caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
8fcb1263f42657ecbc355beff12500dfbcddee17
clang-p2996/llvm/test/CodeGen/AMDGPU/indirect-call.ll
Aaditya 11b0401926 [AMDGPU] Restore SP from saved-FP or saved-BP (#124007) 
Currently, the AMDGPU backend bumps the Stack Pointer 
by fixed size offsets in the prolog of device functions, and 
restores it by the same amount in the epilog.
Prolog:
sp += frameSize

Epilog:
sp -= frameSize

If a function has dynamic stack realignment,
Prolog:
sp += frameSize + max_alignment

Epilog:
sp -= frameSize + max_alignment

These calculations are not optimal in case of dynamic 
stack realignment, and completely fail in case of 
dynamic stack readjustment.
This patch uses the saved Frame Pointer to restore SP. 
Prolog:
fp = sp
sp += frameSize

Epilog:
sp = fp

In case of dynamic stack realignment, SP is restored from 
the saved Base Pointer. 
Prolog:
fp = sp + (max_alignment - 1)
fp = fp & (-max_alignment)
bp = sp
sp += frameSize + max_alignment

Epilog:
sp = bp

(Note: The presence of BP has been enforced in case of any 
dynamic stack realignment.)

---------

Co-authored-by: Pravin Jagtap <Pravin.Jagtap@amd.com>
Co-authored-by: Matt Arsenault <arsenm2@gmail.com>
2025-01-24 19:13:40 +05:30

1631 lines
68 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn-amd-amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -verify-machineinstrs -global-isel < %s | FileCheck -check-prefix=GISEL %s
@gv.fptr0 = external hidden unnamed_addr addrspace(4) constant ptr, align 4
@gv.fptr1 = external hidden unnamed_addr addrspace(4) constant ptr, align 4
define amdgpu_kernel void @test_indirect_call_sgpr_ptr(i8) {
; GCN-LABEL: test_indirect_call_sgpr_ptr:
; GCN: ; %bb.0:
; GCN-NEXT: s_mov_b32 s32, 0
; GCN-NEXT: s_mov_b32 flat_scratch_lo, s13
; GCN-NEXT: s_add_i32 s12, s12, s17
; GCN-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
; GCN-NEXT: s_add_u32 s0, s0, s17
; GCN-NEXT: s_addc_u32 s1, s1, 0
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_getpc_b64 s[14:15]
; GCN-NEXT: s_add_u32 s14, s14, gv.fptr0@rel32@lo+4
; GCN-NEXT: s_addc_u32 s15, s15, gv.fptr0@rel32@hi+12
; GCN-NEXT: v_lshlrev_b32_e32 v2, 20, v2
; GCN-NEXT: s_load_dwordx2 s[18:19], s[14:15], 0x0
; GCN-NEXT: s_add_u32 s8, s8, 8
; GCN-NEXT: s_addc_u32 s9, s9, 0
; GCN-NEXT: v_lshlrev_b32_e32 v1, 10, v1
; GCN-NEXT: v_or_b32_e32 v0, v0, v1
; GCN-NEXT: v_or_b32_e32 v31, v0, v2
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_swappc_b64 s[30:31], s[18:19]
; GCN-NEXT: s_endpgm
;
; GISEL-LABEL: test_indirect_call_sgpr_ptr:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_mov_b32 s32, 0
; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13
; GISEL-NEXT: s_add_i32 s12, s12, s17
; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
; GISEL-NEXT: s_add_u32 s0, s0, s17
; GISEL-NEXT: s_addc_u32 s1, s1, 0
; GISEL-NEXT: s_mov_b32 s13, s15
; GISEL-NEXT: s_mov_b32 s12, s14
; GISEL-NEXT: s_getpc_b64 s[14:15]
; GISEL-NEXT: s_add_u32 s14, s14, gv.fptr0@rel32@lo+4
; GISEL-NEXT: s_addc_u32 s15, s15, gv.fptr0@rel32@hi+12
; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1
; GISEL-NEXT: s_load_dwordx2 s[18:19], s[14:15], 0x0
; GISEL-NEXT: v_or_b32_e32 v0, v0, v1
; GISEL-NEXT: s_add_u32 s8, s8, 8
; GISEL-NEXT: s_addc_u32 s9, s9, 0
; GISEL-NEXT: v_lshlrev_b32_e32 v1, 20, v2
; GISEL-NEXT: v_or_b32_e32 v31, v0, v1
; GISEL-NEXT: s_mov_b32 s14, s16
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19]
; GISEL-NEXT: s_endpgm
%fptr = load ptr, ptr addrspace(4) @gv.fptr0
call void %fptr()
ret void
}
define amdgpu_kernel void @test_indirect_call_sgpr_ptr_arg(i8) {
; GCN-LABEL: test_indirect_call_sgpr_ptr_arg:
; GCN: ; %bb.0:
; GCN-NEXT: s_mov_b32 s32, 0
; GCN-NEXT: s_mov_b32 flat_scratch_lo, s13
; GCN-NEXT: s_add_i32 s12, s12, s17
; GCN-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
; GCN-NEXT: s_add_u32 s0, s0, s17
; GCN-NEXT: s_addc_u32 s1, s1, 0
; GCN-NEXT: s_mov_b32 s13, s15
; GCN-NEXT: s_mov_b32 s12, s14
; GCN-NEXT: s_getpc_b64 s[14:15]
; GCN-NEXT: s_add_u32 s14, s14, gv.fptr1@rel32@lo+4
; GCN-NEXT: s_addc_u32 s15, s15, gv.fptr1@rel32@hi+12
; GCN-NEXT: v_lshlrev_b32_e32 v2, 20, v2
; GCN-NEXT: v_lshlrev_b32_e32 v1, 10, v1
; GCN-NEXT: s_load_dwordx2 s[18:19], s[14:15], 0x0
; GCN-NEXT: s_add_u32 s8, s8, 8
; GCN-NEXT: s_addc_u32 s9, s9, 0
; GCN-NEXT: v_or_b32_e32 v0, v0, v1
; GCN-NEXT: v_or_b32_e32 v31, v0, v2
; GCN-NEXT: v_mov_b32_e32 v0, 0x7b
; GCN-NEXT: s_mov_b32 s14, s16
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_swappc_b64 s[30:31], s[18:19]
; GCN-NEXT: s_endpgm
;
; GISEL-LABEL: test_indirect_call_sgpr_ptr_arg:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_mov_b32 s32, 0
; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13
; GISEL-NEXT: s_add_i32 s12, s12, s17
; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
; GISEL-NEXT: s_add_u32 s0, s0, s17
; GISEL-NEXT: s_addc_u32 s1, s1, 0
; GISEL-NEXT: s_mov_b32 s13, s15
; GISEL-NEXT: s_mov_b32 s12, s14
; GISEL-NEXT: s_getpc_b64 s[14:15]
; GISEL-NEXT: s_add_u32 s14, s14, gv.fptr1@rel32@lo+4
; GISEL-NEXT: s_addc_u32 s15, s15, gv.fptr1@rel32@hi+12
; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1
; GISEL-NEXT: v_lshlrev_b32_e32 v2, 20, v2
; GISEL-NEXT: s_load_dwordx2 s[18:19], s[14:15], 0x0
; GISEL-NEXT: v_or_b32_e32 v0, v0, v1
; GISEL-NEXT: s_add_u32 s8, s8, 8
; GISEL-NEXT: s_addc_u32 s9, s9, 0
; GISEL-NEXT: v_or_b32_e32 v31, v0, v2
; GISEL-NEXT: v_mov_b32_e32 v0, 0x7b
; GISEL-NEXT: s_mov_b32 s14, s16
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19]
; GISEL-NEXT: s_endpgm
%fptr = load ptr, ptr addrspace(4) @gv.fptr1
call void %fptr(i32 123)
ret void
}
define void @test_indirect_call_vgpr_ptr(ptr %fptr) {
; GCN-LABEL: test_indirect_call_vgpr_ptr:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s16, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[18:19], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[18:19]
; GCN-NEXT: v_writelane_b32 v40, s16, 18
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: s_mov_b32 s42, s15
; GCN-NEXT: s_mov_b32 s43, s14
; GCN-NEXT: s_mov_b32 s44, s13
; GCN-NEXT: s_mov_b32 s45, s12
; GCN-NEXT: s_mov_b64 s[34:35], s[10:11]
; GCN-NEXT: s_mov_b64 s[36:37], s[8:9]
; GCN-NEXT: s_mov_b64 s[38:39], s[6:7]
; GCN-NEXT: s_mov_b64 s[40:41], s[4:5]
; GCN-NEXT: s_mov_b64 s[46:47], exec
; GCN-NEXT: .LBB2_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s16, v0
; GCN-NEXT: v_readfirstlane_b32 s17, v1
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GCN-NEXT: s_and_saveexec_b64 s[48:49], vcc
; GCN-NEXT: s_mov_b64 s[4:5], s[40:41]
; GCN-NEXT: s_mov_b64 s[6:7], s[38:39]
; GCN-NEXT: s_mov_b64 s[8:9], s[36:37]
; GCN-NEXT: s_mov_b64 s[10:11], s[34:35]
; GCN-NEXT: s_mov_b32 s12, s45
; GCN-NEXT: s_mov_b32 s13, s44
; GCN-NEXT: s_mov_b32 s14, s43
; GCN-NEXT: s_mov_b32 s15, s42
; GCN-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GCN-NEXT: ; implicit-def: $vgpr0_vgpr1
; GCN-NEXT: ; implicit-def: $vgpr31
; GCN-NEXT: s_xor_b64 exec, exec, s[48:49]
; GCN-NEXT: s_cbranch_execnz .LBB2_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[46:47]
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: v_readlane_b32 s4, v40, 18
; GCN-NEXT: s_or_saveexec_b64 s[6:7], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: s_mov_b32 s33, s4
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s16, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[18:19], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[18:19]
; GISEL-NEXT: v_writelane_b32 v40, s16, 18
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: s_mov_b32 s42, s15
; GISEL-NEXT: s_mov_b32 s43, s14
; GISEL-NEXT: s_mov_b32 s44, s13
; GISEL-NEXT: s_mov_b32 s45, s12
; GISEL-NEXT: s_mov_b64 s[34:35], s[10:11]
; GISEL-NEXT: s_mov_b64 s[36:37], s[8:9]
; GISEL-NEXT: s_mov_b64 s[38:39], s[6:7]
; GISEL-NEXT: s_mov_b64 s[40:41], s[4:5]
; GISEL-NEXT: s_mov_b64 s[46:47], exec
; GISEL-NEXT: .LBB2_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s16, v0
; GISEL-NEXT: v_readfirstlane_b32 s17, v1
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GISEL-NEXT: s_and_saveexec_b64 s[48:49], vcc
; GISEL-NEXT: s_mov_b64 s[4:5], s[40:41]
; GISEL-NEXT: s_mov_b64 s[6:7], s[38:39]
; GISEL-NEXT: s_mov_b64 s[8:9], s[36:37]
; GISEL-NEXT: s_mov_b64 s[10:11], s[34:35]
; GISEL-NEXT: s_mov_b32 s12, s45
; GISEL-NEXT: s_mov_b32 s13, s44
; GISEL-NEXT: s_mov_b32 s14, s43
; GISEL-NEXT: s_mov_b32 s15, s42
; GISEL-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: ; implicit-def: $vgpr31
; GISEL-NEXT: s_xor_b64 exec, exec, s[48:49]
; GISEL-NEXT: s_cbranch_execnz .LBB2_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[46:47]
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: v_readlane_b32 s4, v40, 18
; GISEL-NEXT: s_or_saveexec_b64 s[6:7], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: s_mov_b32 s33, s4
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
call void %fptr()
ret void
}
define void @test_indirect_call_vgpr_ptr_arg(ptr %fptr) {
; GCN-LABEL: test_indirect_call_vgpr_ptr_arg:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s16, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[18:19], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[18:19]
; GCN-NEXT: v_writelane_b32 v40, s16, 18
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: s_mov_b32 s42, s15
; GCN-NEXT: s_mov_b32 s43, s14
; GCN-NEXT: s_mov_b32 s44, s13
; GCN-NEXT: s_mov_b32 s45, s12
; GCN-NEXT: s_mov_b64 s[34:35], s[10:11]
; GCN-NEXT: s_mov_b64 s[36:37], s[8:9]
; GCN-NEXT: s_mov_b64 s[38:39], s[6:7]
; GCN-NEXT: s_mov_b64 s[40:41], s[4:5]
; GCN-NEXT: s_mov_b64 s[46:47], exec
; GCN-NEXT: v_mov_b32_e32 v2, 0x7b
; GCN-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s16, v0
; GCN-NEXT: v_readfirstlane_b32 s17, v1
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GCN-NEXT: s_and_saveexec_b64 s[48:49], vcc
; GCN-NEXT: s_mov_b64 s[4:5], s[40:41]
; GCN-NEXT: s_mov_b64 s[6:7], s[38:39]
; GCN-NEXT: s_mov_b64 s[8:9], s[36:37]
; GCN-NEXT: s_mov_b64 s[10:11], s[34:35]
; GCN-NEXT: s_mov_b32 s12, s45
; GCN-NEXT: s_mov_b32 s13, s44
; GCN-NEXT: s_mov_b32 s14, s43
; GCN-NEXT: s_mov_b32 s15, s42
; GCN-NEXT: v_mov_b32_e32 v0, v2
; GCN-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GCN-NEXT: ; implicit-def: $vgpr0_vgpr1
; GCN-NEXT: ; implicit-def: $vgpr31
; GCN-NEXT: ; implicit-def: $vgpr2
; GCN-NEXT: s_xor_b64 exec, exec, s[48:49]
; GCN-NEXT: s_cbranch_execnz .LBB3_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[46:47]
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: v_readlane_b32 s4, v40, 18
; GCN-NEXT: s_or_saveexec_b64 s[6:7], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: s_mov_b32 s33, s4
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr_arg:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s16, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[18:19], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[18:19]
; GISEL-NEXT: v_writelane_b32 v40, s16, 18
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: s_mov_b32 s42, s15
; GISEL-NEXT: s_mov_b32 s43, s14
; GISEL-NEXT: s_mov_b32 s44, s13
; GISEL-NEXT: s_mov_b32 s45, s12
; GISEL-NEXT: s_mov_b64 s[34:35], s[10:11]
; GISEL-NEXT: s_mov_b64 s[36:37], s[8:9]
; GISEL-NEXT: s_mov_b64 s[38:39], s[6:7]
; GISEL-NEXT: s_mov_b64 s[40:41], s[4:5]
; GISEL-NEXT: s_mov_b64 s[46:47], exec
; GISEL-NEXT: .LBB3_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s16, v0
; GISEL-NEXT: v_readfirstlane_b32 s17, v1
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GISEL-NEXT: s_and_saveexec_b64 s[48:49], vcc
; GISEL-NEXT: v_mov_b32_e32 v0, 0x7b
; GISEL-NEXT: s_mov_b64 s[4:5], s[40:41]
; GISEL-NEXT: s_mov_b64 s[6:7], s[38:39]
; GISEL-NEXT: s_mov_b64 s[8:9], s[36:37]
; GISEL-NEXT: s_mov_b64 s[10:11], s[34:35]
; GISEL-NEXT: s_mov_b32 s12, s45
; GISEL-NEXT: s_mov_b32 s13, s44
; GISEL-NEXT: s_mov_b32 s14, s43
; GISEL-NEXT: s_mov_b32 s15, s42
; GISEL-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: ; implicit-def: $vgpr31
; GISEL-NEXT: s_xor_b64 exec, exec, s[48:49]
; GISEL-NEXT: s_cbranch_execnz .LBB3_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[46:47]
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: v_readlane_b32 s4, v40, 18
; GISEL-NEXT: s_or_saveexec_b64 s[6:7], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: s_mov_b32 s33, s4
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
call void %fptr(i32 123)
ret void
}
define i32 @test_indirect_call_vgpr_ptr_ret(ptr %fptr) {
; GCN-LABEL: test_indirect_call_vgpr_ptr_ret:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s16, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[18:19], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[18:19]
; GCN-NEXT: v_writelane_b32 v40, s16, 18
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: s_mov_b32 s42, s15
; GCN-NEXT: s_mov_b32 s43, s14
; GCN-NEXT: s_mov_b32 s44, s13
; GCN-NEXT: s_mov_b32 s45, s12
; GCN-NEXT: s_mov_b64 s[34:35], s[10:11]
; GCN-NEXT: s_mov_b64 s[36:37], s[8:9]
; GCN-NEXT: s_mov_b64 s[38:39], s[6:7]
; GCN-NEXT: s_mov_b64 s[40:41], s[4:5]
; GCN-NEXT: s_mov_b64 s[46:47], exec
; GCN-NEXT: .LBB4_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s16, v0
; GCN-NEXT: v_readfirstlane_b32 s17, v1
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GCN-NEXT: s_and_saveexec_b64 s[48:49], vcc
; GCN-NEXT: s_mov_b64 s[4:5], s[40:41]
; GCN-NEXT: s_mov_b64 s[6:7], s[38:39]
; GCN-NEXT: s_mov_b64 s[8:9], s[36:37]
; GCN-NEXT: s_mov_b64 s[10:11], s[34:35]
; GCN-NEXT: s_mov_b32 s12, s45
; GCN-NEXT: s_mov_b32 s13, s44
; GCN-NEXT: s_mov_b32 s14, s43
; GCN-NEXT: s_mov_b32 s15, s42
; GCN-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GCN-NEXT: v_mov_b32_e32 v2, v0
; GCN-NEXT: ; implicit-def: $vgpr0_vgpr1
; GCN-NEXT: ; implicit-def: $vgpr31
; GCN-NEXT: s_xor_b64 exec, exec, s[48:49]
; GCN-NEXT: s_cbranch_execnz .LBB4_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[46:47]
; GCN-NEXT: v_add_i32_e32 v0, vcc, 1, v2
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: v_readlane_b32 s4, v40, 18
; GCN-NEXT: s_or_saveexec_b64 s[6:7], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: s_mov_b32 s33, s4
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr_ret:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s16, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[18:19], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[18:19]
; GISEL-NEXT: v_writelane_b32 v40, s16, 18
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: s_mov_b32 s42, s15
; GISEL-NEXT: s_mov_b32 s43, s14
; GISEL-NEXT: s_mov_b32 s44, s13
; GISEL-NEXT: s_mov_b32 s45, s12
; GISEL-NEXT: s_mov_b64 s[34:35], s[10:11]
; GISEL-NEXT: s_mov_b64 s[36:37], s[8:9]
; GISEL-NEXT: s_mov_b64 s[38:39], s[6:7]
; GISEL-NEXT: s_mov_b64 s[40:41], s[4:5]
; GISEL-NEXT: s_mov_b64 s[46:47], exec
; GISEL-NEXT: .LBB4_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s16, v0
; GISEL-NEXT: v_readfirstlane_b32 s17, v1
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GISEL-NEXT: s_and_saveexec_b64 s[48:49], vcc
; GISEL-NEXT: s_mov_b64 s[4:5], s[40:41]
; GISEL-NEXT: s_mov_b64 s[6:7], s[38:39]
; GISEL-NEXT: s_mov_b64 s[8:9], s[36:37]
; GISEL-NEXT: s_mov_b64 s[10:11], s[34:35]
; GISEL-NEXT: s_mov_b32 s12, s45
; GISEL-NEXT: s_mov_b32 s13, s44
; GISEL-NEXT: s_mov_b32 s14, s43
; GISEL-NEXT: s_mov_b32 s15, s42
; GISEL-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GISEL-NEXT: v_mov_b32_e32 v1, v0
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: ; implicit-def: $vgpr31
; GISEL-NEXT: s_xor_b64 exec, exec, s[48:49]
; GISEL-NEXT: s_cbranch_execnz .LBB4_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[46:47]
; GISEL-NEXT: v_add_i32_e32 v0, vcc, 1, v1
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: v_readlane_b32 s4, v40, 18
; GISEL-NEXT: s_or_saveexec_b64 s[6:7], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: s_mov_b32 s33, s4
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
%a = call i32 %fptr()
%b = add i32 %a, 1
ret i32 %b
}
define void @test_indirect_call_vgpr_ptr_in_branch(ptr %fptr, i1 %cond) {
; GCN-LABEL: test_indirect_call_vgpr_ptr_in_branch:
; GCN: ; %bb.0: ; %bb0
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s16, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[18:19], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[18:19]
; GCN-NEXT: v_writelane_b32 v40, s16, 20
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: v_writelane_b32 v40, s50, 18
; GCN-NEXT: v_writelane_b32 v40, s51, 19
; GCN-NEXT: s_mov_b32 s42, s15
; GCN-NEXT: s_mov_b32 s43, s14
; GCN-NEXT: s_mov_b32 s44, s13
; GCN-NEXT: s_mov_b32 s45, s12
; GCN-NEXT: s_mov_b64 s[34:35], s[10:11]
; GCN-NEXT: s_mov_b64 s[36:37], s[8:9]
; GCN-NEXT: s_mov_b64 s[38:39], s[6:7]
; GCN-NEXT: s_mov_b64 s[40:41], s[4:5]
; GCN-NEXT: v_and_b32_e32 v2, 1, v2
; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 1, v2
; GCN-NEXT: s_and_saveexec_b64 s[46:47], vcc
; GCN-NEXT: s_cbranch_execz .LBB5_4
; GCN-NEXT: ; %bb.1: ; %bb1
; GCN-NEXT: s_mov_b64 s[48:49], exec
; GCN-NEXT: .LBB5_2: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s16, v0
; GCN-NEXT: v_readfirstlane_b32 s17, v1
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GCN-NEXT: s_and_saveexec_b64 s[50:51], vcc
; GCN-NEXT: s_mov_b64 s[4:5], s[40:41]
; GCN-NEXT: s_mov_b64 s[6:7], s[38:39]
; GCN-NEXT: s_mov_b64 s[8:9], s[36:37]
; GCN-NEXT: s_mov_b64 s[10:11], s[34:35]
; GCN-NEXT: s_mov_b32 s12, s45
; GCN-NEXT: s_mov_b32 s13, s44
; GCN-NEXT: s_mov_b32 s14, s43
; GCN-NEXT: s_mov_b32 s15, s42
; GCN-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GCN-NEXT: ; implicit-def: $vgpr0_vgpr1
; GCN-NEXT: ; implicit-def: $vgpr31
; GCN-NEXT: s_xor_b64 exec, exec, s[50:51]
; GCN-NEXT: s_cbranch_execnz .LBB5_2
; GCN-NEXT: ; %bb.3:
; GCN-NEXT: s_mov_b64 exec, s[48:49]
; GCN-NEXT: .LBB5_4: ; %bb2
; GCN-NEXT: s_or_b64 exec, exec, s[46:47]
; GCN-NEXT: v_readlane_b32 s51, v40, 19
; GCN-NEXT: v_readlane_b32 s50, v40, 18
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: v_readlane_b32 s4, v40, 20
; GCN-NEXT: s_or_saveexec_b64 s[6:7], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: s_mov_b32 s33, s4
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr_in_branch:
; GISEL: ; %bb.0: ; %bb0
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s16, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[18:19], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[18:19]
; GISEL-NEXT: v_writelane_b32 v40, s16, 20
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: v_writelane_b32 v40, s50, 18
; GISEL-NEXT: v_writelane_b32 v40, s51, 19
; GISEL-NEXT: s_mov_b32 s42, s15
; GISEL-NEXT: s_mov_b32 s43, s14
; GISEL-NEXT: s_mov_b32 s44, s13
; GISEL-NEXT: s_mov_b32 s45, s12
; GISEL-NEXT: s_mov_b64 s[34:35], s[10:11]
; GISEL-NEXT: s_mov_b64 s[36:37], s[8:9]
; GISEL-NEXT: s_mov_b64 s[38:39], s[6:7]
; GISEL-NEXT: s_mov_b64 s[40:41], s[4:5]
; GISEL-NEXT: v_and_b32_e32 v2, 1, v2
; GISEL-NEXT: v_cmp_ne_u32_e32 vcc, 0, v2
; GISEL-NEXT: s_and_saveexec_b64 s[46:47], vcc
; GISEL-NEXT: s_cbranch_execz .LBB5_4
; GISEL-NEXT: ; %bb.1: ; %bb1
; GISEL-NEXT: s_mov_b64 s[48:49], exec
; GISEL-NEXT: .LBB5_2: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s16, v0
; GISEL-NEXT: v_readfirstlane_b32 s17, v1
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
; GISEL-NEXT: s_and_saveexec_b64 s[50:51], vcc
; GISEL-NEXT: s_mov_b64 s[4:5], s[40:41]
; GISEL-NEXT: s_mov_b64 s[6:7], s[38:39]
; GISEL-NEXT: s_mov_b64 s[8:9], s[36:37]
; GISEL-NEXT: s_mov_b64 s[10:11], s[34:35]
; GISEL-NEXT: s_mov_b32 s12, s45
; GISEL-NEXT: s_mov_b32 s13, s44
; GISEL-NEXT: s_mov_b32 s14, s43
; GISEL-NEXT: s_mov_b32 s15, s42
; GISEL-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: ; implicit-def: $vgpr31
; GISEL-NEXT: s_xor_b64 exec, exec, s[50:51]
; GISEL-NEXT: s_cbranch_execnz .LBB5_2
; GISEL-NEXT: ; %bb.3:
; GISEL-NEXT: s_mov_b64 exec, s[48:49]
; GISEL-NEXT: .LBB5_4: ; %bb2
; GISEL-NEXT: s_or_b64 exec, exec, s[46:47]
; GISEL-NEXT: v_readlane_b32 s51, v40, 19
; GISEL-NEXT: v_readlane_b32 s50, v40, 18
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: v_readlane_b32 s4, v40, 20
; GISEL-NEXT: s_or_saveexec_b64 s[6:7], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: s_mov_b32 s33, s4
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
bb0:
br i1 %cond, label %bb1, label %bb2
bb1:
call void %fptr()
br label %bb2
bb2:
ret void
}
define void @test_indirect_call_vgpr_ptr_inreg_arg(ptr %fptr) {
; GCN-LABEL: test_indirect_call_vgpr_ptr_inreg_arg:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s5, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[6:7], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: v_writelane_b32 v40, s50, 18
; GCN-NEXT: v_writelane_b32 v40, s51, 19
; GCN-NEXT: v_writelane_b32 v40, s52, 20
; GCN-NEXT: v_writelane_b32 v40, s53, 21
; GCN-NEXT: v_writelane_b32 v40, s54, 22
; GCN-NEXT: v_writelane_b32 v40, s55, 23
; GCN-NEXT: v_writelane_b32 v40, s56, 24
; GCN-NEXT: v_writelane_b32 v40, s57, 25
; GCN-NEXT: v_writelane_b32 v40, s58, 26
; GCN-NEXT: v_writelane_b32 v40, s59, 27
; GCN-NEXT: v_writelane_b32 v40, s60, 28
; GCN-NEXT: v_writelane_b32 v40, s61, 29
; GCN-NEXT: v_writelane_b32 v40, s62, 30
; GCN-NEXT: v_writelane_b32 v40, s63, 31
; GCN-NEXT: s_mov_b64 s[6:7], exec
; GCN-NEXT: .LBB6_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s8, v0
; GCN-NEXT: v_readfirstlane_b32 s9, v1
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[8:9], v[0:1]
; GCN-NEXT: s_and_saveexec_b64 s[10:11], vcc
; GCN-NEXT: s_movk_i32 s4, 0x7b
; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9]
; GCN-NEXT: ; implicit-def: $vgpr0_vgpr1
; GCN-NEXT: s_xor_b64 exec, exec, s[10:11]
; GCN-NEXT: s_cbranch_execnz .LBB6_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: v_readlane_b32 s63, v40, 31
; GCN-NEXT: v_readlane_b32 s62, v40, 30
; GCN-NEXT: v_readlane_b32 s61, v40, 29
; GCN-NEXT: v_readlane_b32 s60, v40, 28
; GCN-NEXT: v_readlane_b32 s59, v40, 27
; GCN-NEXT: v_readlane_b32 s58, v40, 26
; GCN-NEXT: v_readlane_b32 s57, v40, 25
; GCN-NEXT: v_readlane_b32 s56, v40, 24
; GCN-NEXT: v_readlane_b32 s55, v40, 23
; GCN-NEXT: v_readlane_b32 s54, v40, 22
; GCN-NEXT: v_readlane_b32 s53, v40, 21
; GCN-NEXT: v_readlane_b32 s52, v40, 20
; GCN-NEXT: v_readlane_b32 s51, v40, 19
; GCN-NEXT: v_readlane_b32 s50, v40, 18
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: s_or_saveexec_b64 s[6:7], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[6:7]
; GCN-NEXT: s_mov_b32 s33, s5
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr_inreg_arg:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s5, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[6:7], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: v_writelane_b32 v40, s50, 18
; GISEL-NEXT: v_writelane_b32 v40, s51, 19
; GISEL-NEXT: v_writelane_b32 v40, s52, 20
; GISEL-NEXT: v_writelane_b32 v40, s53, 21
; GISEL-NEXT: v_writelane_b32 v40, s54, 22
; GISEL-NEXT: v_writelane_b32 v40, s55, 23
; GISEL-NEXT: v_writelane_b32 v40, s56, 24
; GISEL-NEXT: v_writelane_b32 v40, s57, 25
; GISEL-NEXT: v_writelane_b32 v40, s58, 26
; GISEL-NEXT: v_writelane_b32 v40, s59, 27
; GISEL-NEXT: v_writelane_b32 v40, s60, 28
; GISEL-NEXT: v_writelane_b32 v40, s61, 29
; GISEL-NEXT: v_writelane_b32 v40, s62, 30
; GISEL-NEXT: v_writelane_b32 v40, s63, 31
; GISEL-NEXT: s_mov_b64 s[6:7], exec
; GISEL-NEXT: .LBB6_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s8, v0
; GISEL-NEXT: v_readfirstlane_b32 s9, v1
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[8:9], v[0:1]
; GISEL-NEXT: s_and_saveexec_b64 s[10:11], vcc
; GISEL-NEXT: s_movk_i32 s4, 0x7b
; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9]
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: s_xor_b64 exec, exec, s[10:11]
; GISEL-NEXT: s_cbranch_execnz .LBB6_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: v_readlane_b32 s63, v40, 31
; GISEL-NEXT: v_readlane_b32 s62, v40, 30
; GISEL-NEXT: v_readlane_b32 s61, v40, 29
; GISEL-NEXT: v_readlane_b32 s60, v40, 28
; GISEL-NEXT: v_readlane_b32 s59, v40, 27
; GISEL-NEXT: v_readlane_b32 s58, v40, 26
; GISEL-NEXT: v_readlane_b32 s57, v40, 25
; GISEL-NEXT: v_readlane_b32 s56, v40, 24
; GISEL-NEXT: v_readlane_b32 s55, v40, 23
; GISEL-NEXT: v_readlane_b32 s54, v40, 22
; GISEL-NEXT: v_readlane_b32 s53, v40, 21
; GISEL-NEXT: v_readlane_b32 s52, v40, 20
; GISEL-NEXT: v_readlane_b32 s51, v40, 19
; GISEL-NEXT: v_readlane_b32 s50, v40, 18
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: s_or_saveexec_b64 s[6:7], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[6:7]
; GISEL-NEXT: s_mov_b32 s33, s5
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
call amdgpu_gfx void %fptr(i32 inreg 123)
ret void
}
define i32 @test_indirect_call_vgpr_ptr_arg_and_reuse(i32 %i, ptr %fptr) {
; GCN-LABEL: test_indirect_call_vgpr_ptr_arg_and_reuse:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s10, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[4:5], -1
; GCN-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: v_writelane_b32 v41, s30, 0
; GCN-NEXT: v_writelane_b32 v41, s31, 1
; GCN-NEXT: v_writelane_b32 v41, s34, 2
; GCN-NEXT: v_writelane_b32 v41, s35, 3
; GCN-NEXT: v_writelane_b32 v41, s36, 4
; GCN-NEXT: v_writelane_b32 v41, s37, 5
; GCN-NEXT: v_writelane_b32 v41, s38, 6
; GCN-NEXT: v_writelane_b32 v41, s39, 7
; GCN-NEXT: v_writelane_b32 v41, s40, 8
; GCN-NEXT: v_writelane_b32 v41, s41, 9
; GCN-NEXT: v_writelane_b32 v41, s42, 10
; GCN-NEXT: v_writelane_b32 v41, s43, 11
; GCN-NEXT: v_writelane_b32 v41, s44, 12
; GCN-NEXT: v_writelane_b32 v41, s45, 13
; GCN-NEXT: v_writelane_b32 v41, s46, 14
; GCN-NEXT: v_writelane_b32 v41, s47, 15
; GCN-NEXT: v_writelane_b32 v41, s48, 16
; GCN-NEXT: v_writelane_b32 v41, s49, 17
; GCN-NEXT: v_writelane_b32 v41, s50, 18
; GCN-NEXT: v_writelane_b32 v41, s51, 19
; GCN-NEXT: v_writelane_b32 v41, s52, 20
; GCN-NEXT: v_writelane_b32 v41, s53, 21
; GCN-NEXT: v_writelane_b32 v41, s54, 22
; GCN-NEXT: v_writelane_b32 v41, s55, 23
; GCN-NEXT: v_writelane_b32 v41, s56, 24
; GCN-NEXT: v_writelane_b32 v41, s57, 25
; GCN-NEXT: v_writelane_b32 v41, s58, 26
; GCN-NEXT: v_writelane_b32 v41, s59, 27
; GCN-NEXT: v_writelane_b32 v41, s60, 28
; GCN-NEXT: v_writelane_b32 v41, s61, 29
; GCN-NEXT: v_writelane_b32 v41, s62, 30
; GCN-NEXT: v_writelane_b32 v41, s63, 31
; GCN-NEXT: v_mov_b32_e32 v40, v0
; GCN-NEXT: s_mov_b64 s[4:5], exec
; GCN-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s6, v1
; GCN-NEXT: v_readfirstlane_b32 s7, v2
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[6:7], v[1:2]
; GCN-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GCN-NEXT: v_mov_b32_e32 v0, v40
; GCN-NEXT: s_swappc_b64 s[30:31], s[6:7]
; GCN-NEXT: ; implicit-def: $vgpr1_vgpr2
; GCN-NEXT: s_xor_b64 exec, exec, s[8:9]
; GCN-NEXT: s_cbranch_execnz .LBB7_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: v_mov_b32_e32 v0, v40
; GCN-NEXT: v_readlane_b32 s63, v41, 31
; GCN-NEXT: v_readlane_b32 s62, v41, 30
; GCN-NEXT: v_readlane_b32 s61, v41, 29
; GCN-NEXT: v_readlane_b32 s60, v41, 28
; GCN-NEXT: v_readlane_b32 s59, v41, 27
; GCN-NEXT: v_readlane_b32 s58, v41, 26
; GCN-NEXT: v_readlane_b32 s57, v41, 25
; GCN-NEXT: v_readlane_b32 s56, v41, 24
; GCN-NEXT: v_readlane_b32 s55, v41, 23
; GCN-NEXT: v_readlane_b32 s54, v41, 22
; GCN-NEXT: v_readlane_b32 s53, v41, 21
; GCN-NEXT: v_readlane_b32 s52, v41, 20
; GCN-NEXT: v_readlane_b32 s51, v41, 19
; GCN-NEXT: v_readlane_b32 s50, v41, 18
; GCN-NEXT: v_readlane_b32 s49, v41, 17
; GCN-NEXT: v_readlane_b32 s48, v41, 16
; GCN-NEXT: v_readlane_b32 s47, v41, 15
; GCN-NEXT: v_readlane_b32 s46, v41, 14
; GCN-NEXT: v_readlane_b32 s45, v41, 13
; GCN-NEXT: v_readlane_b32 s44, v41, 12
; GCN-NEXT: v_readlane_b32 s43, v41, 11
; GCN-NEXT: v_readlane_b32 s42, v41, 10
; GCN-NEXT: v_readlane_b32 s41, v41, 9
; GCN-NEXT: v_readlane_b32 s40, v41, 8
; GCN-NEXT: v_readlane_b32 s39, v41, 7
; GCN-NEXT: v_readlane_b32 s38, v41, 6
; GCN-NEXT: v_readlane_b32 s37, v41, 5
; GCN-NEXT: v_readlane_b32 s36, v41, 4
; GCN-NEXT: v_readlane_b32 s35, v41, 3
; GCN-NEXT: v_readlane_b32 s34, v41, 2
; GCN-NEXT: v_readlane_b32 s31, v41, 1
; GCN-NEXT: v_readlane_b32 s30, v41, 0
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: s_or_saveexec_b64 s[4:5], -1
; GCN-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: s_mov_b32 s33, s10
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr_arg_and_reuse:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s10, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[4:5], -1
; GISEL-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: v_writelane_b32 v41, s30, 0
; GISEL-NEXT: v_writelane_b32 v41, s31, 1
; GISEL-NEXT: v_writelane_b32 v41, s34, 2
; GISEL-NEXT: v_writelane_b32 v41, s35, 3
; GISEL-NEXT: v_writelane_b32 v41, s36, 4
; GISEL-NEXT: v_writelane_b32 v41, s37, 5
; GISEL-NEXT: v_writelane_b32 v41, s38, 6
; GISEL-NEXT: v_writelane_b32 v41, s39, 7
; GISEL-NEXT: v_writelane_b32 v41, s40, 8
; GISEL-NEXT: v_writelane_b32 v41, s41, 9
; GISEL-NEXT: v_writelane_b32 v41, s42, 10
; GISEL-NEXT: v_writelane_b32 v41, s43, 11
; GISEL-NEXT: v_writelane_b32 v41, s44, 12
; GISEL-NEXT: v_writelane_b32 v41, s45, 13
; GISEL-NEXT: v_writelane_b32 v41, s46, 14
; GISEL-NEXT: v_writelane_b32 v41, s47, 15
; GISEL-NEXT: v_writelane_b32 v41, s48, 16
; GISEL-NEXT: v_writelane_b32 v41, s49, 17
; GISEL-NEXT: v_writelane_b32 v41, s50, 18
; GISEL-NEXT: v_writelane_b32 v41, s51, 19
; GISEL-NEXT: v_writelane_b32 v41, s52, 20
; GISEL-NEXT: v_writelane_b32 v41, s53, 21
; GISEL-NEXT: v_writelane_b32 v41, s54, 22
; GISEL-NEXT: v_writelane_b32 v41, s55, 23
; GISEL-NEXT: v_writelane_b32 v41, s56, 24
; GISEL-NEXT: v_writelane_b32 v41, s57, 25
; GISEL-NEXT: v_writelane_b32 v41, s58, 26
; GISEL-NEXT: v_writelane_b32 v41, s59, 27
; GISEL-NEXT: v_writelane_b32 v41, s60, 28
; GISEL-NEXT: v_writelane_b32 v41, s61, 29
; GISEL-NEXT: v_writelane_b32 v41, s62, 30
; GISEL-NEXT: v_writelane_b32 v41, s63, 31
; GISEL-NEXT: v_mov_b32_e32 v40, v0
; GISEL-NEXT: s_mov_b64 s[4:5], exec
; GISEL-NEXT: .LBB7_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s6, v1
; GISEL-NEXT: v_readfirstlane_b32 s7, v2
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[6:7], v[1:2]
; GISEL-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GISEL-NEXT: v_mov_b32_e32 v0, v40
; GISEL-NEXT: s_swappc_b64 s[30:31], s[6:7]
; GISEL-NEXT: ; implicit-def: $vgpr1
; GISEL-NEXT: s_xor_b64 exec, exec, s[8:9]
; GISEL-NEXT: s_cbranch_execnz .LBB7_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: v_mov_b32_e32 v0, v40
; GISEL-NEXT: v_readlane_b32 s63, v41, 31
; GISEL-NEXT: v_readlane_b32 s62, v41, 30
; GISEL-NEXT: v_readlane_b32 s61, v41, 29
; GISEL-NEXT: v_readlane_b32 s60, v41, 28
; GISEL-NEXT: v_readlane_b32 s59, v41, 27
; GISEL-NEXT: v_readlane_b32 s58, v41, 26
; GISEL-NEXT: v_readlane_b32 s57, v41, 25
; GISEL-NEXT: v_readlane_b32 s56, v41, 24
; GISEL-NEXT: v_readlane_b32 s55, v41, 23
; GISEL-NEXT: v_readlane_b32 s54, v41, 22
; GISEL-NEXT: v_readlane_b32 s53, v41, 21
; GISEL-NEXT: v_readlane_b32 s52, v41, 20
; GISEL-NEXT: v_readlane_b32 s51, v41, 19
; GISEL-NEXT: v_readlane_b32 s50, v41, 18
; GISEL-NEXT: v_readlane_b32 s49, v41, 17
; GISEL-NEXT: v_readlane_b32 s48, v41, 16
; GISEL-NEXT: v_readlane_b32 s47, v41, 15
; GISEL-NEXT: v_readlane_b32 s46, v41, 14
; GISEL-NEXT: v_readlane_b32 s45, v41, 13
; GISEL-NEXT: v_readlane_b32 s44, v41, 12
; GISEL-NEXT: v_readlane_b32 s43, v41, 11
; GISEL-NEXT: v_readlane_b32 s42, v41, 10
; GISEL-NEXT: v_readlane_b32 s41, v41, 9
; GISEL-NEXT: v_readlane_b32 s40, v41, 8
; GISEL-NEXT: v_readlane_b32 s39, v41, 7
; GISEL-NEXT: v_readlane_b32 s38, v41, 6
; GISEL-NEXT: v_readlane_b32 s37, v41, 5
; GISEL-NEXT: v_readlane_b32 s36, v41, 4
; GISEL-NEXT: v_readlane_b32 s35, v41, 3
; GISEL-NEXT: v_readlane_b32 s34, v41, 2
; GISEL-NEXT: v_readlane_b32 s31, v41, 1
; GISEL-NEXT: v_readlane_b32 s30, v41, 0
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: s_or_saveexec_b64 s[4:5], -1
; GISEL-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: s_mov_b32 s33, s10
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
call amdgpu_gfx void %fptr(i32 %i)
ret i32 %i
}
; Use a variable inside a waterfall loop and use the return variable after the loop.
; TODO The argument and return variable could be in the same physical register, but the register
; allocator is not able to do that because the return value clashes with the liverange of an
; IMPLICIT_DEF of the argument.
define i32 @test_indirect_call_vgpr_ptr_arg_and_return(i32 %i, ptr %fptr) {
; GCN-LABEL: test_indirect_call_vgpr_ptr_arg_and_return:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s10, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[4:5], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: v_writelane_b32 v40, s50, 18
; GCN-NEXT: v_writelane_b32 v40, s51, 19
; GCN-NEXT: v_writelane_b32 v40, s52, 20
; GCN-NEXT: v_writelane_b32 v40, s53, 21
; GCN-NEXT: v_writelane_b32 v40, s54, 22
; GCN-NEXT: v_writelane_b32 v40, s55, 23
; GCN-NEXT: v_writelane_b32 v40, s56, 24
; GCN-NEXT: v_writelane_b32 v40, s57, 25
; GCN-NEXT: v_writelane_b32 v40, s58, 26
; GCN-NEXT: v_writelane_b32 v40, s59, 27
; GCN-NEXT: v_writelane_b32 v40, s60, 28
; GCN-NEXT: v_writelane_b32 v40, s61, 29
; GCN-NEXT: v_writelane_b32 v40, s62, 30
; GCN-NEXT: v_writelane_b32 v40, s63, 31
; GCN-NEXT: s_mov_b64 s[4:5], exec
; GCN-NEXT: .LBB8_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s8, v1
; GCN-NEXT: v_readfirstlane_b32 s9, v2
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[8:9], v[1:2]
; GCN-NEXT: s_and_saveexec_b64 s[6:7], vcc
; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9]
; GCN-NEXT: v_mov_b32_e32 v3, v0
; GCN-NEXT: ; implicit-def: $vgpr1_vgpr2
; GCN-NEXT: ; implicit-def: $vgpr0
; GCN-NEXT: s_xor_b64 exec, exec, s[6:7]
; GCN-NEXT: s_cbranch_execnz .LBB8_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: v_mov_b32_e32 v0, v3
; GCN-NEXT: v_readlane_b32 s63, v40, 31
; GCN-NEXT: v_readlane_b32 s62, v40, 30
; GCN-NEXT: v_readlane_b32 s61, v40, 29
; GCN-NEXT: v_readlane_b32 s60, v40, 28
; GCN-NEXT: v_readlane_b32 s59, v40, 27
; GCN-NEXT: v_readlane_b32 s58, v40, 26
; GCN-NEXT: v_readlane_b32 s57, v40, 25
; GCN-NEXT: v_readlane_b32 s56, v40, 24
; GCN-NEXT: v_readlane_b32 s55, v40, 23
; GCN-NEXT: v_readlane_b32 s54, v40, 22
; GCN-NEXT: v_readlane_b32 s53, v40, 21
; GCN-NEXT: v_readlane_b32 s52, v40, 20
; GCN-NEXT: v_readlane_b32 s51, v40, 19
; GCN-NEXT: v_readlane_b32 s50, v40, 18
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: s_or_saveexec_b64 s[4:5], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: s_mov_b32 s33, s10
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_call_vgpr_ptr_arg_and_return:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s10, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[4:5], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: v_writelane_b32 v40, s50, 18
; GISEL-NEXT: v_writelane_b32 v40, s51, 19
; GISEL-NEXT: v_writelane_b32 v40, s52, 20
; GISEL-NEXT: v_writelane_b32 v40, s53, 21
; GISEL-NEXT: v_writelane_b32 v40, s54, 22
; GISEL-NEXT: v_writelane_b32 v40, s55, 23
; GISEL-NEXT: v_writelane_b32 v40, s56, 24
; GISEL-NEXT: v_writelane_b32 v40, s57, 25
; GISEL-NEXT: v_writelane_b32 v40, s58, 26
; GISEL-NEXT: v_writelane_b32 v40, s59, 27
; GISEL-NEXT: v_writelane_b32 v40, s60, 28
; GISEL-NEXT: v_writelane_b32 v40, s61, 29
; GISEL-NEXT: v_writelane_b32 v40, s62, 30
; GISEL-NEXT: v_writelane_b32 v40, s63, 31
; GISEL-NEXT: s_mov_b64 s[4:5], exec
; GISEL-NEXT: .LBB8_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s8, v1
; GISEL-NEXT: v_readfirstlane_b32 s9, v2
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[8:9], v[1:2]
; GISEL-NEXT: s_and_saveexec_b64 s[6:7], vcc
; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9]
; GISEL-NEXT: v_mov_b32_e32 v2, v0
; GISEL-NEXT: ; implicit-def: $vgpr1
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: s_xor_b64 exec, exec, s[6:7]
; GISEL-NEXT: s_cbranch_execnz .LBB8_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: v_mov_b32_e32 v0, v2
; GISEL-NEXT: v_readlane_b32 s63, v40, 31
; GISEL-NEXT: v_readlane_b32 s62, v40, 30
; GISEL-NEXT: v_readlane_b32 s61, v40, 29
; GISEL-NEXT: v_readlane_b32 s60, v40, 28
; GISEL-NEXT: v_readlane_b32 s59, v40, 27
; GISEL-NEXT: v_readlane_b32 s58, v40, 26
; GISEL-NEXT: v_readlane_b32 s57, v40, 25
; GISEL-NEXT: v_readlane_b32 s56, v40, 24
; GISEL-NEXT: v_readlane_b32 s55, v40, 23
; GISEL-NEXT: v_readlane_b32 s54, v40, 22
; GISEL-NEXT: v_readlane_b32 s53, v40, 21
; GISEL-NEXT: v_readlane_b32 s52, v40, 20
; GISEL-NEXT: v_readlane_b32 s51, v40, 19
; GISEL-NEXT: v_readlane_b32 s50, v40, 18
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: s_or_saveexec_b64 s[4:5], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: s_mov_b32 s33, s10
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
%ret = call amdgpu_gfx i32 %fptr(i32 %i)
ret i32 %ret
}
; Calling a vgpr can never be a tail call.
define void @test_indirect_tail_call_vgpr_ptr(ptr %fptr) {
; GCN-LABEL: test_indirect_tail_call_vgpr_ptr:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-NEXT: s_mov_b32 s10, s33
; GCN-NEXT: s_mov_b32 s33, s32
; GCN-NEXT: s_or_saveexec_b64 s[4:5], -1
; GCN-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: s_addk_i32 s32, 0x400
; GCN-NEXT: v_writelane_b32 v40, s30, 0
; GCN-NEXT: v_writelane_b32 v40, s31, 1
; GCN-NEXT: v_writelane_b32 v40, s34, 2
; GCN-NEXT: v_writelane_b32 v40, s35, 3
; GCN-NEXT: v_writelane_b32 v40, s36, 4
; GCN-NEXT: v_writelane_b32 v40, s37, 5
; GCN-NEXT: v_writelane_b32 v40, s38, 6
; GCN-NEXT: v_writelane_b32 v40, s39, 7
; GCN-NEXT: v_writelane_b32 v40, s40, 8
; GCN-NEXT: v_writelane_b32 v40, s41, 9
; GCN-NEXT: v_writelane_b32 v40, s42, 10
; GCN-NEXT: v_writelane_b32 v40, s43, 11
; GCN-NEXT: v_writelane_b32 v40, s44, 12
; GCN-NEXT: v_writelane_b32 v40, s45, 13
; GCN-NEXT: v_writelane_b32 v40, s46, 14
; GCN-NEXT: v_writelane_b32 v40, s47, 15
; GCN-NEXT: v_writelane_b32 v40, s48, 16
; GCN-NEXT: v_writelane_b32 v40, s49, 17
; GCN-NEXT: v_writelane_b32 v40, s50, 18
; GCN-NEXT: v_writelane_b32 v40, s51, 19
; GCN-NEXT: v_writelane_b32 v40, s52, 20
; GCN-NEXT: v_writelane_b32 v40, s53, 21
; GCN-NEXT: v_writelane_b32 v40, s54, 22
; GCN-NEXT: v_writelane_b32 v40, s55, 23
; GCN-NEXT: v_writelane_b32 v40, s56, 24
; GCN-NEXT: v_writelane_b32 v40, s57, 25
; GCN-NEXT: v_writelane_b32 v40, s58, 26
; GCN-NEXT: v_writelane_b32 v40, s59, 27
; GCN-NEXT: v_writelane_b32 v40, s60, 28
; GCN-NEXT: v_writelane_b32 v40, s61, 29
; GCN-NEXT: v_writelane_b32 v40, s62, 30
; GCN-NEXT: v_writelane_b32 v40, s63, 31
; GCN-NEXT: s_mov_b64 s[4:5], exec
; GCN-NEXT: .LBB9_1: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: v_readfirstlane_b32 s6, v0
; GCN-NEXT: v_readfirstlane_b32 s7, v1
; GCN-NEXT: v_cmp_eq_u64_e32 vcc, s[6:7], v[0:1]
; GCN-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GCN-NEXT: s_swappc_b64 s[30:31], s[6:7]
; GCN-NEXT: ; implicit-def: $vgpr0_vgpr1
; GCN-NEXT: s_xor_b64 exec, exec, s[8:9]
; GCN-NEXT: s_cbranch_execnz .LBB9_1
; GCN-NEXT: ; %bb.2:
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: v_readlane_b32 s63, v40, 31
; GCN-NEXT: v_readlane_b32 s62, v40, 30
; GCN-NEXT: v_readlane_b32 s61, v40, 29
; GCN-NEXT: v_readlane_b32 s60, v40, 28
; GCN-NEXT: v_readlane_b32 s59, v40, 27
; GCN-NEXT: v_readlane_b32 s58, v40, 26
; GCN-NEXT: v_readlane_b32 s57, v40, 25
; GCN-NEXT: v_readlane_b32 s56, v40, 24
; GCN-NEXT: v_readlane_b32 s55, v40, 23
; GCN-NEXT: v_readlane_b32 s54, v40, 22
; GCN-NEXT: v_readlane_b32 s53, v40, 21
; GCN-NEXT: v_readlane_b32 s52, v40, 20
; GCN-NEXT: v_readlane_b32 s51, v40, 19
; GCN-NEXT: v_readlane_b32 s50, v40, 18
; GCN-NEXT: v_readlane_b32 s49, v40, 17
; GCN-NEXT: v_readlane_b32 s48, v40, 16
; GCN-NEXT: v_readlane_b32 s47, v40, 15
; GCN-NEXT: v_readlane_b32 s46, v40, 14
; GCN-NEXT: v_readlane_b32 s45, v40, 13
; GCN-NEXT: v_readlane_b32 s44, v40, 12
; GCN-NEXT: v_readlane_b32 s43, v40, 11
; GCN-NEXT: v_readlane_b32 s42, v40, 10
; GCN-NEXT: v_readlane_b32 s41, v40, 9
; GCN-NEXT: v_readlane_b32 s40, v40, 8
; GCN-NEXT: v_readlane_b32 s39, v40, 7
; GCN-NEXT: v_readlane_b32 s38, v40, 6
; GCN-NEXT: v_readlane_b32 s37, v40, 5
; GCN-NEXT: v_readlane_b32 s36, v40, 4
; GCN-NEXT: v_readlane_b32 s35, v40, 3
; GCN-NEXT: v_readlane_b32 s34, v40, 2
; GCN-NEXT: v_readlane_b32 s31, v40, 1
; GCN-NEXT: v_readlane_b32 s30, v40, 0
; GCN-NEXT: s_mov_b32 s32, s33
; GCN-NEXT: s_or_saveexec_b64 s[4:5], -1
; GCN-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GCN-NEXT: s_mov_b64 exec, s[4:5]
; GCN-NEXT: s_mov_b32 s33, s10
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-LABEL: test_indirect_tail_call_vgpr_ptr:
; GISEL: ; %bb.0:
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GISEL-NEXT: s_mov_b32 s10, s33
; GISEL-NEXT: s_mov_b32 s33, s32
; GISEL-NEXT: s_or_saveexec_b64 s[4:5], -1
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: s_addk_i32 s32, 0x400
; GISEL-NEXT: v_writelane_b32 v40, s30, 0
; GISEL-NEXT: v_writelane_b32 v40, s31, 1
; GISEL-NEXT: v_writelane_b32 v40, s34, 2
; GISEL-NEXT: v_writelane_b32 v40, s35, 3
; GISEL-NEXT: v_writelane_b32 v40, s36, 4
; GISEL-NEXT: v_writelane_b32 v40, s37, 5
; GISEL-NEXT: v_writelane_b32 v40, s38, 6
; GISEL-NEXT: v_writelane_b32 v40, s39, 7
; GISEL-NEXT: v_writelane_b32 v40, s40, 8
; GISEL-NEXT: v_writelane_b32 v40, s41, 9
; GISEL-NEXT: v_writelane_b32 v40, s42, 10
; GISEL-NEXT: v_writelane_b32 v40, s43, 11
; GISEL-NEXT: v_writelane_b32 v40, s44, 12
; GISEL-NEXT: v_writelane_b32 v40, s45, 13
; GISEL-NEXT: v_writelane_b32 v40, s46, 14
; GISEL-NEXT: v_writelane_b32 v40, s47, 15
; GISEL-NEXT: v_writelane_b32 v40, s48, 16
; GISEL-NEXT: v_writelane_b32 v40, s49, 17
; GISEL-NEXT: v_writelane_b32 v40, s50, 18
; GISEL-NEXT: v_writelane_b32 v40, s51, 19
; GISEL-NEXT: v_writelane_b32 v40, s52, 20
; GISEL-NEXT: v_writelane_b32 v40, s53, 21
; GISEL-NEXT: v_writelane_b32 v40, s54, 22
; GISEL-NEXT: v_writelane_b32 v40, s55, 23
; GISEL-NEXT: v_writelane_b32 v40, s56, 24
; GISEL-NEXT: v_writelane_b32 v40, s57, 25
; GISEL-NEXT: v_writelane_b32 v40, s58, 26
; GISEL-NEXT: v_writelane_b32 v40, s59, 27
; GISEL-NEXT: v_writelane_b32 v40, s60, 28
; GISEL-NEXT: v_writelane_b32 v40, s61, 29
; GISEL-NEXT: v_writelane_b32 v40, s62, 30
; GISEL-NEXT: v_writelane_b32 v40, s63, 31
; GISEL-NEXT: s_mov_b64 s[4:5], exec
; GISEL-NEXT: .LBB9_1: ; =>This Inner Loop Header: Depth=1
; GISEL-NEXT: v_readfirstlane_b32 s6, v0
; GISEL-NEXT: v_readfirstlane_b32 s7, v1
; GISEL-NEXT: v_cmp_eq_u64_e32 vcc, s[6:7], v[0:1]
; GISEL-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GISEL-NEXT: s_swappc_b64 s[30:31], s[6:7]
; GISEL-NEXT: ; implicit-def: $vgpr0
; GISEL-NEXT: s_xor_b64 exec, exec, s[8:9]
; GISEL-NEXT: s_cbranch_execnz .LBB9_1
; GISEL-NEXT: ; %bb.2:
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: v_readlane_b32 s63, v40, 31
; GISEL-NEXT: v_readlane_b32 s62, v40, 30
; GISEL-NEXT: v_readlane_b32 s61, v40, 29
; GISEL-NEXT: v_readlane_b32 s60, v40, 28
; GISEL-NEXT: v_readlane_b32 s59, v40, 27
; GISEL-NEXT: v_readlane_b32 s58, v40, 26
; GISEL-NEXT: v_readlane_b32 s57, v40, 25
; GISEL-NEXT: v_readlane_b32 s56, v40, 24
; GISEL-NEXT: v_readlane_b32 s55, v40, 23
; GISEL-NEXT: v_readlane_b32 s54, v40, 22
; GISEL-NEXT: v_readlane_b32 s53, v40, 21
; GISEL-NEXT: v_readlane_b32 s52, v40, 20
; GISEL-NEXT: v_readlane_b32 s51, v40, 19
; GISEL-NEXT: v_readlane_b32 s50, v40, 18
; GISEL-NEXT: v_readlane_b32 s49, v40, 17
; GISEL-NEXT: v_readlane_b32 s48, v40, 16
; GISEL-NEXT: v_readlane_b32 s47, v40, 15
; GISEL-NEXT: v_readlane_b32 s46, v40, 14
; GISEL-NEXT: v_readlane_b32 s45, v40, 13
; GISEL-NEXT: v_readlane_b32 s44, v40, 12
; GISEL-NEXT: v_readlane_b32 s43, v40, 11
; GISEL-NEXT: v_readlane_b32 s42, v40, 10
; GISEL-NEXT: v_readlane_b32 s41, v40, 9
; GISEL-NEXT: v_readlane_b32 s40, v40, 8
; GISEL-NEXT: v_readlane_b32 s39, v40, 7
; GISEL-NEXT: v_readlane_b32 s38, v40, 6
; GISEL-NEXT: v_readlane_b32 s37, v40, 5
; GISEL-NEXT: v_readlane_b32 s36, v40, 4
; GISEL-NEXT: v_readlane_b32 s35, v40, 3
; GISEL-NEXT: v_readlane_b32 s34, v40, 2
; GISEL-NEXT: v_readlane_b32 s31, v40, 1
; GISEL-NEXT: v_readlane_b32 s30, v40, 0
; GISEL-NEXT: s_mov_b32 s32, s33
; GISEL-NEXT: s_or_saveexec_b64 s[4:5], -1
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GISEL-NEXT: s_mov_b64 exec, s[4:5]
; GISEL-NEXT: s_mov_b32 s33, s10
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_setpc_b64 s[30:31]
tail call amdgpu_gfx void %fptr()
ret void
}
!llvm.module.flags = !{!0}
!0 = !{i32 1, !"amdhsa_code_object_version", i32 400}
Reference in New Issue View Git Blame Copy Permalink