6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-04-09 01:38:03 +00:00
Code Issues Projects Releases Wiki Activity

Shrink ADDS, ADC, RSB, and SUBS.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@78776 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Evan Cheng 2009-08-12 01:49:45 +00:00
parent 394346ba3a
commit 05c269c645
10 changed files with 164 additions and 49 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

132
lib/Target/ARM/Thumb2SizeReduction.cpp
View File

@ -45,29 +45,31 @@ namespace {
uint8_t Imm2Limit; // Limit of immediate field when it's two-address
unsigned LowRegs1 : 1; // Only possible if low-registers are used
unsigned LowRegs2 : 1; // Only possible if low-registers are used (2addr)
unsigned PredCC1 : 1; // 0 - If predicated, cc is on and vice versa.
unsigned PredCC1 : 2; // 0 - If predicated, cc is on and vice versa.
// 1 - No cc field.
// 2 - Always set CPSR.
unsigned PredCC2 : 1;
unsigned Special : 1; // Needs to be dealt with specially
};
static const ReduceEntry ReduceTable[] = {
// Wide, Narrow1, Narrow2, imm1,imm2, lo1, lo2, P/C, S
{ ARM::t2ADCrr, ARM::tADC, 0, 0, 0, 1, 0, 0,0, 0 },
// FIXME: t2ADDS variants.
{ ARM::t2ADCrr, 0, ARM::tADC, 0, 0, 0, 1, 0,0, 0 },
{ ARM::t2ADDri, ARM::tADDi3, ARM::tADDi8, 3, 8, 1, 1, 0,0, 0 },
{ ARM::t2ADDrr, ARM::tADDrr, ARM::tADDhirr, 0, 0, 1, 0, 0,1, 0 },
// Note: immediate scale is 4.
{ ARM::t2ADDrSPi,ARM::tADDrSPi,0, 8, 0, 1, 0, 1,0, 0 },
{ ARM::t2ADDSri,ARM::tADDi3, ARM::tADDi8, 3, 8, 1, 1, 2,2, 1 },
{ ARM::t2ADDSrr,ARM::tADDrr, 0, 0, 0, 1, 0, 2,0, 1 },
{ ARM::t2ANDrr, 0, ARM::tAND, 0, 0, 0, 1, 0,0, 0 },
{ ARM::t2ASRri, ARM::tASRri, 0, 5, 0, 1, 0, 0,0, 0 },
{ ARM::t2ASRrr, 0, ARM::tASRrr, 0, 0, 0, 1, 0,0, 0 },
{ ARM::t2BICrr, 0, ARM::tBIC, 0, 0, 0, 1, 0,0, 0 },
{ ARM::t2CMNrr, ARM::tCMN, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2CMPri, ARM::tCMPi8, 0, 8, 0, 1, 0, 1,0, 0 },
{ ARM::t2CMPrr, ARM::tCMPhir, 0, 0, 0, 0, 0, 1,0, 0 },
{ ARM::t2CMPzri,ARM::tCMPzi8, 0, 8, 0, 1, 0, 1,0, 0 },
{ ARM::t2CMPzrr,ARM::tCMPzhir,0, 0, 0, 0, 0, 1,0, 0 },
{ ARM::t2CMNrr, ARM::tCMN, 0, 0, 0, 1, 0, 2,0, 0 },
{ ARM::t2CMPri, ARM::tCMPi8, 0, 8, 0, 1, 0, 2,0, 0 },
{ ARM::t2CMPrr, ARM::tCMPhir, 0, 0, 0, 0, 0, 2,0, 0 },
{ ARM::t2CMPzri,ARM::tCMPzi8, 0, 8, 0, 1, 0, 2,0, 0 },
{ ARM::t2CMPzrr,ARM::tCMPzhir,0, 0, 0, 0, 0, 2,0, 0 },
{ ARM::t2EORrr, 0, ARM::tEOR, 0, 0, 0, 1, 0,0, 0 },
{ ARM::t2LSLri, ARM::tLSLri, 0, 5, 0, 1, 0, 0,0, 0 },
{ ARM::t2LSLrr, 0, ARM::tLSLrr, 0, 0, 0, 1, 0,0, 0 },
@ -84,13 +86,16 @@ namespace {
{ ARM::t2REV16, ARM::tREV16, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2REVSH, ARM::tREVSH, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2RORrr, 0, ARM::tROR, 0, 0, 0, 1, 0,0, 0 },
// FIXME: T2RSBri immediate must be zero. Also need entry for T2RSBS
//{ ARM::t2RSBri, ARM::tRSB, 0, 0, 0, 1, 0, 0,0, 0 },
{ ARM::t2RSBri, ARM::tRSB, 0, 0, 0, 1, 0, 0,0, 1 },
{ ARM::t2RSBSri,ARM::tRSB, 0, 0, 0, 1, 0, 2,0, 1 },
{ ARM::t2SBCrr, 0, ARM::tSBC, 0, 0, 0, 1, 0,0, 0 },
{ ARM::t2SUBri, ARM::tSUBi3, ARM::tSUBi8, 3, 8, 1, 1, 0,0, 0 },
{ ARM::t2SUBrr, ARM::tSUBrr, 0, 0, 0, 1, 0, 0,0, 0 },
{ ARM::t2SUBSri,ARM::tSUBi3, ARM::tSUBi8, 3, 8, 1, 1, 2,2, 0 },
{ ARM::t2SUBSrr,ARM::tSUBrr, 0, 0, 0, 1, 0, 2,0, 0 },
{ ARM::t2SXTBr, ARM::tSXTB, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2SXTHr, ARM::tSXTH, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2TSTrr, ARM::tTST, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2TSTrr, ARM::tTST, 0, 0, 0, 1, 0, 2,0, 0 },
{ ARM::t2UXTBr, ARM::tUXTB, 0, 0, 0, 1, 0, 1,0, 0 },
{ ARM::t2UXTHr, ARM::tUXTH, 0, 0, 0, 1, 0, 1,0, 0 },
@ -133,6 +138,10 @@ namespace {
/// ReduceOpcodeMap - Maps wide opcode to index of entry in ReduceTable.
DenseMap<unsigned, unsigned> ReduceOpcodeMap;
bool VerifyPredAndCC(MachineInstr *MI, const ReduceEntry &Entry,
bool is2Addr, ARMCC::CondCodes Pred,
bool LiveCPSR, bool &HasCC, bool &CCDead);
bool ReduceLoadStore(MachineBasicBlock &MBB, MachineInstr *MI,
const ReduceEntry &Entry);
@ -165,9 +174,17 @@ Thumb2SizeReduce::Thumb2SizeReduce() : MachineFunctionPass(&ID) {
}
}
static bool VerifyPredAndCC(MachineInstr *MI, const ReduceEntry &Entry,
bool is2Addr, ARMCC::CondCodes Pred,
bool LiveCPSR, bool &HasCC, bool &CCDead) {
static bool HasImplicitCPSRDef(const TargetInstrDesc &TID) {
for (const unsigned *Regs = TID.ImplicitDefs; *Regs; ++Regs)
if (*Regs == ARM::CPSR)
return true;
return false;
}
bool
Thumb2SizeReduce::VerifyPredAndCC(MachineInstr *MI, const ReduceEntry &Entry,
bool is2Addr, ARMCC::CondCodes Pred,
bool LiveCPSR, bool &HasCC, bool &CCDead) {
if ((is2Addr && Entry.PredCC2 == 0) ||
(!is2Addr && Entry.PredCC1 == 0)) {
if (Pred == ARMCC::AL) {
@ -188,6 +205,16 @@ static bool VerifyPredAndCC(MachineInstr *MI, const ReduceEntry &Entry,
if (HasCC)
return false;
}
} else if ((is2Addr && Entry.PredCC2 == 2) ||
(!is2Addr && Entry.PredCC1 == 2)) {
/// Old opcode has an optional def of CPSR.
if (HasCC)
return true;
// If both old opcode does not implicit CPSR def, then it's not ok since
// these new opcodes CPSR def is not meant to be thrown away. e.g. CMP.
if (!HasImplicitCPSRDef(MI->getDesc()))
return false;
HasCC = true;
} else {
// 16-bit instruction does not set CPSR.
if (HasCC)
@ -340,6 +367,33 @@ Thumb2SizeReduce::ReduceSpecial(MachineBasicBlock &MBB, MachineInstr *MI,
const TargetInstrDesc &TID = MI->getDesc();
if (TID.mayLoad() || TID.mayStore())
return ReduceLoadStore(MBB, MI, Entry);
unsigned Opc = MI->getOpcode();
switch (Opc) {
default: break;
case ARM::t2ADDSri:
case ARM::t2ADDSrr: {
unsigned PredReg = 0;
if (getInstrPredicate(MI, PredReg) == ARMCC::AL) {
switch (Opc) {
default: break;
case ARM::t2ADDSri: {
if (ReduceTo2Addr(MBB, MI, Entry, LiveCPSR))
return true;
// fallthrough
}
case ARM::t2ADDSrr:
return ReduceToNarrow(MBB, MI, Entry, LiveCPSR);
}
}
break;
}
case ARM::t2RSBri:
case ARM::t2RSBSri:
if (MI->getOperand(2).getImm() == 0)
return ReduceToNarrow(MBB, MI, Entry, LiveCPSR);
break;
}
return false;
}
@ -395,9 +449,9 @@ Thumb2SizeReduce::ReduceTo2Addr(MachineBasicBlock &MBB, MachineInstr *MI,
// Add the 16-bit instruction.
DebugLoc dl = MI->getDebugLoc();
MachineInstrBuilder MIB = BuildMI(MBB, *MI, dl, TII->get(Entry.NarrowOpc2));
MachineInstrBuilder MIB = BuildMI(MBB, *MI, dl, NewTID);
MIB.addOperand(MI->getOperand(0));
if (HasCC)
if (HasCC && NewTID.hasOptionalDef())
AddDefaultT1CC(MIB, CCDead);
// Transfer the rest of operands.
@ -475,9 +529,9 @@ Thumb2SizeReduce::ReduceToNarrow(MachineBasicBlock &MBB, MachineInstr *MI,
// Add the 16-bit instruction.
DebugLoc dl = MI->getDebugLoc();
MachineInstrBuilder MIB = BuildMI(MBB, *MI, dl, TII->get(Entry.NarrowOpc1));
MachineInstrBuilder MIB = BuildMI(MBB, *MI, dl, NewTID);
MIB.addOperand(MI->getOperand(0));
if (HasCC)
if (HasCC && NewTID.hasOptionalDef())
AddDefaultT1CC(MIB, CCDead);
// Transfer the rest of operands.
@ -485,16 +539,26 @@ Thumb2SizeReduce::ReduceToNarrow(MachineBasicBlock &MBB, MachineInstr *MI,
for (unsigned i = 1, e = MI->getNumOperands(); i != e; ++i) {
if (i < NumOps && TID.OpInfo[i].isOptionalDef())
continue;
if ((TID.getOpcode() == ARM::t2RSBSri ||
TID.getOpcode() == ARM::t2RSBri) && i == 2)
// Skip the zero immediate operand, it's now implicit.
continue;
bool isPred = (i < NumOps && TID.OpInfo[i].isPredicate());
if (SkipPred && isPred)
continue;
const MachineOperand &MO = MI->getOperand(i);
if (Scale > 1 && !isPred && MO.isImm())
MIB.addImm(MO.getImm() / Scale);
else
else {
if (MO.isReg() && MO.isImplicit() && MO.getReg() == ARM::CPSR)
// Skip implicit def of CPSR. Either it's modeled as an optional
// def now or it's already an implicit def on the new instruction.
continue;
MIB.addOperand(MO);
}
}
if (!TID.isPredicable() && NewTID.isPredicable())
AddDefaultPred(MIB);
DOUT << "Converted 32-bit: " << *MI << " to 16-bit: " << *MIB;
@ -503,20 +567,28 @@ Thumb2SizeReduce::ReduceToNarrow(MachineBasicBlock &MBB, MachineInstr *MI,
return true;
}
static bool UpdateCPSRLiveness(MachineInstr &MI, bool LiveCPSR) {
static bool UpdateCPSRDef(MachineInstr &MI, bool LiveCPSR) {
bool HasDef = false;
for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
const MachineOperand &MO = MI.getOperand(i);
if (!MO.isReg() || MO.isUndef())
if (!MO.isReg() || MO.isUndef() || MO.isUse())
continue;
if (MO.getReg() != ARM::CPSR)
continue;
if (MO.isDef()) {
if (!MO.isDead())
HasDef = true;
continue;
}
if (!MO.isDead())
HasDef = true;
}
return HasDef || LiveCPSR;
}
static bool UpdateCPSRUse(MachineInstr &MI, bool LiveCPSR) {
for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
const MachineOperand &MO = MI.getOperand(i);
if (!MO.isReg() || MO.isUndef() || MO.isDef())
continue;
if (MO.getReg() != ARM::CPSR)
continue;
assert(LiveCPSR && "CPSR liveness tracking is wrong!");
if (MO.isKill()) {
LiveCPSR = false;
@ -524,7 +596,7 @@ static bool UpdateCPSRLiveness(MachineInstr &MI, bool LiveCPSR) {
}
}
return HasDef || LiveCPSR;
return LiveCPSR;
}
bool Thumb2SizeReduce::ReduceMBB(MachineBasicBlock &MBB) {
@ -546,6 +618,8 @@ bool Thumb2SizeReduce::ReduceMBB(MachineBasicBlock &MBB) {
NextMII = next(MII);
MachineInstr *MI = &*MII;
LiveCPSR = UpdateCPSRUse(*MI, LiveCPSR);
unsigned Opcode = MI->getOpcode();
DenseMap<unsigned, unsigned>::iterator OPI = ReduceOpcodeMap.find(Opcode);
if (OPI != ReduceOpcodeMap.end()) {
@ -574,7 +648,7 @@ bool Thumb2SizeReduce::ReduceMBB(MachineBasicBlock &MBB) {
}
ProcessNext:
LiveCPSR = UpdateCPSRLiveness(*MI, LiveCPSR);
LiveCPSR = UpdateCPSRDef(*MI, LiveCPSR);
}
return Modified;

12
test/CodeGen/Thumb2/carry.ll
View File

@ -1,15 +1,21 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep "subs\\.w r" | count 2
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep "adc\\.w r"
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep "sbc\\.w r" | count 2
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
define i64 @f1(i64 %a, i64 %b) {
entry:
; CHECK: f1:
; CHECK: subs r0, r0, r2
; CHECK: sbcs r1, r3
%tmp = sub i64 %a, %b
ret i64 %tmp
}
define i64 @f2(i64 %a, i64 %b) {
entry:
; CHECK: f2:
; CHECK: adds r0, r0, r0
; CHECK: adcs r1, r1
; CHECK: subs r0, r0, r2
; CHECK: sbcs r1, r3
%tmp1 = shl i64 %a, 1
%tmp2 = sub i64 %tmp1, %b
ret i64 %tmp2

18
test/CodeGen/Thumb2/thumb2-adc.ll
View File

@ -1,32 +1,48 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {adc\\W*r\[0-9\],\\W*r\[0-9\],\\W*#\[0-9\]*} | grep {#171\\|#1179666\\|#872428544\\|#1448498774\\|#66846720} | count 5
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
; 734439407618 = 0x000000ab00000002
define i64 @f1(i64 %a) {
; CHECK: f1:
; CHECK: adds r0, #2
%tmp = add i64 %a, 734439407618
ret i64 %tmp
}
; 5066626890203138 = 0x0012001200000002
define i64 @f2(i64 %a) {
; CHECK: f2:
; CHECK: adds r0, #2
%tmp = add i64 %a, 5066626890203138
ret i64 %tmp
}
; 3747052064576897026 = 0x3400340000000002
define i64 @f3(i64 %a) {
; CHECK: f3:
; CHECK: adds r0, #2
%tmp = add i64 %a, 3747052064576897026
ret i64 %tmp
}
; 6221254862626095106 = 0x5656565600000002
define i64 @f4(i64 %a) {
; CHECK: f4:
; CHECK: adds r0, #2
%tmp = add i64 %a, 6221254862626095106
ret i64 %tmp
}
; 287104476244869122 = 0x03fc000000000002
define i64 @f5(i64 %a) {
; CHECK: f5:
; CHECK: adds r0, #2
%tmp = add i64 %a, 287104476244869122
ret i64 %tmp
}
define i64 @f6(i64 %a, i64 %b) {
; CHECK: f6:
; CHECK: adds r0, r0, r2
%tmp = add i64 %a, %b
ret i64 %tmp
}

6
test/CodeGen/Thumb2/thumb2-adc2.ll
View File

@ -1,6 +0,0 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {adc\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*r\[0-9\]*} | count 1
define i64 @f1(i64 %a, i64 %b) {
%tmp = add i64 %a, %b
ret i64 %tmp
}

17
test/CodeGen/Thumb2/thumb2-add4.ll
View File

@ -1,31 +1,46 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {adds\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*#\[0-9\]*} | grep {#171\\|#1179666\\|#872428544\\|#1448498774\\|#66846720} | count 5
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
; 171 = 0x000000ab
define i64 @f1(i64 %a) {
; CHECK: f1:
; CHECK: adds r0, #171
; CHECK: adc r1, r1, #0
%tmp = add i64 %a, 171
ret i64 %tmp
}
; 1179666 = 0x00120012
define i64 @f2(i64 %a) {
; CHECK: f2:
; CHECK: adds.w r0, r0, #1179666
; CHECK: adc r1, r1, #0
%tmp = add i64 %a, 1179666
ret i64 %tmp
}
; 872428544 = 0x34003400
define i64 @f3(i64 %a) {
; CHECK: f3:
; CHECK: adds.w r0, r0, #872428544
; CHECK: adc r1, r1, #0
%tmp = add i64 %a, 872428544
ret i64 %tmp
}
; 1448498774 = 0x56565656
define i64 @f4(i64 %a) {
; CHECK: f4:
; CHECK: adds.w r0, r0, #1448498774
; CHECK: adc r1, r1, #0
%tmp = add i64 %a, 1448498774
ret i64 %tmp
}
; 66846720 = 0x03fc0000
define i64 @f5(i64 %a) {
; CHECK: f5:
; CHECK: adds.w r0, r0, #66846720
; CHECK: adc r1, r1, #0
%tmp = add i64 %a, 66846720
ret i64 %tmp
}

5
test/CodeGen/Thumb2/thumb2-add6.ll
View File

@ -1,6 +1,9 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {adds\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*r\[0-9\]} | count 1
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
define i64 @f1(i64 %a, i64 %b) {
; CHECK: f1:
; CHECK: adds r0, r0, r2
; CHECK: adcs r1, r3
%tmp = add i64 %a, %b
ret i64 %tmp
}

4
test/CodeGen/Thumb2/thumb2-neg.ll
View File

@ -1,6 +1,8 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {rsb\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*#0} | count 1
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
define i32 @f1(i32 %a) {
; CHECK: f1:
; CHECK: rsbs r0, r0, #0
%tmp = sub i32 0, %a
ret i32 %tmp
}

8
test/CodeGen/Thumb2/thumb2-sbc.ll Normal file
View File

@ -0,0 +1,8 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
define i64 @f1(i64 %a, i64 %b) {
; CHECK: f1:
; CHECK: subs r0, r0, r2
%tmp = sub i64 %a, %b
ret i64 %tmp
}

6
test/CodeGen/Thumb2/thumb2-sbc2.ll
View File

@ -1,6 +0,0 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {sbc\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*r\[0-9\]*} | count 1
define i64 @f1(i64 %a, i64 %b) {
%tmp = sub i64 %a, %b
ret i64 %tmp
}

5
test/CodeGen/Thumb2/thumb2-sub5.ll
View File

@ -1,6 +1,9 @@
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | grep {subs\\.w\\W*r\[0-9\],\\W*r\[0-9\],\\W*r\[0-9\]} | count 1
; RUN: llvm-as < %s | llc -march=thumb -mattr=+thumb2 | FileCheck %s
define i64 @f1(i64 %a, i64 %b) {
; CHECK: f1:
; CHECK: subs r0, r0, r2
; CHECK: sbcs r1, r3
%tmp = sub i64 %a, %b
ret i64 %tmp
}