AMDGPU: Add more tests for atomicrmw handling

Add agent scope copies of atomicrmw atomics tests.
Expand testing for the undo identity atomicrmw case.
Test 16-bit atomic expansions.

Author: arsenm

llvm/test/CodeGen/AMDGPU/idemponent-atomics.ll

@@ -76,6 +76,26 @@ entry:
   ret i32 %val
 }
 
+define i32 @global_agent_acquire_release_idempotent_or__no_fine_grained(ptr addrspace(1) %in) {
+; GFX940-LABEL: global_agent_acquire_release_idempotent_or__no_fine_grained:
+; GFX940:       ; %bb.0: ; %entry
+; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX940-NEXT:    v_mov_b32_e32 v2, 0
+; GFX940-NEXT:    buffer_wbl2 sc1
+; GFX940-NEXT:    global_atomic_or v0, v[0:1], v2, off sc0
+; GFX940-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NEXT:    buffer_inv sc1
+; GFX940-NEXT:    s_setpc_b64 s[30:31]
+; OPT-LABEL: @global_agent_acquire_release_idempotent_or__no_fine_grained(
+; OPT-NEXT:  entry:
+; OPT-NEXT:    [[VAL:%.*]] = atomicrmw or ptr addrspace(1) [[IN:%.*]], i32 0 syncscope("agent-one-as") acq_rel, align 4, !amdgpu.no.fine.grained.memory [[META0:![0-9]+]]
+; OPT-NEXT:    ret i32 [[VAL]]
+;
+entry:
+  %val = atomicrmw or ptr addrspace(1) %in, i32 0 syncscope("agent-one-as") acq_rel, align 4, !amdgpu.no.fine.grained.memory !0
+  ret i32 %val
+}
+
 define i32 @global_agent_seq_cst_idempotent_or(ptr addrspace(1) %in) {
 ; GFX940-LABEL: global_agent_seq_cst_idempotent_or:
 ; GFX940:       ; %bb.0: ; %entry
@@ -113,6 +133,23 @@ entry:
   ret i32 %val
 }
 
+define i32 @global_agent_monotonic_idempotent_add__no_fine_grained(ptr addrspace(1) %in) {
+; GFX940-LABEL: global_agent_monotonic_idempotent_add__no_fine_grained:
+; GFX940:       ; %bb.0: ; %entry
+; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX940-NEXT:    global_load_dword v0, v[0:1], off sc0
+; GFX940-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NEXT:    s_setpc_b64 s[30:31]
+; OPT-LABEL: @global_agent_monotonic_idempotent_add__no_fine_grained(
+; OPT-NEXT:  entry:
+; OPT-NEXT:    [[VAL:%.*]] = load atomic i32, ptr addrspace(1) [[IN:%.*]] syncscope("workgroup") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META0]]
+; OPT-NEXT:    ret i32 [[VAL]]
+;
+entry:
+  %val = atomicrmw add ptr addrspace(1) %in, i32 0 syncscope("workgroup") monotonic, align 4, !amdgpu.no.fine.grained.memory !0
+  ret i32 %val
+}
+
 define i32 @global_agent_monotonic_idempotent_sub(ptr addrspace(1) %in) {
 ; GFX940-LABEL: global_agent_monotonic_idempotent_sub:
 ; GFX940:       ; %bb.0: ; %entry
@@ -130,6 +167,23 @@ entry:
   ret i32 %val
 }
 
+define i32 @global_agent_monotonic_idempotent_sub__no_fine_grained(ptr addrspace(1) %in) {
+; GFX940-LABEL: global_agent_monotonic_idempotent_sub__no_fine_grained:
+; GFX940:       ; %bb.0: ; %entry
+; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX940-NEXT:    global_load_dword v0, v[0:1], off
+; GFX940-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NEXT:    s_setpc_b64 s[30:31]
+; OPT-LABEL: @global_agent_monotonic_idempotent_sub__no_fine_grained(
+; OPT-NEXT:  entry:
+; OPT-NEXT:    [[VAL:%.*]] = load atomic i32, ptr addrspace(1) [[IN:%.*]] syncscope("wavefront") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META0]]
+; OPT-NEXT:    ret i32 [[VAL]]
+;
+entry:
+  %val = atomicrmw sub ptr addrspace(1) %in, i32 0 syncscope("wavefront") monotonic, align 4, !amdgpu.no.fine.grained.memory !0
+  ret i32 %val
+}
+
 define i32 @global_system_monotonic_idempotent_xor(ptr addrspace(1) %in) {
 ; GFX940-LABEL: global_system_monotonic_idempotent_xor:
 ; GFX940:       ; %bb.0: ; %entry
@@ -147,6 +201,23 @@ entry:
   ret i32 %val
 }
 
+define i32 @global_system_monotonic_idempotent_xor__no_fine_grained(ptr addrspace(1) %in) {
+; GFX940-LABEL: global_system_monotonic_idempotent_xor__no_fine_grained:
+; GFX940:       ; %bb.0: ; %entry
+; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX940-NEXT:    global_load_dword v0, v[0:1], off sc0 sc1
+; GFX940-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NEXT:    s_setpc_b64 s[30:31]
+; OPT-LABEL: @global_system_monotonic_idempotent_xor__no_fine_grained(
+; OPT-NEXT:  entry:
+; OPT-NEXT:    [[VAL:%.*]] = load atomic i32, ptr addrspace(1) [[IN:%.*]] monotonic, align 4, !amdgpu.no.fine.grained.memory [[META0]]
+; OPT-NEXT:    ret i32 [[VAL]]
+;
+entry:
+  %val = atomicrmw xor ptr addrspace(1) %in, i32 0 monotonic, align 4, !amdgpu.no.fine.grained.memory !0
+  ret i32 %val
+}
+
 define i32 @global_agent_monotonic_idempotent_and(ptr addrspace(1) %in) {
 ; GFX940-LABEL: global_agent_monotonic_idempotent_and:
 ; GFX940:       ; %bb.0: ; %entry
@@ -163,3 +234,22 @@ entry:
   %val = atomicrmw and ptr addrspace(1) %in, i32 -1 syncscope("singlethread") monotonic, align 4
   ret i32 %val
 }
+
+define i32 @global_agent_monotonic_idempotent_and_no_fined_grain(ptr addrspace(1) %in) {
+; GFX940-LABEL: global_agent_monotonic_idempotent_and_no_fined_grain:
+; GFX940:       ; %bb.0: ; %entry
+; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX940-NEXT:    global_load_dword v0, v[0:1], off
+; GFX940-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NEXT:    s_setpc_b64 s[30:31]
+; OPT-LABEL: @global_agent_monotonic_idempotent_and_no_fined_grain(
+; OPT-NEXT:  entry:
+; OPT-NEXT:    [[VAL:%.*]] = load atomic i32, ptr addrspace(1) [[IN:%.*]] syncscope("singlethread") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META0]]
+; OPT-NEXT:    ret i32 [[VAL]]
+;
+entry:
+  %val = atomicrmw and ptr addrspace(1) %in, i32 -1 syncscope("singlethread") monotonic, align 4, !amdgpu.no.fine.grained.memory !0
+  ret i32 %val
+}
+
+!0 = !{}