[AMDGPU] Select VMEM prefetch for llvm.prefetch on gfx1250 (#150493)

We have a choice to use a scalar or vector prefetch for an uniform
pointer. Since we do not have scalar stores our scalar cache is
practically readonly. The rw argument of the prefetch intrinsic is
used to force vector operation even for an uniform case. On GFX12
scalar prefetch will be used anyway, it is still useful but it will
only bring data to L2.

2 files changed, 905 insertions, 33 deletions

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.prefetch.ll b/llvm/test/CodeGen/AMDGPU/llvm.prefetch.ll
index 429b3b8..6e24a6a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.prefetch.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.prefetch.ll
@@ -1,36 +1,54 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GCN,GFX1250,GL2-ONLY %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefixes=GCN,SPREFETCH,GFX1250-SPREFETCH,GFX1250-SPREFETCH-SDAG %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 -mattr=+safe-cu-prefetch < %s | FileCheck --check-prefixes=GCN,GFX1250,SAFE-CU %s
 ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GCN,NOSPREFETCH %s
-; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefixes=GCN,SPREFETCH,SPREFETCH-SDAG %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefixes=GCN,SPREFETCH,GFX12-SPREFETCH,SPREFETCH-SDAG %s
 ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck --check-prefixes=GCN,NOSPREFETCH %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GCN,GFX1250,GL2-ONLY %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefixes=GCN,SPREFETCH,GFX1250-SPREFETCH,GFX1250-SPREFETCH-GISEL %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 -mattr=+safe-cu-prefetch < %s | FileCheck --check-prefixes=GCN,GFX1250,SAFE-CU %s
 ; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GCN,NOSPREFETCH %s
-; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefixes=GCN,SPREFETCH,SPREFETCH-GISEL %s
+; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefixes=GCN,SPREFETCH,GFX12-SPREFETCH,SPREFETCH-GISEL %s
 ; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck --check-prefixes=GCN,NOSPREFETCH %s
 
 ; Scalar data prefetch
 
 define amdgpu_ps void @prefetch_data_sgpr(ptr addrspace(4) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_data_sgpr:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_data_sgpr:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_data_sgpr:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   tail call void @llvm.prefetch.p4(ptr addrspace(4) %ptr, i32 0, i32 0, i32 1)
   ret void
 }
 
 define amdgpu_ps void @prefetch_data_sgpr_offset(ptr addrspace(4) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_data_sgpr_offset:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_data_sgpr_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] offset:512 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_data_sgpr_offset:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x200, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   %gep = getelementptr float, ptr addrspace(4) %ptr, i32 128
   tail call void @llvm.prefetch.p4(ptr addrspace(4) %gep, i32 0, i32 0, i32 1)
@@ -40,14 +58,20 @@ entry:
 ; Check large offsets
 
 define amdgpu_ps void @prefetch_data_sgpr_max_offset(ptr addrspace(4) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_data_sgpr_max_offset:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_data_sgpr_max_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] offset:8388607 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_data_sgpr_max_offset:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x7fffff, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_max_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   %gep = getelementptr i8, ptr addrspace(4) %ptr, i32 8388607
   tail call void @llvm.prefetch.p4(ptr addrspace(4) %gep, i32 0, i32 0, i32 1)
@@ -55,6 +79,20 @@ entry:
 }
 
 define amdgpu_ps void @prefetch_data_sgpr_min_offset(ptr addrspace(4) inreg %ptr) {
+; GFX1250-LABEL: prefetch_data_sgpr_min_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] offset:-8388608 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-SDAG-LABEL: prefetch_data_sgpr_min_offset:
+; GFX1250-SPREFETCH-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_mov_b64 s[2:3], lit64(0xffffffffff800000)
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_add_nc_u64 s[0:1], s[0:1], s[2:3]
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_endpgm
+;
 ; NOSPREFETCH-LABEL: prefetch_data_sgpr_min_offset:
 ; NOSPREFETCH:       ; %bb.0: ; %entry
 ; NOSPREFETCH-NEXT:    s_endpgm
@@ -68,6 +106,13 @@ define amdgpu_ps void @prefetch_data_sgpr_min_offset(ptr addrspace(4) inreg %ptr
 ; SPREFETCH-SDAG-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
 ; SPREFETCH-SDAG-NEXT:    s_endpgm
 ;
+; GFX1250-SPREFETCH-GISEL-LABEL: prefetch_data_sgpr_min_offset:
+; GFX1250-SPREFETCH-GISEL:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0xff800000
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_ci_u32 s1, s1, -1
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_endpgm
+;
 ; SPREFETCH-GISEL-LABEL: prefetch_data_sgpr_min_offset:
 ; SPREFETCH-GISEL:       ; %bb.0: ; %entry
 ; SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0xff800000
@@ -81,6 +126,18 @@ entry:
 }
 
 define amdgpu_ps void @prefetch_data_sgpr_too_large_offset(ptr addrspace(4) inreg %ptr) {
+; GFX1250-LABEL: prefetch_data_sgpr_too_large_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0x800000
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-SDAG-LABEL: prefetch_data_sgpr_too_large_offset:
+; GFX1250-SPREFETCH-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 0x800000
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_endpgm
+;
 ; NOSPREFETCH-LABEL: prefetch_data_sgpr_too_large_offset:
 ; NOSPREFETCH:       ; %bb.0: ; %entry
 ; NOSPREFETCH-NEXT:    s_endpgm
@@ -91,6 +148,13 @@ define amdgpu_ps void @prefetch_data_sgpr_too_large_offset(ptr addrspace(4) inre
 ; SPREFETCH-SDAG-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
 ; SPREFETCH-SDAG-NEXT:    s_endpgm
 ;
+; GFX1250-SPREFETCH-GISEL-LABEL: prefetch_data_sgpr_too_large_offset:
+; GFX1250-SPREFETCH-GISEL:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0x800000
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_ci_u32 s1, s1, 0
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_endpgm
+;
 ; SPREFETCH-GISEL-LABEL: prefetch_data_sgpr_too_large_offset:
 ; SPREFETCH-GISEL:       ; %bb.0: ; %entry
 ; SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0x800000
@@ -105,15 +169,113 @@ entry:
 
 ; Check divergent address
 
-define amdgpu_ps void @prefetch_data_vgpr(ptr addrspace(1) %ptr) {
-; GCN-LABEL: prefetch_data_vgpr:
-; GCN:       ; %bb.0: ; %entry
-; GCN-NEXT:    s_endpgm
+define amdgpu_ps void @prefetch_data_vgpr_global(ptr addrspace(1) %ptr) {
+; GFX1250-LABEL: prefetch_data_vgpr_global:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    global_prefetch_b8 v[0:1], off scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_global:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v[0:1], off scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_global:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_global:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
 entry:
   tail call void @llvm.prefetch.p1(ptr addrspace(1) %ptr, i32 0, i32 0, i32 1)
   ret void
 }
 
+define amdgpu_ps void @prefetch_data_vgpr_flat(ptr %ptr) {
+; GFX1250-LABEL: prefetch_data_vgpr_flat:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_flat:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_flat:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_flat:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.pf(ptr %ptr, i32 0, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_sgpr_vgpr_offset_global(ptr addrspace(1) inreg %ptr, i32 %offset) {
+; GFX1250-LABEL: prefetch_data_sgpr_vgpr_offset_global:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_sgpr_vgpr_offset_global:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_vgpr_offset_global:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_sgpr_vgpr_offset_global:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+; GFX12-LABEL: prefetch_data_sgpr_vgpr_offset_global:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    s_endpgm
+; GFX11-LABEL: prefetch_data_sgpr_vgpr_offset_global:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_endpgm
+entry:
+  %gep = getelementptr i8, ptr addrspace(1) %ptr, i32 %offset
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %gep, i32 0, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_sgpr_vgpr_offset_flat(ptr inreg %ptr, i32 %offset) {
+; GFX1250-LABEL: prefetch_data_sgpr_vgpr_offset_flat:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    flat_prefetch_b8 v0, s[0:1] offset:128 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_sgpr_vgpr_offset_flat:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v0, s[0:1] offset:128 scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_vgpr_offset_flat:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_sgpr_vgpr_offset_flat:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+; GFX12-LABEL: prefetch_data_sgpr_vgpr_offset_flat:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    s_endpgm
+; GFX11-LABEL: prefetch_data_sgpr_vgpr_offset_flat:
+; GFX11:       ; %bb.0: ; %entry
+; GFX11-NEXT:    s_endpgm
+entry:
+  %gep1 = getelementptr i8, ptr %ptr, i32 %offset
+  %gep2 = getelementptr i8, ptr %gep1, i32 128
+  tail call void @llvm.prefetch.pf(ptr %gep2, i32 0, i32 0, i32 1)
+  ret void
+}
+
 ; Check LDS and Scratch, we cannot prefetch it
 
 define amdgpu_ps void @prefetch_data_lds(ptr addrspace(3) inreg %ptr) {
@@ -137,43 +299,59 @@ entry:
 ; Check supported address spaces
 
 define amdgpu_ps void @prefetch_data_sgpr_flat(ptr inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_data_sgpr_flat:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_data_sgpr_flat:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    flat_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_data_sgpr_flat:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_flat:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   tail call void @llvm.prefetch.pf(ptr %ptr, i32 0, i32 0, i32 1)
   ret void
 }
 
 define amdgpu_ps void @prefetch_data_sgpr_global(ptr addrspace(1) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_data_sgpr_global:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_data_sgpr_global:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_data_sgpr_global:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_global:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   tail call void @llvm.prefetch.p1(ptr addrspace(1) %ptr, i32 0, i32 0, i32 1)
   ret void
 }
 
 define amdgpu_ps void @prefetch_data_sgpr_constant_32bit(ptr addrspace(6) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_data_sgpr_constant_32bit:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_data_sgpr_constant_32bit:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_data_sgpr_constant_32bit:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_mov_b32 s1, 0
 ; SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_constant_32bit:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   tail call void @llvm.prefetch.p6(ptr addrspace(6) %ptr, i32 0, i32 0, i32 1)
   ret void
@@ -182,28 +360,36 @@ entry:
 ; I$ prefetch
 
 define amdgpu_ps void @prefetch_inst_sgpr(ptr addrspace(4) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_inst_sgpr:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_inst_sgpr:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_inst_sgpr:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_inst_sgpr:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   tail call void @llvm.prefetch.p4(ptr addrspace(4) %ptr, i32 0, i32 0, i32 0)
   ret void
 }
 
 define amdgpu_ps void @prefetch_inst_sgpr_offset(ptr addrspace(4) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_inst_sgpr_offset:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_inst_sgpr_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_inst_sgpr_offset:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_inst s[0:1], 0x80, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_inst_sgpr_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   %gep = getelementptr i8, ptr addrspace(4) %ptr, i32 128
   tail call void @llvm.prefetch.p4(ptr addrspace(4) %gep, i32 0, i32 0, i32 0)
@@ -213,14 +399,18 @@ entry:
 ; Check large offsets
 
 define amdgpu_ps void @prefetch_inst_sgpr_max_offset(ptr addrspace(4) inreg %ptr) {
-; NOSPREFETCH-LABEL: prefetch_inst_sgpr_max_offset:
-; NOSPREFETCH:       ; %bb.0: ; %entry
-; NOSPREFETCH-NEXT:    s_endpgm
+; GFX1250-LABEL: prefetch_inst_sgpr_max_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_endpgm
 ;
 ; SPREFETCH-LABEL: prefetch_inst_sgpr_max_offset:
 ; SPREFETCH:       ; %bb.0: ; %entry
 ; SPREFETCH-NEXT:    s_prefetch_inst s[0:1], 0x7fffff, null, 0
 ; SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_inst_sgpr_max_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
 entry:
   %gep = getelementptr i8, ptr addrspace(4) %ptr, i32 8388607
   tail call void @llvm.prefetch.p4(ptr addrspace(4) %gep, i32 0, i32 0, i32 0)
@@ -228,6 +418,18 @@ entry:
 }
 
 define amdgpu_ps void @prefetch_inst_sgpr_min_offset(ptr addrspace(4) inreg %ptr) {
+; GFX1250-LABEL: prefetch_inst_sgpr_min_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-SDAG-LABEL: prefetch_inst_sgpr_min_offset:
+; GFX1250-SPREFETCH-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_mov_b64 s[2:3], lit64(0xffffffffff800000)
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_add_nc_u64 s[0:1], s[0:1], s[2:3]
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_endpgm
+;
 ; NOSPREFETCH-LABEL: prefetch_inst_sgpr_min_offset:
 ; NOSPREFETCH:       ; %bb.0: ; %entry
 ; NOSPREFETCH-NEXT:    s_endpgm
@@ -241,6 +443,13 @@ define amdgpu_ps void @prefetch_inst_sgpr_min_offset(ptr addrspace(4) inreg %ptr
 ; SPREFETCH-SDAG-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
 ; SPREFETCH-SDAG-NEXT:    s_endpgm
 ;
+; GFX1250-SPREFETCH-GISEL-LABEL: prefetch_inst_sgpr_min_offset:
+; GFX1250-SPREFETCH-GISEL:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0xff800000
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_ci_u32 s1, s1, -1
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_endpgm
+;
 ; SPREFETCH-GISEL-LABEL: prefetch_inst_sgpr_min_offset:
 ; SPREFETCH-GISEL:       ; %bb.0: ; %entry
 ; SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0xff800000
@@ -254,6 +463,16 @@ entry:
 }
 
 define amdgpu_ps void @prefetch_inst_sgpr_too_large_offset(ptr addrspace(4) inreg %ptr) {
+; GFX1250-LABEL: prefetch_inst_sgpr_too_large_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-SDAG-LABEL: prefetch_inst_sgpr_too_large_offset:
+; GFX1250-SPREFETCH-SDAG:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 0x800000
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-SDAG-NEXT:    s_endpgm
+;
 ; NOSPREFETCH-LABEL: prefetch_inst_sgpr_too_large_offset:
 ; NOSPREFETCH:       ; %bb.0: ; %entry
 ; NOSPREFETCH-NEXT:    s_endpgm
@@ -264,6 +483,13 @@ define amdgpu_ps void @prefetch_inst_sgpr_too_large_offset(ptr addrspace(4) inre
 ; SPREFETCH-SDAG-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
 ; SPREFETCH-SDAG-NEXT:    s_endpgm
 ;
+; GFX1250-SPREFETCH-GISEL-LABEL: prefetch_inst_sgpr_too_large_offset:
+; GFX1250-SPREFETCH-GISEL:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0x800000
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_add_co_ci_u32 s1, s1, 0
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_prefetch_inst s[0:1], 0x0, null, 0
+; GFX1250-SPREFETCH-GISEL-NEXT:    s_endpgm
+;
 ; SPREFETCH-GISEL-LABEL: prefetch_inst_sgpr_too_large_offset:
 ; SPREFETCH-GISEL:       ; %bb.0: ; %entry
 ; SPREFETCH-GISEL-NEXT:    s_add_co_u32 s0, s0, 0x800000
@@ -276,6 +502,282 @@ entry:
   ret void
 }
 
+; Check cache locality
+
+define amdgpu_ps void @prefetch_data_vgpr_flat_dev(ptr %ptr) {
+; GFX1250-LABEL: prefetch_data_vgpr_flat_dev:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_DEV
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_flat_dev:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_DEV
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_flat_dev:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_flat_dev:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.pf(ptr %ptr, i32 0, i32 1, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_vgpr_flat_se(ptr %ptr) {
+; GFX1250-LABEL: prefetch_data_vgpr_flat_se:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_SE
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_flat_se:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_SE
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_flat_se:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_flat_se:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.pf(ptr %ptr, i32 0, i32 2, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_vgpr_flat_cu(ptr %ptr) {
+; GL2-ONLY-LABEL: prefetch_data_vgpr_flat_cu:
+; GL2-ONLY:       ; %bb.0: ; %entry
+; GL2-ONLY-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_SE
+; GL2-ONLY-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_flat_cu:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v[0:1] scope:SCOPE_SE
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; SAFE-CU-LABEL: prefetch_data_vgpr_flat_cu:
+; SAFE-CU:       ; %bb.0: ; %entry
+; SAFE-CU-NEXT:    flat_prefetch_b8 v[0:1]
+; SAFE-CU-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_flat_cu:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_flat_cu:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.pf(ptr %ptr, i32 0, i32 3, i32 1)
+  ret void
+}
+
+; flat offset
+
+define amdgpu_ps void @prefetch_data_vgpr_flat_offset(ptr %ptr) {
+; GFX1250-LABEL: prefetch_data_vgpr_flat_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    flat_prefetch_b8 v[0:1] offset:512 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_flat_offset:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v[0:1] offset:512 scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_flat_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_flat_offset:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  %gep = getelementptr float, ptr %ptr, i32 128
+  tail call void @llvm.prefetch.pf(ptr %gep, i32 0, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_vgpr_global_offset(ptr addrspace(1) %ptr) {
+; GFX1250-LABEL: prefetch_data_vgpr_global_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    global_prefetch_b8 v[0:1], off offset:512 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_global_offset:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v[0:1], off offset:512 scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_global_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_global_offset:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  %gep = getelementptr float, ptr addrspace(1) %ptr, i32 128
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %gep, i32 0, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_vgpr_global_saddr(ptr addrspace(1) inreg %ptr, i32 %voffset) {
+; GFX1250-LABEL: prefetch_data_vgpr_global_saddr:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_global_saddr:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_global_saddr:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_global_saddr:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  %gep = getelementptr i8, ptr addrspace(1) %ptr, i32 %voffset
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %gep, i32 0, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_vgpr_global_saddr_offset(ptr addrspace(1) inreg %ptr, i32 %voffset) {
+; GFX1250-LABEL: prefetch_data_vgpr_global_saddr_offset:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] offset:128 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_vgpr_global_saddr_offset:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v0, s[0:1] offset:128 scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_vgpr_global_saddr_offset:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_vgpr_global_saddr_offset:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  %gep1 = getelementptr i8, ptr addrspace(1) %ptr, i32 %voffset
+  %gep2 = getelementptr i8, ptr addrspace(1) %gep1, i32 128
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %gep2, i32 0, i32 0, i32 1)
+  ret void
+}
+
+; Cannot prefetch I$ with flat or global instructions.
+
+define amdgpu_ps void @prefetch_inst_vgpr_global(ptr addrspace(1) %ptr) {
+; GCN-LABEL: prefetch_inst_vgpr_global:
+; GCN:       ; %bb.0: ; %entry
+; GCN-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %ptr, i32 0, i32 0, i32 0)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_inst_vgpr_flat(ptr %ptr) {
+; GCN-LABEL: prefetch_inst_vgpr_flat:
+; GCN:       ; %bb.0: ; %entry
+; GCN-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.pf(ptr %ptr, i32 0, i32 0, i32 0)
+  ret void
+}
+
+; Force vector prefetch for uniform address with rw = 1 argument.
+
+define amdgpu_ps void @prefetch_data_sgpr_flat_force_vector(ptr inreg %ptr) {
+; GFX1250-LABEL: prefetch_data_sgpr_flat_force_vector:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    flat_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_sgpr_flat_force_vector:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-SPREFETCH-NEXT:    flat_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_flat_force_vector:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_sgpr_flat_force_vector:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.pf(ptr %ptr, i32 1, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_sgpr_global_force_vector(ptr addrspace(1) inreg %ptr) {
+; GFX1250-LABEL: prefetch_data_sgpr_global_force_vector:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_sgpr_global_force_vector:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v0, s[0:1] scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_global_force_vector:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_sgpr_global_force_vector:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x0, null, 0
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %ptr, i32 1, i32 0, i32 1)
+  ret void
+}
+
+define amdgpu_ps void @prefetch_data_sgpr_global_saddr_force_vector(ptr addrspace(1) inreg %ptr) {
+; GFX1250-LABEL: prefetch_data_sgpr_global_saddr_force_vector:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[0:1] offset:1024 scope:SCOPE_SYS
+; GFX1250-NEXT:    s_endpgm
+;
+; GFX1250-SPREFETCH-LABEL: prefetch_data_sgpr_global_saddr_force_vector:
+; GFX1250-SPREFETCH:       ; %bb.0: ; %entry
+; GFX1250-SPREFETCH-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-SPREFETCH-NEXT:    global_prefetch_b8 v0, s[0:1] offset:1024 scope:SCOPE_SYS
+; GFX1250-SPREFETCH-NEXT:    s_endpgm
+;
+; NOSPREFETCH-LABEL: prefetch_data_sgpr_global_saddr_force_vector:
+; NOSPREFETCH:       ; %bb.0: ; %entry
+; NOSPREFETCH-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: prefetch_data_sgpr_global_saddr_force_vector:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_prefetch_data s[0:1], 0x400, null, 0
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+entry:
+  %gep = getelementptr i8, ptr addrspace(1) %ptr, i32 1024
+  tail call void @llvm.prefetch.p1(ptr addrspace(1) %gep, i32 1, i32 0, i32 1)
+  ret void
+}
+
 declare void @llvm.prefetch.pf(ptr nocapture readonly, i32, i32, i32)
 declare void @llvm.prefetch.p1(ptr addrspace(1) nocapture readonly, i32, i32, i32)
 declare void @llvm.prefetch.p3(ptr addrspace(3) nocapture readonly, i32, i32, i32)
diff --git a/llvm/test/CodeGen/AMDGPU/loop-prefetch-data.ll b/llvm/test/CodeGen/AMDGPU/loop-prefetch-data.ll
index 874dece..1e6b77e 100644
--- a/llvm/test/CodeGen/AMDGPU/loop-prefetch-data.ll
+++ b/llvm/test/CodeGen/AMDGPU/loop-prefetch-data.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -amdgpu-loop-prefetch < %s | FileCheck --check-prefix=GFX12 %s
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -amdgpu-loop-prefetch -mattr=+safe-smem-prefetch < %s | FileCheck --check-prefix=GFX12-SPREFETCH %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1250 -amdgpu-loop-prefetch < %s | FileCheck --check-prefix=GFX1250 %s
 
 define amdgpu_kernel void @copy_flat(ptr nocapture %d, ptr nocapture readonly %s, i32 %n) {
 ; GFX12-LABEL: copy_flat:
@@ -55,6 +56,33 @@ define amdgpu_kernel void @copy_flat(ptr nocapture %d, ptr nocapture readonly %s
 ; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB0_2
 ; GFX12-SPREFETCH-NEXT:  .LBB0_3: ; %for.end
 ; GFX12-SPREFETCH-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: copy_flat:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_load_b32 s6, s[4:5], 0x34
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_cmp_eq_u32 s6, 0
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB0_3
+; GFX1250-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX1250-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[2:3], s[2:3], 0xb0
+; GFX1250-NEXT:  .LBB0_2: ; %for.body
+; GFX1250-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX1250-NEXT:    flat_load_b128 v[2:5], v0, s[2:3] offset:-176
+; GFX1250-NEXT:    flat_prefetch_b8 v0, s[2:3] scope:SCOPE_SE
+; GFX1250-NEXT:    s_add_co_i32 s6, s6, -1
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[2:3], s[2:3], 16
+; GFX1250-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-NEXT:    flat_store_b128 v0, v[2:5], s[0:1]
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 16
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB0_2
+; GFX1250-NEXT:  .LBB0_3: ; %for.end
+; GFX1250-NEXT:    s_endpgm
 entry:
   %cmp6.not = icmp eq i32 %n, 0
   br i1 %cmp6.not, label %for.end, label %for.body
@@ -123,6 +151,33 @@ define amdgpu_kernel void @copy_global(ptr addrspace(1) nocapture %d, ptr addrsp
 ; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB1_2
 ; GFX12-SPREFETCH-NEXT:  .LBB1_3: ; %for.end
 ; GFX12-SPREFETCH-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: copy_global:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_load_b32 s6, s[4:5], 0x34
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_cmp_eq_u32 s6, 0
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB1_3
+; GFX1250-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX1250-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[2:3], s[2:3], 0xb0
+; GFX1250-NEXT:  .LBB1_2: ; %for.body
+; GFX1250-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX1250-NEXT:    global_load_b128 v[2:5], v0, s[2:3] offset:-176
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[2:3] scope:SCOPE_SE
+; GFX1250-NEXT:    s_add_co_i32 s6, s6, -1
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[2:3], s[2:3], 16
+; GFX1250-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX1250-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-NEXT:    global_store_b128 v0, v[2:5], s[0:1]
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 16
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB1_2
+; GFX1250-NEXT:  .LBB1_3: ; %for.end
+; GFX1250-NEXT:    s_endpgm
 entry:
   %cmp6.not = icmp eq i32 %n, 0
   br i1 %cmp6.not, label %for.end, label %for.body
@@ -193,6 +248,34 @@ define amdgpu_kernel void @copy_constant(ptr addrspace(1) nocapture %d, ptr addr
 ; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB2_2
 ; GFX12-SPREFETCH-NEXT:  .LBB2_3: ; %for.end
 ; GFX12-SPREFETCH-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: copy_constant:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_load_b32 s6, s[4:5], 0x34
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_cmp_eq_u32 s6, 0
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB2_3
+; GFX1250-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX1250-NEXT:    s_load_b128 s[0:3], s[4:5], 0x24
+; GFX1250-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-NEXT:  .LBB2_2: ; %for.body
+; GFX1250-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    global_prefetch_b8 v0, s[2:3] offset:176 scope:SCOPE_SE
+; GFX1250-NEXT:    s_load_b128 s[8:11], s[2:3], 0x0
+; GFX1250-NEXT:    s_add_co_i32 s6, s6, -1
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[2:3], s[2:3], 16
+; GFX1250-NEXT:    s_cmp_lg_u32 s6, 0
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], s[8:9]
+; GFX1250-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX1250-NEXT:    global_store_b128 v0, v[2:5], s[0:1]
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 16
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB2_2
+; GFX1250-NEXT:  .LBB2_3: ; %for.end
+; GFX1250-NEXT:    s_endpgm
 entry:
   %cmp6.not = icmp eq i32 %n, 0
   br i1 %cmp6.not, label %for.end, label %for.body
@@ -262,6 +345,29 @@ define amdgpu_kernel void @copy_local(ptr addrspace(3) nocapture %d, ptr addrspa
 ; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB3_1
 ; GFX12-SPREFETCH-NEXT:  .LBB3_2: ; %for.end
 ; GFX12-SPREFETCH-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: copy_local:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_load_b96 s[0:2], s[4:5], 0x24
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_cmp_eq_u32 s2, 0
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB3_2
+; GFX1250-NEXT:  .LBB3_1: ; %for.body
+; GFX1250-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX1250-NEXT:    v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v4, s0
+; GFX1250-NEXT:    s_add_co_i32 s2, s2, -1
+; GFX1250-NEXT:    s_add_co_i32 s0, s0, 16
+; GFX1250-NEXT:    s_add_co_i32 s1, s1, 16
+; GFX1250-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset0:2 offset1:3
+; GFX1250-NEXT:    ds_load_2addr_b32 v[2:3], v2 offset1:1
+; GFX1250-NEXT:    s_cmp_lg_u32 s2, 0
+; GFX1250-NEXT:    s_wait_dscnt 0x1
+; GFX1250-NEXT:    ds_store_2addr_b32 v4, v0, v1 offset0:2 offset1:3
+; GFX1250-NEXT:    s_wait_dscnt 0x1
+; GFX1250-NEXT:    ds_store_2addr_b32 v4, v2, v3 offset1:1
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB3_1
+; GFX1250-NEXT:  .LBB3_2: ; %for.end
+; GFX1250-NEXT:    s_endpgm
 entry:
   %cmp6.not = icmp eq i32 %n, 0
   br i1 %cmp6.not, label %for.end, label %for.body
@@ -280,3 +386,267 @@ for.body:                                         ; preds = %entry, %for.body
 for.end:                                          ; preds = %for.body, %entry
   ret void
 }
+
+define amdgpu_kernel void @copy_flat_divergent(ptr nocapture %d, ptr nocapture readonly %s, i32 %n) {
+; GFX12-LABEL: copy_flat_divergent:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    s_load_b32 s0, s[4:5], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_cmp_eq_u32 s0, 0
+; GFX12-NEXT:    s_cbranch_scc1 .LBB4_3
+; GFX12-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX12-NEXT:    s_load_b128 s[4:7], s[4:5], 0x24
+; GFX12-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_add_co_u32 v2, s1, s6, v0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v3, null, s7, 0, s1
+; GFX12-NEXT:    v_add_co_u32 v0, s1, s4, v0
+; GFX12-NEXT:    v_add_co_u32 v2, vcc_lo, 0xb0, v2
+; GFX12-NEXT:    s_wait_alu 0xf1ff
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v1, null, s5, 0, s1
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-NEXT:  .LBB4_2: ; %for.body
+; GFX12-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX12-NEXT:    flat_load_b128 v[4:7], v[2:3] offset:-176
+; GFX12-NEXT:    v_add_co_u32 v2, vcc_lo, v2, 16
+; GFX12-NEXT:    s_wait_alu 0xfffd
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-NEXT:    s_add_co_i32 s0, s0, -1
+; GFX12-NEXT:    s_wait_alu 0xfffe
+; GFX12-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    flat_store_b128 v[0:1], v[4:7]
+; GFX12-NEXT:    v_add_co_u32 v0, vcc_lo, v0, 16
+; GFX12-NEXT:    s_wait_alu 0xfffd
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX12-NEXT:    s_cbranch_scc1 .LBB4_2
+; GFX12-NEXT:  .LBB4_3: ; %for.end
+; GFX12-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: copy_flat_divergent:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_load_b32 s0, s[4:5], 0x34
+; GFX12-SPREFETCH-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SPREFETCH-NEXT:    s_cmp_eq_u32 s0, 0
+; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB4_3
+; GFX12-SPREFETCH-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX12-SPREFETCH-NEXT:    s_load_b128 s[4:7], s[4:5], 0x24
+; GFX12-SPREFETCH-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX12-SPREFETCH-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-SPREFETCH-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX12-SPREFETCH-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v2, s1, s6, v0
+; GFX12-SPREFETCH-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v3, null, s7, 0, s1
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v0, s1, s4, v0
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v2, vcc_lo, 0xb0, v2
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xf1ff
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v1, null, s5, 0, s1
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-SPREFETCH-NEXT:  .LBB4_2: ; %for.body
+; GFX12-SPREFETCH-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX12-SPREFETCH-NEXT:    flat_load_b128 v[4:7], v[2:3] offset:-176
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v2, vcc_lo, v2, 16
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xfffd
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-SPREFETCH-NEXT:    s_add_co_i32 s0, s0, -1
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xfffe
+; GFX12-SPREFETCH-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX12-SPREFETCH-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SPREFETCH-NEXT:    flat_store_b128 v[0:1], v[4:7]
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v0, vcc_lo, v0, 16
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xfffd
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB4_2
+; GFX12-SPREFETCH-NEXT:  .LBB4_3: ; %for.end
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: copy_flat_divergent:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_load_b32 s0, s[4:5], 0x34
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_cmp_eq_u32 s0, 0
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB4_3
+; GFX1250-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX1250-NEXT:    s_load_b128 s[4:7], s[4:5], 0x24
+; GFX1250-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1250-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_lshlrev_b32 v0, 4, v0
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[2:3], s[6:7], v[0:1]
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[0:1], s[4:5], v[0:1]
+; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[2:3], 0xb0, v[2:3]
+; GFX1250-NEXT:  .LBB4_2: ; %for.body
+; GFX1250-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX1250-NEXT:    flat_load_b128 v[4:7], v[2:3] offset:-176
+; GFX1250-NEXT:    flat_prefetch_b8 v[2:3] scope:SCOPE_SE
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[2:3], 16, v[2:3]
+; GFX1250-NEXT:    s_add_co_i32 s0, s0, -1
+; GFX1250-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX1250-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-NEXT:    flat_store_b128 v[0:1], v[4:7]
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[0:1], 16, v[0:1]
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB4_2
+; GFX1250-NEXT:  .LBB4_3: ; %for.end
+; GFX1250-NEXT:    s_endpgm
+entry:
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %s.tid = getelementptr inbounds <4 x i32>, ptr %s, i32 %tid
+  %d.tid = getelementptr inbounds <4 x i32>, ptr %d, i32 %tid
+  %cmp6.not = icmp eq i32 %n, 0
+  br i1 %cmp6.not, label %for.end, label %for.body
+
+for.body:                                         ; preds = %entry, %for.body
+  %i.07 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
+  %idxprom = zext i32 %i.07 to i64
+  %arrayidx = getelementptr inbounds <4 x i32>, ptr %s.tid, i64 %idxprom
+  %ld = load <4 x i32>, ptr %arrayidx, align 4
+  %arrayidx2 = getelementptr inbounds <4 x i32>, ptr %d.tid, i64 %idxprom
+  store <4 x i32> %ld, ptr %arrayidx2, align 4
+  %inc = add nuw i32 %i.07, 1
+  %exitcond.not = icmp eq i32 %inc, %n
+  br i1 %exitcond.not, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body, %entry
+  ret void
+}
+
+define amdgpu_kernel void @copy_global_divergent(ptr addrspace(1) nocapture %d, ptr addrspace(1) nocapture readonly %s, i32 %n) {
+; GFX12-LABEL: copy_global_divergent:
+; GFX12:       ; %bb.0: ; %entry
+; GFX12-NEXT:    s_load_b32 s0, s[4:5], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    s_cmp_eq_u32 s0, 0
+; GFX12-NEXT:    s_cbranch_scc1 .LBB5_3
+; GFX12-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX12-NEXT:    s_load_b128 s[4:7], s[4:5], 0x24
+; GFX12-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_add_co_u32 v2, s1, s6, v0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v3, null, s7, 0, s1
+; GFX12-NEXT:    v_add_co_u32 v0, s1, s4, v0
+; GFX12-NEXT:    v_add_co_u32 v2, vcc_lo, 0xb0, v2
+; GFX12-NEXT:    s_wait_alu 0xf1ff
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v1, null, s5, 0, s1
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-NEXT:  .LBB5_2: ; %for.body
+; GFX12-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX12-NEXT:    global_load_b128 v[4:7], v[2:3], off offset:-176
+; GFX12-NEXT:    v_add_co_u32 v2, vcc_lo, v2, 16
+; GFX12-NEXT:    s_wait_alu 0xfffd
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-NEXT:    s_add_co_i32 s0, s0, -1
+; GFX12-NEXT:    s_wait_alu 0xfffe
+; GFX12-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX12-NEXT:    s_wait_loadcnt 0x0
+; GFX12-NEXT:    global_store_b128 v[0:1], v[4:7], off
+; GFX12-NEXT:    v_add_co_u32 v0, vcc_lo, v0, 16
+; GFX12-NEXT:    s_wait_alu 0xfffd
+; GFX12-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX12-NEXT:    s_cbranch_scc1 .LBB5_2
+; GFX12-NEXT:  .LBB5_3: ; %for.end
+; GFX12-NEXT:    s_endpgm
+;
+; GFX12-SPREFETCH-LABEL: copy_global_divergent:
+; GFX12-SPREFETCH:       ; %bb.0: ; %entry
+; GFX12-SPREFETCH-NEXT:    s_load_b32 s0, s[4:5], 0x34
+; GFX12-SPREFETCH-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SPREFETCH-NEXT:    s_cmp_eq_u32 s0, 0
+; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB5_3
+; GFX12-SPREFETCH-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX12-SPREFETCH-NEXT:    s_load_b128 s[4:7], s[4:5], 0x24
+; GFX12-SPREFETCH-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX12-SPREFETCH-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-SPREFETCH-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX12-SPREFETCH-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v2, s1, s6, v0
+; GFX12-SPREFETCH-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v3, null, s7, 0, s1
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v0, s1, s4, v0
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v2, vcc_lo, 0xb0, v2
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xf1ff
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v1, null, s5, 0, s1
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-SPREFETCH-NEXT:  .LBB5_2: ; %for.body
+; GFX12-SPREFETCH-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX12-SPREFETCH-NEXT:    global_load_b128 v[4:7], v[2:3], off offset:-176
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v2, vcc_lo, v2, 16
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xfffd
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX12-SPREFETCH-NEXT:    s_add_co_i32 s0, s0, -1
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xfffe
+; GFX12-SPREFETCH-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX12-SPREFETCH-NEXT:    s_wait_loadcnt 0x0
+; GFX12-SPREFETCH-NEXT:    global_store_b128 v[0:1], v[4:7], off
+; GFX12-SPREFETCH-NEXT:    v_add_co_u32 v0, vcc_lo, v0, 16
+; GFX12-SPREFETCH-NEXT:    s_wait_alu 0xfffd
+; GFX12-SPREFETCH-NEXT:    v_add_co_ci_u32_e64 v1, null, 0, v1, vcc_lo
+; GFX12-SPREFETCH-NEXT:    s_cbranch_scc1 .LBB5_2
+; GFX12-SPREFETCH-NEXT:  .LBB5_3: ; %for.end
+; GFX12-SPREFETCH-NEXT:    s_endpgm
+;
+; GFX1250-LABEL: copy_global_divergent:
+; GFX1250:       ; %bb.0: ; %entry
+; GFX1250-NEXT:    s_load_b32 s0, s[4:5], 0x34
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    s_cmp_eq_u32 s0, 0
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB5_3
+; GFX1250-NEXT:  ; %bb.1: ; %for.body.preheader
+; GFX1250-NEXT:    s_load_b128 s[4:7], s[4:5], 0x24
+; GFX1250-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1250-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_lshlrev_b32 v0, 4, v0
+; GFX1250-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[2:3], s[6:7], v[0:1]
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[0:1], s[4:5], v[0:1]
+; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[2:3], 0xb0, v[2:3]
+; GFX1250-NEXT:  .LBB5_2: ; %for.body
+; GFX1250-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX1250-NEXT:    global_load_b128 v[4:7], v[2:3], off offset:-176
+; GFX1250-NEXT:    global_prefetch_b8 v[2:3], off scope:SCOPE_SE
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[2:3], 16, v[2:3]
+; GFX1250-NEXT:    s_add_co_i32 s0, s0, -1
+; GFX1250-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX1250-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX1250-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-NEXT:    global_store_b128 v[0:1], v[4:7], off
+; GFX1250-NEXT:    s_wait_xcnt 0x0
+; GFX1250-NEXT:    v_add_nc_u64_e32 v[0:1], 16, v[0:1]
+; GFX1250-NEXT:    s_cbranch_scc1 .LBB5_2
+; GFX1250-NEXT:  .LBB5_3: ; %for.end
+; GFX1250-NEXT:    s_endpgm
+entry:
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %s.tid = getelementptr inbounds <4 x i32>, ptr addrspace(1) %s, i32 %tid
+  %d.tid = getelementptr inbounds <4 x i32>, ptr addrspace(1) %d, i32 %tid
+  %cmp6.not = icmp eq i32 %n, 0
+  br i1 %cmp6.not, label %for.end, label %for.body
+
+for.body:                                         ; preds = %entry, %for.body
+  %i.07 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
+  %idxprom = zext i32 %i.07 to i64
+  %arrayidx = getelementptr inbounds <4 x i32>, ptr addrspace(1) %s.tid, i64 %idxprom
+  %ld = load <4 x i32>, ptr addrspace(1) %arrayidx, align 4
+  %arrayidx2 = getelementptr inbounds <4 x i32>, ptr addrspace(1) %d.tid, i64 %idxprom
+  store <4 x i32> %ld, ptr addrspace(1) %arrayidx2, align 4
+  %inc = add nuw i32 %i.07, 1
+  %exitcond.not = icmp eq i32 %inc, %n
+  br i1 %exitcond.not, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body, %entry
+  ret void
+}
+
+declare i32 @llvm.amdgcn.workitem.id.x()