04fff547e2
Currently the return address ABI registers s[30:31], which fall in the call clobbered register range, are added as a live-in on the function entry to preserve its value when we have calls so that it gets saved and restored around the calls. But the DWARF unwind information (CFI) needs to track where the return address resides in a frame and the above approach makes it difficult to track the return address when the CFI information is emitted during the frame lowering, due to the involvment of understanding the control flow. This patch moves the return address ABI registers s[30:31] into callee saved registers range and stops adding live-in for return address registers, so that the CFI machinery will know where the return address resides when CSR save/restore happen during the frame lowering. And doing the above poses an issue that now the return instruction uses undefined register `sgpr30_sgpr31`. This is resolved by hiding the return address register use by the return instruction through the `SI_RETURN` pseudo instruction, which doesn't take any input operands, until the `SI_RETURN` pseudo gets lowered to the `S_SETPC_B64_return` during the `expandPostRAPseudo()`. As an added benefit, this patch simplifies overall return instruction handling. Note: The AMDGPU CFI changes are there only in the downstream code and another version of this patch will be posted for review for the downstream code. Reviewed By: arsenm, ronlieb Differential Revision: https://reviews.llvm.org/D114652
170 lines
7.4 KiB
LLVM
170 lines
7.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=SDAG -enable-var-scope %s
|
|
; RUN: llc -global-isel -mtriple=amdgcn--amdpal -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GISEL -enable-var-scope %s
|
|
|
|
declare void @extern_c_func()
|
|
|
|
define amdgpu_gfx void @gfx_func() {
|
|
; SDAG-LABEL: gfx_func:
|
|
; SDAG: ; %bb.0:
|
|
; SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; SDAG-NEXT: s_or_saveexec_b64 s[34:35], -1
|
|
; SDAG-NEXT: buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; SDAG-NEXT: s_mov_b64 exec, s[34:35]
|
|
; SDAG-NEXT: v_writelane_b32 v40, s33, 28
|
|
; SDAG-NEXT: v_writelane_b32 v40, s4, 0
|
|
; SDAG-NEXT: v_writelane_b32 v40, s5, 1
|
|
; SDAG-NEXT: v_writelane_b32 v40, s6, 2
|
|
; SDAG-NEXT: v_writelane_b32 v40, s7, 3
|
|
; SDAG-NEXT: v_writelane_b32 v40, s8, 4
|
|
; SDAG-NEXT: v_writelane_b32 v40, s9, 5
|
|
; SDAG-NEXT: v_writelane_b32 v40, s10, 6
|
|
; SDAG-NEXT: v_writelane_b32 v40, s11, 7
|
|
; SDAG-NEXT: v_writelane_b32 v40, s12, 8
|
|
; SDAG-NEXT: v_writelane_b32 v40, s13, 9
|
|
; SDAG-NEXT: v_writelane_b32 v40, s14, 10
|
|
; SDAG-NEXT: v_writelane_b32 v40, s15, 11
|
|
; SDAG-NEXT: v_writelane_b32 v40, s16, 12
|
|
; SDAG-NEXT: v_writelane_b32 v40, s17, 13
|
|
; SDAG-NEXT: v_writelane_b32 v40, s18, 14
|
|
; SDAG-NEXT: v_writelane_b32 v40, s19, 15
|
|
; SDAG-NEXT: v_writelane_b32 v40, s20, 16
|
|
; SDAG-NEXT: v_writelane_b32 v40, s21, 17
|
|
; SDAG-NEXT: v_writelane_b32 v40, s22, 18
|
|
; SDAG-NEXT: v_writelane_b32 v40, s23, 19
|
|
; SDAG-NEXT: s_mov_b32 s33, s32
|
|
; SDAG-NEXT: s_addk_i32 s32, 0x400
|
|
; SDAG-NEXT: v_writelane_b32 v40, s24, 20
|
|
; SDAG-NEXT: v_writelane_b32 v40, s25, 21
|
|
; SDAG-NEXT: s_getpc_b64 s[34:35]
|
|
; SDAG-NEXT: s_add_u32 s34, s34, extern_c_func@gotpcrel32@lo+4
|
|
; SDAG-NEXT: s_addc_u32 s35, s35, extern_c_func@gotpcrel32@hi+12
|
|
; SDAG-NEXT: v_writelane_b32 v40, s26, 22
|
|
; SDAG-NEXT: s_load_dwordx2 s[34:35], s[34:35], 0x0
|
|
; SDAG-NEXT: v_writelane_b32 v40, s27, 23
|
|
; SDAG-NEXT: v_writelane_b32 v40, s28, 24
|
|
; SDAG-NEXT: v_writelane_b32 v40, s29, 25
|
|
; SDAG-NEXT: v_writelane_b32 v40, s30, 26
|
|
; SDAG-NEXT: s_mov_b64 s[8:9], 0
|
|
; SDAG-NEXT: v_writelane_b32 v40, s31, 27
|
|
; SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; SDAG-NEXT: s_swappc_b64 s[30:31], s[34:35]
|
|
; SDAG-NEXT: v_readlane_b32 s31, v40, 27
|
|
; SDAG-NEXT: v_readlane_b32 s30, v40, 26
|
|
; SDAG-NEXT: v_readlane_b32 s29, v40, 25
|
|
; SDAG-NEXT: v_readlane_b32 s28, v40, 24
|
|
; SDAG-NEXT: v_readlane_b32 s27, v40, 23
|
|
; SDAG-NEXT: v_readlane_b32 s26, v40, 22
|
|
; SDAG-NEXT: v_readlane_b32 s25, v40, 21
|
|
; SDAG-NEXT: v_readlane_b32 s24, v40, 20
|
|
; SDAG-NEXT: v_readlane_b32 s23, v40, 19
|
|
; SDAG-NEXT: v_readlane_b32 s22, v40, 18
|
|
; SDAG-NEXT: v_readlane_b32 s21, v40, 17
|
|
; SDAG-NEXT: v_readlane_b32 s20, v40, 16
|
|
; SDAG-NEXT: v_readlane_b32 s19, v40, 15
|
|
; SDAG-NEXT: v_readlane_b32 s18, v40, 14
|
|
; SDAG-NEXT: v_readlane_b32 s17, v40, 13
|
|
; SDAG-NEXT: v_readlane_b32 s16, v40, 12
|
|
; SDAG-NEXT: v_readlane_b32 s15, v40, 11
|
|
; SDAG-NEXT: v_readlane_b32 s14, v40, 10
|
|
; SDAG-NEXT: v_readlane_b32 s13, v40, 9
|
|
; SDAG-NEXT: v_readlane_b32 s12, v40, 8
|
|
; SDAG-NEXT: v_readlane_b32 s11, v40, 7
|
|
; SDAG-NEXT: v_readlane_b32 s10, v40, 6
|
|
; SDAG-NEXT: v_readlane_b32 s9, v40, 5
|
|
; SDAG-NEXT: v_readlane_b32 s8, v40, 4
|
|
; SDAG-NEXT: v_readlane_b32 s7, v40, 3
|
|
; SDAG-NEXT: v_readlane_b32 s6, v40, 2
|
|
; SDAG-NEXT: v_readlane_b32 s5, v40, 1
|
|
; SDAG-NEXT: v_readlane_b32 s4, v40, 0
|
|
; SDAG-NEXT: s_addk_i32 s32, 0xfc00
|
|
; SDAG-NEXT: v_readlane_b32 s33, v40, 28
|
|
; SDAG-NEXT: s_or_saveexec_b64 s[34:35], -1
|
|
; SDAG-NEXT: buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; SDAG-NEXT: s_mov_b64 exec, s[34:35]
|
|
; SDAG-NEXT: s_waitcnt vmcnt(0)
|
|
; SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GISEL-LABEL: gfx_func:
|
|
; GISEL: ; %bb.0:
|
|
; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GISEL-NEXT: s_or_saveexec_b64 s[34:35], -1
|
|
; GISEL-NEXT: buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GISEL-NEXT: s_mov_b64 exec, s[34:35]
|
|
; GISEL-NEXT: v_writelane_b32 v40, s33, 28
|
|
; GISEL-NEXT: v_writelane_b32 v40, s4, 0
|
|
; GISEL-NEXT: v_writelane_b32 v40, s5, 1
|
|
; GISEL-NEXT: v_writelane_b32 v40, s6, 2
|
|
; GISEL-NEXT: v_writelane_b32 v40, s7, 3
|
|
; GISEL-NEXT: v_writelane_b32 v40, s8, 4
|
|
; GISEL-NEXT: v_writelane_b32 v40, s9, 5
|
|
; GISEL-NEXT: v_writelane_b32 v40, s10, 6
|
|
; GISEL-NEXT: v_writelane_b32 v40, s11, 7
|
|
; GISEL-NEXT: v_writelane_b32 v40, s12, 8
|
|
; GISEL-NEXT: v_writelane_b32 v40, s13, 9
|
|
; GISEL-NEXT: v_writelane_b32 v40, s14, 10
|
|
; GISEL-NEXT: v_writelane_b32 v40, s15, 11
|
|
; GISEL-NEXT: v_writelane_b32 v40, s16, 12
|
|
; GISEL-NEXT: v_writelane_b32 v40, s17, 13
|
|
; GISEL-NEXT: v_writelane_b32 v40, s18, 14
|
|
; GISEL-NEXT: v_writelane_b32 v40, s19, 15
|
|
; GISEL-NEXT: v_writelane_b32 v40, s20, 16
|
|
; GISEL-NEXT: v_writelane_b32 v40, s21, 17
|
|
; GISEL-NEXT: v_writelane_b32 v40, s22, 18
|
|
; GISEL-NEXT: v_writelane_b32 v40, s23, 19
|
|
; GISEL-NEXT: s_mov_b32 s33, s32
|
|
; GISEL-NEXT: s_addk_i32 s32, 0x400
|
|
; GISEL-NEXT: v_writelane_b32 v40, s24, 20
|
|
; GISEL-NEXT: v_writelane_b32 v40, s25, 21
|
|
; GISEL-NEXT: s_getpc_b64 s[34:35]
|
|
; GISEL-NEXT: s_add_u32 s34, s34, extern_c_func@gotpcrel32@lo+4
|
|
; GISEL-NEXT: s_addc_u32 s35, s35, extern_c_func@gotpcrel32@hi+12
|
|
; GISEL-NEXT: v_writelane_b32 v40, s26, 22
|
|
; GISEL-NEXT: s_load_dwordx2 s[34:35], s[34:35], 0x0
|
|
; GISEL-NEXT: v_writelane_b32 v40, s27, 23
|
|
; GISEL-NEXT: v_writelane_b32 v40, s28, 24
|
|
; GISEL-NEXT: v_writelane_b32 v40, s29, 25
|
|
; GISEL-NEXT: v_writelane_b32 v40, s30, 26
|
|
; GISEL-NEXT: s_mov_b64 s[8:9], s[4:5]
|
|
; GISEL-NEXT: v_writelane_b32 v40, s31, 27
|
|
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GISEL-NEXT: s_swappc_b64 s[30:31], s[34:35]
|
|
; GISEL-NEXT: v_readlane_b32 s31, v40, 27
|
|
; GISEL-NEXT: v_readlane_b32 s30, v40, 26
|
|
; GISEL-NEXT: v_readlane_b32 s29, v40, 25
|
|
; GISEL-NEXT: v_readlane_b32 s28, v40, 24
|
|
; GISEL-NEXT: v_readlane_b32 s27, v40, 23
|
|
; GISEL-NEXT: v_readlane_b32 s26, v40, 22
|
|
; GISEL-NEXT: v_readlane_b32 s25, v40, 21
|
|
; GISEL-NEXT: v_readlane_b32 s24, v40, 20
|
|
; GISEL-NEXT: v_readlane_b32 s23, v40, 19
|
|
; GISEL-NEXT: v_readlane_b32 s22, v40, 18
|
|
; GISEL-NEXT: v_readlane_b32 s21, v40, 17
|
|
; GISEL-NEXT: v_readlane_b32 s20, v40, 16
|
|
; GISEL-NEXT: v_readlane_b32 s19, v40, 15
|
|
; GISEL-NEXT: v_readlane_b32 s18, v40, 14
|
|
; GISEL-NEXT: v_readlane_b32 s17, v40, 13
|
|
; GISEL-NEXT: v_readlane_b32 s16, v40, 12
|
|
; GISEL-NEXT: v_readlane_b32 s15, v40, 11
|
|
; GISEL-NEXT: v_readlane_b32 s14, v40, 10
|
|
; GISEL-NEXT: v_readlane_b32 s13, v40, 9
|
|
; GISEL-NEXT: v_readlane_b32 s12, v40, 8
|
|
; GISEL-NEXT: v_readlane_b32 s11, v40, 7
|
|
; GISEL-NEXT: v_readlane_b32 s10, v40, 6
|
|
; GISEL-NEXT: v_readlane_b32 s9, v40, 5
|
|
; GISEL-NEXT: v_readlane_b32 s8, v40, 4
|
|
; GISEL-NEXT: v_readlane_b32 s7, v40, 3
|
|
; GISEL-NEXT: v_readlane_b32 s6, v40, 2
|
|
; GISEL-NEXT: v_readlane_b32 s5, v40, 1
|
|
; GISEL-NEXT: v_readlane_b32 s4, v40, 0
|
|
; GISEL-NEXT: s_addk_i32 s32, 0xfc00
|
|
; GISEL-NEXT: v_readlane_b32 s33, v40, 28
|
|
; GISEL-NEXT: s_or_saveexec_b64 s[34:35], -1
|
|
; GISEL-NEXT: buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GISEL-NEXT: s_mov_b64 exec, s[34:35]
|
|
; GISEL-NEXT: s_waitcnt vmcnt(0)
|
|
; GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
call void @extern_c_func()
|
|
ret void
|
|
}
|