49 lines · plain
1; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=GCN %s2 3; GatherAllAliases gives up on trying to analyze cases where the4; pointer may have been loaded from an aliased store, so make sure5; that this works and allows moving the stores to a better chain to6; allow them to be merged merged when it's clear the pointer is loaded7; from constant/invariant memory.8 9; GCN-LABEL: {{^}}test_merge_store_constant_i16_invariant_global_pointer_load:10; GCN-DAG: buffer_load_dwordx2 [[PTR:v\[[0-9]+:[0-9]+\]]],11; GCN-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0x1c8007b12; GCN: buffer_store_dword [[K]], [[PTR]]13define void @test_merge_store_constant_i16_invariant_global_pointer_load(ptr addrspace(1) dereferenceable(4096) nonnull %in) #0 {14 %ptr = load ptr addrspace(1), ptr addrspace(1) %in, !invariant.load !015 %ptr.1 = getelementptr i16, ptr addrspace(1) %ptr, i64 116 store i16 123, ptr addrspace(1) %ptr, align 417 store i16 456, ptr addrspace(1) %ptr.118 ret void19}20 21; GCN-LABEL: {{^}}test_merge_store_constant_i16_invariant_constant_pointer_load:22; GCN: s_load_dwordx2 s[[[SPTR_LO:[0-9]+]]:[[SPTR_HI:[0-9]+]]]23; GCN: v_mov_b32_e32 [[K:v[0-9]+]], 0x1c8007b24; GCN: buffer_store_dword [[K]], off, s[[[SPTR_LO]]:25define amdgpu_kernel void @test_merge_store_constant_i16_invariant_constant_pointer_load(ptr addrspace(4) dereferenceable(4096) nonnull %in) #0 {26 %ptr = load ptr addrspace(1), ptr addrspace(4) %in, !invariant.load !027 %ptr.1 = getelementptr i16, ptr addrspace(1) %ptr, i64 128 store i16 123, ptr addrspace(1) %ptr, align 429 store i16 456, ptr addrspace(1) %ptr.130 ret void31}32 33; Invariant global load should be equivalently handled to constant.34; GCN-LABEL: {{^}}test_merge_store_global_i16_invariant_uniform_global_pointer_load:35; GCN: s_load_dwordx2 s[[[SPTR_LO:[0-9]+]]:[[SPTR_HI:[0-9]+]]]36; GCN: v_mov_b32_e32 [[K:v[0-9]+]], 0x1c8007b37; GCN: buffer_store_dword [[K]], off, s[[[SPTR_LO]]:38define amdgpu_kernel void @test_merge_store_global_i16_invariant_uniform_global_pointer_load(ptr addrspace(1) dereferenceable(4096) nonnull %in) #0 {39 %ptr = load ptr addrspace(1), ptr addrspace(1) %in, !invariant.load !040 %ptr.1 = getelementptr i16, ptr addrspace(1) %ptr, i64 141 store i16 123, ptr addrspace(1) %ptr, align 442 store i16 456, ptr addrspace(1) %ptr.143 ret void44}45 46!0 = !{}47 48attributes #0 = { nounwind }49