For Mips 16, add the optimization where the 16 bit form of addiu sp can be used

if the offset fits in 11 bits. This makes use of the fact that the abi
requires sp to be 8 byte aligned so the actual offset can fit in 8
bits. It will be shifted left and sign extended before being actually used.
The assembler or direct object emitter will shift right the 11 bit
signed field by 3 bits. We don't need to deal with that here.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@175073 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Reed Kotler 2013-02-13 20:28:27 +00:00
parent f098620095
commit 6b9d461780
4 changed files with 77 additions and 3 deletions

View File

@ -184,7 +184,7 @@ void Mips16InstrInfo::makeFrame(unsigned SP, int64_t FrameSize,
int64_t Remainder = FrameSize - Base;
BuildMI(MBB, I, DL, get(Mips::SaveRaF16)). addImm(Base);
if (isInt<16>(-Remainder))
BuildMI(MBB, I, DL, get(Mips::AddiuSpImmX16)). addImm(-Remainder);
BuildAddiuSpImm(MBB, I, DL, -Remainder);
else
adjustStackPtrBig(SP, -Remainder, MBB, I, Mips::V0, Mips::V1);
}
@ -225,7 +225,7 @@ void Mips16InstrInfo::restoreFrame(unsigned SP, int64_t FrameSize,
// returns largest possible n bit unsigned integer
int64_t Remainder = FrameSize - Base;
if (isInt<16>(Remainder))
BuildMI(MBB, I, DL, get(Mips::AddiuSpImmX16)). addImm(Remainder);
BuildAddiuSpImm(MBB, I, DL, Remainder);
else
adjustStackPtrBig(SP, Remainder, MBB, I, Mips::A0, Mips::A1);
BuildMI(MBB, I, DL, get(Mips::RestoreRaF16)). addImm(Base);
@ -299,7 +299,7 @@ void Mips16InstrInfo::adjustStackPtr(unsigned SP, int64_t Amount,
MachineBasicBlock::iterator I) const {
DebugLoc DL = I != MBB.end() ? I->getDebugLoc() : DebugLoc();
if (isInt<16>(Amount)) // need to change to addiu sp, ....and isInt<16>
BuildMI(MBB, I, DL, get(Mips::AddiuSpImmX16)). addImm(Amount);
BuildAddiuSpImm(MBB, I, DL, Amount);
else
adjustStackPtrBigUnrestricted(SP, Amount, MBB, I);
}
@ -400,6 +400,15 @@ void Mips16InstrInfo::ExpandRetRA16(MachineBasicBlock &MBB,
BuildMI(MBB, I, I->getDebugLoc(), get(Opc));
}
void Mips16InstrInfo::BuildAddiuSpImm(
MachineBasicBlock &MBB,
MachineBasicBlock::iterator II, DebugLoc DL, int64_t Imm) const {
if (validSpImm8(Imm))
BuildMI(MBB, II, DL, get(Mips::AddiuSpImm16)).addImm(Imm);
else
BuildMI(MBB, II, DL, get(Mips::AddiuSpImmX16)).addImm(Imm);
}
const MipsInstrInfo *llvm::createMips16InstrInfo(MipsTargetMachine &TM) {
return new Mips16InstrInfo(TM);
}

View File

@ -86,6 +86,18 @@ public:
MachineBasicBlock::iterator II, DebugLoc DL,
unsigned &NewImm) const;
static bool validSpImm8(int offset) {
return ((offset & 7) == 0) && isInt<11>(offset);
}
//
// build the proper one based on the Imm field
//
void BuildAddiuSpImm(MachineBasicBlock &MBB,
MachineBasicBlock::iterator II, DebugLoc DL,
int64_t Imm) const;
private:
virtual unsigned GetAnalyzableBrOpc(unsigned Opc) const;

View File

@ -31,6 +31,21 @@ def mem16_ea : Operand<i32> {
let EncoderMethod = "getMemEncoding";
}
//
//
// I8 instruction format
//
class FI816_ins_base<bits<3> _func, string asmstr,
string asmstr2, InstrItinClass itin>:
FI816<_func, (outs), (ins simm16:$imm), !strconcat(asmstr, asmstr2),
[], itin>;
class FI816_SP_ins<bits<3> _func, string asmstr,
InstrItinClass itin>:
FI816_ins_base<_func, asmstr, "\t$$sp, $imm # 16 bit inst", itin>;
//
// RI instruction format
//
@ -451,6 +466,13 @@ def AddiuRxPcImmX16: FEXT_RI16_PC_ins<0b00001, "addiu", IIAlu>;
// Purpose: Add Immediate Unsigned Word (2-Operand, SP-Relative, Extended)
// To add a constant to the stack pointer.
//
def AddiuSpImm16
: FI816_SP_ins<0b011, "addiu", IIAlu> {
let Defs = [SP];
let Uses = [SP];
let AddedComplexity = 5;
}
def AddiuSpImmX16
: FEXT_I816_SP_ins<0b011, "addiu", IIAlu> {
let Defs = [SP];

View File

@ -0,0 +1,31 @@
; RUN: llc -march=mipsel -mcpu=mips16 -relocation-model=static < %s | FileCheck %s -check-prefix=16
@i = global i32 25, align 4
@.str = private unnamed_addr constant [5 x i8] c"%i \0A\00", align 1
define void @p(i32* %i) nounwind {
entry:
ret void
}
define void @foo() nounwind {
entry:
%y = alloca [512 x i32], align 4
%x = alloca i32, align 8
%zz = alloca i32, align 4
%z = alloca i32, align 4
%0 = load i32* @i, align 4
%arrayidx = getelementptr inbounds [512 x i32]* %y, i32 0, i32 10
store i32 %0, i32* %arrayidx, align 4
%1 = load i32* @i, align 4
store i32 %1, i32* %x, align 8
call void @p(i32* %x)
%arrayidx1 = getelementptr inbounds [512 x i32]* %y, i32 0, i32 10
call void @p(i32* %arrayidx1)
ret void
}
; 16: save $ra, $s0, $s1, 2040
; 16: addiu $sp, -48 # 16 bit inst
; 16: addiu $sp, 48 # 16 bit inst
; 16: restore $ra, $s0, $s1, 2040