[AMDGPU] Fix setting nontemporal in memory legalizer (#83815)

Iterator MI can advance in insertWait() but we need original instruction
to set temporal hint. Just move it before handling volatile.

Author: mbrkusanin

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

@@ -2358,6 +2358,11 @@ bool SIGfx12CacheControl::enableVolatileAndOrNonTemporal(
 
   bool Changed = false;
 
+  if (IsNonTemporal) {
+    // Set non-temporal hint for all cache levels.
+    Changed |= setTH(MI, AMDGPU::CPol::TH_NT);
+  }
+
   if (IsVolatile) {
     Changed |= setScope(MI, AMDGPU::CPol::SCOPE_SYS);
 
@@ -2370,11 +2375,6 @@ bool SIGfx12CacheControl::enableVolatileAndOrNonTemporal(
                           Position::AFTER);
   }
 
-  if (IsNonTemporal) {
-    // Set non-temporal hint for all cache levels.
-    Changed |= setTH(MI, AMDGPU::CPol::TH_NT);
-  }
-
   return Changed;
 }
 

llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll

@@ -582,5 +582,170 @@ entry:
   ret void
 }
 
+define amdgpu_kernel void @flat_nontemporal_volatile_load(
+; GFX7-LABEL: flat_nontemporal_volatile_load:
+; GFX7:       ; %bb.0: ; %entry
+; GFX7-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    v_mov_b32_e32 v0, s0
+; GFX7-NEXT:    v_mov_b32_e32 v1, s1
+; GFX7-NEXT:    flat_load_dword v2, v[0:1] glc
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
+; GFX7-NEXT:    v_mov_b32_e32 v0, s2
+; GFX7-NEXT:    v_mov_b32_e32 v1, s3
+; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    flat_store_dword v[0:1], v2
+; GFX7-NEXT:    s_endpgm
+;
+; GFX10-WGP-LABEL: flat_nontemporal_volatile_load:
+; GFX10-WGP:       ; %bb.0: ; %entry
+; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-WGP-NEXT:    flat_load_dword v2, v[0:1] glc dlc
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-WGP-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-WGP-NEXT:    flat_store_dword v[0:1], v2
+; GFX10-WGP-NEXT:    s_endpgm
+;
+; GFX10-CU-LABEL: flat_nontemporal_volatile_load:
+; GFX10-CU:       ; %bb.0: ; %entry
+; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-CU-NEXT:    flat_load_dword v2, v[0:1] glc dlc
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-CU-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-CU-NEXT:    flat_store_dword v[0:1], v2
+; GFX10-CU-NEXT:    s_endpgm
+;
+; SKIP-CACHE-INV-LABEL: flat_nontemporal_volatile_load:
+; SKIP-CACHE-INV:       ; %bb.0: ; %entry
+; SKIP-CACHE-INV-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s0
+; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s1
+; SKIP-CACHE-INV-NEXT:    flat_load_dword v2, v[0:1] glc
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
+; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v0, s2
+; SKIP-CACHE-INV-NEXT:    v_mov_b32_e32 v1, s3
+; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    flat_store_dword v[0:1], v2
+; SKIP-CACHE-INV-NEXT:    s_endpgm
+;
+; GFX90A-NOTTGSPLIT-LABEL: flat_nontemporal_volatile_load:
+; GFX90A-NOTTGSPLIT:       ; %bb.0: ; %entry
+; GFX90A-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX90A-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
+; GFX90A-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
+; GFX90A-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] glc
+; GFX90A-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
+; GFX90A-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
+; GFX90A-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX90A-NOTTGSPLIT-NEXT:    s_endpgm
+;
+; GFX90A-TGSPLIT-LABEL: flat_nontemporal_volatile_load:
+; GFX90A-TGSPLIT:       ; %bb.0: ; %entry
+; GFX90A-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX90A-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
+; GFX90A-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
+; GFX90A-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] glc
+; GFX90A-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
+; GFX90A-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
+; GFX90A-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX90A-TGSPLIT-NEXT:    s_endpgm
+;
+; GFX940-NOTTGSPLIT-LABEL: flat_nontemporal_volatile_load:
+; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
+; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
+; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
+; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc0 sc1
+; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
+; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
+; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
+; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
+;
+; GFX940-TGSPLIT-LABEL: flat_nontemporal_volatile_load:
+; GFX940-TGSPLIT:       ; %bb.0: ; %entry
+; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
+; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
+; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc0 sc1
+; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
+; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
+; GFX940-TGSPLIT-NEXT:    s_endpgm
+;
+; GFX11-WGP-LABEL: flat_nontemporal_volatile_load:
+; GFX11-WGP:       ; %bb.0: ; %entry
+; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1] glc dlc
+; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_endpgm
+;
+; GFX11-CU-LABEL: flat_nontemporal_volatile_load:
+; GFX11-CU:       ; %bb.0: ; %entry
+; GFX11-CU-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX11-CU-NEXT:    flat_load_b32 v2, v[0:1] glc dlc
+; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-CU-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_endpgm
+;
+; GFX12-WGP-LABEL: flat_nontemporal_volatile_load:
+; GFX12-WGP:       ; %bb.0: ; %entry
+; GFX12-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX12-WGP-NEXT:    s_wait_kmcnt 0x0
+; GFX12-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-WGP-NEXT:    flat_load_b32 v2, v[0:1] th:TH_LOAD_NT scope:SCOPE_SYS
+; GFX12-WGP-NEXT:    s_wait_loadcnt 0x0
+; GFX12-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-WGP-NEXT:    s_wait_dscnt 0x0
+; GFX12-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX12-WGP-NEXT:    s_endpgm
+;
+; GFX12-CU-LABEL: flat_nontemporal_volatile_load:
+; GFX12-CU:       ; %bb.0: ; %entry
+; GFX12-CU-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX12-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX12-CU-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
+; GFX12-CU-NEXT:    flat_load_b32 v2, v[0:1] th:TH_LOAD_NT scope:SCOPE_SYS
+; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX12-CU-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-CU-NEXT:    s_wait_dscnt 0x0
+; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX12-CU-NEXT:    s_endpgm
+    ptr %in, ptr %out) {
+entry:
+  %val = load volatile i32, ptr %in, align 4, !nontemporal !0
+  store i32 %val, ptr %out
+  ret void
+}
+
 !0 = !{i32 1}
 declare i32 @llvm.amdgcn.workitem.id.x()

llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll

@@ -576,5 +576,163 @@ entry:
   ret void
 }
 
+define amdgpu_kernel void @global_nontemporal_volatile_load(
+; GFX6-LABEL: global_nontemporal_volatile_load:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX6-NEXT:    s_mov_b32 s7, 0x100f000
+; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    s_mov_b32 s4, s0
+; GFX6-NEXT:    s_mov_b32 s5, s1
+; GFX6-NEXT:    buffer_load_dword v0, off, s[4:7], 0 glc
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
+; GFX6-NEXT:    s_mov_b32 s4, s2
+; GFX6-NEXT:    s_mov_b32 s5, s3
+; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX7-LABEL: global_nontemporal_volatile_load:
+; GFX7:       ; %bb.0: ; %entry
+; GFX7-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    v_mov_b32_e32 v0, s0
+; GFX7-NEXT:    v_mov_b32_e32 v1, s1
+; GFX7-NEXT:    flat_load_dword v2, v[0:1] glc
+; GFX7-NEXT:    s_waitcnt vmcnt(0)
+; GFX7-NEXT:    v_mov_b32_e32 v0, s2
+; GFX7-NEXT:    v_mov_b32_e32 v1, s3
+; GFX7-NEXT:    flat_store_dword v[0:1], v2
+; GFX7-NEXT:    s_endpgm
+;
+; GFX10-WGP-LABEL: global_nontemporal_volatile_load:
+; GFX10-WGP:       ; %bb.0: ; %entry
+; GFX10-WGP-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-WGP-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-WGP-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-WGP-NEXT:    global_load_dword v1, v0, s[0:1] glc dlc
+; GFX10-WGP-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-WGP-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX10-WGP-NEXT:    s_endpgm
+;
+; GFX10-CU-LABEL: global_nontemporal_volatile_load:
+; GFX10-CU:       ; %bb.0: ; %entry
+; GFX10-CU-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX10-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-CU-NEXT:    global_load_dword v1, v0, s[0:1] glc dlc
+; GFX10-CU-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-CU-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX10-CU-NEXT:    s_endpgm
+;
+; SKIP-CACHE-INV-LABEL: global_nontemporal_volatile_load:
+; SKIP-CACHE-INV:       ; %bb.0: ; %entry
+; SKIP-CACHE-INV-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; SKIP-CACHE-INV-NEXT:    s_mov_b32 s7, 0xf000
+; SKIP-CACHE-INV-NEXT:    s_mov_b32 s6, -1
+; SKIP-CACHE-INV-NEXT:    s_waitcnt lgkmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_mov_b32 s4, s0
+; SKIP-CACHE-INV-NEXT:    s_mov_b32 s5, s1
+; SKIP-CACHE-INV-NEXT:    buffer_load_dword v0, off, s[4:7], 0 glc
+; SKIP-CACHE-INV-NEXT:    s_waitcnt vmcnt(0)
+; SKIP-CACHE-INV-NEXT:    s_mov_b32 s4, s2
+; SKIP-CACHE-INV-NEXT:    s_mov_b32 s5, s3
+; SKIP-CACHE-INV-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; SKIP-CACHE-INV-NEXT:    s_endpgm
+;
+; GFX90A-NOTTGSPLIT-LABEL: global_nontemporal_volatile_load:
+; GFX90A-NOTTGSPLIT:       ; %bb.0: ; %entry
+; GFX90A-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX90A-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] glc
+; GFX90A-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX90A-NOTTGSPLIT-NEXT:    s_endpgm
+;
+; GFX90A-TGSPLIT-LABEL: global_nontemporal_volatile_load:
+; GFX90A-TGSPLIT:       ; %bb.0: ; %entry
+; GFX90A-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX90A-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] glc
+; GFX90A-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX90A-TGSPLIT-NEXT:    s_endpgm
+;
+; GFX940-NOTTGSPLIT-LABEL: global_nontemporal_volatile_load:
+; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
+; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
+; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc0 sc1
+; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
+; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
+;
+; GFX940-TGSPLIT-LABEL: global_nontemporal_volatile_load:
+; GFX940-TGSPLIT:       ; %bb.0: ; %entry
+; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
+; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
+; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc0 sc1
+; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
+; GFX940-TGSPLIT-NEXT:    s_endpgm
+;
+; GFX11-WGP-LABEL: global_nontemporal_volatile_load:
+; GFX11-WGP:       ; %bb.0: ; %entry
+; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc dlc
+; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_nop 0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-WGP-NEXT:    s_endpgm
+;
+; GFX11-CU-LABEL: global_nontemporal_volatile_load:
+; GFX11-CU:       ; %bb.0: ; %entry
+; GFX11-CU-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX11-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1] glc dlc
+; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_nop 0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-CU-NEXT:    s_endpgm
+;
+; GFX12-WGP-LABEL: global_nontemporal_volatile_load:
+; GFX12-WGP:       ; %bb.0: ; %entry
+; GFX12-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX12-WGP-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-WGP-NEXT:    s_wait_kmcnt 0x0
+; GFX12-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] th:TH_LOAD_NT scope:SCOPE_SYS
+; GFX12-WGP-NEXT:    s_wait_loadcnt 0x0
+; GFX12-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX12-WGP-NEXT:    s_nop 0
+; GFX12-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-WGP-NEXT:    s_endpgm
+;
+; GFX12-CU-LABEL: global_nontemporal_volatile_load:
+; GFX12-CU:       ; %bb.0: ; %entry
+; GFX12-CU-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX12-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX12-CU-NEXT:    global_load_b32 v1, v0, s[0:1] th:TH_LOAD_NT scope:SCOPE_SYS
+; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX12-CU-NEXT:    s_nop 0
+; GFX12-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-CU-NEXT:    s_endpgm
+    ptr addrspace(1) %in, ptr addrspace(1) %out) {
+entry:
+  %val = load volatile i32, ptr addrspace(1) %in, align 4, !nontemporal !0
+  store i32 %val, ptr addrspace(1) %out
+  ret void
+}
+
 !0 = !{i32 1}
 declare i32 @llvm.amdgcn.workitem.id.x()