caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

AMDGPU: Materialize frame index before add

Browse Source 
It isn't generally safe to fold the frame index
directly into the operand since it will possibly
not be an inline immediate after it is expanded.

This surprisingly seems to produce better code, since
the FI doesn't prevent folding other immediate operands.

llvm-svn: 288185
This commit is contained in:
Matt Arsenault
2016-11-29 19:20:48 +00:00
parent ff8bb49bf4
commit f96eeec005
2 changed files with 10 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
View File

@@ -245,12 +245,17 @@ void SIRegisterInfo::materializeFrameBaseRegister(MachineBasicBlock *MBB,
unsigned UnusedCarry = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
unsigned OffsetReg = MRI.createVirtualRegister(&AMDGPU::SReg_32_XM0RegClass);
unsigned FIReg = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
BuildMI(*MBB, Ins, DL, TII->get(AMDGPU::S_MOV_B32), OffsetReg)
.addImm(Offset);
BuildMI(*MBB, Ins, DL, TII->get(AMDGPU::V_MOV_B32_e32), FIReg)
.addFrameIndex(FrameIdx);
BuildMI(*MBB, Ins, DL, TII->get(AMDGPU::V_ADD_I32_e64), BaseReg)
.addReg(UnusedCarry, RegState::Define | RegState::Dead)
.addReg(OffsetReg, RegState::Kill)
.addFrameIndex(FrameIdx);
.addReg(FIReg);
}
void SIRegisterInfo::resolveFrameIndex(MachineInstr &MI, unsigned BaseReg,

7
llvm/test/CodeGen/AMDGPU/local-stack-slot-bug.ll
View File

@@ -7,15 +7,16 @@
;
; CHECK-LABEL: {{^}}main:
; CHECK: v_mov_b32_e32 [[BASE_FI:v[0-9]+]], 0{{$}}
; FIXME: add 0?
; CHECK-DAG: s_movk_i32 [[K0:s[0-9]+]], 0x140
; CHECK-DAG: v_add_i32_e64 [[ADD_K0:v[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, [[K0]], 0
; CHECK-DAG: v_add_i32_e32 [[ADD_K0:v[0-9]+]], vcc, 0x140, [[BASE_FI]]
; CHECK-DAG: v_lshlrev_b32_e32 [[BYTES:v[0-9]+]], 2, v0
; CHECK-DAG: buffer_store_dword {{v[0-9]+}}, [[ADD_K0]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}}
; CHECK-DAG: v_add_i32_e32 [[HI_OFF:v[0-9]+]], vcc, 0x200, [[BYTES]]
; CHECK-DAG: v_add_i32_e32 [[LO_OFF:v[0-9]+]], vcc, 0, [[BYTES]]
; CHECK-DAG: v_add_i32_e32 [[LO_OFF:v[0-9]+]], vcc, [[BASE_FI]], [[BYTES]]
; CHECK: buffer_load_dword {{v[0-9]+}}, [[LO_OFF]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen
; CHECK: buffer_load_dword {{v[0-9]+}}, [[HI_OFF]], {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen