Updated LLVM to fix bad disassembly of operands

and occasionally failure to recognize instructions.
This problem affects ARM and Thumb BLX instructions.

llvm-svn: 156307
This commit is contained in:
Sean Callanan 2012-05-07 16:50:20 +00:00
parent e1f6554ed8
commit bcaa94db6c
1 changed files with 191 additions and 172 deletions

View File

@ -41,150 +41,6 @@ Index: lib/Target/ARM/ARMInstrNEON.td
[]>;
// Vector Move Operations.
Index: lib/Target/ARM/ARMInstrVFP.td
===================================================================
--- lib/Target/ARM/ARMInstrVFP.td (revision 152265)
+++ lib/Target/ARM/ARMInstrVFP.td (working copy)
@@ -818,7 +818,29 @@
// FP to Fixed-Point:
-def VTOSHS : AVConv1XI<0b11101, 0b11, 0b1110, 0b1010, 0,
+// Single Precision register
+class AVConv1XInsS_Encode<bits<5> op1, bits<2> op2, bits<4> op3, bits<4> op4, bit op5,
+ dag oops, dag iops, InstrItinClass itin, string opc, string asm,
+ list<dag> pattern>
+ : AVConv1XI<op1, op2, op3, op4, op5, oops, iops, itin, opc, asm, pattern> {
+ bits<5> dst;
+ // if dp_operation then UInt(D:Vd) else UInt(Vd:D);
+ let Inst{22} = dst{0};
+ let Inst{15-12} = dst{4-1};
+}
+
+// Double Precision register
+class AVConv1XInsD_Encode<bits<5> op1, bits<2> op2, bits<4> op3, bits<4> op4, bit op5,
+ dag oops, dag iops, InstrItinClass itin, string opc, string asm,
+ list<dag> pattern>
+ : AVConv1XI<op1, op2, op3, op4, op5, oops, iops, itin, opc, asm, pattern> {
+ bits<5> dst;
+ // if dp_operation then UInt(D:Vd) else UInt(Vd:D);
+ let Inst{22} = dst{4};
+ let Inst{15-12} = dst{3-0};
+}
+
+def VTOSHS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1110, 0b1010, 0,
(outs SPR:$dst), (ins SPR:$a, fbits16:$fbits),
IIC_fpCVTSI, "vcvt", ".s16.f32\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -826,7 +848,7 @@
let D = VFPNeonA8Domain;
}
-def VTOUHS : AVConv1XI<0b11101, 0b11, 0b1111, 0b1010, 0,
+def VTOUHS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1111, 0b1010, 0,
(outs SPR:$dst), (ins SPR:$a, fbits16:$fbits),
IIC_fpCVTSI, "vcvt", ".u16.f32\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -834,7 +856,7 @@
let D = VFPNeonA8Domain;
}
-def VTOSLS : AVConv1XI<0b11101, 0b11, 0b1110, 0b1010, 1,
+def VTOSLS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1110, 0b1010, 1,
(outs SPR:$dst), (ins SPR:$a, fbits32:$fbits),
IIC_fpCVTSI, "vcvt", ".s32.f32\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -842,7 +864,7 @@
let D = VFPNeonA8Domain;
}
-def VTOULS : AVConv1XI<0b11101, 0b11, 0b1111, 0b1010, 1,
+def VTOULS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1111, 0b1010, 1,
(outs SPR:$dst), (ins SPR:$a, fbits32:$fbits),
IIC_fpCVTSI, "vcvt", ".u32.f32\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -850,25 +872,25 @@
let D = VFPNeonA8Domain;
}
-def VTOSHD : AVConv1XI<0b11101, 0b11, 0b1110, 0b1011, 0,
+def VTOSHD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1110, 0b1011, 0,
(outs DPR:$dst), (ins DPR:$a, fbits16:$fbits),
IIC_fpCVTDI, "vcvt", ".s16.f64\t$dst, $a, $fbits", []>;
-def VTOUHD : AVConv1XI<0b11101, 0b11, 0b1111, 0b1011, 0,
+def VTOUHD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1111, 0b1011, 0,
(outs DPR:$dst), (ins DPR:$a, fbits16:$fbits),
IIC_fpCVTDI, "vcvt", ".u16.f64\t$dst, $a, $fbits", []>;
-def VTOSLD : AVConv1XI<0b11101, 0b11, 0b1110, 0b1011, 1,
+def VTOSLD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1110, 0b1011, 1,
(outs DPR:$dst), (ins DPR:$a, fbits32:$fbits),
IIC_fpCVTDI, "vcvt", ".s32.f64\t$dst, $a, $fbits", []>;
-def VTOULD : AVConv1XI<0b11101, 0b11, 0b1111, 0b1011, 1,
+def VTOULD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1111, 0b1011, 1,
(outs DPR:$dst), (ins DPR:$a, fbits32:$fbits),
IIC_fpCVTDI, "vcvt", ".u32.f64\t$dst, $a, $fbits", []>;
// Fixed-Point to FP:
-def VSHTOS : AVConv1XI<0b11101, 0b11, 0b1010, 0b1010, 0,
+def VSHTOS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1010, 0b1010, 0,
(outs SPR:$dst), (ins SPR:$a, fbits16:$fbits),
IIC_fpCVTIS, "vcvt", ".f32.s16\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -876,7 +898,7 @@
let D = VFPNeonA8Domain;
}
-def VUHTOS : AVConv1XI<0b11101, 0b11, 0b1011, 0b1010, 0,
+def VUHTOS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1011, 0b1010, 0,
(outs SPR:$dst), (ins SPR:$a, fbits16:$fbits),
IIC_fpCVTIS, "vcvt", ".f32.u16\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -884,7 +906,7 @@
let D = VFPNeonA8Domain;
}
-def VSLTOS : AVConv1XI<0b11101, 0b11, 0b1010, 0b1010, 1,
+def VSLTOS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1010, 0b1010, 1,
(outs SPR:$dst), (ins SPR:$a, fbits32:$fbits),
IIC_fpCVTIS, "vcvt", ".f32.s32\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -892,7 +914,7 @@
let D = VFPNeonA8Domain;
}
-def VULTOS : AVConv1XI<0b11101, 0b11, 0b1011, 0b1010, 1,
+def VULTOS : AVConv1XInsS_Encode<0b11101, 0b11, 0b1011, 0b1010, 1,
(outs SPR:$dst), (ins SPR:$a, fbits32:$fbits),
IIC_fpCVTIS, "vcvt", ".f32.u32\t$dst, $a, $fbits", []> {
// Some single precision VFP instructions may be executed on both NEON and
@@ -900,19 +922,19 @@
let D = VFPNeonA8Domain;
}
-def VSHTOD : AVConv1XI<0b11101, 0b11, 0b1010, 0b1011, 0,
+def VSHTOD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1010, 0b1011, 0,
(outs DPR:$dst), (ins DPR:$a, fbits16:$fbits),
IIC_fpCVTID, "vcvt", ".f64.s16\t$dst, $a, $fbits", []>;
-def VUHTOD : AVConv1XI<0b11101, 0b11, 0b1011, 0b1011, 0,
+def VUHTOD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1011, 0b1011, 0,
(outs DPR:$dst), (ins DPR:$a, fbits16:$fbits),
IIC_fpCVTID, "vcvt", ".f64.u16\t$dst, $a, $fbits", []>;
-def VSLTOD : AVConv1XI<0b11101, 0b11, 0b1010, 0b1011, 1,
+def VSLTOD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1010, 0b1011, 1,
(outs DPR:$dst), (ins DPR:$a, fbits32:$fbits),
IIC_fpCVTID, "vcvt", ".f64.s32\t$dst, $a, $fbits", []>;
-def VULTOD : AVConv1XI<0b11101, 0b11, 0b1011, 0b1011, 1,
+def VULTOD : AVConv1XInsD_Encode<0b11101, 0b11, 0b1011, 0b1011, 1,
(outs DPR:$dst), (ins DPR:$a, fbits32:$fbits),
IIC_fpCVTID, "vcvt", ".f64.u32\t$dst, $a, $fbits", []>;
Index: lib/Target/ARM/ARMInstrThumb2.td
===================================================================
--- lib/Target/ARM/ARMInstrThumb2.td (revision 152265)
@ -197,6 +53,142 @@ Index: lib/Target/ARM/ARMInstrThumb2.td
}
let isNotDuplicable = 1, isIndirectBranch = 1 in {
Index: lib/Target/ARM/ARMInstrThumb.td
===================================================================
--- lib/Target/ARM/ARMInstrThumb.td (revision 152265)
+++ lib/Target/ARM/ARMInstrThumb.td (working copy)
@@ -413,11 +413,11 @@
"bl${p}\t$func",
[(ARMtcall tglobaladdr:$func)]>,
Requires<[IsThumb, IsNotIOS]> {
- bits<22> func;
- let Inst{26} = func{21};
+ bits<24> func;
+ let Inst{26} = func{23};
let Inst{25-16} = func{20-11};
- let Inst{13} = 1;
- let Inst{11} = 1;
+ let Inst{13} = func{22};
+ let Inst{11} = func{21};
let Inst{10-0} = func{10-0};
}
@@ -427,10 +427,11 @@
"blx${p}\t$func",
[(ARMcall tglobaladdr:$func)]>,
Requires<[IsThumb, HasV5T, IsNotIOS]> {
- bits<21> func;
+ bits<24> func;
+ let Inst{26} = func{23};
let Inst{25-16} = func{20-11};
- let Inst{13} = 1;
- let Inst{11} = 1;
+ let Inst{13} = func{22};
+ let Inst{11} = func{21};
let Inst{10-1} = func{10-1};
let Inst{0} = 0; // func{0} is assumed zero
}
Index: lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
===================================================================
--- lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp (revision 152265)
+++ lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp (working copy)
@@ -397,39 +397,65 @@
return swapped;
}
case ARM::fixup_arm_thumb_bl: {
- // The value doesn't encode the low bit (always zero) and is offset by
- // four. The value is encoded into disjoint bit positions in the destination
- // opcode. x = unchanged, I = immediate value bit, S = sign extension bit
- //
- // BL: xxxxxSIIIIIIIIII xxxxxIIIIIIIIIII
- //
- // Note that the halfwords are stored high first, low second; so we need
- // to transpose the fixup value here to map properly.
- unsigned isNeg = (int64_t(Value - 4) < 0) ? 1 : 0;
- uint32_t Binary = 0;
- Value = 0x3fffff & ((Value - 4) >> 1);
- Binary = (Value & 0x7ff) << 16; // Low imm11 value.
- Binary |= (Value & 0x1ffc00) >> 11; // High imm10 value.
- Binary |= isNeg << 10; // Sign bit.
- return Binary;
+ // The value doesn't encode the low bit (always zero) and is offset by
+ // four. The 32-bit immediate value is encoded as
+ // imm32 = SignExtend(S:I1:I2:imm10:imm11:0)
+ // where I1 = NOT(J1 ^ S) and I2 = NOT(J2 ^ S).
+ // The value is encoded into disjoint bit positions in the destination
+ // opcode. x = unchanged, I = immediate value bit, S = sign extension bit,
+ // J = either J1 or J2 bit
+ //
+ // BL: xxxxxSIIIIIIIIII xxJxJIIIIIIIIIII
+ //
+ // Note that the halfwords are stored high first, low second; so we need
+ // to transpose the fixup value here to map properly.
+ uint32_t offset = (Value - 4) >> 1;
+ uint32_t signBit = (offset & 0x800000) >> 23;
+ uint32_t I1Bit = (offset & 0x400000) >> 22;
+ uint32_t J1Bit = (I1Bit ^ 0x1) ^ signBit;
+ uint32_t I2Bit = (offset & 0x200000) >> 21;
+ uint32_t J2Bit = (I2Bit ^ 0x1) ^ signBit;
+ uint32_t imm10Bits = (offset & 0x1FF800) >> 11;
+ uint32_t imm11Bits = (offset & 0x000007FF);
+
+ uint32_t Binary = 0;
+ uint32_t firstHalf = (((uint16_t)signBit << 10) | (uint16_t)imm10Bits);
+ uint32_t secondHalf = (((uint16_t)J1Bit << 13) | ((uint16_t)J2Bit << 11) |
+ (uint16_t)imm11Bits);
+ Binary |= secondHalf << 16;
+ Binary |= firstHalf;
+ return Binary;
+
}
case ARM::fixup_arm_thumb_blx: {
- // The value doesn't encode the low two bits (always zero) and is offset by
- // four (see fixup_arm_thumb_cp). The value is encoded into disjoint bit
- // positions in the destination opcode. x = unchanged, I = immediate value
- // bit, S = sign extension bit, 0 = zero.
- //
- // BLX: xxxxxSIIIIIIIIII xxxxxIIIIIIIIII0
- //
- // Note that the halfwords are stored high first, low second; so we need
- // to transpose the fixup value here to map properly.
- unsigned isNeg = (int64_t(Value-4) < 0) ? 1 : 0;
- uint32_t Binary = 0;
- Value = 0xfffff & ((Value - 2) >> 2);
- Binary = (Value & 0x3ff) << 17; // Low imm10L value.
- Binary |= (Value & 0xffc00) >> 10; // High imm10H value.
- Binary |= isNeg << 10; // Sign bit.
- return Binary;
+ // The value doesn't encode the low two bits (always zero) and is offset by
+ // four (see fixup_arm_thumb_cp). The 32-bit immediate value is encoded as
+ // imm32 = SignExtend(S:I1:I2:imm10H:imm10L:00)
+ // where I1 = NOT(J1 ^ S) and I2 = NOT(J2 ^ S).
+ // The value is encoded into disjoint bit positions in the destination
+ // opcode. x = unchanged, I = immediate value bit, S = sign extension bit,
+ // J = either J1 or J2 bit, 0 = zero.
+ //
+ // BLX: xxxxxSIIIIIIIIII xxJxJIIIIIIIIII0
+ //
+ // Note that the halfwords are stored high first, low second; so we need
+ // to transpose the fixup value here to map properly.
+ uint32_t offset = (Value - 2) >> 2;
+ uint32_t signBit = (offset & 0x400000) >> 22;
+ uint32_t I1Bit = (offset & 0x200000) >> 21;
+ uint32_t J1Bit = (I1Bit ^ 0x1) ^ signBit;
+ uint32_t I2Bit = (offset & 0x100000) >> 20;
+ uint32_t J2Bit = (I2Bit ^ 0x1) ^ signBit;
+ uint32_t imm10HBits = (offset & 0xFFC00) >> 10;
+ uint32_t imm10LBits = (offset & 0x3FF);
+
+ uint32_t Binary = 0;
+ uint32_t firstHalf = (((uint16_t)signBit << 10) | (uint16_t)imm10HBits);
+ uint32_t secondHalf = (((uint16_t)J1Bit << 13) | ((uint16_t)J2Bit << 11) |
+ ((uint16_t)imm10LBits) << 1);
+ Binary |= secondHalf << 16;
+ Binary |= firstHalf;
+ return Binary;
}
case ARM::fixup_arm_thumb_cp:
// Offset by 4, and don't encode the low two bits. Two bytes of that
Index: lib/Target/ARM/Disassembler/ARMDisassembler.cpp
===================================================================
--- lib/Target/ARM/Disassembler/ARMDisassembler.cpp (revision 152265)
@ -293,33 +285,7 @@ Index: lib/Target/ARM/Disassembler/ARMDisassembler.cpp
break;
}
@@ -2555,7 +2605,6 @@
unsigned Rm = fieldFromInstruction32(Insn, 0, 4);
unsigned align = fieldFromInstruction32(Insn, 4, 1);
unsigned size = 1 << fieldFromInstruction32(Insn, 6, 2);
- unsigned pred = fieldFromInstruction32(Insn, 22, 4);
align *= 2*size;
switch (Inst.getOpcode()) {
@@ -2586,16 +2635,11 @@
return MCDisassembler::Fail;
Inst.addOperand(MCOperand::CreateImm(align));
- if (Rm == 0xD)
- Inst.addOperand(MCOperand::CreateReg(0));
- else if (Rm != 0xF) {
+ if (Rm != 0xD && Rm != 0xF) {
if (!Check(S, DecodeGPRRegisterClass(Inst, Rm, Address, Decoder)))
return MCDisassembler::Fail;
}
- if (!Check(S, DecodePredicateOperand(Inst, pred, Address, Decoder)))
- return MCDisassembler::Fail;
-
return S;
}
@@ -2837,19 +2881,25 @@
@@ -2837,19 +2887,25 @@
static DecodeStatus DecodeThumbBROperand(llvm::MCInst &Inst, unsigned Val,
uint64_t Address, const void *Decoder) {
@ -348,7 +314,35 @@ Index: lib/Target/ARM/Disassembler/ARMDisassembler.cpp
return MCDisassembler::Success;
}
@@ -3271,7 +3321,9 @@
@@ -3162,10 +3218,25 @@
static DecodeStatus DecodeThumbBLXOffset(llvm::MCInst &Inst, unsigned Val,
uint64_t Address, const void *Decoder) {
+ // Val is passed in as S:J1:J2:imm10H:imm10L:0
+ // Note only one trailing zero not two. Also the J1 and J2 values are from
+ // the encoded instruction. So here change to I1 and I2 values via:
+ // I1 = NOT(J1 EOR S);
+ // I2 = NOT(J2 EOR S);
+ // and build the imm32 with two trailing zeros as documented:
+ // imm32 = SignExtend(S:I1:I2:imm10H:imm10L:00, 32);
+ unsigned S = (Val >> 23) & 1;
+ unsigned J1 = (Val >> 22) & 1;
+ unsigned J2 = (Val >> 21) & 1;
+ unsigned I1 = !(J1 ^ S);
+ unsigned I2 = !(J2 ^ S);
+ unsigned tmp = (Val & ~0x600000) | (I1 << 22) | (I2 << 21);
+ int imm32 = SignExtend32<25>(tmp << 1);
+
if (!tryAddingSymbolicOperand(Address,
- (Address & ~2u) + SignExtend32<22>(Val << 1) + 4,
+ (Address & ~2u) + imm32 + 4,
true, 4, Inst, Decoder))
- Inst.addOperand(MCOperand::CreateImm(SignExtend32<22>(Val << 1)));
+ Inst.addOperand(MCOperand::CreateImm(imm32));
return MCDisassembler::Success;
}
@@ -3271,15 +3342,32 @@
static DecodeStatus
DecodeThumbBCCTargetOperand(llvm::MCInst &Inst, unsigned Val,
uint64_t Address, const void *Decoder){
@ -359,3 +353,28 @@ Index: lib/Target/ARM/Disassembler/ARMDisassembler.cpp
return MCDisassembler::Success;
}
static DecodeStatus DecodeThumbBLTargetOperand(llvm::MCInst &Inst, unsigned Val,
uint64_t Address, const void *Decoder){
- if (!tryAddingSymbolicOperand(Address, Address + SignExtend32<22>(Val<<1) + 4,
+ // Val is passed in as S:J1:J2:imm10:imm11
+ // Note no trailing zero after imm11. Also the J1 and J2 values are from
+ // the encoded instruction. So here change to I1 and I2 values via:
+ // I1 = NOT(J1 EOR S);
+ // I2 = NOT(J2 EOR S);
+ // and build the imm32 with one trailing zero as documented:
+ // imm32 = SignExtend(S:I1:I2:imm10:imm11:0, 32);
+ unsigned S = (Val >> 23) & 1;
+ unsigned J1 = (Val >> 22) & 1;
+ unsigned J2 = (Val >> 21) & 1;
+ unsigned I1 = !(J1 ^ S);
+ unsigned I2 = !(J2 ^ S);
+ unsigned tmp = (Val & ~0x600000) | (I1 << 22) | (I2 << 21);
+ int imm32 = SignExtend32<25>(tmp << 1);
+
+ if (!tryAddingSymbolicOperand(Address, Address + imm32 + 4,
true, 4, Inst, Decoder))
- Inst.addOperand(MCOperand::CreateImm(SignExtend32<22>(Val << 1)));
+ Inst.addOperand(MCOperand::CreateImm(imm32));
return MCDisassembler::Success;
}