llvm-project

f557f803 - [CIR][AMDGPU][OpenCL] Classify __hip_atomic_*/__opencl_atomic_* operations (#220856)

[CIR][AMDGPU][OpenCL] Classify __hip_atomic_*/__opencl_atomic_* operations (#220856) load, store, exchange, compare_exchange_strong/weak, and every fetch op were missing from the operand-classification switches, so they fell to the default and reported an unimplemented atomic. They take the same operands as their __c11_atomic_*/__atomic_* counterparts, so they belong with them. This is the ground work for implementing more AMDGPU atomics. --------- Signed-off-by: Steffen Holst Larsen <sholstla@amd.com>