[llvm] 372115f - [AMDGPU] Precommit test for i8 vector CopyToReg handling patch
Jeffrey Byrnes via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 13 11:27:56 PDT 2023
Author: Jeffrey Byrnes
Date: 2023-09-13T11:27:15-07:00
New Revision: 372115fadddcfc3dcac7909f9ce56d36d26bb429
URL: https://github.com/llvm/llvm-project/commit/372115fadddcfc3dcac7909f9ce56d36d26bb429
DIFF: https://github.com/llvm/llvm-project/commit/372115fadddcfc3dcac7909f9ce56d36d26bb429.diff
LOG: [AMDGPU] Precommit test for i8 vector CopyToReg handling patch
Adds test to show impact on cross block CopyToReg & CopyFromReg handling for n x i8, and shows NFC on CC
Differential Revision: https://reviews.llvm.org/D159303
Change-Id: Ib6d9802dbebe8e3245e4ccfd4a6f23357de8c480
Added:
llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
Modified:
llvm/test/CodeGen/AMDGPU/calling-conventions.ll
llvm/test/CodeGen/AMDGPU/function-args.ll
llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AMDGPU/calling-conventions.ll b/llvm/test/CodeGen/AMDGPU/calling-conventions.ll
index c2c639b91db5c5e..cb89841b58f9787 100644
--- a/llvm/test/CodeGen/AMDGPU/calling-conventions.ll
+++ b/llvm/test/CodeGen/AMDGPU/calling-conventions.ll
@@ -955,4 +955,1131 @@ define amdgpu_ps i16 @ret_ps_mesa_i16() {
ret i16 123
}
+define amdgpu_kernel void @amd_kernel_i8(i8 %arg0) {
+; SI-LABEL: amd_kernel_i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dword s0, s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_add_i32 s0, s0, s0
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dword s0, s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: flat_store_byte v[0:1], v0
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b32 s0, s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_add_i32 s0, s0, s0
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-NEXT: v_mov_b32_e32 v0, s0
+; GFX11-NEXT: global_store_b8 v[0:1], v0, off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add i8 %arg0, %arg0
+ store i8 %add, ptr addrspace(1) undef
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v2i8(<2 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v2i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dword s1, s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s0, 0
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_bfe_u32 s2, s1, 0x80008
+; SI-NEXT: s_add_i32 s1, s1, s1
+; SI-NEXT: s_and_b32 s1, s1, 0xff
+; SI-NEXT: s_add_i32 s2, s2, s2
+; SI-NEXT: s_lshl_b32 s2, s2, 8
+; SI-NEXT: s_or_b32 s4, s1, s2
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_mov_b32 s1, s0
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v2i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dword s0, s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_add_u32_sdwa v2, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_mov_b32_e32 v3, s0
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: flat_store_short v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v2i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b32 s0, s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s0
+; GFX11-NEXT: v_add_nc_u16 v1, s0, s0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v0
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <2 x i8> %arg0, %arg0
+ store <2 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v4i8(<4 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v4i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dword s1, s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s0, 0
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_lshr_b32 s2, s1, 16
+; SI-NEXT: s_lshr_b32 s4, s1, 24
+; SI-NEXT: s_bfe_u32 s5, s1, 0x80008
+; SI-NEXT: s_add_i32 s1, s1, s1
+; SI-NEXT: s_add_i32 s4, s4, s4
+; SI-NEXT: s_add_i32 s2, s2, s2
+; SI-NEXT: s_and_b32 s1, s1, 0xff
+; SI-NEXT: s_add_i32 s5, s5, s5
+; SI-NEXT: s_lshl_b32 s4, s4, 24
+; SI-NEXT: s_and_b32 s2, s2, 0xff
+; SI-NEXT: s_lshl_b32 s5, s5, 8
+; SI-NEXT: s_lshl_b32 s2, s2, 16
+; SI-NEXT: s_or_b32 s1, s1, s5
+; SI-NEXT: s_or_b32 s2, s4, s2
+; SI-NEXT: s_and_b32 s1, s1, 0xffff
+; SI-NEXT: s_or_b32 s4, s1, s2
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_mov_b32 s1, s0
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v4i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dword s0, s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s1, s0, 24
+; VI-NEXT: s_lshr_b32 s2, s0, 16
+; VI-NEXT: s_add_i32 s2, s2, s2
+; VI-NEXT: s_add_i32 s1, s1, s1
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_lshlrev_b16_e64 v1, 8, s1
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v2, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v1, s0
+; VI-NEXT: v_or_b32_sdwa v3, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v0, 0
+; VI-NEXT: v_mov_b32_e32 v1, 0
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: flat_store_dword v[0:1], v2
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v4i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b32 s0, s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s0
+; GFX11-NEXT: s_lshr_b32 s1, s0, 16
+; GFX11-NEXT: s_lshr_b32 s2, s0, 24
+; GFX11-NEXT: v_add_nc_u16 v1, s0, s0
+; GFX11-NEXT: v_add_nc_u16 v2, s2, s2
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: v_add_nc_u16 v3, s1, s1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-NEXT: v_lshlrev_b16 v2, 8, v2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX11-NEXT: v_and_b32_e32 v3, 0xff, v3
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v0, v1, v0
+; GFX11-NEXT: v_or_b32_e32 v1, v3, v2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_and_b32_e32 v2, 0xffff, v0
+; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_lshlrev_b32 v3, 16, v1
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-NEXT: global_store_b32 v[0:1], v2, off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <4 x i8> %arg0, %arg0
+ store <4 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v3i8(<3 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v3i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dword s4, s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s1, 0
+; SI-NEXT: s_mov_b32 s0, 2
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_lshr_b32 s5, s4, 16
+; SI-NEXT: s_bfe_u32 s6, s4, 0x80008
+; SI-NEXT: s_add_i32 s4, s4, s4
+; SI-NEXT: s_and_b32 s4, s4, 0xff
+; SI-NEXT: s_add_i32 s6, s6, s6
+; SI-NEXT: s_add_i32 s5, s5, s5
+; SI-NEXT: s_lshl_b32 s6, s6, 8
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: s_or_b32 s4, s4, s6
+; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT: s_mov_b32 s0, s1
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_short v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v3i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dword s0, s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v2, 0
+; VI-NEXT: v_mov_b32_e32 v3, 0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s1, s0, 16
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_mov_b32_e32 v1, s0
+; VI-NEXT: s_add_i32 s1, s1, s1
+; VI-NEXT: v_or_b32_sdwa v4, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v0, 2
+; VI-NEXT: v_mov_b32_e32 v1, 0
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: flat_store_byte v[0:1], v5
+; VI-NEXT: flat_store_short v[2:3], v4
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v3i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b32 s0, s[0:1], 0x24
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: v_mov_b32_e32 v3, 0
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s0
+; GFX11-NEXT: v_add_nc_u16 v1, s0, s0
+; GFX11-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-NEXT: v_add_nc_u16 v6, s0, s0
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v1
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v0
+; GFX11-NEXT: v_mov_b32_e32 v0, 2
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b8 v[0:1], v6, off
+; GFX11-NEXT: global_store_b16 v[2:3], v4, off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <3 x i8> %arg0, %arg0
+ store <3 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v5i8(<5 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v5i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s1, 0
+; SI-NEXT: s_mov_b32 s0, 4
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_lshr_b32 s6, s4, 16
+; SI-NEXT: s_lshr_b32 s7, s4, 24
+; SI-NEXT: s_bfe_u32 s8, s4, 0x80008
+; SI-NEXT: s_add_i32 s4, s4, s4
+; SI-NEXT: s_add_i32 s5, s5, s5
+; SI-NEXT: s_add_i32 s7, s7, s7
+; SI-NEXT: s_add_i32 s6, s6, s6
+; SI-NEXT: s_and_b32 s4, s4, 0xff
+; SI-NEXT: s_add_i32 s8, s8, s8
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: s_lshl_b32 s5, s7, 24
+; SI-NEXT: s_and_b32 s6, s6, 0xff
+; SI-NEXT: s_lshl_b32 s7, s8, 8
+; SI-NEXT: s_lshl_b32 s6, s6, 16
+; SI-NEXT: s_or_b32 s4, s4, s7
+; SI-NEXT: s_or_b32 s5, s5, s6
+; SI-NEXT: s_and_b32 s4, s4, 0xffff
+; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0
+; SI-NEXT: s_mov_b32 s0, s1
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v5i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s2, s0, 24
+; VI-NEXT: s_lshr_b32 s3, s0, 16
+; VI-NEXT: s_add_i32 s3, s3, s3
+; VI-NEXT: s_add_i32 s2, s2, s2
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_lshlrev_b16_e64 v1, 8, s2
+; VI-NEXT: v_mov_b32_e32 v2, s3
+; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v2, s0
+; VI-NEXT: s_and_b32 s1, s1, 0xff
+; VI-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_add_i32 s1, s1, s1
+; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v0, 4
+; VI-NEXT: v_mov_b32_e32 v1, 0
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: v_mov_b32_e32 v2, 0
+; VI-NEXT: v_mov_b32_e32 v3, 0
+; VI-NEXT: flat_store_byte v[0:1], v5
+; VI-NEXT: flat_store_dword v[2:3], v4
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v5i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s0
+; GFX11-NEXT: s_lshr_b32 s2, s0, 16
+; GFX11-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-NEXT: v_add_nc_u16 v1, s0, s0
+; GFX11-NEXT: v_add_nc_u16 v2, s3, s3
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: v_add_nc_u16 v3, s2, s2
+; GFX11-NEXT: v_add_nc_u16 v6, s1, s1
+; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-NEXT: v_lshlrev_b16 v2, 8, v2
+; GFX11-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX11-NEXT: v_and_b32_e32 v3, 0xff, v3
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v0, v1, v0
+; GFX11-NEXT: v_or_b32_e32 v1, v3, v2
+; GFX11-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_and_b32 v4, 0xffff, v0
+; GFX11-NEXT: v_dual_mov_b32 v0, 4 :: v_dual_lshlrev_b32 v5, 16, v1
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b8 v[0:1], v6, off
+; GFX11-NEXT: global_store_b32 v[2:3], v4, off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <5 x i8> %arg0, %arg0
+ store <5 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v8i8(<8 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v8i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s0, 0
+; SI-NEXT: s_mov_b32 s3, 0xf000
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_lshr_b32 s1, s4, 16
+; SI-NEXT: s_lshr_b32 s2, s4, 24
+; SI-NEXT: s_lshr_b32 s6, s5, 16
+; SI-NEXT: s_lshr_b32 s7, s5, 24
+; SI-NEXT: s_bfe_u32 s8, s4, 0x80008
+; SI-NEXT: s_bfe_u32 s9, s5, 0x80008
+; SI-NEXT: s_add_i32 s5, s5, s5
+; SI-NEXT: s_add_i32 s4, s4, s4
+; SI-NEXT: s_add_i32 s7, s7, s7
+; SI-NEXT: s_add_i32 s6, s6, s6
+; SI-NEXT: s_and_b32 s5, s5, 0xff
+; SI-NEXT: s_add_i32 s9, s9, s9
+; SI-NEXT: s_add_i32 s2, s2, s2
+; SI-NEXT: s_add_i32 s1, s1, s1
+; SI-NEXT: s_and_b32 s4, s4, 0xff
+; SI-NEXT: s_add_i32 s8, s8, s8
+; SI-NEXT: s_lshl_b32 s7, s7, 24
+; SI-NEXT: s_and_b32 s6, s6, 0xff
+; SI-NEXT: s_lshl_b32 s9, s9, 8
+; SI-NEXT: s_lshl_b32 s2, s2, 24
+; SI-NEXT: s_and_b32 s1, s1, 0xff
+; SI-NEXT: s_lshl_b32 s8, s8, 8
+; SI-NEXT: s_lshl_b32 s6, s6, 16
+; SI-NEXT: s_or_b32 s5, s5, s9
+; SI-NEXT: s_lshl_b32 s1, s1, 16
+; SI-NEXT: s_or_b32 s4, s4, s8
+; SI-NEXT: s_or_b32 s6, s7, s6
+; SI-NEXT: s_and_b32 s5, s5, 0xffff
+; SI-NEXT: s_or_b32 s1, s2, s1
+; SI-NEXT: s_and_b32 s2, s4, 0xffff
+; SI-NEXT: s_or_b32 s4, s5, s6
+; SI-NEXT: s_or_b32 s5, s2, s1
+; SI-NEXT: s_mov_b32 s2, -1
+; SI-NEXT: s_mov_b32 s1, s0
+; SI-NEXT: v_mov_b32_e32 v0, s5
+; SI-NEXT: v_mov_b32_e32 v1, s4
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v8i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s2, s1, 24
+; VI-NEXT: s_lshr_b32 s3, s1, 16
+; VI-NEXT: s_add_i32 s3, s3, s3
+; VI-NEXT: s_add_i32 s2, s2, s2
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_add_i32 s1, s1, s1
+; VI-NEXT: v_lshlrev_b16_e64 v2, 8, s2
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: v_add_u32_sdwa v1, vcc, v1, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v3, s1
+; VI-NEXT: s_lshr_b32 s4, s0, 24
+; VI-NEXT: s_lshr_b32 s5, s0, 16
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_add_i32 s5, s5, s5
+; VI-NEXT: s_add_i32 s4, s4, s4
+; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v2, s0
+; VI-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v2, 8, s4
+; VI-NEXT: v_mov_b32_e32 v3, s5
+; VI-NEXT: v_or_b32_sdwa v4, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v2, 0
+; VI-NEXT: v_mov_b32_e32 v3, 0
+; VI-NEXT: v_or_b32_sdwa v0, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v8i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s0
+; GFX11-NEXT: v_lshrrev_b16 v1, 8, s1
+; GFX11-NEXT: s_lshr_b32 s2, s0, 16
+; GFX11-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-NEXT: s_lshr_b32 s4, s1, 16
+; GFX11-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-NEXT: v_add_nc_u16 v2, s1, s1
+; GFX11-NEXT: v_add_nc_u16 v3, s0, s0
+; GFX11-NEXT: v_add_nc_u16 v4, s5, s5
+; GFX11-NEXT: v_add_nc_u16 v5, s4, s4
+; GFX11-NEXT: v_add_nc_u16 v6, s3, s3
+; GFX11-NEXT: v_add_nc_u16 v7, s2, s2
+; GFX11-NEXT: v_add_nc_u16 v1, v1, v1
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_and_b32_e32 v3, 0xff, v3
+; GFX11-NEXT: v_lshlrev_b16 v4, 8, v4
+; GFX11-NEXT: v_and_b32_e32 v5, 0xff, v5
+; GFX11-NEXT: v_lshlrev_b16 v6, 8, v6
+; GFX11-NEXT: v_and_b32_e32 v7, 0xff, v7
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX11-NEXT: v_or_b32_e32 v4, v5, v4
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_or_b32_e32 v5, v7, v6
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_or_b32_e32 v0, v3, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v2, 16, v4
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_lshlrev_b32_e32 v4, 16, v5
+; GFX11-NEXT: v_and_b32_e32 v3, 0xffff, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_and_b32 v5, 0xffff, v0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_or_b32_e32 v3, v3, v2
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-NEXT: v_or_b32_e32 v2, v5, v4
+; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <8 x i8> %arg0, %arg0
+ store <8 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v16i8(<16 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v16i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s4, 0
+; SI-NEXT: s_mov_b32 s7, 0xf000
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_lshr_b32 s5, s0, 16
+; SI-NEXT: s_lshr_b32 s6, s0, 24
+; SI-NEXT: s_lshr_b32 s8, s1, 16
+; SI-NEXT: s_lshr_b32 s9, s1, 24
+; SI-NEXT: s_lshr_b32 s10, s2, 16
+; SI-NEXT: s_lshr_b32 s11, s2, 24
+; SI-NEXT: s_lshr_b32 s12, s3, 16
+; SI-NEXT: s_lshr_b32 s13, s3, 24
+; SI-NEXT: s_bfe_u32 s14, s0, 0x80008
+; SI-NEXT: s_bfe_u32 s15, s1, 0x80008
+; SI-NEXT: s_bfe_u32 s16, s2, 0x80008
+; SI-NEXT: s_bfe_u32 s17, s3, 0x80008
+; SI-NEXT: s_add_i32 s3, s3, s3
+; SI-NEXT: s_add_i32 s2, s2, s2
+; SI-NEXT: s_add_i32 s1, s1, s1
+; SI-NEXT: s_add_i32 s0, s0, s0
+; SI-NEXT: s_add_i32 s13, s13, s13
+; SI-NEXT: s_add_i32 s12, s12, s12
+; SI-NEXT: s_and_b32 s3, s3, 0xff
+; SI-NEXT: s_add_i32 s17, s17, s17
+; SI-NEXT: s_add_i32 s11, s11, s11
+; SI-NEXT: s_add_i32 s10, s10, s10
+; SI-NEXT: s_and_b32 s2, s2, 0xff
+; SI-NEXT: s_add_i32 s16, s16, s16
+; SI-NEXT: s_add_i32 s9, s9, s9
+; SI-NEXT: s_add_i32 s8, s8, s8
+; SI-NEXT: s_and_b32 s1, s1, 0xff
+; SI-NEXT: s_add_i32 s15, s15, s15
+; SI-NEXT: s_add_i32 s6, s6, s6
+; SI-NEXT: s_add_i32 s5, s5, s5
+; SI-NEXT: s_and_b32 s0, s0, 0xff
+; SI-NEXT: s_add_i32 s14, s14, s14
+; SI-NEXT: s_lshl_b32 s13, s13, 24
+; SI-NEXT: s_and_b32 s12, s12, 0xff
+; SI-NEXT: s_lshl_b32 s17, s17, 8
+; SI-NEXT: s_lshl_b32 s11, s11, 24
+; SI-NEXT: s_and_b32 s10, s10, 0xff
+; SI-NEXT: s_lshl_b32 s16, s16, 8
+; SI-NEXT: s_lshl_b32 s9, s9, 24
+; SI-NEXT: s_and_b32 s8, s8, 0xff
+; SI-NEXT: s_lshl_b32 s15, s15, 8
+; SI-NEXT: s_lshl_b32 s6, s6, 24
+; SI-NEXT: s_and_b32 s5, s5, 0xff
+; SI-NEXT: s_lshl_b32 s14, s14, 8
+; SI-NEXT: s_lshl_b32 s12, s12, 16
+; SI-NEXT: s_or_b32 s3, s3, s17
+; SI-NEXT: s_lshl_b32 s10, s10, 16
+; SI-NEXT: s_or_b32 s2, s2, s16
+; SI-NEXT: s_lshl_b32 s8, s8, 16
+; SI-NEXT: s_or_b32 s1, s1, s15
+; SI-NEXT: s_lshl_b32 s5, s5, 16
+; SI-NEXT: s_or_b32 s0, s0, s14
+; SI-NEXT: s_or_b32 s12, s13, s12
+; SI-NEXT: s_and_b32 s3, s3, 0xffff
+; SI-NEXT: s_or_b32 s10, s11, s10
+; SI-NEXT: s_and_b32 s2, s2, 0xffff
+; SI-NEXT: s_or_b32 s8, s9, s8
+; SI-NEXT: s_and_b32 s1, s1, 0xffff
+; SI-NEXT: s_or_b32 s5, s6, s5
+; SI-NEXT: s_and_b32 s0, s0, 0xffff
+; SI-NEXT: s_or_b32 s3, s3, s12
+; SI-NEXT: s_or_b32 s2, s2, s10
+; SI-NEXT: s_or_b32 s1, s1, s8
+; SI-NEXT: s_or_b32 s0, s0, s5
+; SI-NEXT: s_mov_b32 s6, -1
+; SI-NEXT: s_mov_b32 s5, s4
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: v_mov_b32_e32 v1, s1
+; SI-NEXT: v_mov_b32_e32 v2, s2
+; SI-NEXT: v_mov_b32_e32 v3, s3
+; SI-NEXT: buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v16i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s4, s3, 24
+; VI-NEXT: s_lshr_b32 s5, s3, 16
+; VI-NEXT: s_add_i32 s5, s5, s5
+; VI-NEXT: s_add_i32 s4, s4, s4
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: s_add_i32 s3, s3, s3
+; VI-NEXT: v_lshlrev_b16_e64 v4, 8, s4
+; VI-NEXT: v_mov_b32_e32 v5, s5
+; VI-NEXT: s_lshr_b32 s6, s2, 24
+; VI-NEXT: s_lshr_b32 s7, s2, 16
+; VI-NEXT: v_add_u32_sdwa v3, vcc, v3, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v5, s3
+; VI-NEXT: s_add_i32 s7, s7, s7
+; VI-NEXT: s_add_i32 s6, s6, s6
+; VI-NEXT: v_or_b32_sdwa v3, v5, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: s_add_i32 s2, s2, s2
+; VI-NEXT: v_or_b32_sdwa v3, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v4, 8, s6
+; VI-NEXT: v_mov_b32_e32 v5, s7
+; VI-NEXT: s_lshr_b32 s8, s1, 24
+; VI-NEXT: s_lshr_b32 s9, s1, 16
+; VI-NEXT: v_add_u32_sdwa v2, vcc, v2, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v5, s2
+; VI-NEXT: s_add_i32 s9, s9, s9
+; VI-NEXT: s_add_i32 s8, s8, s8
+; VI-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_add_i32 s1, s1, s1
+; VI-NEXT: v_or_b32_sdwa v2, v2, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v4, 8, s8
+; VI-NEXT: v_mov_b32_e32 v5, s9
+; VI-NEXT: s_lshr_b32 s10, s0, 24
+; VI-NEXT: s_lshr_b32 s11, s0, 16
+; VI-NEXT: v_add_u32_sdwa v1, vcc, v1, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v5, s1
+; VI-NEXT: s_add_i32 s11, s11, s11
+; VI-NEXT: s_add_i32 s10, s10, s10
+; VI-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_or_b32_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v4, 8, s10
+; VI-NEXT: v_mov_b32_e32 v5, s11
+; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v6, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v4, s0
+; VI-NEXT: v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v4, 0
+; VI-NEXT: v_mov_b32_e32 v5, 0
+; VI-NEXT: v_or_b32_sdwa v0, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v16i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: s_lshr_b32 s10, s3, 16
+; GFX11-NEXT: s_lshr_b32 s11, s3, 24
+; GFX11-NEXT: v_lshrrev_b16 v2, 8, s2
+; GFX11-NEXT: v_lshrrev_b16 v3, 8, s3
+; GFX11-NEXT: v_add_nc_u16 v7, s11, s11
+; GFX11-NEXT: v_add_nc_u16 v8, s10, s10
+; GFX11-NEXT: v_add_nc_u16 v4, s3, s3
+; GFX11-NEXT: v_add_nc_u16 v5, s2, s2
+; GFX11-NEXT: v_add_nc_u16 v3, v3, v3
+; GFX11-NEXT: v_add_nc_u16 v2, v2, v2
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v8, 0xff, v8
+; GFX11-NEXT: s_lshr_b32 s7, s1, 24
+; GFX11-NEXT: v_lshrrev_b16 v1, 8, s1
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s0
+; GFX11-NEXT: v_add_nc_u16 v11, s7, s7
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v4
+; GFX11-NEXT: v_and_b32_e32 v5, 0xff, v5
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_lshlrev_b16 v2, 8, v2
+; GFX11-NEXT: v_or_b32_e32 v7, v8, v7
+; GFX11-NEXT: s_lshr_b32 s6, s1, 16
+; GFX11-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-NEXT: s_lshr_b32 s5, s0, 24
+; GFX11-NEXT: s_lshr_b32 s8, s2, 16
+; GFX11-NEXT: s_lshr_b32 s9, s2, 24
+; GFX11-NEXT: v_add_nc_u16 v6, s1, s1
+; GFX11-NEXT: v_add_nc_u16 v12, s6, s6
+; GFX11-NEXT: v_add_nc_u16 v1, v1, v1
+; GFX11-NEXT: v_add_nc_u16 v9, s9, s9
+; GFX11-NEXT: v_add_nc_u16 v10, s8, s8
+; GFX11-NEXT: v_or_b32_e32 v3, v4, v3
+; GFX11-NEXT: v_or_b32_e32 v2, v5, v2
+; GFX11-NEXT: v_lshlrev_b16 v4, 8, v11
+; GFX11-NEXT: v_lshlrev_b32_e32 v5, 16, v7
+; GFX11-NEXT: v_add_nc_u16 v7, s0, s0
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: v_add_nc_u16 v8, s5, s5
+; GFX11-NEXT: v_add_nc_u16 v11, s4, s4
+; GFX11-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v12, 0xff, v12
+; GFX11-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX11-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-NEXT: v_and_b32_e32 v7, 0xff, v7
+; GFX11-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX11-NEXT: v_lshlrev_b16 v8, 8, v8
+; GFX11-NEXT: v_and_b32_e32 v11, 0xff, v11
+; GFX11-NEXT: v_or_b32_e32 v1, v6, v1
+; GFX11-NEXT: v_or_b32_e32 v4, v12, v4
+; GFX11-NEXT: v_or_b32_e32 v9, v10, v9
+; GFX11-NEXT: v_or_b32_e32 v0, v7, v0
+; GFX11-NEXT: v_or_b32_e32 v6, v11, v8
+; GFX11-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; GFX11-NEXT: v_lshlrev_b32_e32 v7, 16, v9
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; GFX11-NEXT: v_or_b32_e32 v3, v3, v5
+; GFX11-NEXT: v_or_b32_e32 v1, v1, v4
+; GFX11-NEXT: v_mov_b32_e32 v4, 0
+; GFX11-NEXT: v_or_b32_e32 v2, v2, v7
+; GFX11-NEXT: v_mov_b32_e32 v5, 0
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v6
+; GFX11-NEXT: global_store_b128 v[4:5], v[0:3], off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <16 x i8> %arg0, %arg0
+ store <16 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
+define amdgpu_kernel void @amd_kernel_v32i8(<32 x i8> %arg0) {
+; SI-LABEL: amd_kernel_v32i8:
+; SI: ; %bb.0: ; %entry
+; SI-NEXT: s_load_dwordx8 s[0:7], s[0:1], 0x9
+; SI-NEXT: s_mov_b32 s9, 0
+; SI-NEXT: s_mov_b32 s8, 16
+; SI-NEXT: s_mov_b32 s11, 0xf000
+; SI-NEXT: s_mov_b32 s10, -1
+; SI-NEXT: s_waitcnt lgkmcnt(0)
+; SI-NEXT: s_lshr_b32 s12, s4, 16
+; SI-NEXT: s_lshr_b32 s13, s4, 24
+; SI-NEXT: s_lshr_b32 s14, s5, 16
+; SI-NEXT: s_lshr_b32 s15, s5, 24
+; SI-NEXT: s_lshr_b32 s16, s6, 16
+; SI-NEXT: s_lshr_b32 s17, s6, 24
+; SI-NEXT: s_lshr_b32 s18, s7, 16
+; SI-NEXT: s_lshr_b32 s19, s7, 24
+; SI-NEXT: s_lshr_b32 s20, s0, 16
+; SI-NEXT: s_lshr_b32 s21, s0, 24
+; SI-NEXT: s_lshr_b32 s22, s1, 16
+; SI-NEXT: s_lshr_b32 s23, s1, 24
+; SI-NEXT: s_lshr_b32 s24, s2, 16
+; SI-NEXT: s_lshr_b32 s25, s2, 24
+; SI-NEXT: s_lshr_b32 s26, s3, 16
+; SI-NEXT: s_lshr_b32 s27, s3, 24
+; SI-NEXT: s_bfe_u32 s28, s4, 0x80008
+; SI-NEXT: s_bfe_u32 s29, s5, 0x80008
+; SI-NEXT: s_bfe_u32 s30, s6, 0x80008
+; SI-NEXT: s_bfe_u32 s31, s7, 0x80008
+; SI-NEXT: s_bfe_u32 s33, s0, 0x80008
+; SI-NEXT: s_bfe_u32 s34, s1, 0x80008
+; SI-NEXT: s_bfe_u32 s35, s2, 0x80008
+; SI-NEXT: s_bfe_u32 s36, s3, 0x80008
+; SI-NEXT: s_add_i32 s3, s3, s3
+; SI-NEXT: s_add_i32 s2, s2, s2
+; SI-NEXT: s_add_i32 s1, s1, s1
+; SI-NEXT: s_add_i32 s0, s0, s0
+; SI-NEXT: s_add_i32 s7, s7, s7
+; SI-NEXT: s_add_i32 s6, s6, s6
+; SI-NEXT: s_add_i32 s5, s5, s5
+; SI-NEXT: s_add_i32 s4, s4, s4
+; SI-NEXT: s_add_i32 s27, s27, s27
+; SI-NEXT: s_add_i32 s26, s26, s26
+; SI-NEXT: s_and_b32 s3, s3, 0xff
+; SI-NEXT: s_add_i32 s36, s36, s36
+; SI-NEXT: s_add_i32 s25, s25, s25
+; SI-NEXT: s_add_i32 s24, s24, s24
+; SI-NEXT: s_and_b32 s2, s2, 0xff
+; SI-NEXT: s_add_i32 s35, s35, s35
+; SI-NEXT: s_add_i32 s23, s23, s23
+; SI-NEXT: s_add_i32 s22, s22, s22
+; SI-NEXT: s_and_b32 s1, s1, 0xff
+; SI-NEXT: s_add_i32 s34, s34, s34
+; SI-NEXT: s_add_i32 s21, s21, s21
+; SI-NEXT: s_add_i32 s20, s20, s20
+; SI-NEXT: s_and_b32 s0, s0, 0xff
+; SI-NEXT: s_add_i32 s33, s33, s33
+; SI-NEXT: s_add_i32 s19, s19, s19
+; SI-NEXT: s_add_i32 s18, s18, s18
+; SI-NEXT: s_and_b32 s7, s7, 0xff
+; SI-NEXT: s_add_i32 s31, s31, s31
+; SI-NEXT: s_add_i32 s17, s17, s17
+; SI-NEXT: s_add_i32 s16, s16, s16
+; SI-NEXT: s_and_b32 s6, s6, 0xff
+; SI-NEXT: s_add_i32 s30, s30, s30
+; SI-NEXT: s_add_i32 s15, s15, s15
+; SI-NEXT: s_add_i32 s14, s14, s14
+; SI-NEXT: s_and_b32 s5, s5, 0xff
+; SI-NEXT: s_add_i32 s29, s29, s29
+; SI-NEXT: s_add_i32 s13, s13, s13
+; SI-NEXT: s_add_i32 s12, s12, s12
+; SI-NEXT: s_and_b32 s4, s4, 0xff
+; SI-NEXT: s_add_i32 s28, s28, s28
+; SI-NEXT: s_lshl_b32 s27, s27, 24
+; SI-NEXT: s_and_b32 s26, s26, 0xff
+; SI-NEXT: s_lshl_b32 s36, s36, 8
+; SI-NEXT: s_lshl_b32 s25, s25, 24
+; SI-NEXT: s_and_b32 s24, s24, 0xff
+; SI-NEXT: s_lshl_b32 s35, s35, 8
+; SI-NEXT: s_lshl_b32 s23, s23, 24
+; SI-NEXT: s_and_b32 s22, s22, 0xff
+; SI-NEXT: s_lshl_b32 s34, s34, 8
+; SI-NEXT: s_lshl_b32 s21, s21, 24
+; SI-NEXT: s_and_b32 s20, s20, 0xff
+; SI-NEXT: s_lshl_b32 s33, s33, 8
+; SI-NEXT: s_lshl_b32 s19, s19, 24
+; SI-NEXT: s_and_b32 s18, s18, 0xff
+; SI-NEXT: s_lshl_b32 s31, s31, 8
+; SI-NEXT: s_lshl_b32 s17, s17, 24
+; SI-NEXT: s_and_b32 s16, s16, 0xff
+; SI-NEXT: s_lshl_b32 s30, s30, 8
+; SI-NEXT: s_lshl_b32 s15, s15, 24
+; SI-NEXT: s_and_b32 s14, s14, 0xff
+; SI-NEXT: s_lshl_b32 s29, s29, 8
+; SI-NEXT: s_lshl_b32 s13, s13, 24
+; SI-NEXT: s_and_b32 s12, s12, 0xff
+; SI-NEXT: s_lshl_b32 s28, s28, 8
+; SI-NEXT: s_lshl_b32 s26, s26, 16
+; SI-NEXT: s_or_b32 s3, s3, s36
+; SI-NEXT: s_lshl_b32 s24, s24, 16
+; SI-NEXT: s_or_b32 s2, s2, s35
+; SI-NEXT: s_lshl_b32 s22, s22, 16
+; SI-NEXT: s_or_b32 s1, s1, s34
+; SI-NEXT: s_lshl_b32 s20, s20, 16
+; SI-NEXT: s_or_b32 s0, s0, s33
+; SI-NEXT: s_lshl_b32 s18, s18, 16
+; SI-NEXT: s_or_b32 s7, s7, s31
+; SI-NEXT: s_lshl_b32 s16, s16, 16
+; SI-NEXT: s_or_b32 s6, s6, s30
+; SI-NEXT: s_lshl_b32 s14, s14, 16
+; SI-NEXT: s_or_b32 s5, s5, s29
+; SI-NEXT: s_lshl_b32 s12, s12, 16
+; SI-NEXT: s_or_b32 s4, s4, s28
+; SI-NEXT: s_or_b32 s26, s27, s26
+; SI-NEXT: s_and_b32 s3, s3, 0xffff
+; SI-NEXT: s_or_b32 s24, s25, s24
+; SI-NEXT: s_and_b32 s2, s2, 0xffff
+; SI-NEXT: s_or_b32 s22, s23, s22
+; SI-NEXT: s_and_b32 s1, s1, 0xffff
+; SI-NEXT: s_or_b32 s20, s21, s20
+; SI-NEXT: s_and_b32 s0, s0, 0xffff
+; SI-NEXT: s_or_b32 s18, s19, s18
+; SI-NEXT: s_and_b32 s7, s7, 0xffff
+; SI-NEXT: s_or_b32 s16, s17, s16
+; SI-NEXT: s_and_b32 s6, s6, 0xffff
+; SI-NEXT: s_or_b32 s14, s15, s14
+; SI-NEXT: s_and_b32 s5, s5, 0xffff
+; SI-NEXT: s_or_b32 s12, s13, s12
+; SI-NEXT: s_and_b32 s4, s4, 0xffff
+; SI-NEXT: s_or_b32 s3, s3, s26
+; SI-NEXT: s_or_b32 s2, s2, s24
+; SI-NEXT: s_or_b32 s1, s1, s22
+; SI-NEXT: s_or_b32 s7, s7, s18
+; SI-NEXT: s_or_b32 s6, s6, s16
+; SI-NEXT: s_or_b32 s5, s5, s14
+; SI-NEXT: s_or_b32 s4, s4, s12
+; SI-NEXT: s_or_b32 s0, s0, s20
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: v_mov_b32_e32 v1, s5
+; SI-NEXT: v_mov_b32_e32 v2, s6
+; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v0, s0
+; SI-NEXT: v_mov_b32_e32 v1, s1
+; SI-NEXT: v_mov_b32_e32 v2, s2
+; SI-NEXT: v_mov_b32_e32 v3, s3
+; SI-NEXT: s_mov_b32 s8, s9
+; SI-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0
+; SI-NEXT: s_endpgm
+;
+; VI-LABEL: amd_kernel_v32i8:
+; VI: ; %bb.0: ; %entry
+; VI-NEXT: s_load_dwordx8 s[0:7], s[0:1], 0x24
+; VI-NEXT: v_mov_b32_e32 v10, 0
+; VI-NEXT: v_mov_b32_e32 v11, 0
+; VI-NEXT: s_waitcnt lgkmcnt(0)
+; VI-NEXT: s_lshr_b32 s8, s3, 24
+; VI-NEXT: s_lshr_b32 s9, s3, 16
+; VI-NEXT: s_add_i32 s9, s9, s9
+; VI-NEXT: s_add_i32 s8, s8, s8
+; VI-NEXT: v_mov_b32_e32 v3, s3
+; VI-NEXT: s_add_i32 s3, s3, s3
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s8
+; VI-NEXT: v_mov_b32_e32 v9, s9
+; VI-NEXT: s_lshr_b32 s10, s2, 24
+; VI-NEXT: s_lshr_b32 s11, s2, 16
+; VI-NEXT: v_add_u32_sdwa v3, vcc, v3, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s3
+; VI-NEXT: s_add_i32 s11, s11, s11
+; VI-NEXT: s_add_i32 s10, s10, s10
+; VI-NEXT: v_or_b32_sdwa v3, v9, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v2, s2
+; VI-NEXT: s_add_i32 s2, s2, s2
+; VI-NEXT: v_or_b32_sdwa v3, v3, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s10
+; VI-NEXT: v_mov_b32_e32 v9, s11
+; VI-NEXT: s_lshr_b32 s12, s1, 24
+; VI-NEXT: s_lshr_b32 s13, s1, 16
+; VI-NEXT: v_add_u32_sdwa v2, vcc, v2, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s2
+; VI-NEXT: v_mov_b32_e32 v0, s4
+; VI-NEXT: s_add_i32 s13, s13, s13
+; VI-NEXT: s_add_i32 s12, s12, s12
+; VI-NEXT: v_or_b32_sdwa v2, v9, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_add_u32_sdwa v4, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_mov_b32_e32 v0, s5
+; VI-NEXT: v_mov_b32_e32 v1, s1
+; VI-NEXT: s_add_i32 s1, s1, s1
+; VI-NEXT: v_or_b32_sdwa v2, v2, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s12
+; VI-NEXT: v_mov_b32_e32 v9, s13
+; VI-NEXT: s_lshr_b32 s14, s0, 24
+; VI-NEXT: s_lshr_b32 s15, s0, 16
+; VI-NEXT: v_add_u32_sdwa v5, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_mov_b32_e32 v0, s6
+; VI-NEXT: v_add_u32_sdwa v1, vcc, v1, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s1
+; VI-NEXT: v_add_u32_sdwa v6, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_mov_b32_e32 v0, s7
+; VI-NEXT: s_add_i32 s15, s15, s15
+; VI-NEXT: s_add_i32 s14, s14, s14
+; VI-NEXT: v_or_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_add_u32_sdwa v7, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_mov_b32_e32 v0, s0
+; VI-NEXT: s_add_i32 s0, s0, s0
+; VI-NEXT: v_or_b32_sdwa v1, v1, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s14
+; VI-NEXT: v_mov_b32_e32 v9, s15
+; VI-NEXT: s_lshr_b32 s16, s7, 24
+; VI-NEXT: s_lshr_b32 s17, s7, 16
+; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s0
+; VI-NEXT: s_add_i32 s17, s17, s17
+; VI-NEXT: s_add_i32 s16, s16, s16
+; VI-NEXT: v_or_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_add_i32 s7, s7, s7
+; VI-NEXT: v_or_b32_sdwa v0, v0, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s16
+; VI-NEXT: v_mov_b32_e32 v9, s17
+; VI-NEXT: s_lshr_b32 s18, s6, 24
+; VI-NEXT: s_lshr_b32 s19, s6, 16
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s7
+; VI-NEXT: s_add_i32 s19, s19, s19
+; VI-NEXT: s_add_i32 s18, s18, s18
+; VI-NEXT: v_or_b32_sdwa v7, v9, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_add_i32 s6, s6, s6
+; VI-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s18
+; VI-NEXT: v_mov_b32_e32 v9, s19
+; VI-NEXT: s_lshr_b32 s20, s5, 24
+; VI-NEXT: s_lshr_b32 s21, s5, 16
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s6
+; VI-NEXT: s_add_i32 s21, s21, s21
+; VI-NEXT: s_add_i32 s20, s20, s20
+; VI-NEXT: v_or_b32_sdwa v6, v9, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_add_i32 s5, s5, s5
+; VI-NEXT: v_or_b32_sdwa v6, v6, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s20
+; VI-NEXT: v_mov_b32_e32 v9, s21
+; VI-NEXT: s_lshr_b32 s22, s4, 24
+; VI-NEXT: s_lshr_b32 s23, s4, 16
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s5
+; VI-NEXT: s_add_i32 s23, s23, s23
+; VI-NEXT: s_add_i32 s22, s22, s22
+; VI-NEXT: v_or_b32_sdwa v5, v9, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_add_i32 s4, s4, s4
+; VI-NEXT: v_or_b32_sdwa v5, v5, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e64 v8, 8, s22
+; VI-NEXT: v_mov_b32_e32 v9, s23
+; VI-NEXT: v_or_b32_sdwa v8, v9, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v9, s4
+; VI-NEXT: v_or_b32_sdwa v4, v9, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v4, v4, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v8, 16
+; VI-NEXT: v_mov_b32_e32 v9, 0
+; VI-NEXT: flat_store_dwordx4 v[8:9], v[4:7]
+; VI-NEXT: flat_store_dwordx4 v[10:11], v[0:3]
+; VI-NEXT: s_endpgm
+;
+; GFX11-LABEL: amd_kernel_v32i8:
+; GFX11: ; %bb.0: ; %entry
+; GFX11-NEXT: s_load_b256 s[0:7], s[0:1], 0x24
+; GFX11-NEXT: s_waitcnt lgkmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v3, 8, s2
+; GFX11-NEXT: v_lshrrev_b16 v7, 8, s3
+; GFX11-NEXT: s_lshr_b32 s21, s3, 16
+; GFX11-NEXT: s_lshr_b32 s22, s3, 24
+; GFX11-NEXT: v_add_nc_u16 v8, s3, s3
+; GFX11-NEXT: v_add_nc_u16 v9, s2, s2
+; GFX11-NEXT: v_add_nc_u16 v7, v7, v7
+; GFX11-NEXT: v_add_nc_u16 v10, s22, s22
+; GFX11-NEXT: v_add_nc_u16 v11, s21, s21
+; GFX11-NEXT: v_add_nc_u16 v3, v3, v3
+; GFX11-NEXT: v_lshrrev_b16 v2, 8, s1
+; GFX11-NEXT: v_and_b32_e32 v8, 0xff, v8
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v9, 0xff, v9
+; GFX11-NEXT: v_lshlrev_b16 v10, 8, v10
+; GFX11-NEXT: v_and_b32_e32 v11, 0xff, v11
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: s_lshr_b32 s18, s1, 16
+; GFX11-NEXT: s_lshr_b32 s19, s1, 24
+; GFX11-NEXT: s_lshr_b32 s20, s2, 24
+; GFX11-NEXT: s_lshr_b32 s2, s2, 16
+; GFX11-NEXT: v_or_b32_e32 v7, v8, v7
+; GFX11-NEXT: v_add_nc_u16 v8, s20, s20
+; GFX11-NEXT: v_or_b32_e32 v10, v11, v10
+; GFX11-NEXT: v_or_b32_e32 v3, v9, v3
+; GFX11-NEXT: v_add_nc_u16 v9, s2, s2
+; GFX11-NEXT: v_add_nc_u16 v11, s1, s1
+; GFX11-NEXT: v_add_nc_u16 v2, v2, v2
+; GFX11-NEXT: v_add_nc_u16 v12, s19, s19
+; GFX11-NEXT: v_add_nc_u16 v13, s18, s18
+; GFX11-NEXT: v_lshrrev_b16 v1, 8, s0
+; GFX11-NEXT: v_lshlrev_b16 v8, 8, v8
+; GFX11-NEXT: v_and_b32_e32 v9, 0xff, v9
+; GFX11-NEXT: v_and_b32_e32 v11, 0xff, v11
+; GFX11-NEXT: v_lshlrev_b16 v2, 8, v2
+; GFX11-NEXT: v_lshlrev_b16 v12, 8, v12
+; GFX11-NEXT: v_and_b32_e32 v13, 0xff, v13
+; GFX11-NEXT: v_and_b32_e32 v14, 0xffff, v3
+; GFX11-NEXT: v_or_b32_e32 v3, v9, v8
+; GFX11-NEXT: v_or_b32_e32 v2, v11, v2
+; GFX11-NEXT: v_add_nc_u16 v9, s0, s0
+; GFX11-NEXT: v_or_b32_e32 v8, v13, v12
+; GFX11-NEXT: v_add_nc_u16 v1, v1, v1
+; GFX11-NEXT: v_lshrrev_b16 v6, 8, s7
+; GFX11-NEXT: v_and_b32_e32 v7, 0xffff, v7
+; GFX11-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; GFX11-NEXT: v_lshlrev_b32_e32 v11, 16, v3
+; GFX11-NEXT: v_and_b32_e32 v12, 0xffff, v2
+; GFX11-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; GFX11-NEXT: v_and_b32_e32 v9, 0xff, v9
+; GFX11-NEXT: v_lshlrev_b16 v13, 8, v1
+; GFX11-NEXT: v_lshrrev_b16 v5, 8, s6
+; GFX11-NEXT: s_lshr_b32 s14, s7, 16
+; GFX11-NEXT: s_lshr_b32 s15, s7, 24
+; GFX11-NEXT: s_lshr_b32 s16, s0, 16
+; GFX11-NEXT: s_lshr_b32 s17, s0, 24
+; GFX11-NEXT: v_or_b32_e32 v3, v7, v10
+; GFX11-NEXT: v_or_b32_e32 v2, v14, v11
+; GFX11-NEXT: v_add_nc_u16 v7, s7, s7
+; GFX11-NEXT: v_or_b32_e32 v1, v12, v8
+; GFX11-NEXT: v_or_b32_e32 v8, v9, v13
+; GFX11-NEXT: v_add_nc_u16 v9, s17, s17
+; GFX11-NEXT: v_add_nc_u16 v10, s16, s16
+; GFX11-NEXT: v_add_nc_u16 v6, v6, v6
+; GFX11-NEXT: v_add_nc_u16 v11, s15, s15
+; GFX11-NEXT: v_add_nc_u16 v12, s14, s14
+; GFX11-NEXT: v_add_nc_u16 v13, s6, s6
+; GFX11-NEXT: v_add_nc_u16 v5, v5, v5
+; GFX11-NEXT: v_and_b32_e32 v7, 0xff, v7
+; GFX11-NEXT: v_lshlrev_b16 v6, 8, v6
+; GFX11-NEXT: v_lshlrev_b16 v11, 8, v11
+; GFX11-NEXT: v_and_b32_e32 v12, 0xff, v12
+; GFX11-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX11-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-NEXT: v_lshrrev_b16 v0, 8, s4
+; GFX11-NEXT: v_lshrrev_b16 v4, 8, s5
+; GFX11-NEXT: v_and_b32_e32 v13, 0xff, v13
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX11-NEXT: s_lshr_b32 s12, s6, 16
+; GFX11-NEXT: s_lshr_b32 s13, s6, 24
+; GFX11-NEXT: s_lshr_b32 s8, s4, 16
+; GFX11-NEXT: s_lshr_b32 s9, s4, 24
+; GFX11-NEXT: s_lshr_b32 s10, s5, 16
+; GFX11-NEXT: s_lshr_b32 s11, s5, 24
+; GFX11-NEXT: v_or_b32_e32 v6, v7, v6
+; GFX11-NEXT: v_or_b32_e32 v7, v12, v11
+; GFX11-NEXT: v_and_b32_e32 v12, 0xffff, v8
+; GFX11-NEXT: v_or_b32_e32 v8, v10, v9
+; GFX11-NEXT: v_add_nc_u16 v9, s13, s13
+; GFX11-NEXT: v_add_nc_u16 v10, s12, s12
+; GFX11-NEXT: v_or_b32_e32 v5, v13, v5
+; GFX11-NEXT: v_add_nc_u16 v11, s5, s5
+; GFX11-NEXT: v_add_nc_u16 v4, v4, v4
+; GFX11-NEXT: v_add_nc_u16 v13, s11, s11
+; GFX11-NEXT: v_add_nc_u16 v14, s10, s10
+; GFX11-NEXT: v_add_nc_u16 v15, s4, s4
+; GFX11-NEXT: v_add_nc_u16 v0, v0, v0
+; GFX11-NEXT: v_add_nc_u16 v16, s9, s9
+; GFX11-NEXT: v_add_nc_u16 v17, s8, s8
+; GFX11-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX11-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-NEXT: v_and_b32_e32 v11, 0xff, v11
+; GFX11-NEXT: v_lshlrev_b16 v4, 8, v4
+; GFX11-NEXT: v_lshlrev_b16 v13, 8, v13
+; GFX11-NEXT: v_and_b32_e32 v14, 0xff, v14
+; GFX11-NEXT: v_and_b32_e32 v15, 0xff, v15
+; GFX11-NEXT: v_lshlrev_b16 v0, 8, v0
+; GFX11-NEXT: v_lshlrev_b16 v16, 8, v16
+; GFX11-NEXT: v_and_b32_e32 v17, 0xff, v17
+; GFX11-NEXT: v_or_b32_e32 v9, v10, v9
+; GFX11-NEXT: v_or_b32_e32 v4, v11, v4
+; GFX11-NEXT: v_or_b32_e32 v10, v14, v13
+; GFX11-NEXT: v_or_b32_e32 v0, v15, v0
+; GFX11-NEXT: v_or_b32_e32 v11, v17, v16
+; GFX11-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; GFX11-NEXT: v_lshlrev_b32_e32 v7, 16, v7
+; GFX11-NEXT: v_and_b32_e32 v5, 0xffff, v5
+; GFX11-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; GFX11-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; GFX11-NEXT: v_lshlrev_b32_e32 v13, 16, v8
+; GFX11-NEXT: v_or_b32_e32 v7, v6, v7
+; GFX11-NEXT: v_or_b32_e32 v6, v5, v9
+; GFX11-NEXT: v_mov_b32_e32 v8, 16
+; GFX11-NEXT: v_mov_b32_e32 v9, 0
+; GFX11-NEXT: v_or_b32_e32 v5, v4, v10
+; GFX11-NEXT: v_or_b32_e32 v4, v0, v11
+; GFX11-NEXT: v_mov_b32_e32 v10, 0
+; GFX11-NEXT: v_mov_b32_e32 v11, 0
+; GFX11-NEXT: v_or_b32_e32 v0, v12, v13
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b128 v[8:9], v[4:7], off
+; GFX11-NEXT: global_store_b128 v[10:11], v[0:3], off
+; GFX11-NEXT: s_nop 0
+; GFX11-NEXT: s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT: s_endpgm
+entry:
+ %add = add <32 x i8> %arg0, %arg0
+ store <32 x i8> %add, ptr addrspace(1) null
+ ret void
+}
+
attributes #0 = { nounwind noinline }
diff --git a/llvm/test/CodeGen/AMDGPU/function-args.ll b/llvm/test/CodeGen/AMDGPU/function-args.ll
index 8edefdd1dd15cc8..01dcc26566663be 100644
--- a/llvm/test/CodeGen/AMDGPU/function-args.ll
+++ b/llvm/test/CodeGen/AMDGPU/function-args.ll
@@ -920,6 +920,50 @@ define void @void_func_v16i64(<16 x i64> %arg0) #0 {
ret void
}
+define void @void_func_v2i8(<2 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v2i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: s_mov_b32 s4, 0
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: s_mov_b32 s5, s4
+; CI-NEXT: buffer_store_short v0, off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v2i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: s_mov_b32 s4, 0
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: s_mov_b32 s5, s4
+; GFX89-NEXT: buffer_store_short v0, off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v2i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: s_mov_b32 s0, 0
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: s_mov_b32 s1, s0
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: buffer_store_b16 v0, off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <2 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
define void @void_func_v2i16(<2 x i16> %arg0) #0 {
; CI-LABEL: void_func_v2i16:
; CI: ; %bb.0:
@@ -953,6 +997,628 @@ define void @void_func_v2i16(<2 x i16> %arg0) #0 {
ret void
}
+define void @void_func_v3i8(<3 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v3i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: s_mov_b32 s5, 0
+; CI-NEXT: s_mov_b32 s4, 2
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: buffer_store_byte v2, off, s[4:7], 0
+; CI-NEXT: s_mov_b32 s4, s5
+; CI-NEXT: buffer_store_short v0, off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v3i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: s_mov_b32 s5, 0
+; GFX89-NEXT: s_mov_b32 s4, 2
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: buffer_store_byte v2, off, s[4:7], 0
+; GFX89-NEXT: s_mov_b32 s4, s5
+; GFX89-NEXT: buffer_store_short v0, off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v3i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: s_mov_b32 s1, 0
+; GFX11-NEXT: s_mov_b32 s0, 2
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: buffer_store_b8 v2, off, s[0:3], 0
+; GFX11-NEXT: s_mov_b32 s0, s1
+; GFX11-NEXT: buffer_store_b16 v0, off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <3 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
+define void @void_func_v4i8(<4 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v4i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: v_or_b32_e32 v2, v3, v2
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: s_mov_b32 s4, 0
+; CI-NEXT: v_or_b32_e32 v0, v0, v2
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: s_mov_b32 s5, s4
+; CI-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v4i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX89-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s4, 0
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: s_mov_b32 s5, s4
+; GFX89-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v4i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: s_mov_b32 s0, 0
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: s_mov_b32 s1, s0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: buffer_store_b32 v0, off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <4 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
+define void @void_func_v5i8(<5 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v5i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: v_or_b32_e32 v2, v3, v2
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: s_mov_b32 s5, 0
+; CI-NEXT: s_mov_b32 s4, 4
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: v_or_b32_e32 v0, v0, v2
+; CI-NEXT: buffer_store_byte v4, off, s[4:7], 0
+; CI-NEXT: s_mov_b32 s4, s5
+; CI-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v5i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX89-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s5, 0
+; GFX89-NEXT: s_mov_b32 s4, 4
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: buffer_store_byte v4, off, s[4:7], 0
+; GFX89-NEXT: s_mov_b32 s4, s5
+; GFX89-NEXT: buffer_store_dword v0, off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v5i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: s_mov_b32 s1, 0
+; GFX11-NEXT: s_mov_b32 s0, 4
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: buffer_store_b8 v4, off, s[0:3], 0
+; GFX11-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-NEXT: s_mov_b32 s0, s1
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: buffer_store_b32 v0, off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <5 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
+define void @void_func_v8i8(<8 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v8i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; CI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; CI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
+; CI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; CI-NEXT: v_or_b32_e32 v4, v4, v5
+; CI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: v_or_b32_e32 v6, v7, v6
+; CI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; CI-NEXT: v_or_b32_e32 v2, v3, v2
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: s_mov_b32 s4, 0
+; CI-NEXT: v_or_b32_e32 v4, v4, v6
+; CI-NEXT: v_or_b32_e32 v3, v0, v2
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: s_mov_b32 s5, s4
+; CI-NEXT: buffer_store_dwordx2 v[3:4], off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v8i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v5, 8, v7
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX89-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s4, 0
+; GFX89-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v3, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: s_mov_b32 s5, s4
+; GFX89-NEXT: buffer_store_dwordx2 v[3:4], off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v8i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v4
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: v_or_b32_e32 v5, v6, v7
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: s_mov_b32 s0, 0
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xffff, v4
+; GFX11-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: v_lshlrev_b32_e32 v4, 16, v1
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: s_mov_b32 s1, s0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v4
+; GFX11-NEXT: buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <8 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
+define void @void_func_v16i8(<16 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v16i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_and_b32_e32 v14, 0xff, v14
+; CI-NEXT: v_lshlrev_b32_e32 v13, 8, v13
+; CI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; CI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; CI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; CI-NEXT: v_and_b32_e32 v8, 0xff, v8
+; CI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; CI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; CI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: v_lshlrev_b32_e32 v15, 24, v15
+; CI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; CI-NEXT: v_or_b32_e32 v12, v12, v13
+; CI-NEXT: v_lshlrev_b32_e32 v11, 24, v11
+; CI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; CI-NEXT: v_or_b32_e32 v8, v8, v9
+; CI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
+; CI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; CI-NEXT: v_or_b32_e32 v4, v4, v5
+; CI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: v_or_b32_e32 v14, v15, v14
+; CI-NEXT: v_and_b32_e32 v12, 0xffff, v12
+; CI-NEXT: v_or_b32_e32 v10, v11, v10
+; CI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; CI-NEXT: v_or_b32_e32 v6, v7, v6
+; CI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; CI-NEXT: v_or_b32_e32 v2, v3, v2
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: s_mov_b32 s4, 0
+; CI-NEXT: v_or_b32_e32 v12, v12, v14
+; CI-NEXT: v_or_b32_e32 v11, v8, v10
+; CI-NEXT: v_or_b32_e32 v10, v4, v6
+; CI-NEXT: v_or_b32_e32 v9, v0, v2
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: s_mov_b32 s5, s4
+; CI-NEXT: buffer_store_dwordx4 v[9:12], off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v16i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v13, 8, v13
+; GFX89-NEXT: v_lshlrev_b16_e32 v9, 8, v9
+; GFX89-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: v_or_b32_sdwa v12, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v13, 8, v15
+; GFX89-NEXT: v_or_b32_sdwa v8, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v9, 8, v11
+; GFX89-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v5, 8, v7
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX89-NEXT: v_or_b32_sdwa v13, v14, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v9, v10, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s4, 0
+; GFX89-NEXT: v_or_b32_sdwa v12, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v11, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v10, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: s_mov_b32 s5, s4
+; GFX89-NEXT: buffer_store_dwordx4 v[9:12], off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v16i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v13, 8, v13
+; GFX11-NEXT: v_and_b32_e32 v12, 0xff, v12
+; GFX11-NEXT: v_lshlrev_b16 v15, 8, v15
+; GFX11-NEXT: v_and_b32_e32 v14, 0xff, v14
+; GFX11-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX11-NEXT: v_and_b32_e32 v8, 0xff, v8
+; GFX11-NEXT: v_lshlrev_b16 v11, 8, v11
+; GFX11-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v4
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_or_b32_e32 v12, v12, v13
+; GFX11-NEXT: v_or_b32_e32 v13, v14, v15
+; GFX11-NEXT: v_or_b32_e32 v8, v8, v9
+; GFX11-NEXT: v_or_b32_e32 v10, v10, v11
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: v_or_b32_e32 v5, v6, v7
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: v_and_b32_e32 v9, 0xffff, v12
+; GFX11-NEXT: v_lshlrev_b32_e32 v12, 16, v13
+; GFX11-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; GFX11-NEXT: v_lshlrev_b32_e32 v2, 16, v10
+; GFX11-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v6, 16, v1
+; GFX11-NEXT: v_or_b32_e32 v3, v9, v12
+; GFX11-NEXT: v_or_b32_e32 v2, v8, v2
+; GFX11-NEXT: v_or_b32_e32 v1, v4, v5
+; GFX11-NEXT: s_mov_b32 s0, 0
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v6
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: s_mov_b32 s1, s0
+; GFX11-NEXT: buffer_store_b128 v[0:3], off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <16 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
+define void @void_func_v32i8(<32 x i8> %arg0) #0 {
+; CI-LABEL: void_func_v32i8:
+; CI: ; %bb.0:
+; CI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; CI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; CI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; CI-NEXT: v_or_b32_e32 v2, v3, v2
+; CI-NEXT: buffer_load_dword v3, off, s[0:3], s32
+; CI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; CI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; CI-NEXT: v_or_b32_e32 v4, v4, v5
+; CI-NEXT: v_lshlrev_b32_e32 v5, 24, v7
+; CI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; CI-NEXT: v_lshlrev_b32_e32 v13, 8, v13
+; CI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; CI-NEXT: v_or_b32_e32 v5, v5, v6
+; CI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; CI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; CI-NEXT: v_or_b32_e32 v12, v12, v13
+; CI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; CI-NEXT: v_and_b32_e32 v8, 0xff, v8
+; CI-NEXT: v_and_b32_e32 v13, 0xff, v14
+; CI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; CI-NEXT: v_or_b32_e32 v7, v4, v5
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v29
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v28
+; CI-NEXT: v_and_b32_e32 v6, 0xff, v26
+; CI-NEXT: v_or_b32_e32 v8, v8, v9
+; CI-NEXT: v_lshlrev_b32_e32 v9, 24, v15
+; CI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
+; CI-NEXT: v_lshlrev_b32_e32 v11, 24, v11
+; CI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; CI-NEXT: v_or_b32_e32 v1, v4, v1
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v30
+; CI-NEXT: v_lshlrev_b32_e32 v5, 24, v27
+; CI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: v_or_b32_e32 v9, v9, v13
+; CI-NEXT: v_or_b32_e32 v10, v11, v10
+; CI-NEXT: v_and_b32_e32 v11, 0xffff, v12
+; CI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; CI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; CI-NEXT: v_or_b32_e32 v5, v5, v6
+; CI-NEXT: v_or_b32_e32 v6, v0, v2
+; CI-NEXT: v_or_b32_e32 v9, v11, v9
+; CI-NEXT: v_or_b32_e32 v8, v8, v10
+; CI-NEXT: v_lshlrev_b32_e32 v10, 8, v25
+; CI-NEXT: v_and_b32_e32 v11, 0xff, v24
+; CI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; CI-NEXT: s_mov_b32 s5, 0
+; CI-NEXT: s_mov_b32 s4, 16
+; CI-NEXT: s_mov_b32 s7, 0xf000
+; CI-NEXT: s_mov_b32 s6, -1
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: v_lshlrev_b32_e32 v0, 24, v3
+; CI-NEXT: v_or_b32_e32 v0, v0, v4
+; CI-NEXT: v_or_b32_e32 v3, v1, v0
+; CI-NEXT: v_or_b32_e32 v0, v11, v10
+; CI-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; CI-NEXT: v_and_b32_e32 v1, 0xff, v22
+; CI-NEXT: v_or_b32_e32 v2, v0, v5
+; CI-NEXT: v_lshlrev_b32_e32 v0, 24, v23
+; CI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; CI-NEXT: v_or_b32_e32 v0, v0, v1
+; CI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v20
+; CI-NEXT: v_or_b32_e32 v1, v4, v1
+; CI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; CI-NEXT: v_and_b32_e32 v4, 0xff, v18
+; CI-NEXT: v_or_b32_e32 v1, v1, v0
+; CI-NEXT: v_lshlrev_b32_e32 v0, 24, v19
+; CI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; CI-NEXT: v_or_b32_e32 v0, v0, v4
+; CI-NEXT: v_lshlrev_b32_e32 v4, 8, v17
+; CI-NEXT: v_and_b32_e32 v5, 0xff, v16
+; CI-NEXT: v_or_b32_e32 v4, v5, v4
+; CI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; CI-NEXT: v_or_b32_e32 v0, v4, v0
+; CI-NEXT: buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; CI-NEXT: s_mov_b32 s4, s5
+; CI-NEXT: buffer_store_dwordx4 v[6:9], off, s[4:7], 0
+; CI-NEXT: s_waitcnt vmcnt(0)
+; CI-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX89-LABEL: void_func_v32i8:
+; GFX89: ; %bb.0:
+; GFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v13, 8, v13
+; GFX89-NEXT: v_or_b32_sdwa v12, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v13, 8, v15
+; GFX89-NEXT: v_or_b32_sdwa v13, v14, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: buffer_load_ubyte v14, off, s[0:3], s32
+; GFX89-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX89-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v5, 8, v7
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX89-NEXT: v_lshlrev_b16_e32 v9, 8, v9
+; GFX89-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v6, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v29
+; GFX89-NEXT: v_or_b32_sdwa v8, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v9, 8, v11
+; GFX89-NEXT: v_or_b32_sdwa v7, v28, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v25
+; GFX89-NEXT: v_or_b32_sdwa v9, v10, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v10, v24, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v27
+; GFX89-NEXT: v_lshlrev_b16_e32 v2, 8, v23
+; GFX89-NEXT: v_or_b32_sdwa v11, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_lshlrev_b16_e32 v1, 8, v21
+; GFX89-NEXT: v_lshlrev_b16_e32 v3, 8, v17
+; GFX89-NEXT: v_lshlrev_b16_e32 v15, 8, v19
+; GFX89-NEXT: v_or_b32_sdwa v19, v22, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v2, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v17, v20, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v16, v16, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v15, v18, v15 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: s_mov_b32 s5, 0
+; GFX89-NEXT: s_mov_b32 s4, 16
+; GFX89-NEXT: s_mov_b32 s7, 0xf000
+; GFX89-NEXT: s_mov_b32 s6, -1
+; GFX89-NEXT: v_or_b32_sdwa v1, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v0, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v6, v10, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v5, v17, v19 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v4, v16, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v3, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: v_lshlrev_b16_e32 v8, 8, v14
+; GFX89-NEXT: v_or_b32_sdwa v8, v30, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX89-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX89-NEXT: buffer_store_dwordx4 v[4:7], off, s[4:7], 0
+; GFX89-NEXT: s_mov_b32 s4, s5
+; GFX89-NEXT: buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; GFX89-NEXT: s_waitcnt vmcnt(0)
+; GFX89-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: void_func_v32i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: scratch_load_u8 v31, off, s32
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX11-NEXT: v_and_b32_e32 v8, 0xff, v8
+; GFX11-NEXT: v_lshlrev_b16 v11, 8, v11
+; GFX11-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-NEXT: v_lshlrev_b16 v17, 8, v17
+; GFX11-NEXT: v_and_b32_e32 v16, 0xff, v16
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: v_lshlrev_b16 v13, 8, v13
+; GFX11-NEXT: v_and_b32_e32 v12, 0xff, v12
+; GFX11-NEXT: v_lshlrev_b16 v15, 8, v15
+; GFX11-NEXT: v_and_b32_e32 v14, 0xff, v14
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v4
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX11-NEXT: v_lshlrev_b16 v29, 8, v29
+; GFX11-NEXT: v_and_b32_e32 v28, 0xff, v28
+; GFX11-NEXT: v_and_b32_e32 v30, 0xff, v30
+; GFX11-NEXT: v_lshlrev_b16 v25, 8, v25
+; GFX11-NEXT: v_and_b32_e32 v24, 0xff, v24
+; GFX11-NEXT: v_lshlrev_b16 v27, 8, v27
+; GFX11-NEXT: v_and_b32_e32 v26, 0xff, v26
+; GFX11-NEXT: v_lshlrev_b16 v21, 8, v21
+; GFX11-NEXT: v_and_b32_e32 v20, 0xff, v20
+; GFX11-NEXT: v_lshlrev_b16 v23, 8, v23
+; GFX11-NEXT: v_and_b32_e32 v22, 0xff, v22
+; GFX11-NEXT: v_lshlrev_b16 v19, 8, v19
+; GFX11-NEXT: v_and_b32_e32 v18, 0xff, v18
+; GFX11-NEXT: v_or_b32_e32 v8, v8, v9
+; GFX11-NEXT: v_or_b32_e32 v9, v10, v11
+; GFX11-NEXT: v_or_b32_e32 v11, v16, v17
+; GFX11-NEXT: v_lshlrev_b32_e32 v17, 16, v1
+; GFX11-NEXT: v_or_b32_e32 v12, v12, v13
+; GFX11-NEXT: v_or_b32_e32 v13, v14, v15
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: v_or_b32_e32 v5, v6, v7
+; GFX11-NEXT: v_or_b32_e32 v2, v28, v29
+; GFX11-NEXT: v_or_b32_e32 v3, v24, v25
+; GFX11-NEXT: v_or_b32_e32 v6, v26, v27
+; GFX11-NEXT: v_or_b32_e32 v7, v20, v21
+; GFX11-NEXT: v_or_b32_e32 v10, v22, v23
+; GFX11-NEXT: v_or_b32_e32 v14, v18, v19
+; GFX11-NEXT: v_and_b32_e32 v15, 0xffff, v4
+; GFX11-NEXT: v_lshlrev_b32_e32 v16, 16, v5
+; GFX11-NEXT: v_and_b32_e32 v18, 0xffff, v2
+; GFX11-NEXT: v_and_b32_e32 v4, 0xffff, v3
+; GFX11-NEXT: v_lshlrev_b32_e32 v5, 16, v6
+; GFX11-NEXT: v_and_b32_e32 v7, 0xffff, v7
+; GFX11-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; GFX11-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; GFX11-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; GFX11-NEXT: v_and_b32_e32 v12, 0xffff, v12
+; GFX11-NEXT: v_lshlrev_b32_e32 v13, 16, v13
+; GFX11-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; GFX11-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_or_b32_e32 v6, v4, v5
+; GFX11-NEXT: v_or_b32_e32 v5, v7, v10
+; GFX11-NEXT: v_or_b32_e32 v4, v11, v14
+; GFX11-NEXT: v_or_b32_e32 v3, v12, v13
+; GFX11-NEXT: v_or_b32_e32 v2, v8, v9
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v17
+; GFX11-NEXT: s_mov_b32 s1, 0
+; GFX11-NEXT: s_mov_b32 s0, 16
+; GFX11-NEXT: s_mov_b32 s3, 0x31016000
+; GFX11-NEXT: s_mov_b32 s2, -1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v31
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_or_b32_e32 v1, v30, v1
+; GFX11-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_or_b32_e32 v7, v18, v1
+; GFX11-NEXT: v_or_b32_e32 v1, v15, v16
+; GFX11-NEXT: buffer_store_b128 v[4:7], off, s[0:3], 0
+; GFX11-NEXT: s_mov_b32 s0, s1
+; GFX11-NEXT: buffer_store_b128 v[0:3], off, s[0:3], 0
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+ store <32 x i8> %arg0, ptr addrspace(1) null
+ ret void
+}
+
define void @void_func_v3i16(<3 x i16> %arg0) #0 {
; CI-LABEL: void_func_v3i16:
; CI: ; %bb.0:
@@ -3499,8 +4165,8 @@ define void @void_func_v3i32_wasted_reg(<3 x i32> %arg0, i32 %arg1) #0 {
}
; Check there is no crash.
-define void @void_func_v16i8(<16 x i8> %arg0) #0 {
-; CIGFX89-LABEL: void_func_v16i8:
+define void @void_func_volatile_v16i8(<16 x i8> %arg0) #0 {
+; CIGFX89-LABEL: void_func_volatile_v16i8:
; CIGFX89: ; %bb.0:
; CIGFX89-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CIGFX89-NEXT: s_mov_b32 s7, 0xf000
@@ -3539,7 +4205,7 @@ define void @void_func_v16i8(<16 x i8> %arg0) #0 {
; CIGFX89-NEXT: s_waitcnt vmcnt(0)
; CIGFX89-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-LABEL: void_func_v16i8:
+; GFX11-LABEL: void_func_volatile_v16i8:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: s_mov_b32 s3, 0x31016000
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
index 439670d65b16d91..c01634231403241 100644
--- a/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
@@ -11,6 +11,21 @@ declare hidden amdgpu_gfx void @external_void_func_i1_zeroext(i1 zeroext) #0
declare hidden amdgpu_gfx void @external_void_func_i8(i8) #0
declare hidden amdgpu_gfx void @external_void_func_i8_signext(i8 signext) #0
declare hidden amdgpu_gfx void @external_void_func_i8_zeroext(i8 zeroext) #0
+declare hidden amdgpu_gfx void @external_void_func_v2i8(<2 x i8>) #0
+declare hidden amdgpu_gfx void @external_void_func_v3i8(<3 x i8>) #0
+declare hidden amdgpu_gfx void @external_void_func_v4i8(<4 x i8>) #0
+declare hidden amdgpu_gfx void @external_void_func_v5i8(<5 x i8>) #0
+declare hidden amdgpu_gfx void @external_void_func_v8i8(<8 x i8>) #0
+declare hidden amdgpu_gfx void @external_void_func_v16i8(<8 x i8>) #0
+declare hidden amdgpu_gfx void @external_void_func_v32i8(<32 x i8>) #0
+
+declare hidden amdgpu_gfx i8 @external_void_func_i8_ret(i8) #0
+declare hidden amdgpu_gfx <2 x i8> @external_void_func_v2i8_ret(<2 x i8>) #0
+declare hidden amdgpu_gfx <3 x i8> @external_void_func_v3i8_ret(<3 x i8>) #0
+declare hidden amdgpu_gfx <4 x i8> @external_void_func_v4i8_ret(<4 x i8>) #0
+declare hidden amdgpu_gfx <5 x i8> @external_void_func_v5i8_ret(<5 x i8>) #0
+declare hidden amdgpu_gfx <8 x i8> @external_void_func_v8i8_ret(<8 x i8>) #0
+declare hidden amdgpu_gfx <32 x i8> @external_void_func_v32i8_ret(<32 x i8>) #0
declare hidden amdgpu_gfx void @external_void_func_i16(i16) #0
declare hidden amdgpu_gfx void @external_void_func_i16_signext(i16 signext) #0
@@ -91,8 +106,6 @@ declare hidden amdgpu_gfx void @external_void_func_struct_i8_i32({ i8, i32 }) #0
declare hidden amdgpu_gfx void @external_void_func_byval_struct_i8_i32(ptr addrspace(5) byval({ i8, i32 })) #0
declare hidden amdgpu_gfx void @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32(ptr addrspace(5) sret({ i8, i32 }), ptr addrspace(5) byval({ i8, i32 })) #0
-declare hidden amdgpu_gfx void @external_void_func_v16i8(<16 x i8>) #0
-
define amdgpu_gfx void @test_call_external_void_func_i1_imm() #0 {
; GFX9-LABEL: test_call_external_void_func_i1_imm:
; GFX9: ; %bb.0:
@@ -3040,6 +3053,2749 @@ define amdgpu_gfx void @test_call_external_void_func_v3f64_imm() #0 {
ret void
}
+define amdgpu_gfx void @test_call_external_void_func_v2i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: global_load_ushort v0, v[0:1], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v2i8 at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v2i8 at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b16_e32 v1, 8, v0
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v2i8 at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v2i8 at rel32@hi+12
+; GFX10-NEXT: global_load_ushort v0, v[0:1], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b16 v1, 8, v0
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v2i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v2i8 at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v2i8 at rel32@hi+12
+; GFX11-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v1, 8, v0
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v2i8:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v2i8 at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v2i8 at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_ushort v0, v[0:1], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b16 v1, 8, v0
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <2 x i8>, ptr addrspace(1) null
+ call amdgpu_gfx void @external_void_func_v2i8(<2 x i8> %val)
+ ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v3i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: global_load_dword v0, v[0:1], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v3i8 at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v3i8 at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v3i8 at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v3i8 at rel32@hi+12
+; GFX10-NEXT: global_load_dword v0, v[0:1], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v3i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v3i8 at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v3i8 at rel32@hi+12
+; GFX11-NEXT: global_load_b32 v0, v[0:1], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v3i8:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v3i8 at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v3i8 at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dword v0, v[0:1], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <3 x i8>, ptr addrspace(1) null
+ call amdgpu_gfx void @external_void_func_v3i8(<3 x i8> %val)
+ ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v4i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: global_load_dword v0, v[0:1], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v4i8 at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v4i8 at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v4i8 at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v4i8 at rel32@hi+12
+; GFX10-NEXT: global_load_dword v0, v[0:1], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v4i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v4i8 at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v4i8 at rel32@hi+12
+; GFX11-NEXT: global_load_b32 v0, v[0:1], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v4i8:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v4i8 at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v4i8 at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dword v0, v[0:1], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <4 x i8>, ptr addrspace(1) null
+ call amdgpu_gfx void @external_void_func_v4i8(<4 x i8> %val)
+ ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v5i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v5i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: global_load_dwordx2 v[5:6], v[0:1], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v5i8 at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v5i8 at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
+; GFX9-NEXT: v_mov_b32_e32 v4, v6
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v5i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v5i8 at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v5i8 at rel32@hi+12
+; GFX10-NEXT: global_load_dwordx2 v[5:6], v[0:1], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX10-NEXT: v_mov_b32_e32 v0, v5
+; GFX10-NEXT: v_mov_b32_e32 v4, v6
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v5i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v5i8 at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v5i8 at rel32@hi+12
+; GFX11-NEXT: global_load_b64 v[5:6], v[0:1], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_mov_b32_e32 v0, v5
+; GFX11-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX11-NEXT: v_mov_b32_e32 v4, v6
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v5i8:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v5i8 at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v5i8 at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dwordx2 v[5:6], v[0:1], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, v5
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, v6
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <5 x i8>, ptr addrspace(1) null
+ call amdgpu_gfx void @external_void_func_v5i8(<5 x i8> %val)
+ ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v8i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v8i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v8i8 at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v8i8 at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX9-NEXT: v_mov_b32_e32 v4, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v8
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v8i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v8i8 at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v8i8 at rel32@hi+12
+; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v8
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v8i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v8i8 at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v8i8 at rel32@hi+12
+; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX11-NEXT: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v1, v8
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v8i8:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v8i8 at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v8i8 at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, v8
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <8 x i8>, ptr addrspace(1) null
+ call amdgpu_gfx void @external_void_func_v8i8(<8 x i8> %val)
+ ret void
+}
+
+define amdgpu_gfx void @test_call_external_void_func_v32i8() #0 {
+; GFX9-LABEL: test_call_external_void_func_v32i8:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NEXT: v_mov_b32_e32 v4, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: v_mov_b32_e32 v5, 0
+; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: global_load_dwordx4 v[16:19], v[4:5], off
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v32i8 at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v32i8 at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX9-NEXT: v_mov_b32_e32 v4, v1
+; GFX9-NEXT: v_mov_b32_e32 v8, v2
+; GFX9-NEXT: v_mov_b32_e32 v12, v3
+; GFX9-NEXT: v_mov_b32_e32 v20, v17
+; GFX9-NEXT: v_mov_b32_e32 v24, v18
+; GFX9-NEXT: v_mov_b32_e32 v28, v19
+; GFX9-NEXT: v_mov_b32_e32 v1, v35
+; GFX9-NEXT: v_mov_b32_e32 v2, v36
+; GFX9-NEXT: v_mov_b32_e32 v3, v37
+; GFX9-NEXT: v_mov_b32_e32 v17, v32
+; GFX9-NEXT: v_mov_b32_e32 v18, v33
+; GFX9-NEXT: v_mov_b32_e32 v19, v34
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v32i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-NEXT: v_mov_b32_e32 v4, 16
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_mov_b32_e32 v5, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v32i8 at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v32i8 at rel32@hi+12
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-NEXT: global_load_dwordx4 v[16:19], v[4:5], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(1)
+; GFX10-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX10-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX10-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX10-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX10-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX10-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX10-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX10-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX10-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX10-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX10-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX10-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-NEXT: v_mov_b32_e32 v8, v2
+; GFX10-NEXT: v_mov_b32_e32 v12, v3
+; GFX10-NEXT: v_mov_b32_e32 v20, v17
+; GFX10-NEXT: v_mov_b32_e32 v24, v18
+; GFX10-NEXT: v_mov_b32_e32 v28, v19
+; GFX10-NEXT: v_mov_b32_e32 v1, v35
+; GFX10-NEXT: v_mov_b32_e32 v2, v36
+; GFX10-NEXT: v_mov_b32_e32 v3, v37
+; GFX10-NEXT: v_mov_b32_e32 v17, v32
+; GFX10-NEXT: v_mov_b32_e32 v18, v33
+; GFX10-NEXT: v_mov_b32_e32 v19, v34
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v32i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v4, 16
+; GFX11-NEXT: v_mov_b32_e32 v5, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v32i8 at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v32i8 at rel32@hi+12
+; GFX11-NEXT: global_load_b128 v[16:19], v[4:5], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(1)
+; GFX11-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX11-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX11-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX11-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX11-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX11-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX11-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX11-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX11-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX11-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX11-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX11-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX11-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX11-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX11-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX11-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX11-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX11-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX11-NEXT: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v1, v35
+; GFX11-NEXT: v_mov_b32_e32 v8, v2
+; GFX11-NEXT: v_mov_b32_e32 v12, v3
+; GFX11-NEXT: v_mov_b32_e32 v20, v17
+; GFX11-NEXT: v_mov_b32_e32 v24, v18
+; GFX11-NEXT: v_dual_mov_b32 v28, v19 :: v_dual_mov_b32 v19, v34
+; GFX11-NEXT: v_dual_mov_b32 v2, v36 :: v_dual_mov_b32 v3, v37
+; GFX11-NEXT: v_dual_mov_b32 v17, v32 :: v_dual_mov_b32 v18, v33
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v32i8:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, 16
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v5, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v32i8 at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v32i8 at rel32@hi+12
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
+; GFX10-SCRATCH-NEXT: global_load_dwordx4 v[16:19], v[4:5], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(1)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v8, v2
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v12, v3
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v20, v17
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v24, v18
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v28, v19
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, v35
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v2, v36
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v3, v37
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v17, v32
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v18, v33
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v19, v34
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <32 x i8>, ptr addrspace(1) null
+ call amdgpu_gfx void @external_void_func_v32i8(<32 x i8> %val)
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: global_load_ubyte v0, v[41:42], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_i8_ret at rel32@hi+12
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: global_store_byte v[41:42], v0, off
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_i8_ret at rel32@hi+12
+; GFX10-NEXT: global_load_ubyte v0, v[41:42], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: global_store_byte v[41:42], v0, off
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v42, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_mov_b32_e32 v42, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_u8 v0, v[41:42], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: global_store_b8 v[41:42], v0, off
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_load_b32 v42, off, s33
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_ubyte v0, v[41:42], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: global_store_byte v[41:42], v0, off
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load i8, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx i8 @external_void_func_i8_ret(i8 %val)
+ store i8 %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_v2i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_v2i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: global_load_ushort v0, v[41:42], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v2i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v2i8_ret at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b16_e32 v1, 8, v0
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_short v[41:42], v0, off
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v2i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v2i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v2i8_ret at rel32@hi+12
+; GFX10-NEXT: global_load_ushort v0, v[41:42], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b16 v1, 8, v0
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: global_store_short v[41:42], v0, off
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v2i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v42, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_mov_b32_e32 v42, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v2i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v2i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_u16 v0, v[41:42], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b16 v1, 8, v0
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_4)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: global_store_b16 v[41:42], v0, off
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_load_b32 v42, off, s33
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v2i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v2i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v2i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_ushort v0, v[41:42], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b16 v1, 8, v0
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: global_store_short v[41:42], v0, off
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <2 x i8>, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx <2 x i8> @external_void_func_v2i8_ret(<2 x i8> %val)
+ store <2 x i8> %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_v3i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_v3i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: global_load_dword v0, v[41:42], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v3i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v3i8_ret at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_mov_b32_e32 v3, 2
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_byte v[3:4], v2, off
+; GFX9-NEXT: global_store_short v[41:42], v0, off
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v3i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v3i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v3i8_ret at rel32@hi+12
+; GFX10-NEXT: global_load_dword v0, v[41:42], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_mov_b32_e32 v3, 2
+; GFX10-NEXT: v_mov_b32_e32 v4, 0
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: global_store_byte v[3:4], v2, off
+; GFX10-NEXT: global_store_short v[41:42], v0, off
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v3i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v42, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_mov_b32_e32 v42, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v3i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v3i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_b32 v0, v[41:42], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v0
+; GFX11-NEXT: v_mov_b32_e32 v0, 2
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_or_b32_e32 v3, v4, v3
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b8 v[0:1], v2, off
+; GFX11-NEXT: global_store_b16 v[41:42], v3, off
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_load_b32 v42, off, s33
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v3i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v3i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v3i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dword v0, v[41:42], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v3, 2
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: global_store_byte v[3:4], v2, off
+; GFX10-SCRATCH-NEXT: global_store_short v[41:42], v0, off
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <3 x i8>, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx <3 x i8> @external_void_func_v3i8_ret(<3 x i8> %val)
+ store <3 x i8> %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_v4i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_v4i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: global_load_dword v0, v[41:42], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v4i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v4i8_ret at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_dword v[41:42], v0, off
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v4i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v4i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v4i8_ret at rel32@hi+12
+; GFX10-NEXT: global_load_dword v0, v[41:42], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v1, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: global_store_dword v[41:42], v0, off
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v4i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v42, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_mov_b32_e32 v42, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v4i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v4i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_b32 v0, v[41:42], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: global_store_b32 v[41:42], v0, off
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_load_b32 v42, off, s33
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v4i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v4i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v4i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dword v0, v[41:42], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v1, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: global_store_dword v[41:42], v0, off
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <4 x i8>, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx <4 x i8> @external_void_func_v4i8_ret(<4 x i8> %val)
+ store <4 x i8> %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_v5i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_v5i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: global_load_dwordx2 v[5:6], v[41:42], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v5i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v5i8_ret at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX9-NEXT: v_mov_b32_e32 v0, v5
+; GFX9-NEXT: v_mov_b32_e32 v4, v6
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_or_b32_sdwa v5, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v0, 8, v3
+; GFX9-NEXT: v_or_b32_sdwa v2, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v0, 4
+; GFX9-NEXT: v_mov_b32_e32 v1, 0
+; GFX9-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_byte v[0:1], v4, off
+; GFX9-NEXT: global_store_dword v[41:42], v2, off
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v5i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v5i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v5i8_ret at rel32@hi+12
+; GFX10-NEXT: global_load_dwordx2 v[5:6], v[41:42], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX10-NEXT: v_mov_b32_e32 v0, v5
+; GFX10-NEXT: v_mov_b32_e32 v4, v6
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: v_or_b32_sdwa v5, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v0, 4
+; GFX10-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: global_store_byte v[0:1], v4, off
+; GFX10-NEXT: global_store_dword v[41:42], v2, off
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v5i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v42, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_mov_b32_e32 v42, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v5i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v5i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_b64 v[5:6], v[41:42], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_mov_b32_e32 v0, v5
+; GFX11-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX11-NEXT: v_mov_b32_e32 v4, v6
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT: v_and_b32_e32 v2, 0xffff, v0
+; GFX11-NEXT: v_dual_mov_b32 v0, 4 :: v_dual_lshlrev_b32 v3, 16, v1
+; GFX11-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b8 v[0:1], v4, off
+; GFX11-NEXT: global_store_b32 v[41:42], v2, off
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_load_b32 v42, off, s33
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v5i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v5i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v5i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dwordx2 v[5:6], v[41:42], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b64 v[3:4], 24, v[5:6]
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v1, 8, v5
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, v5
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, v6
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v5, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v0, 4
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, 0
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: global_store_byte v[0:1], v4, off
+; GFX10-SCRATCH-NEXT: global_store_dword v[41:42], v2, off
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <5 x i8>, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx <5 x i8> @external_void_func_v5i8_ret(<5 x i8> %val)
+ store <5 x i8> %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_v8i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_v8i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: global_load_dwordx2 v[0:1], v[41:42], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x400
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v8i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v8i8_ret at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX9-NEXT: v_mov_b32_e32 v4, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v8
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v5, 8, v7
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX9-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v3, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_dwordx2 v[41:42], v[3:4], off
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xfc00
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v8i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x200
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v8i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v8i8_ret at rel32@hi+12
+; GFX10-NEXT: global_load_dwordx2 v[0:1], v[41:42], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v8
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX10-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v5, v6, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: v_or_b32_sdwa v1, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: global_store_dwordx2 v[41:42], v[0:1], off
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfe00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v8i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v42, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_mov_b32_e32 v42, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 16
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v8i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v8i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_b64 v[0:1], v[41:42], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX11-NEXT: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v1, v8
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v4
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: v_or_b32_e32 v5, v6, v7
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xffff, v4
+; GFX11-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_lshlrev_b32_e32 v4, 16, v1
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: v_or_b32_e32 v1, v2, v3
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v4
+; GFX11-NEXT: global_store_b64 v[41:42], v[0:1], off
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: scratch_load_b32 v42, off, s33
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_add_i32 s32, s32, -16
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v8i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 16
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v8i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v8i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: global_load_dwordx2 v[0:1], v[41:42], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v8, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, v8
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v5, v6, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v1, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: global_store_dwordx2 v[41:42], v[0:1], off
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:8 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, -16
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <8 x i8>, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx <8 x i8> @external_void_func_v8i8_ret(<8 x i8> %val)
+ store <8 x i8> %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+define amdgpu_gfx void @test_call_external_void_func_v32i8_ret() #0 {
+; GFX9-LABEL: test_call_external_void_func_v32i8_ret:
+; GFX9: ; %bb.0:
+; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT: s_mov_b32 s34, s33
+; GFX9-NEXT: s_mov_b32 s33, s32
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Spill
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX9-NEXT: v_mov_b32_e32 v41, 0
+; GFX9-NEXT: v_mov_b32_e32 v43, 16
+; GFX9-NEXT: v_mov_b32_e32 v42, 0
+; GFX9-NEXT: v_mov_b32_e32 v44, 0
+; GFX9-NEXT: global_load_dwordx4 v[0:3], v[41:42], off
+; GFX9-NEXT: global_load_dwordx4 v[16:19], v[43:44], off
+; GFX9-NEXT: v_writelane_b32 v40, s34, 2
+; GFX9-NEXT: s_addk_i32 s32, 0x800
+; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: s_getpc_b64 s[34:35]
+; GFX9-NEXT: s_add_u32 s34, s34, external_void_func_v3i8_ret at rel32@lo+4
+; GFX9-NEXT: s_addc_u32 s35, s35, external_void_func_v3i8_ret at rel32@hi+12
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX9-NEXT: v_mov_b32_e32 v4, v1
+; GFX9-NEXT: v_mov_b32_e32 v8, v2
+; GFX9-NEXT: v_mov_b32_e32 v12, v3
+; GFX9-NEXT: v_mov_b32_e32 v20, v17
+; GFX9-NEXT: v_mov_b32_e32 v24, v18
+; GFX9-NEXT: v_mov_b32_e32 v28, v19
+; GFX9-NEXT: v_mov_b32_e32 v1, v35
+; GFX9-NEXT: v_mov_b32_e32 v2, v36
+; GFX9-NEXT: v_mov_b32_e32 v3, v37
+; GFX9-NEXT: v_mov_b32_e32 v17, v32
+; GFX9-NEXT: v_mov_b32_e32 v18, v33
+; GFX9-NEXT: v_mov_b32_e32 v19, v34
+; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v3
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v5, 8, v7
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v6, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v0, 8, v29
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v31
+; GFX9-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v3, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v0, 8, v25
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v27
+; GFX9-NEXT: v_or_b32_sdwa v0, v24, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v0, 8, v21
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v23
+; GFX9-NEXT: v_or_b32_sdwa v0, v20, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v22, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v13, 8, v13
+; GFX9-NEXT: v_lshlrev_b16_e32 v9, 8, v9
+; GFX9-NEXT: v_or_b32_sdwa v7, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v0, 8, v17
+; GFX9-NEXT: v_lshlrev_b16_e32 v4, 8, v19
+; GFX9-NEXT: v_or_b32_sdwa v12, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v13, 8, v15
+; GFX9-NEXT: v_or_b32_sdwa v8, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v9, 8, v11
+; GFX9-NEXT: v_or_b32_sdwa v0, v16, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v4, v18, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v13, v14, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v10, v10, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v9, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v8, v8, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: global_store_dwordx4 v[43:44], v[0:3], off
+; GFX9-NEXT: global_store_dwordx4 v[41:42], v[6:9], off
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s33 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Reload
+; GFX9-NEXT: v_readlane_b32 s31, v40, 1
+; GFX9-NEXT: v_readlane_b32 s30, v40, 0
+; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Reload
+; GFX9-NEXT: s_mov_b64 exec, s[36:37]
+; GFX9-NEXT: s_addk_i32 s32, 0xf800
+; GFX9-NEXT: s_mov_b32 s33, s34
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: test_call_external_void_func_v32i8_ret:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_mov_b32 s34, s33
+; GFX10-NEXT: s_mov_b32 s33, s32
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Spill
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
+; GFX10-NEXT: buffer_store_dword v44, off, s[0:3], s33 ; 4-byte Folded Spill
+; GFX10-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-NEXT: v_mov_b32_e32 v43, 16
+; GFX10-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-NEXT: v_mov_b32_e32 v44, 0
+; GFX10-NEXT: v_writelane_b32 v40, s34, 2
+; GFX10-NEXT: s_addk_i32 s32, 0x400
+; GFX10-NEXT: s_getpc_b64 s[34:35]
+; GFX10-NEXT: s_add_u32 s34, s34, external_void_func_v3i8_ret at rel32@lo+4
+; GFX10-NEXT: s_addc_u32 s35, s35, external_void_func_v3i8_ret at rel32@hi+12
+; GFX10-NEXT: s_clause 0x1
+; GFX10-NEXT: global_load_dwordx4 v[0:3], v[41:42], off
+; GFX10-NEXT: global_load_dwordx4 v[16:19], v[43:44], off
+; GFX10-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-NEXT: s_waitcnt vmcnt(1)
+; GFX10-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX10-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX10-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX10-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX10-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX10-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX10-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX10-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX10-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX10-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX10-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX10-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX10-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX10-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-NEXT: v_mov_b32_e32 v8, v2
+; GFX10-NEXT: v_mov_b32_e32 v12, v3
+; GFX10-NEXT: v_mov_b32_e32 v20, v17
+; GFX10-NEXT: v_mov_b32_e32 v24, v18
+; GFX10-NEXT: v_mov_b32_e32 v28, v19
+; GFX10-NEXT: v_mov_b32_e32 v1, v35
+; GFX10-NEXT: v_mov_b32_e32 v2, v36
+; GFX10-NEXT: v_mov_b32_e32 v3, v37
+; GFX10-NEXT: v_mov_b32_e32 v17, v32
+; GFX10-NEXT: v_mov_b32_e32 v18, v33
+; GFX10-NEXT: v_mov_b32_e32 v19, v34
+; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
+; GFX10-NEXT: v_lshlrev_b16 v13, 8, v13
+; GFX10-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX10-NEXT: v_lshlrev_b16 v11, 8, v11
+; GFX10-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX10-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX10-NEXT: v_or_b32_sdwa v12, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_lshlrev_b16 v13, 8, v15
+; GFX10-NEXT: v_or_b32_sdwa v8, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v9, v10, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v7, v6, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v13, v14, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-NEXT: v_or_b32_sdwa v5, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_lshlrev_b16 v8, 8, v31
+; GFX10-NEXT: v_or_b32_sdwa v4, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_lshlrev_b16 v7, 8, v29
+; GFX10-NEXT: v_lshlrev_b16 v9, 8, v25
+; GFX10-NEXT: v_or_b32_sdwa v6, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_lshlrev_b16 v1, 8, v3
+; GFX10-NEXT: v_or_b32_sdwa v3, v28, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v7, v30, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v8, v24, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_lshlrev_b16 v9, 8, v27
+; GFX10-NEXT: v_lshlrev_b16 v10, 8, v21
+; GFX10-NEXT: v_lshlrev_b16 v11, 8, v23
+; GFX10-NEXT: v_lshlrev_b16 v12, 8, v17
+; GFX10-NEXT: v_lshlrev_b16 v13, 8, v19
+; GFX10-NEXT: v_or_b32_sdwa v9, v26, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v14, v20, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v11, v22, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v12, v16, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v13, v18, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v10, v3, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v9, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v8, v14, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v7, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: v_or_b32_sdwa v3, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT: global_store_dwordx4 v[43:44], v[7:10], off
+; GFX10-NEXT: global_store_dwordx4 v[41:42], v[3:6], off
+; GFX10-NEXT: s_clause 0x3
+; GFX10-NEXT: buffer_load_dword v44, off, s[0:3], s33
+; GFX10-NEXT: buffer_load_dword v43, off, s[0:3], s33 offset:4
+; GFX10-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:8
+; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:12
+; GFX10-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-NEXT: v_readlane_b32 s34, v40, 2
+; GFX10-NEXT: s_or_saveexec_b32 s35, -1
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Reload
+; GFX10-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-NEXT: s_mov_b32 exec_lo, s35
+; GFX10-NEXT: s_addk_i32 s32, 0xfc00
+; GFX10-NEXT: s_mov_b32 s33, s34
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: test_call_external_void_func_v32i8_ret:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT: s_mov_b32 s0, s33
+; GFX11-NEXT: s_mov_b32 s33, s32
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_store_b32 off, v40, s33 offset:16 ; 4-byte Folded Spill
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_clause 0x3
+; GFX11-NEXT: scratch_store_b32 off, v41, s33 offset:12
+; GFX11-NEXT: scratch_store_b32 off, v42, s33 offset:8
+; GFX11-NEXT: scratch_store_b32 off, v43, s33 offset:4
+; GFX11-NEXT: scratch_store_b32 off, v44, s33
+; GFX11-NEXT: v_mov_b32_e32 v41, 0
+; GFX11-NEXT: v_dual_mov_b32 v42, 0 :: v_dual_mov_b32 v43, 16
+; GFX11-NEXT: v_mov_b32_e32 v44, 0
+; GFX11-NEXT: v_writelane_b32 v40, s0, 2
+; GFX11-NEXT: s_add_i32 s32, s32, 32
+; GFX11-NEXT: global_load_b128 v[0:3], v[41:42], off
+; GFX11-NEXT: s_getpc_b64 s[0:1]
+; GFX11-NEXT: s_add_u32 s0, s0, external_void_func_v3i8_ret at rel32@lo+4
+; GFX11-NEXT: s_addc_u32 s1, s1, external_void_func_v3i8_ret at rel32@hi+12
+; GFX11-NEXT: global_load_b128 v[16:19], v[43:44], off
+; GFX11-NEXT: v_writelane_b32 v40, s30, 0
+; GFX11-NEXT: v_writelane_b32 v40, s31, 1
+; GFX11-NEXT: s_waitcnt vmcnt(1)
+; GFX11-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX11-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX11-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX11-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX11-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX11-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX11-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX11-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX11-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX11-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX11-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX11-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX11-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX11-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX11-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX11-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX11-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX11-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX11-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX11-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX11-NEXT: v_dual_mov_b32 v4, v1 :: v_dual_mov_b32 v1, v35
+; GFX11-NEXT: v_mov_b32_e32 v8, v2
+; GFX11-NEXT: v_mov_b32_e32 v12, v3
+; GFX11-NEXT: v_mov_b32_e32 v20, v17
+; GFX11-NEXT: v_mov_b32_e32 v24, v18
+; GFX11-NEXT: v_dual_mov_b32 v28, v19 :: v_dual_mov_b32 v19, v34
+; GFX11-NEXT: v_dual_mov_b32 v2, v36 :: v_dual_mov_b32 v3, v37
+; GFX11-NEXT: v_dual_mov_b32 v17, v32 :: v_dual_mov_b32 v18, v33
+; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX11-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX11-NEXT: v_and_b32_e32 v8, 0xff, v8
+; GFX11-NEXT: v_lshlrev_b16 v11, 8, v11
+; GFX11-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX11-NEXT: v_and_b32_e32 v4, 0xff, v4
+; GFX11-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX11-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX11-NEXT: v_lshlrev_b16 v13, 8, v13
+; GFX11-NEXT: v_and_b32_e32 v12, 0xff, v12
+; GFX11-NEXT: v_or_b32_e32 v8, v8, v9
+; GFX11-NEXT: v_or_b32_e32 v9, v10, v11
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v5
+; GFX11-NEXT: v_or_b32_e32 v5, v6, v7
+; GFX11-NEXT: v_or_b32_e32 v12, v12, v13
+; GFX11-NEXT: v_lshlrev_b16 v13, 8, v15
+; GFX11-NEXT: v_and_b32_e32 v14, 0xff, v14
+; GFX11-NEXT: v_and_b32_e32 v7, 0xffff, v8
+; GFX11-NEXT: v_lshlrev_b32_e32 v8, 16, v9
+; GFX11-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-NEXT: v_lshlrev_b32_e32 v9, 16, v5
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
+; GFX11-NEXT: v_or_b32_e32 v13, v14, v13
+; GFX11-NEXT: v_or_b32_e32 v5, v7, v8
+; GFX11-NEXT: v_or_b32_e32 v4, v4, v9
+; GFX11-NEXT: v_and_b32_e32 v7, 0xff, v28
+; GFX11-NEXT: v_or_b32_e32 v0, v0, v1
+; GFX11-NEXT: v_lshlrev_b16 v1, 8, v29
+; GFX11-NEXT: v_lshlrev_b16 v8, 8, v31
+; GFX11-NEXT: v_and_b32_e32 v9, 0xff, v30
+; GFX11-NEXT: v_lshlrev_b16 v10, 8, v25
+; GFX11-NEXT: v_and_b32_e32 v11, 0xff, v24
+; GFX11-NEXT: v_and_b32_e32 v12, 0xffff, v12
+; GFX11-NEXT: v_lshlrev_b32_e32 v6, 16, v13
+; GFX11-NEXT: v_lshlrev_b16 v3, 8, v3
+; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-NEXT: v_or_b32_e32 v1, v7, v1
+; GFX11-NEXT: v_or_b32_e32 v7, v9, v8
+; GFX11-NEXT: v_or_b32_e32 v8, v11, v10
+; GFX11-NEXT: v_or_b32_e32 v6, v12, v6
+; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-NEXT: v_and_b32_e32 v9, 0xff, v26
+; GFX11-NEXT: v_lshlrev_b32_e32 v3, 16, v7
+; GFX11-NEXT: v_and_b32_e32 v7, 0xffff, v8
+; GFX11-NEXT: v_lshlrev_b16 v8, 8, v27
+; GFX11-NEXT: v_lshlrev_b16 v10, 8, v21
+; GFX11-NEXT: v_and_b32_e32 v11, 0xff, v20
+; GFX11-NEXT: v_lshlrev_b16 v12, 8, v23
+; GFX11-NEXT: v_and_b32_e32 v13, 0xff, v22
+; GFX11-NEXT: v_lshlrev_b16 v14, 8, v17
+; GFX11-NEXT: v_and_b32_e32 v15, 0xff, v16
+; GFX11-NEXT: v_lshlrev_b16 v16, 8, v19
+; GFX11-NEXT: v_and_b32_e32 v17, 0xff, v18
+; GFX11-NEXT: v_or_b32_e32 v8, v9, v8
+; GFX11-NEXT: v_or_b32_e32 v9, v11, v10
+; GFX11-NEXT: v_or_b32_e32 v10, v13, v12
+; GFX11-NEXT: v_or_b32_e32 v11, v15, v14
+; GFX11-NEXT: v_or_b32_e32 v12, v17, v16
+; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; GFX11-NEXT: v_and_b32_e32 v13, 0xffff, v9
+; GFX11-NEXT: v_lshlrev_b32_e32 v14, 16, v10
+; GFX11-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; GFX11-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; GFX11-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; GFX11-NEXT: v_or_b32_e32 v10, v1, v3
+; GFX11-NEXT: v_or_b32_e32 v9, v7, v8
+; GFX11-NEXT: v_or_b32_e32 v8, v13, v14
+; GFX11-NEXT: v_or_b32_e32 v7, v11, v12
+; GFX11-NEXT: v_or_b32_e32 v3, v0, v2
+; GFX11-NEXT: s_clause 0x1
+; GFX11-NEXT: global_store_b128 v[43:44], v[7:10], off
+; GFX11-NEXT: global_store_b128 v[41:42], v[3:6], off
+; GFX11-NEXT: s_clause 0x3
+; GFX11-NEXT: scratch_load_b32 v44, off, s33
+; GFX11-NEXT: scratch_load_b32 v43, off, s33 offset:4
+; GFX11-NEXT: scratch_load_b32 v42, off, s33 offset:8
+; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:12
+; GFX11-NEXT: v_readlane_b32 s31, v40, 1
+; GFX11-NEXT: v_readlane_b32 s30, v40, 0
+; GFX11-NEXT: v_readlane_b32 s0, v40, 2
+; GFX11-NEXT: s_or_saveexec_b32 s1, -1
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 offset:16 ; 4-byte Folded Reload
+; GFX11-NEXT: s_mov_b32 exec_lo, s1
+; GFX11-NEXT: s_addk_i32 s32, 0xffe0
+; GFX11-NEXT: s_mov_b32 s33, s0
+; GFX11-NEXT: s_waitcnt vmcnt(0)
+; GFX11-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-SCRATCH-LABEL: test_call_external_void_func_v32i8_ret:
+; GFX10-SCRATCH: ; %bb.0:
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SCRATCH-NEXT: s_mov_b32 s0, s33
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s32
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v40, s33 offset:16 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v41, s33 offset:12 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v42, s33 offset:8 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v43, s33 offset:4 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: scratch_store_dword off, v44, s33 ; 4-byte Folded Spill
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v41, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v43, 16
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v42, 0
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v44, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s0, 2
+; GFX10-SCRATCH-NEXT: s_add_i32 s32, s32, 32
+; GFX10-SCRATCH-NEXT: s_getpc_b64 s[0:1]
+; GFX10-SCRATCH-NEXT: s_add_u32 s0, s0, external_void_func_v3i8_ret at rel32@lo+4
+; GFX10-SCRATCH-NEXT: s_addc_u32 s1, s1, external_void_func_v3i8_ret at rel32@hi+12
+; GFX10-SCRATCH-NEXT: s_clause 0x1
+; GFX10-SCRATCH-NEXT: global_load_dwordx4 v[0:3], v[41:42], off
+; GFX10-SCRATCH-NEXT: global_load_dwordx4 v[16:19], v[43:44], off
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s30, 0
+; GFX10-SCRATCH-NEXT: v_writelane_b32 v40, s31, 1
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(1)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v35, 8, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v36, 16, v0
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v37, 24, v0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v32, 8, v16
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v33, 16, v16
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v13, 8, v3
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v14, 16, v3
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v15, 24, v3
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v21, 8, v17
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v23, 24, v17
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v25, 8, v18
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v26, 16, v18
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v27, 24, v18
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v29, 8, v19
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v30, 16, v19
+; GFX10-SCRATCH-NEXT: v_lshrrev_b32_e32 v31, 24, v19
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v4, v1
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v8, v2
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v12, v3
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v20, v17
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v24, v18
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v28, v19
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v1, v35
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v2, v36
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v3, v37
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v17, v32
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v18, v33
+; GFX10-SCRATCH-NEXT: v_mov_b32_e32 v19, v34
+; GFX10-SCRATCH-NEXT: s_swappc_b64 s[30:31], s[0:1]
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v13, 8, v13
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v9, 8, v9
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v11, 8, v11
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v5, 8, v5
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v7, 8, v7
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v12, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v13, 8, v15
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v8, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v9, v10, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v7, v6, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v13, v14, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v1
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v5, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v8, 8, v31
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v4, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v7, 8, v29
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v9, 8, v25
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v6, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v1, 8, v3
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v3, v28, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v7, v30, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v8, v24, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v9, 8, v27
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v10, 8, v21
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v11, 8, v23
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v12, 8, v17
+; GFX10-SCRATCH-NEXT: v_lshlrev_b16 v13, 8, v19
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v9, v26, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v14, v20, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v11, v22, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v12, v16, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v13, v18, v13 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v10, v3, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v9, v8, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v8, v14, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v7, v12, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: v_or_b32_sdwa v3, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-SCRATCH-NEXT: global_store_dwordx4 v[43:44], v[7:10], off
+; GFX10-SCRATCH-NEXT: global_store_dwordx4 v[41:42], v[3:6], off
+; GFX10-SCRATCH-NEXT: s_clause 0x3
+; GFX10-SCRATCH-NEXT: scratch_load_dword v44, off, s33
+; GFX10-SCRATCH-NEXT: scratch_load_dword v43, off, s33 offset:4
+; GFX10-SCRATCH-NEXT: scratch_load_dword v42, off, s33 offset:8
+; GFX10-SCRATCH-NEXT: scratch_load_dword v41, off, s33 offset:12
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s31, v40, 1
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s30, v40, 0
+; GFX10-SCRATCH-NEXT: v_readlane_b32 s0, v40, 2
+; GFX10-SCRATCH-NEXT: s_or_saveexec_b32 s1, -1
+; GFX10-SCRATCH-NEXT: scratch_load_dword v40, off, s33 offset:16 ; 4-byte Folded Reload
+; GFX10-SCRATCH-NEXT: s_waitcnt_depctr 0xffe3
+; GFX10-SCRATCH-NEXT: s_mov_b32 exec_lo, s1
+; GFX10-SCRATCH-NEXT: s_addk_i32 s32, 0xffe0
+; GFX10-SCRATCH-NEXT: s_mov_b32 s33, s0
+; GFX10-SCRATCH-NEXT: s_waitcnt vmcnt(0)
+; GFX10-SCRATCH-NEXT: s_setpc_b64 s[30:31]
+ %val = load <32 x i8>, ptr addrspace(1) null
+ %tmp = call amdgpu_gfx <32 x i8> @external_void_func_v3i8_ret(<32 x i8> %val)
+ store <32 x i8> %tmp, ptr addrspace(1) null
+ ret void
+}
+
+
+
define amdgpu_gfx void @test_call_external_void_func_v2i16() #0 {
; GFX9-LABEL: test_call_external_void_func_v2i16:
; GFX9: ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll b/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
new file mode 100644
index 000000000000000..b22ae9b9c852792
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
@@ -0,0 +1,2002 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -march=amdgcn -mcpu=gfx906 < %s | FileCheck --check-prefix=GFX906 %s
+
+define amdgpu_kernel void @v3i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v3i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX906-NEXT: v_lshlrev_b32_e32 v1, 2, v0
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dword v2, v1, s[4:5]
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v4, 8, v2
+; GFX906-NEXT: s_and_saveexec_b64 s[0:1], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB0_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 2, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v2, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v3, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dword v2, v[2:3], off
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v4, 8, v2
+; GFX906-NEXT: .LBB0_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v4
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_byte v1, v3, s[2:3] offset:2
+; GFX906-NEXT: global_store_short v1, v0, s[2:3]
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <3 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <3 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <3 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <3 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <3 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <3 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+define amdgpu_kernel void @v4i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v4i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX906-NEXT: v_lshlrev_b32_e32 v1, 2, v0
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dword v2, v1, s[4:5]
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v4, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v5, 8, v2
+; GFX906-NEXT: s_and_saveexec_b64 s[0:1], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB1_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 2, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v2, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v3, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dword v2, v[2:3], off
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v4, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v5, 8, v2
+; GFX906-NEXT: .LBB1_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v5
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v3
+; GFX906-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dword v1, v0, s[2:3]
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <4 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <4 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <4 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <4 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <4 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+define amdgpu_kernel void @v5i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v5i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX906-NEXT: v_lshlrev_b32_e32 v1, 3, v0
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dwordx2 v[2:3], v1, s[4:5]
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b64 v[4:5], 24, v[2:3]
+; GFX906-NEXT: v_lshrrev_b32_e32 v5, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v6, 8, v2
+; GFX906-NEXT: s_and_saveexec_b64 s[0:1], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB2_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 3, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v2, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v3, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b64 v[4:5], 24, v[2:3]
+; GFX906-NEXT: v_lshrrev_b32_e32 v5, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v6, 8, v2
+; GFX906-NEXT: .LBB2_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v6
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v4
+; GFX906-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_byte v1, v3, s[2:3] offset:4
+; GFX906-NEXT: global_store_dword v1, v0, s[2:3]
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <5 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <5 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <5 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <5 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <5 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <5 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+define amdgpu_kernel void @v8i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v8i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX906-NEXT: v_lshlrev_b32_e32 v1, 3, v0
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dwordx2 v[2:3], v1, s[4:5]
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v4, 24, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v5, 16, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v6, 8, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v7, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v8, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX906-NEXT: s_and_saveexec_b64 s[0:1], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB3_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 3, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v2, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v3, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dwordx2 v[2:3], v[2:3], off
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v4, 24, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v5, 16, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v6, 8, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v7, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v8, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; GFX906-NEXT: .LBB3_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v9
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v7
+; GFX906-NEXT: v_or_b32_sdwa v2, v8, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v6
+; GFX906-NEXT: v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v3, 8, v4
+; GFX906-NEXT: v_or_b32_sdwa v3, v5, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx2 v1, v[2:3], s[2:3]
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <8 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <8 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <8 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <8 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <8 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <8 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+define amdgpu_kernel void @v16i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v16i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX906-NEXT: v_lshlrev_b32_e32 v1, 4, v0
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dwordx4 v[2:5], v1, s[4:5]
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v6, 24, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v7, 16, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v8, 8, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v9, 24, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v10, 16, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v11, 8, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v12, 24, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v13, 16, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v14, 8, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v15, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v16, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v17, 8, v2
+; GFX906-NEXT: s_and_saveexec_b64 s[0:1], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB4_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 4, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v2, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v3, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dwordx4 v[2:5], v[2:3], off
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v6, 24, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v7, 16, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v8, 8, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v9, 24, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v10, 16, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v11, 8, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v12, 24, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v13, 16, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v14, 8, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v15, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v16, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v17, 8, v2
+; GFX906-NEXT: .LBB4_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v17
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v15
+; GFX906-NEXT: v_or_b32_sdwa v2, v16, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v14
+; GFX906-NEXT: v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v3, 8, v12
+; GFX906-NEXT: v_or_b32_sdwa v3, v13, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v11
+; GFX906-NEXT: v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v4, 8, v9
+; GFX906-NEXT: v_or_b32_sdwa v4, v10, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v8
+; GFX906-NEXT: v_or_b32_sdwa v0, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v5, 8, v6
+; GFX906-NEXT: v_or_b32_sdwa v5, v7, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3]
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <16 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <16 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <16 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <16 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <16 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <16 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+define amdgpu_kernel void @v32i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v32i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: v_lshlrev_b32_e32 v1, 5, v0
+; GFX906-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x34
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dwordx4 v[2:5], v1, s[4:5] offset:16
+; GFX906-NEXT: global_load_dwordx4 v[6:9], v1, s[4:5]
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v10, 24, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v11, 16, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v12, 8, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v13, 24, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v14, 16, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v15, 8, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v16, 24, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v17, 16, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v18, 8, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v19, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v20, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v21, 8, v2
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v22, 24, v9
+; GFX906-NEXT: v_lshrrev_b32_e32 v23, 16, v9
+; GFX906-NEXT: v_lshrrev_b32_e32 v24, 8, v9
+; GFX906-NEXT: v_lshrrev_b32_e32 v25, 24, v8
+; GFX906-NEXT: v_lshrrev_b32_e32 v26, 16, v8
+; GFX906-NEXT: v_lshrrev_b32_e32 v27, 8, v8
+; GFX906-NEXT: v_lshrrev_b32_e32 v28, 24, v7
+; GFX906-NEXT: v_lshrrev_b32_e32 v29, 16, v7
+; GFX906-NEXT: v_lshrrev_b32_e32 v30, 8, v7
+; GFX906-NEXT: v_lshrrev_b32_e32 v32, 24, v6
+; GFX906-NEXT: v_lshrrev_b32_e32 v33, 16, v6
+; GFX906-NEXT: v_lshrrev_b32_e32 v31, 8, v6
+; GFX906-NEXT: s_and_saveexec_b64 s[2:3], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB5_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 5, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v10, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v11, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dwordx4 v[2:5], v[10:11], off offset:16
+; GFX906-NEXT: global_load_dwordx4 v[6:9], v[10:11], off
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v10, 24, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v11, 16, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v12, 8, v5
+; GFX906-NEXT: v_lshrrev_b32_e32 v13, 24, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v14, 16, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v15, 8, v4
+; GFX906-NEXT: v_lshrrev_b32_e32 v16, 24, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v17, 16, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v18, 8, v3
+; GFX906-NEXT: v_lshrrev_b32_e32 v19, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v20, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v21, 8, v2
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshrrev_b32_e32 v22, 24, v9
+; GFX906-NEXT: v_lshrrev_b32_e32 v23, 16, v9
+; GFX906-NEXT: v_lshrrev_b32_e32 v24, 8, v9
+; GFX906-NEXT: v_lshrrev_b32_e32 v25, 24, v8
+; GFX906-NEXT: v_lshrrev_b32_e32 v26, 16, v8
+; GFX906-NEXT: v_lshrrev_b32_e32 v27, 8, v8
+; GFX906-NEXT: v_lshrrev_b32_e32 v28, 24, v7
+; GFX906-NEXT: v_lshrrev_b32_e32 v29, 16, v7
+; GFX906-NEXT: v_lshrrev_b32_e32 v30, 8, v7
+; GFX906-NEXT: v_lshrrev_b32_e32 v32, 24, v6
+; GFX906-NEXT: v_lshrrev_b32_e32 v33, 16, v6
+; GFX906-NEXT: v_lshrrev_b32_e32 v31, 8, v6
+; GFX906-NEXT: .LBB5_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX906-NEXT: v_lshlrev_b16_e32 v28, 8, v28
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v32
+; GFX906-NEXT: v_or_b32_sdwa v28, v29, v28 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v29, 8, v31
+; GFX906-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v6, v6, v29 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v6, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v27
+; GFX906-NEXT: v_or_b32_sdwa v0, v8, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v8, 8, v25
+; GFX906-NEXT: v_or_b32_sdwa v8, v26, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v8, v0, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v24
+; GFX906-NEXT: v_or_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v9, 8, v22
+; GFX906-NEXT: v_or_b32_sdwa v9, v23, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v9, v0, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v21
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v19
+; GFX906-NEXT: v_or_b32_sdwa v2, v20, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v18
+; GFX906-NEXT: v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v3, 8, v16
+; GFX906-NEXT: v_or_b32_sdwa v3, v17, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v15
+; GFX906-NEXT: v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v4, 8, v13
+; GFX906-NEXT: v_or_b32_sdwa v4, v14, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v12
+; GFX906-NEXT: v_lshlrev_b16_e32 v29, 8, v30
+; GFX906-NEXT: v_or_b32_sdwa v0, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v5, 8, v10
+; GFX906-NEXT: v_or_b32_sdwa v7, v7, v29 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v11, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v7, v7, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[6:9], s[0:1]
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[0:1] offset:16
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <32 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <32 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <32 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <32 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <32 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <32 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+define amdgpu_kernel void @v256i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1) %src2, ptr addrspace(1) nocapture %dst) {
+; GFX906-LABEL: v256i8_liveout:
+; GFX906: ; %bb.0: ; %entry
+; GFX906-NEXT: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
+; GFX906-NEXT: s_mov_b32 s9, SCRATCH_RSRC_DWORD1
+; GFX906-NEXT: s_mov_b32 s10, -1
+; GFX906-NEXT: s_mov_b32 s11, 0xe00000
+; GFX906-NEXT: s_add_u32 s8, s8, s3
+; GFX906-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX906-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX906-NEXT: v_lshlrev_b32_e32 v2, 3, v0
+; GFX906-NEXT: s_addc_u32 s9, s9, 0
+; GFX906-NEXT: s_waitcnt lgkmcnt(0)
+; GFX906-NEXT: global_load_dwordx4 v[18:21], v2, s[4:5] offset:240
+; GFX906-NEXT: global_load_dwordx4 v[6:9], v2, s[4:5] offset:224
+; GFX906-NEXT: global_load_dwordx4 v[10:13], v2, s[4:5] offset:208
+; GFX906-NEXT: global_load_dwordx4 v[14:17], v2, s[4:5] offset:192
+; GFX906-NEXT: v_mov_b32_e32 v1, 0
+; GFX906-NEXT: v_cmp_gt_u32_e32 vcc, 15, v0
+; GFX906-NEXT: s_waitcnt vmcnt(3)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v21
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:20 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v21
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:24 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v21
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:28 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v20
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:32 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v20
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:36 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v20
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:40 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v19
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:44 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v19
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:48 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v19
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:52 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v18
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:56 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v18
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:60 ; 4-byte Folded Spill
+; GFX906-NEXT: buffer_store_dword v18, off, s[8:11], 0 offset:4 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: buffer_store_dword v19, off, s[8:11], 0 offset:8 ; 4-byte Folded Spill
+; GFX906-NEXT: buffer_store_dword v20, off, s[8:11], 0 offset:12 ; 4-byte Folded Spill
+; GFX906-NEXT: buffer_store_dword v21, off, s[8:11], 0 offset:16 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v18
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:64 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v9
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:68 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v9
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:72 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v9
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:76 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v8
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:80 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v8
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:84 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v8
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:88 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v7
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:92 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v7
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:96 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v7
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:100 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v6
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:104 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v6
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:108 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v6
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:112 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v13
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:116 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v13
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:120 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v13
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:124 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v12
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:128 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v12
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:132 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v12
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:136 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v11
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:140 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v11
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:144 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v11
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:148 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v10
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:152 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v10
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:156 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v10
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:160 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v17
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:164 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v17
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:168 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v17
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:180 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v16
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:172 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v16
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:176 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v16
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:192 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v15
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:184 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v15
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:188 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v15
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:204 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v14
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:196 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v14
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:200 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v14
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:208 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[18:21], v2, s[4:5] offset:176
+; GFX906-NEXT: global_load_dwordx4 v[22:25], v2, s[4:5] offset:160
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v21
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:212 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v21
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:216 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v21
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:228 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v20
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:220 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v20
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:224 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v20
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:240 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v19
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:232 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v19
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:236 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v19
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:252 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v18
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:244 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v18
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:248 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v18
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:256 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v25
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:260 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v25
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:264 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v25
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:276 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v24
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:268 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v24
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:272 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v24
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:288 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v23
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:280 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v23
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:284 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v23
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:300 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v22
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:292 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v22
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:296 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v22
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:304 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[26:29], v2, s[4:5] offset:144
+; GFX906-NEXT: global_load_dwordx4 v[30:33], v2, s[4:5] offset:128
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v29
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:308 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v29
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:312 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v29
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:324 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v28
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:316 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v28
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:320 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v28
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:336 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v27
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:328 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v27
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:332 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v27
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:348 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v26
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:340 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v26
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:344 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v26
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:352 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v33
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:356 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v33
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:360 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v33
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:372 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v32
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:364 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v32
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:368 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v32
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:384 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v31
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:376 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v31
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:380 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v31
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:396 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v30
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:388 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v30
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:392 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v30
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:400 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[34:37], v2, s[4:5] offset:112
+; GFX906-NEXT: global_load_dwordx4 v[38:41], v2, s[4:5] offset:96
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v37
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:404 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v37
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:408 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v37
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:420 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v36
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:412 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v36
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:416 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v36
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:432 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v35
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:424 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v35
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:428 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v35
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:444 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v34
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:436 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v34
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:440 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v34
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:448 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v41
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:452 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v41
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:456 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v41
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:468 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v40
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:460 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v40
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:464 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v40
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:480 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v39
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:472 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v39
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:476 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v39
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:492 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v38
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:484 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v38
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:488 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v38
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:496 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[42:45], v2, s[4:5] offset:80
+; GFX906-NEXT: global_load_dwordx4 v[46:49], v2, s[4:5] offset:64
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v45
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:500 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v45
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:504 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v45
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:516 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v44
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:508 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v44
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:512 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v44
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:528 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v43
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:520 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v43
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:524 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v43
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:540 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v42
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:532 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v42
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:536 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v42
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:544 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v49
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:548 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v49
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:552 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v49
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:564 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v48
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:556 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v48
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:560 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v48
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:576 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v47
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:568 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v47
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:572 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v47
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:588 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v46
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:580 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v46
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:584 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v46
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:592 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[50:53], v2, s[4:5] offset:48
+; GFX906-NEXT: global_load_dwordx4 v[54:57], v2, s[4:5] offset:32
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v53
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:596 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v53
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:600 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v53
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:612 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v52
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:604 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v52
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:608 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v52
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:624 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v51
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:616 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v51
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:620 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v51
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:636 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v50
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:628 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v50
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:632 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v50
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:640 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v57
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:644 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v57
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:648 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v57
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:660 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v56
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:652 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v56
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:656 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v56
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:672 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v55
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:664 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v55
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:668 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v55
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:684 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 24, v54
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:676 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 16, v54
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:680 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v3, 8, v54
+; GFX906-NEXT: buffer_store_dword v3, off, s[8:11], 0 offset:688 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[58:61], v2, s[4:5] offset:16
+; GFX906-NEXT: s_nop 0
+; GFX906-NEXT: global_load_dwordx4 v[2:5], v2, s[4:5]
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v61
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:692 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v61
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:696 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v61
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:708 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v60
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:700 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v60
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:704 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v60
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:720 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v59
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:712 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v59
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:716 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v59
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:732 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v58
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:724 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v58
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:728 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v58
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:736 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v5
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:740 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v5
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:744 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v5
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:756 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v4
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:748 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v4
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:752 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v4
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:768 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v3
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:760 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v3
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:764 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 24, v2
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:772 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v63, 8, v3
+; GFX906-NEXT: buffer_store_dword v62, off, s[8:11], 0 offset:776 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v2
+; GFX906-NEXT: s_and_saveexec_b64 s[0:1], vcc
+; GFX906-NEXT: s_cbranch_execz .LBB6_2
+; GFX906-NEXT: ; %bb.1: ; %bb.1
+; GFX906-NEXT: v_lshlrev_b64 v[2:3], 3, v[0:1]
+; GFX906-NEXT: v_mov_b32_e32 v0, s7
+; GFX906-NEXT: v_add_co_u32_e32 v2, vcc, s6, v2
+; GFX906-NEXT: v_addc_co_u32_e32 v3, vcc, v0, v3, vcc
+; GFX906-NEXT: global_load_dwordx4 v[18:21], v[2:3], off offset:240
+; GFX906-NEXT: global_load_dwordx4 v[6:9], v[2:3], off offset:224
+; GFX906-NEXT: global_load_dwordx4 v[10:13], v[2:3], off offset:208
+; GFX906-NEXT: global_load_dwordx4 v[14:17], v[2:3], off offset:192
+; GFX906-NEXT: s_waitcnt vmcnt(3)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v21
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:20 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v21
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:24 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v21
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:28 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v20
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:32 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v20
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:36 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v20
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:40 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v19
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:44 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v19
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:48 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v19
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:52 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v18
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:56 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v18
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:60 ; 4-byte Folded Spill
+; GFX906-NEXT: buffer_store_dword v18, off, s[8:11], 0 offset:4 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: buffer_store_dword v19, off, s[8:11], 0 offset:8 ; 4-byte Folded Spill
+; GFX906-NEXT: buffer_store_dword v20, off, s[8:11], 0 offset:12 ; 4-byte Folded Spill
+; GFX906-NEXT: buffer_store_dword v21, off, s[8:11], 0 offset:16 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v18
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:64 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v9
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:68 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v9
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:72 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v9
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:76 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v8
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:80 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v8
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:84 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v8
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:88 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v7
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:92 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v7
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:96 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v7
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:100 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v6
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:104 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v6
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:108 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v6
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:112 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v13
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:116 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v13
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:120 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v13
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:124 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v12
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:128 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v12
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:132 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v12
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:136 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v11
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:140 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v11
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:144 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v11
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:148 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v10
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:152 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v10
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:156 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v10
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:160 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v17
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:164 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v17
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:168 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v17
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:180 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v16
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:172 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v16
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:176 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v16
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:192 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v15
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:184 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v15
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:188 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v15
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:204 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v14
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:196 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v14
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:200 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v14
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:208 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[18:21], v[2:3], off offset:176
+; GFX906-NEXT: global_load_dwordx4 v[22:25], v[2:3], off offset:160
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v21
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:212 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v21
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:216 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v21
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:228 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v20
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:220 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v20
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:224 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v20
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:240 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v19
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:232 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v19
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:236 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v19
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:252 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v18
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:244 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v18
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:248 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v18
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:256 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v25
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:260 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v25
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:264 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v25
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:276 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v24
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:268 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v24
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:272 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v24
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:288 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v23
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:280 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v23
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:284 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v23
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:300 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v22
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:292 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v22
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:296 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v22
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:304 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[26:29], v[2:3], off offset:144
+; GFX906-NEXT: global_load_dwordx4 v[30:33], v[2:3], off offset:128
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v29
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:308 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v29
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:312 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v29
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:324 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v28
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:316 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v28
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:320 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v28
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:336 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v27
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:328 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v27
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:332 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v27
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:348 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v26
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:340 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v26
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:344 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v26
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:352 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v33
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:356 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v33
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:360 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v33
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:372 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v32
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:364 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v32
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:368 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v32
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:384 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v31
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:376 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v31
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:380 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v31
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:396 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v30
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:388 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v30
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:392 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v30
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:400 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[34:37], v[2:3], off offset:112
+; GFX906-NEXT: global_load_dwordx4 v[38:41], v[2:3], off offset:96
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v37
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:404 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v37
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:408 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v37
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:420 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v36
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:412 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v36
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:416 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v36
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:432 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v35
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:424 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v35
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:428 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v35
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:444 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v34
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:436 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v34
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:440 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v34
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:448 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v41
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:452 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v41
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:456 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v41
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:468 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v40
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:460 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v40
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:464 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v40
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:480 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v39
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:472 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v39
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:476 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v39
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:492 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v38
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:484 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v38
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:488 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v38
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:496 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[42:45], v[2:3], off offset:80
+; GFX906-NEXT: global_load_dwordx4 v[46:49], v[2:3], off offset:64
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v45
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:500 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v45
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:504 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v45
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:516 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v44
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:508 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v44
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:512 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v44
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:528 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v43
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:520 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v43
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:524 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v43
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:540 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v42
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:532 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v42
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:536 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v42
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:544 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v49
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:548 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v49
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:552 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v49
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:564 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v48
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:556 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v48
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:560 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v48
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:576 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v47
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:568 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v47
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:572 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v47
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:588 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v46
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:580 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v46
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:584 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v46
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:592 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[50:53], v[2:3], off offset:48
+; GFX906-NEXT: global_load_dwordx4 v[54:57], v[2:3], off offset:32
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v53
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:596 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v53
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:600 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v53
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:612 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v52
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:604 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v52
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:608 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v52
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:624 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v51
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:616 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v51
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:620 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v51
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:636 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v50
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:628 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v50
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:632 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v50
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:640 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v57
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:644 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v57
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:648 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v57
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:660 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v56
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:652 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v56
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:656 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v56
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:672 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v55
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:664 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v55
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:668 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v55
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:684 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v54
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:676 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v54
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:680 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v54
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:688 ; 4-byte Folded Spill
+; GFX906-NEXT: global_load_dwordx4 v[58:61], v[2:3], off offset:16
+; GFX906-NEXT: s_nop 0
+; GFX906-NEXT: global_load_dwordx4 v[2:5], v[2:3], off
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v61
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:692 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v61
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:696 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v61
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:708 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v60
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:700 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v60
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:704 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v60
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:720 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v59
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:712 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v59
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:716 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v59
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:732 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v58
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:724 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v58
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:728 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v58
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:736 ; 4-byte Folded Spill
+; GFX906-NEXT: s_waitcnt vmcnt(12)
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v5
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:740 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v5
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:744 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v5
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:756 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v4
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:748 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v4
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:752 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 8, v4
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:768 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v3
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:760 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v3
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:764 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 24, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v63, 8, v3
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:772 ; 4-byte Folded Spill
+; GFX906-NEXT: v_lshrrev_b32_e32 v0, 16, v2
+; GFX906-NEXT: v_lshrrev_b32_e32 v62, 8, v2
+; GFX906-NEXT: buffer_store_dword v0, off, s[8:11], 0 offset:776 ; 4-byte Folded Spill
+; GFX906-NEXT: .LBB6_2: ; %bb.2
+; GFX906-NEXT: s_or_b64 exec, exec, s[0:1]
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v62
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v63
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:768 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v62, off, s[8:11], 0 offset:776 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v63, off, s[8:11], 0 offset:764 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:756 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:772 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v62, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v62, off, s[8:11], 0 offset:760 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:748 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v62, 8, v62
+; GFX906-NEXT: v_or_b32_sdwa v62, v63, v62 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v62 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v62, off, s[8:11], 0 offset:752 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:740 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v62, off, s[8:11], 0 offset:744 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3]
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:732 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:736 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v59, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:720 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v59, off, s[8:11], 0 offset:716 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v60, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:708 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v58, off, s[8:11], 0 offset:728 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v61, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:724 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v58, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v58, off, s[8:11], 0 offset:712 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:700 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v58, 8, v58
+; GFX906-NEXT: v_or_b32_sdwa v58, v59, v58 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v58, off, s[8:11], 0 offset:704 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:692 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v58, off, s[8:11], 0 offset:696 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:16
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:684 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:688 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v55, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:672 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v55, off, s[8:11], 0 offset:668 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v56, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:660 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v54, off, s[8:11], 0 offset:680 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v57, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:676 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v54, off, s[8:11], 0 offset:664 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:652 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v54, 8, v54
+; GFX906-NEXT: v_or_b32_sdwa v54, v55, v54 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v54 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v54, off, s[8:11], 0 offset:656 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:644 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v54, off, s[8:11], 0 offset:648 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:32
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:636 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:640 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v51, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:624 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v51, off, s[8:11], 0 offset:620 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v50, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v52, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:612 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v50, off, s[8:11], 0 offset:632 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v53, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:628 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v50, off, s[8:11], 0 offset:616 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:604 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v50, 8, v50
+; GFX906-NEXT: v_or_b32_sdwa v50, v51, v50 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v50 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v50, off, s[8:11], 0 offset:608 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v50, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:596 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v50, off, s[8:11], 0 offset:600 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v50, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:48
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:588 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:592 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v47, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:576 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v47, off, s[8:11], 0 offset:572 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v48, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:564 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v46, off, s[8:11], 0 offset:584 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v49, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:580 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v46, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v46, off, s[8:11], 0 offset:568 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:556 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v46, 8, v46
+; GFX906-NEXT: v_or_b32_sdwa v46, v47, v46 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v46, off, s[8:11], 0 offset:560 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:548 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v46, off, s[8:11], 0 offset:552 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:64
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:540 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:544 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v43, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:528 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v43, off, s[8:11], 0 offset:524 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v44, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:516 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v42, off, s[8:11], 0 offset:536 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v45, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:532 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v42, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v42, off, s[8:11], 0 offset:520 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:508 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v42, 8, v42
+; GFX906-NEXT: v_or_b32_sdwa v42, v43, v42 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v42, off, s[8:11], 0 offset:512 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:500 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v42, off, s[8:11], 0 offset:504 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:80
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:492 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:496 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v39, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:480 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v39, off, s[8:11], 0 offset:476 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v40, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:468 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v38, off, s[8:11], 0 offset:488 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v41, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:484 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v38, off, s[8:11], 0 offset:472 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:460 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v38, 8, v38
+; GFX906-NEXT: v_or_b32_sdwa v38, v39, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v38 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v38, off, s[8:11], 0 offset:464 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:452 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v38, off, s[8:11], 0 offset:456 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:96
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:444 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:448 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v35, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:432 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v35, off, s[8:11], 0 offset:428 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v34, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v36, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:420 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v34, off, s[8:11], 0 offset:440 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v37, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:436 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v34, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v34, off, s[8:11], 0 offset:424 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:412 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v34, 8, v34
+; GFX906-NEXT: v_or_b32_sdwa v34, v35, v34 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v34, off, s[8:11], 0 offset:416 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v34, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:404 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v34, off, s[8:11], 0 offset:408 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v34, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:112
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:396 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:400 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v31, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:384 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v31, off, s[8:11], 0 offset:380 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v32, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:372 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v30, off, s[8:11], 0 offset:392 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v33, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:388 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v30, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v30, off, s[8:11], 0 offset:376 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:364 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v30, 8, v30
+; GFX906-NEXT: v_or_b32_sdwa v30, v31, v30 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v30 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v30, off, s[8:11], 0 offset:368 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:356 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v30, off, s[8:11], 0 offset:360 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:128
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:348 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:352 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v27, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:336 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v27, off, s[8:11], 0 offset:332 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v26, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v28, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:324 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v26, off, s[8:11], 0 offset:344 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v29, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:340 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v26, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v26, off, s[8:11], 0 offset:328 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:316 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v26, 8, v26
+; GFX906-NEXT: v_or_b32_sdwa v26, v27, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v26, off, s[8:11], 0 offset:320 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v26, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:308 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v26, off, s[8:11], 0 offset:312 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v26, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:144
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:300 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:304 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v23, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:288 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v23, off, s[8:11], 0 offset:284 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v22, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v24, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:276 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v22, off, s[8:11], 0 offset:296 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v25, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:292 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v22, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v22, off, s[8:11], 0 offset:280 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:268 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v22, 8, v22
+; GFX906-NEXT: v_or_b32_sdwa v22, v23, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v22, off, s[8:11], 0 offset:272 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v22, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:260 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v22, off, s[8:11], 0 offset:264 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v22, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:160
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:252 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:256 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:240 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v19, off, s[8:11], 0 offset:236 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v20, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:228 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v18, off, s[8:11], 0 offset:248 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v21, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:244 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v18, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v18, off, s[8:11], 0 offset:232 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:220 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v18, 8, v18
+; GFX906-NEXT: v_or_b32_sdwa v18, v19, v18 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v18, off, s[8:11], 0 offset:224 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:212 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v18, off, s[8:11], 0 offset:216 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:176
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:204 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:208 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v3, v15, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:192 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v15, off, s[8:11], 0 offset:188 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(2)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v14, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v4, v16, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:180 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v14, off, s[8:11], 0 offset:200 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v5, v17, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:196 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v14, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v14, off, s[8:11], 0 offset:184 ; 4-byte Folded Reload
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:172 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v14, 8, v14
+; GFX906-NEXT: v_or_b32_sdwa v14, v15, v14 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v3, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v14, off, s[8:11], 0 offset:176 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v14, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:164 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v14, off, s[8:11], 0 offset:168 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_or_b32_sdwa v0, v14, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:192
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:160 ; 4-byte Folded Reload
+; GFX906-NEXT: s_nop 0
+; GFX906-NEXT: buffer_load_dword v3, off, s[8:11], 0 offset:156 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:152 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v4, off, s[8:11], 0 offset:144 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v5, off, s[8:11], 0 offset:132 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(4)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v10, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v10, off, s[8:11], 0 offset:120 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(3)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:148 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v3, off, s[8:11], 0 offset:140 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v3, 8, v3
+; GFX906-NEXT: v_or_b32_sdwa v0, v11, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:136 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v4, off, s[8:11], 0 offset:128 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v4, 8, v4
+; GFX906-NEXT: v_or_b32_sdwa v0, v12, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:124 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v5, off, s[8:11], 0 offset:116 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX906-NEXT: v_or_b32_sdwa v0, v13, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v10, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:208
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:112 ; 4-byte Folded Reload
+; GFX906-NEXT: s_nop 0
+; GFX906-NEXT: buffer_load_dword v3, off, s[8:11], 0 offset:108 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:104 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v4, off, s[8:11], 0 offset:96 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v5, off, s[8:11], 0 offset:84 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(4)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v6, off, s[8:11], 0 offset:72 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(3)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:100 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v3, off, s[8:11], 0 offset:92 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v3, 8, v3
+; GFX906-NEXT: v_or_b32_sdwa v0, v7, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:88 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v4, off, s[8:11], 0 offset:80 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v4, 8, v4
+; GFX906-NEXT: v_or_b32_sdwa v0, v8, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:76 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v5, off, s[8:11], 0 offset:68 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX906-NEXT: v_or_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:224
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:64 ; 4-byte Folded Reload
+; GFX906-NEXT: s_nop 0
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:4 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v3, off, s[8:11], 0 offset:8 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v4, off, s[8:11], 0 offset:12 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v5, off, s[8:11], 0 offset:16 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v6, off, s[8:11], 0 offset:60 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(5)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v2, off, s[8:11], 0 offset:56 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; GFX906-NEXT: v_or_b32_sdwa v2, v6, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:52 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v6, off, s[8:11], 0 offset:48 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v3, off, s[8:11], 0 offset:44 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v3, 8, v3
+; GFX906-NEXT: v_or_b32_sdwa v3, v6, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:40 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v6, off, s[8:11], 0 offset:36 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v4, off, s[8:11], 0 offset:32 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v4, 8, v4
+; GFX906-NEXT: v_or_b32_sdwa v4, v6, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v4, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v0, off, s[8:11], 0 offset:28 ; 4-byte Folded Reload
+; GFX906-NEXT: buffer_load_dword v6, off, s[8:11], 0 offset:24 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(1)
+; GFX906-NEXT: v_lshlrev_b16_e32 v0, 8, v0
+; GFX906-NEXT: v_or_b32_sdwa v0, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: buffer_load_dword v5, off, s[8:11], 0 offset:20 ; 4-byte Folded Reload
+; GFX906-NEXT: s_waitcnt vmcnt(0)
+; GFX906-NEXT: v_lshlrev_b16_e32 v5, 8, v5
+; GFX906-NEXT: v_or_b32_sdwa v5, v6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX906-NEXT: v_or_b32_sdwa v5, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX906-NEXT: global_store_dwordx4 v1, v[2:5], s[2:3] offset:240
+; GFX906-NEXT: s_endpgm
+entry:
+ %idx = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr <8 x i8>, ptr addrspace(1) %src1, i32 %idx
+ %vec1 = load <256 x i8>, ptr addrspace(1) %gep1
+ %gep2 = getelementptr <8 x i8>, ptr addrspace(1) %src2, i32 %idx
+ %vec2 = load <256 x i8>, ptr addrspace(1) %gep2
+ %cmp = icmp ult i32 %idx, 15
+ br i1 %cmp, label %bb.1, label %bb.2
+bb.1:
+ br label %bb.2
+
+bb.2:
+ %tmp5 = phi <256 x i8> [ %vec1, %entry ], [ %vec2, %bb.1 ]
+ store <256 x i8> %tmp5, ptr addrspace(1) %dst, align 4
+ ret void
+}
+
+declare i32 @llvm.amdgcn.workitem.id.x()
+
More information about the llvm-commits
mailing list