Add a sched group mask for LDSDMA instructions (PR #190872)

Shilei Tian via llvm-commits llvm-commits at lists.llvm.org
Wed Apr 8 09:28:09 PDT 2026


================
@@ -1605,6 +1605,7 @@ The AMDGPU backend implements the following LLVM IR intrinsics.
                                                    - 0x0100: All DS read instructions may be scheduled across sched_barrier.
                                                    - 0x0200: All DS write instructions may be scheduled across sched_barrier.
                                                    - 0x0400: All Transcendental (e.g. V_EXP) instructions may be scheduled across sched_barrier.
+                                                   - 0x0800: All LDSDMA instructions may be scheduled across sched_barrier.
----------------
shiltian wrote:

Not gonna say I like this, but apparently downstream code really do and heavily depends on this to build their software pipeline. Something unrelated, but there was even a request that asks the compiler to provide a whole infrastructure to do software pipelining.

https://github.com/llvm/llvm-project/pull/190872


More information about the llvm-commits mailing list