AMDGPU: Add baseline tests for flat offset splitting

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@375364 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 2916 insertions, 0 deletions

diff --git a/test/CodeGen/AMDGPU/offset-split-flat.ll b/test/CodeGen/AMDGPU/offset-split-flat.ll
new file mode 100644
index 00000000000..5763c53ff39
--- /dev/null
+++ b/test/CodeGen/AMDGPU/offset-split-flat.ll
@@ -0,0 +1,1470 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10 %s
+
+; Test splitting flat instruction offsets into the low and high bits
+; when the offset doesn't fit in the offset field.
+
+define i8 @flat_inst_valu_offset_1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:1
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, v0, 1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 1
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:2047
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x7ff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 2047
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 4095
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8191
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_neg_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfffff800, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfffff800, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -2048
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_neg_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfffff000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfffff000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -4096
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_neg_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xffffe000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xffffe000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -8192
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_2x_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_2x_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_2x_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 4095
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_2x_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_2x_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_2x_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8191
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_2x_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_2x_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x3fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_2x_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x3fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 16383
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_2x_neg_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_2x_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfffff000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_2x_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfffff000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -4096
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_2x_neg_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_2x_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xffffe000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_2x_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xffffe000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -8192
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @flat_inst_valu_offset_2x_neg_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_2x_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xffffc000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_2x_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xffffc000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -16384
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2047
+define i8 @flat_inst_valu_offset_64bit_11bit_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_11bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x7ff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_11bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x7ff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8589936639
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2048
+define i8 @flat_inst_valu_offset_64bit_11bit_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_11bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x800, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_11bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x800, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8589936640
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4095
+define i8 @flat_inst_valu_offset_64bit_12bit_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_12bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_12bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8589938687
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4096
+define i8 @flat_inst_valu_offset_64bit_12bit_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_12bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_12bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8589938688
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8191
+define i8 @flat_inst_valu_offset_64bit_13bit_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_13bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_13bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8589942783
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8192
+define i8 @flat_inst_valu_offset_64bit_13bit_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_13bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x2000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_13bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x2000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 8589942784
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2047
+define i8 @flat_inst_valu_offset_64bit_11bit_neg_high_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_11bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x7ff, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_11bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x7ff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854773761
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2048
+define i8 @flat_inst_valu_offset_64bit_11bit_neg_high_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_11bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x800, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_11bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x800, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854773760
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4095
+define i8 @flat_inst_valu_offset_64bit_12bit_neg_high_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_12bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfff, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_12bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854771713
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4096
+define i8 @flat_inst_valu_offset_64bit_12bit_neg_high_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_12bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1000, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_12bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854771712
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8191
+define i8 @flat_inst_valu_offset_64bit_13bit_neg_high_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_13bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_13bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854767617
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8192
+define i8 @flat_inst_valu_offset_64bit_13bit_neg_high_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_valu_offset_64bit_13bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x2000, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: flat_inst_valu_offset_64bit_13bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x2000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854767616
+  %load = load i8, i8* %gep, align 4
+  ret i8 %load
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:1
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 1
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 1
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:2047
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 2047
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 4095
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8191
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_neg_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfffff800
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfffff800
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -2048
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_neg_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfffff000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfffff000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -4096
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_neg_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -8192
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_2x_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_2x_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1] offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_2x_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 4095
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_2x_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_2x_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_2x_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8191
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_2x_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_2x_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x3fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_2x_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x3fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 16383
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_2x_neg_11bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_2x_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfffff000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_2x_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfffff000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -4096
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_2x_neg_12bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_2x_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_2x_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -8192
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+define amdgpu_kernel void @flat_inst_salu_offset_2x_neg_13bit_max(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_2x_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xffffc000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_2x_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xffffc000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -16384
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2047
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_11bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_11bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8589936639
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2048
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_11bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_11bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8589936640
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4095
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_12bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_12bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8589938687
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4096
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_12bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_12bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8589938688
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8191
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_13bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_13bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8589942783
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8192
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_13bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_13bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 8589942784
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2047
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_neg_high_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_11bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_11bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854773761
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2048
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_11bit_neg_high_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_11bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_11bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854773760
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4095
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_neg_high_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_12bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_12bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854771713
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4096
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_12bit_neg_high_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_12bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_12bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854771712
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8191
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_neg_high_split0(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_13bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_13bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854767617
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8192
+define amdgpu_kernel void @flat_inst_salu_offset_64bit_13bit_neg_high_split1(i8* %p) {
+; GFX9-LABEL: flat_inst_salu_offset_64bit_13bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    flat_store_byte v[0:1], v0
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: flat_inst_salu_offset_64bit_13bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    flat_store_byte v[0:1], v0
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8* %p, i64 -9223372036854767616
+  %load = load volatile i8, i8* %gep, align 1
+  store i8 %load, i8* undef
+  ret void
+}
diff --git a/test/CodeGen/AMDGPU/offset-split-global.ll b/test/CodeGen/AMDGPU/offset-split-global.ll
new file mode 100644
index 00000000000..ad32b739bdb
--- /dev/null
+++ b/test/CodeGen/AMDGPU/offset-split-global.ll
@@ -0,0 +1,1446 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10 %s
+
+; Test splitting flat instruction offsets into the low and high bits
+; when the offset doesn't fit in the offset field.
+
+define i8 @global_inst_valu_offset_1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:1
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off offset:1
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 1
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:2047
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off offset:2047
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 2047
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 4095
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8191
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_neg_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:-2048
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off offset:-2048
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -2048
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_neg_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:-4096
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfffff000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -4096
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_neg_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xffffe000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xffffe000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -8192
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_2x_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_2x_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_2x_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 4095
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_2x_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_2x_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_2x_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8191
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_2x_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_2x_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x3fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_2x_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x3fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 16383
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_2x_neg_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_2x_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:-4096
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_2x_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfffff000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -4096
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_2x_neg_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_2x_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xffffe000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_2x_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xffffe000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -8192
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define i8 @global_inst_valu_offset_2x_neg_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_2x_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xffffc000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_2x_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xffffc000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, -1, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -16384
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2047
+define i8 @global_inst_valu_offset_64bit_11bit_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_11bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x7ff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_11bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x7ff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589936639
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2048
+define i8 @global_inst_valu_offset_64bit_11bit_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_11bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x800, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_11bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x800, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589936640
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4095
+define i8 @global_inst_valu_offset_64bit_12bit_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_12bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_12bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589938687
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4096
+define i8 @global_inst_valu_offset_64bit_12bit_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_12bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_12bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589938688
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8191
+define i8 @global_inst_valu_offset_64bit_13bit_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_13bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_13bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589942783
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8192
+define i8 @global_inst_valu_offset_64bit_13bit_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_13bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x2000, v0
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 2, v1, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_13bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x2000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 2, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589942784
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2047
+define i8 @global_inst_valu_offset_64bit_11bit_neg_high_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_11bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x7ff, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_11bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x7ff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854773761
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2048
+define i8 @global_inst_valu_offset_64bit_11bit_neg_high_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_11bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x800, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_11bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x800, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854773760
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4095
+define i8 @global_inst_valu_offset_64bit_12bit_neg_high_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_12bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0xfff, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_12bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0xfff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854771713
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4096
+define i8 @global_inst_valu_offset_64bit_12bit_neg_high_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_12bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1000, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_12bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854771712
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8191
+define i8 @global_inst_valu_offset_64bit_13bit_neg_high_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_13bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x1fff, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_13bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x1fff, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854767617
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8192
+define i8 @global_inst_valu_offset_64bit_13bit_neg_high_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_valu_offset_64bit_13bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, 0x2000, v0
+; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
+; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, v1, v2, vcc
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: global_inst_valu_offset_64bit_13bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    v_add_co_u32_e64 v0, vcc_lo, 0x2000, v0
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0x80000000, v1, vcc_lo
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854767616
+  %load = load i8, i8 addrspace(1)* %gep, align 4
+  ret i8 %load
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:1
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off offset:1
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 1
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:2047
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off offset:2047
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 2047
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 4095
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8191
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_neg_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:-2048
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off offset:-2048
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -2048
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_neg_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:-4096
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfffff000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -4096
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_neg_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -8192
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_2x_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_2x_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:4095
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_2x_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 4095
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_2x_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_2x_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_2x_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8191
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_2x_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_2x_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x3fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_2x_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x3fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 16383
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_2x_neg_11bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_2x_neg_11bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off offset:-4096
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_2x_neg_11bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfffff000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -4096
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_2x_neg_12bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_2x_neg_12bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_2x_neg_12bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xffffe000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -8192
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @global_inst_salu_offset_2x_neg_13bit_max(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_2x_neg_13bit_max:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xffffc000
+; GFX9-NEXT:    s_addc_u32 s1, s1, -1
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_2x_neg_13bit_max:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xffffc000
+; GFX10-NEXT:    s_addc_u32 s1, s1, -1
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -16384
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2047
+define amdgpu_kernel void @global_inst_salu_offset_64bit_11bit_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_11bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_11bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589936639
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 11-bit low-bits (1ull << 33) | 2048
+define amdgpu_kernel void @global_inst_salu_offset_64bit_11bit_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_11bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_11bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589936640
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4095
+define amdgpu_kernel void @global_inst_salu_offset_64bit_12bit_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_12bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_12bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589938687
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 12-bit low-bits (1ull << 33) | 4096
+define amdgpu_kernel void @global_inst_salu_offset_64bit_12bit_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_12bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_12bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589938688
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8191
+define amdgpu_kernel void @global_inst_salu_offset_64bit_13bit_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_13bit_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_13bit_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589942783
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 13-bit low-bits (1ull << 33) | 8192
+define amdgpu_kernel void @global_inst_salu_offset_64bit_13bit_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_13bit_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 2
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_13bit_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 2
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 8589942784
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2047
+define amdgpu_kernel void @global_inst_salu_offset_64bit_11bit_neg_high_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_11bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_11bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x7ff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854773761
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 11-bit low-bits, negative high bits (1ull << 63) | 2048
+define amdgpu_kernel void @global_inst_salu_offset_64bit_11bit_neg_high_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_11bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_11bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x800
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854773760
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4095
+define amdgpu_kernel void @global_inst_salu_offset_64bit_12bit_neg_high_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_12bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_12bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0xfff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854771713
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 12-bit low-bits, negative high bits (1ull << 63) | 4096
+define amdgpu_kernel void @global_inst_salu_offset_64bit_12bit_neg_high_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_12bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_12bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854771712
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8191
+define amdgpu_kernel void @global_inst_salu_offset_64bit_13bit_neg_high_split0(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_13bit_neg_high_split0:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_13bit_neg_high_split0:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x1fff
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854767617
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}
+
+; Fill 13-bit low-bits, negative high bits (1ull << 63) | 8192
+define amdgpu_kernel void @global_inst_salu_offset_64bit_13bit_neg_high_split1(i8 addrspace(1)* %p) {
+; GFX9-LABEL: global_inst_salu_offset_64bit_13bit_neg_high_split1:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX9-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX9-NEXT:    v_mov_b32_e32 v0, s0
+; GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    global_store_byte v[0:1], v0, off
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_inst_salu_offset_64bit_13bit_neg_high_split1:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GFX10-NEXT:    ; implicit-def: $vcc_hi
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_add_u32 s0, s0, 0x2000
+; GFX10-NEXT:    s_addc_u32 s1, s1, 0x80000000
+; GFX10-NEXT:    v_mov_b32_e32 v0, s0
+; GFX10-NEXT:    v_mov_b32_e32 v1, s1
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    global_store_byte v[0:1], v0, off
+; GFX10-NEXT:    s_endpgm
+  %gep = getelementptr i8, i8 addrspace(1)* %p, i64 -9223372036854767616
+  %load = load volatile i8, i8 addrspace(1)* %gep, align 1
+  store i8 %load, i8 addrspace(1)* undef
+  ret void
+}