// Copyright 2014 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#ifndef V8_PPC_CONSTANTS_PPC_H_
#define V8_PPC_CONSTANTS_PPC_H_
#include <stdint.h>
#include "src/base/logging.h"
#include "src/base/macros.h"
#include "src/globals.h"
namespace v8 {
namespace internal {
// Number of registers
const int kNumRegisters = 32;
// FP support.
const int kNumDoubleRegisters = 32;
const int kNoRegister = -1;
// Used in embedded constant pool builder - max reach in bits for
// various load instructions (one less due to unsigned)
const int kLoadPtrMaxReachBits = 15;
const int kLoadDoubleMaxReachBits = 15;
// sign-extend the least significant 16-bits of value <imm>
#define SIGN_EXT_IMM16(imm) ((static_cast<int>(imm) << 16) >> 16)
// sign-extend the least significant 26-bits of value <imm>
#define SIGN_EXT_IMM26(imm) ((static_cast<int>(imm) << 6) >> 6)
// -----------------------------------------------------------------------------
// Conditions.
// Defines constants and accessor classes to assemble, disassemble and
// simulate PPC instructions.
//
// Section references in the code refer to the "PowerPC Microprocessor
// Family: The Programmer.s Reference Guide" from 10/95
// https://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/852569B20050FF778525699600741775/$file/prg.pdf
//
// Constants for specific fields are defined in their respective named enums.
// General constants are in an anonymous enum in class Instr.
enum Condition {
kNoCondition = -1,
eq = 0, // Equal.
ne = 1, // Not equal.
ge = 2, // Greater or equal.
lt = 3, // Less than.
gt = 4, // Greater than.
le = 5, // Less then or equal
unordered = 6, // Floating-point unordered
ordered = 7,
overflow = 8, // Summary overflow
nooverflow = 9,
al = 10 // Always.
};
inline Condition NegateCondition(Condition cond) {
DCHECK(cond != al);
return static_cast<Condition>(cond ^ ne);
}
// Commute a condition such that {a cond b == b cond' a}.
inline Condition CommuteCondition(Condition cond) {
switch (cond) {
case lt:
return gt;
case gt:
return lt;
case ge:
return le;
case le:
return ge;
default:
return cond;
}
}
// -----------------------------------------------------------------------------
// Instructions encoding.
// Instr is merely used by the Assembler to distinguish 32bit integers
// representing instructions from usual 32 bit values.
// Instruction objects are pointers to 32bit values, and provide methods to
// access the various ISA fields.
typedef uint32_t Instr;
#define PPC_XX3_OPCODE_LIST(V) \
/* VSX Scalar Add Double-Precision */ \
V(xsadddp, XSADDDP, 0xF0000100) \
/* VSX Scalar Add Single-Precision */ \
V(xsaddsp, XSADDSP, 0xF0000000) \
/* VSX Scalar Compare Ordered Double-Precision */ \
V(xscmpodp, XSCMPODP, 0xF0000158) \
/* VSX Scalar Compare Unordered Double-Precision */ \
V(xscmpudp, XSCMPUDP, 0xF0000118) \
/* VSX Scalar Copy Sign Double-Precision */ \
V(xscpsgndp, XSCPSGNDP, 0xF0000580) \
/* VSX Scalar Divide Double-Precision */ \
V(xsdivdp, XSDIVDP, 0xF00001C0) \
/* VSX Scalar Divide Single-Precision */ \
V(xsdivsp, XSDIVSP, 0xF00000C0) \
/* VSX Scalar Multiply-Add Type-A Double-Precision */ \
V(xsmaddadp, XSMADDADP, 0xF0000108) \
/* VSX Scalar Multiply-Add Type-A Single-Precision */ \
V(xsmaddasp, XSMADDASP, 0xF0000008) \
/* VSX Scalar Multiply-Add Type-M Double-Precision */ \
V(xsmaddmdp, XSMADDMDP, 0xF0000148) \
/* VSX Scalar Multiply-Add Type-M Single-Precision */ \
V(xsmaddmsp, XSMADDMSP, 0xF0000048) \
/* VSX Scalar Maximum Double-Precision */ \
V(xsmaxdp, XSMAXDP, 0xF0000500) \
/* VSX Scalar Minimum Double-Precision */ \
V(xsmindp, XSMINDP, 0xF0000540) \
/* VSX Scalar Multiply-Subtract Type-A Double-Precision */ \
V(xsmsubadp, XSMSUBADP, 0xF0000188) \
/* VSX Scalar Multiply-Subtract Type-A Single-Precision */ \
V(xsmsubasp, XSMSUBASP, 0xF0000088) \
/* VSX Scalar Multiply-Subtract Type-M Double-Precision */ \
V(xsmsubmdp, XSMSUBMDP, 0xF00001C8) \
/* VSX Scalar Multiply-Subtract Type-M Single-Precision */ \
V(xsmsubmsp, XSMSUBMSP, 0xF00000C8) \
/* VSX Scalar Multiply Double-Precision */ \
V(xsmuldp, XSMULDP, 0xF0000180) \
/* VSX Scalar Multiply Single-Precision */ \
V(xsmulsp, XSMULSP, 0xF0000080) \
/* VSX Scalar Negative Multiply-Add Type-A Double-Precision */ \
V(xsnmaddadp, XSNMADDADP, 0xF0000508) \
/* VSX Scalar Negative Multiply-Add Type-A Single-Precision */ \
V(xsnmaddasp, XSNMADDASP, 0xF0000408) \
/* VSX Scalar Negative Multiply-Add Type-M Double-Precision */ \
V(xsnmaddmdp, XSNMADDMDP, 0xF0000548) \
/* VSX Scalar Negative Multiply-Add Type-M Single-Precision */ \
V(xsnmaddmsp, XSNMADDMSP, 0xF0000448) \
/* VSX Scalar Negative Multiply-Subtract Type-A Double-Precision */ \
V(xsnmsubadp, XSNMSUBADP, 0xF0000588) \
/* VSX Scalar Negative Multiply-Subtract Type-A Single-Precision */ \
V(xsnmsubasp, XSNMSUBASP, 0xF0000488) \
/* VSX Scalar Negative Multiply-Subtract Type-M Double-Precision */ \
V(xsnmsubmdp, XSNMSUBMDP, 0xF00005C8) \
/* VSX Scalar Negative Multiply-Subtract Type-M Single-Precision */ \
V(xsnmsubmsp, XSNMSUBMSP, 0xF00004C8) \
/* VSX Scalar Reciprocal Estimate Double-Precision */ \
V(xsredp, XSREDP, 0xF0000168) \
/* VSX Scalar Reciprocal Estimate Single-Precision */ \
V(xsresp, XSRESP, 0xF0000068) \
/* VSX Scalar Subtract Double-Precision */ \
V(xssubdp, XSSUBDP, 0xF0000140) \
/* VSX Scalar Subtract Single-Precision */ \
V(xssubsp, XSSUBSP, 0xF0000040) \
/* VSX Scalar Test for software Divide Double-Precision */ \
V(xstdivdp, XSTDIVDP, 0xF00001E8) \
/* VSX Vector Add Double-Precision */ \
V(xvadddp, XVADDDP, 0xF0000300) \
/* VSX Vector Add Single-Precision */ \
V(xvaddsp, XVADDSP, 0xF0000200) \
/* VSX Vector Compare Equal To Double-Precision */ \
V(xvcmpeqdp, XVCMPEQDP, 0xF0000318) \
/* VSX Vector Compare Equal To Double-Precision & record CR6 */ \
V(xvcmpeqdpx, XVCMPEQDPx, 0xF0000718) \
/* VSX Vector Compare Equal To Single-Precision */ \
V(xvcmpeqsp, XVCMPEQSP, 0xF0000218) \
/* VSX Vector Compare Equal To Single-Precision & record CR6 */ \
V(xvcmpeqspx, XVCMPEQSPx, 0xF0000618) \
/* VSX Vector Compare Greater Than or Equal To Double-Precision */ \
V(xvcmpgedp, XVCMPGEDP, 0xF0000398) \
/* VSX Vector Compare Greater Than or Equal To Double-Precision & record */ \
/* CR6 */ \
V(xvcmpgedpx, XVCMPGEDPx, 0xF0000798) \
/* VSX Vector Compare Greater Than or Equal To Single-Precision */ \
V(xvcmpgesp, XVCMPGESP, 0xF0000298) \
/* VSX Vector Compare Greater Than or Equal To Single-Precision & record */ \
/* CR6 */ \
V(xvcmpgespx, XVCMPGESPx, 0xF0000698) \
/* VSX Vector Compare Greater Than Double-Precision */ \
V(xvcmpgtdp, XVCMPGTDP, 0xF0000358) \
/* VSX Vector Compare Greater Than Double-Precision & record CR6 */ \
V(xvcmpgtdpx, XVCMPGTDPx, 0xF0000758) \
/* VSX Vector Compare Greater Than Single-Precision */ \
V(xvcmpgtsp, XVCMPGTSP, 0xF0000258) \
/* VSX Vector Compare Greater Than Single-Precision & record CR6 */ \
V(xvcmpgtspx, XVCMPGTSPx, 0xF0000658) \
/* VSX Vector Copy Sign Double-Precision */ \
V(xvcpsgndp, XVCPSGNDP, 0xF0000780) \
/* VSX Vector Copy Sign Single-Precision */ \
V(xvcpsgnsp, XVCPSGNSP, 0xF0000680) \
/* VSX Vector Divide Double-Precision */ \
V(xvdivdp, XVDIVDP, 0xF00003C0) \
/* VSX Vector Divide Single-Precision */ \
V(xvdivsp, XVDIVSP, 0xF00002C0) \
/* VSX Vector Multiply-Add Type-A Double-Precision */ \
V(xvmaddadp, XVMADDADP, 0xF0000308) \
/* VSX Vector Multiply-Add Type-A Single-Precision */ \
V(xvmaddasp, XVMADDASP, 0xF0000208) \
/* VSX Vector Multiply-Add Type-M Double-Precision */ \
V(xvmaddmdp, XVMADDMDP, 0xF0000348) \
/* VSX Vector Multiply-Add Type-M Single-Precision */ \
V(xvmaddmsp, XVMADDMSP, 0xF0000248) \
/* VSX Vector Maximum Double-Precision */ \
V(xvmaxdp, XVMAXDP, 0xF0000700) \
/* VSX Vector Maximum Single-Precision */ \
V(xvmaxsp, XVMAXSP, 0xF0000600) \
/* VSX Vector Minimum Double-Precision */ \
V(xvmindp, XVMINDP, 0xF0000740) \
/* VSX Vector Minimum Single-Precision */ \
V(xvminsp, XVMINSP, 0xF0000640) \
/* VSX Vector Multiply-Subtract Type-A Double-Precision */ \
V(xvmsubadp, XVMSUBADP, 0xF0000388) \
/* VSX Vector Multiply-Subtract Type-A Single-Precision */ \
V(xvmsubasp, XVMSUBASP, 0xF0000288) \
/* VSX Vector Multiply-Subtract Type-M Double-Precision */ \
V(xvmsubmdp, XVMSUBMDP, 0xF00003C8) \
/* VSX Vector Multiply-Subtract Type-M Single-Precision */ \
V(xvmsubmsp, XVMSUBMSP, 0xF00002C8) \
/* VSX Vector Multiply Double-Precision */ \
V(xvmuldp, XVMULDP, 0xF0000380) \
/* VSX Vector Multiply Single-Precision */ \
V(xvmulsp, XVMULSP, 0xF0000280) \
/* VSX Vector Negative Multiply-Add Type-A Double-Precision */ \
V(xvnmaddadp, XVNMADDADP, 0xF0000708) \
/* VSX Vector Negative Multiply-Add Type-A Single-Precision */ \
V(xvnmaddasp, XVNMADDASP, 0xF0000608) \
/* VSX Vector Negative Multiply-Add Type-M Double-Precision */ \
V(xvnmaddmdp, XVNMADDMDP, 0xF0000748) \
/* VSX Vector Negative Multiply-Add Type-M Single-Precision */ \
V(xvnmaddmsp, XVNMADDMSP, 0xF0000648) \
/* VSX Vector Negative Multiply-Subtract Type-A Double-Precision */ \
V(xvnmsubadp, XVNMSUBADP, 0xF0000788) \
/* VSX Vector Negative Multiply-Subtract Type-A Single-Precision */ \
V(xvnmsubasp, XVNMSUBASP, 0xF0000688) \
/* VSX Vector Negative Multiply-Subtract Type-M Double-Precision */ \
V(xvnmsubmdp, XVNMSUBMDP, 0xF00007C8) \
/* VSX Vector Negative Multiply-Subtract Type-M Single-Precision */ \
V(xvnmsubmsp, XVNMSUBMSP, 0xF00006C8) \
/* VSX Vector Reciprocal Estimate Double-Precision */ \
V(xvredp, XVREDP, 0xF0000368) \
/* VSX Vector Reciprocal Estimate Single-Precision */ \
V(xvresp, XVRESP, 0xF0000268) \
/* VSX Vector Subtract Double-Precision */ \
V(xvsubdp, XVSUBDP, 0xF0000340) \
/* VSX Vector Subtract Single-Precision */ \
V(xvsubsp, XVSUBSP, 0xF0000240) \
/* VSX Vector Test for software Divide Double-Precision */ \
V(xvtdivdp, XVTDIVDP, 0xF00003E8) \
/* VSX Vector Test for software Divide Single-Precision */ \
V(xvtdivsp, XVTDIVSP, 0xF00002E8) \
/* VSX Logical AND */ \
V(xxland, XXLAND, 0xF0000410) \
/* VSX Logical AND with Complement */ \
V(xxlandc, XXLANDC, 0xF0000450) \
/* VSX Logical Equivalence */ \
V(xxleqv, XXLEQV, 0xF00005D0) \
/* VSX Logical NAND */ \
V(xxlnand, XXLNAND, 0xF0000590) \
/* VSX Logical NOR */ \
V(xxlnor, XXLNOR, 0xF0000510) \
/* VSX Logical OR */ \
V(xxlor, XXLOR, 0xF0000490) \
/* VSX Logical OR with Complement */ \
V(xxlorc, XXLORC, 0xF0000550) \
/* VSX Logical XOR */ \
V(xxlxor, XXLXOR, 0xF00004D0) \
/* VSX Merge High Word */ \
V(xxmrghw, XXMRGHW, 0xF0000090) \
/* VSX Merge Low Word */ \
V(xxmrglw, XXMRGLW, 0xF0000190) \
/* VSX Permute Doubleword Immediate */ \
V(xxpermdi, XXPERMDI, 0xF0000050) \
/* VSX Shift Left Double by Word Immediate */ \
V(xxsldwi, XXSLDWI, 0xF0000010) \
/* VSX Splat Word */ \
V(xxspltw, XXSPLTW, 0xF0000290)
#define PPC_Z23_OPCODE_LIST(V) \
/* Decimal Quantize */ \
V(dqua, DQUA, 0xEC000006) \
/* Decimal Quantize Immediate */ \
V(dquai, DQUAI, 0xEC000086) \
/* Decimal Quantize Immediate Quad */ \
V(dquaiq, DQUAIQ, 0xFC000086) \
/* Decimal Quantize Quad */ \
V(dquaq, DQUAQ, 0xFC000006) \
/* Decimal Floating Round To FP Integer Without Inexact */ \
V(drintn, DRINTN, 0xEC0001C6) \
/* Decimal Floating Round To FP Integer Without Inexact Quad */ \
V(drintnq, DRINTNQ, 0xFC0001C6) \
/* Decimal Floating Round To FP Integer With Inexact */ \
V(drintx, DRINTX, 0xEC0000C6) \
/* Decimal Floating Round To FP Integer With Inexact Quad */ \
V(drintxq, DRINTXQ, 0xFC0000C6) \
/* Decimal Floating Reround */ \
V(drrnd, DRRND, 0xEC000046) \
/* Decimal Floating Reround Quad */ \
V(drrndq, DRRNDQ, 0xFC000046)
#define PPC_Z22_OPCODE_LIST(V) \
/* Decimal Floating Shift Coefficient Left Immediate */ \
V(dscli, DSCLI, 0xEC000084) \
/* Decimal Floating Shift Coefficient Left Immediate Quad */ \
V(dscliq, DSCLIQ, 0xFC000084) \
/* Decimal Floating Shift Coefficient Right Immediate */ \
V(dscri, DSCRI, 0xEC0000C4) \
/* Decimal Floating Shift Coefficient Right Immediate Quad */ \
V(dscriq, DSCRIQ, 0xFC0000C4) \
/* Decimal Floating Test Data Class */ \
V(dtstdc, DTSTDC, 0xEC000184) \
/* Decimal Floating Test Data Class Quad */ \
V(dtstdcq, DTSTDCQ, 0xFC000184) \
/* Decimal Floating Test Data Group */ \
V(dtstdg, DTSTDG, 0xEC0001C4) \
/* Decimal Floating Test Data Group Quad */ \
V(dtstdgq, DTSTDGQ, 0xFC0001C4)
#define PPC_XX2_OPCODE_LIST(V) \
/* Move To VSR Doubleword */ \
V(mtvsrd, MTVSRD, 0x7C000166) \
/* Move To VSR Word Algebraic */ \
V(mtvsrwa, MTVSRWA, 0x7C0001A6) \
/* Move To VSR Word and Zero */ \
V(mtvsrwz, MTVSRWZ, 0x7C0001E6) \
/* VSX Scalar Absolute Value Double-Precision */ \
V(xsabsdp, XSABSDP, 0xF0000564) \
/* VSX Scalar Convert Double-Precision to Single-Precision */ \
V(xscvdpsp, XSCVDPSP, 0xF0000424) \
/* VSX Scalar Convert Double-Precision to Single-Precision format Non- */ \
/* signalling */ \
V(xscvdpspn, XSCVDPSPN, 0xF000042C) \
/* VSX Scalar Convert Double-Precision to Signed Fixed-Point Doubleword */ \
/* Saturate */ \
V(xscvdpsxds, XSCVDPSXDS, 0xF0000560) \
/* VSX Scalar Convert Double-Precision to Signed Fixed-Point Word */ \
/* Saturate */ \
V(xscvdpsxws, XSCVDPSXWS, 0xF0000160) \
/* VSX Scalar Convert Double-Precision to Unsigned Fixed-Point */ \
/* Doubleword Saturate */ \
V(xscvdpuxds, XSCVDPUXDS, 0xF0000520) \
/* VSX Scalar Convert Double-Precision to Unsigned Fixed-Point Word */ \
/* Saturate */ \
V(xscvdpuxws, XSCVDPUXWS, 0xF0000120) \
/* VSX Scalar Convert Single-Precision to Double-Precision (p=1) */ \
V(xscvspdp, XSCVSPDP, 0xF0000524) \
/* Scalar Convert Single-Precision to Double-Precision format Non- */ \
/* signalling */ \
V(xscvspdpn, XSCVSPDPN, 0xF000052C) \
/* VSX Scalar Convert Signed Fixed-Point Doubleword to Double-Precision */ \
V(xscvsxddp, XSCVSXDDP, 0xF00005E0) \
/* VSX Scalar Convert Signed Fixed-Point Doubleword to Single-Precision */ \
V(xscvsxdsp, XSCVSXDSP, 0xF00004E0) \
/* VSX Scalar Convert Unsigned Fixed-Point Doubleword to Double- */ \
/* Precision */ \
V(xscvuxddp, XSCVUXDDP, 0xF00005A0) \
/* VSX Scalar Convert Unsigned Fixed-Point Doubleword to Single- */ \
/* Precision */ \
V(xscvuxdsp, XSCVUXDSP, 0xF00004A0) \
/* VSX Scalar Negative Absolute Value Double-Precision */ \
V(xsnabsdp, XSNABSDP, 0xF00005A4) \
/* VSX Scalar Negate Double-Precision */ \
V(xsnegdp, XSNEGDP, 0xF00005E4) \
/* VSX Scalar Round to Double-Precision Integer */ \
V(xsrdpi, XSRDPI, 0xF0000124) \
/* VSX Scalar Round to Double-Precision Integer using Current rounding */ \
/* mode */ \
V(xsrdpic, XSRDPIC, 0xF00001AC) \
/* VSX Scalar Round to Double-Precision Integer toward -Infinity */ \
V(xsrdpim, XSRDPIM, 0xF00001E4) \
/* VSX Scalar Round to Double-Precision Integer toward +Infinity */ \
V(xsrdpip, XSRDPIP, 0xF00001A4) \
/* VSX Scalar Round to Double-Precision Integer toward Zero */ \
V(xsrdpiz, XSRDPIZ, 0xF0000164) \
/* VSX Scalar Round to Single-Precision */ \
V(xsrsp, XSRSP, 0xF0000464) \
/* VSX Scalar Reciprocal Square Root Estimate Double-Precision */ \
V(xsrsqrtedp, XSRSQRTEDP, 0xF0000128) \
/* VSX Scalar Reciprocal Square Root Estimate Single-Precision */ \
V(xsrsqrtesp, XSRSQRTESP, 0xF0000028) \
/* VSX Scalar Square Root Double-Precision */ \
V(xssqrtdp, XSSQRTDP, 0xF000012C) \
/* VSX Scalar Square Root Single-Precision */ \
V(xssqrtsp, XSSQRTSP, 0xF000002C) \
/* VSX Scalar Test for software Square Root Double-Precision */ \
V(xstsqrtdp, XSTSQRTDP, 0xF00001A8) \
/* VSX Vector Absolute Value Double-Precision */ \
V(xvabsdp, XVABSDP, 0xF0000764) \
/* VSX Vector Absolute Value Single-Precision */ \
V(xvabssp, XVABSSP, 0xF0000664) \
/* VSX Vector Convert Double-Precision to Single-Precision */ \
V(xvcvdpsp, XVCVDPSP, 0xF0000624) \
/* VSX Vector Convert Double-Precision to Signed Fixed-Point Doubleword */ \
/* Saturate */ \
V(xvcvdpsxds, XVCVDPSXDS, 0xF0000760) \
/* VSX Vector Convert Double-Precision to Signed Fixed-Point Word */ \
/* Saturate */ \
V(xvcvdpsxws, XVCVDPSXWS, 0xF0000360) \
/* VSX Vector Convert Double-Precision to Unsigned Fixed-Point */ \
/* Doubleword Saturate */ \
V(xvcvdpuxds, XVCVDPUXDS, 0xF0000720) \
/* VSX Vector Convert Double-Precision to Unsigned Fixed-Point Word */ \
/* Saturate */ \
V(xvcvdpuxws, XVCVDPUXWS, 0xF0000320) \
/* VSX Vector Convert Single-Precision to Double-Precision */ \
V(xvcvspdp, XVCVSPDP, 0xF0000724) \
/* VSX Vector Convert Single-Precision to Signed Fixed-Point Doubleword */ \
/* Saturate */ \
V(xvcvspsxds, XVCVSPSXDS, 0xF0000660) \
/* VSX Vector Convert Single-Precision to Signed Fixed-Point Word */ \
/* Saturate */ \
V(xvcvspsxws, XVCVSPSXWS, 0xF0000260) \
/* VSX Vector Convert Single-Precision to Unsigned Fixed-Point */ \
/* Doubleword Saturate */ \
V(xvcvspuxds, XVCVSPUXDS, 0xF0000620) \
/* VSX Vector Convert Single-Precision to Unsigned Fixed-Point Word */ \
/* Saturate */ \
V(xvcvspuxws, XVCVSPUXWS, 0xF0000220) \
/* VSX Vector Convert Signed Fixed-Point Doubleword to Double-Precision */ \
V(xvcvsxddp, XVCVSXDDP, 0xF00007E0) \
/* VSX Vector Convert Signed Fixed-Point Doubleword to Single-Precision */ \
V(xvcvsxdsp, XVCVSXDSP, 0xF00006E0) \
/* VSX Vector Convert Signed Fixed-Point Word to Double-Precision */ \
V(xvcvsxwdp, XVCVSXWDP, 0xF00003E0) \
/* VSX Vector Convert Signed Fixed-Point Word to Single-Precision */ \
V(xvcvsxwsp, XVCVSXWSP, 0xF00002E0) \
/* VSX Vector Convert Unsigned Fixed-Point Doubleword to Double- */ \
/* Precision */ \
V(xvcvuxddp, XVCVUXDDP, 0xF00007A0) \
/* VSX Vector Convert Unsigned Fixed-Point Doubleword to Single- */ \
/* Precision */ \
V(xvcvuxdsp, XVCVUXDSP, 0xF00006A0) \
/* VSX Vector Convert Unsigned Fixed-Point Word to Double-Precision */ \
V(xvcvuxwdp, XVCVUXWDP, 0xF00003A0) \
/* VSX Vector Convert Unsigned Fixed-Point Word to Single-Precision */ \
V(xvcvuxwsp, XVCVUXWSP, 0xF00002A0) \
/* VSX Vector Negative Absolute Value Double-Precision */ \
V(xvnabsdp, XVNABSDP, 0xF00007A4) \
/* VSX Vector Negative Absolute Value Single-Precision */ \
V(xvnabssp, XVNABSSP, 0xF00006A4) \
/* VSX Vector Negate Double-Precision */ \
V(xvnegdp, XVNEGDP, 0xF00007E4) \
/* VSX Vector Negate Single-Precision */ \
V(xvnegsp, XVNEGSP, 0xF00006E4) \
/* VSX Vector Round to Double-Precision Integer */ \
V(xvrdpi, XVRDPI, 0xF0000324) \
/* VSX Vector Round to Double-Precision Integer using Current rounding */ \
/* mode */ \
V(xvrdpic, XVRDPIC, 0xF00003AC) \
/* VSX Vector Round to Double-Precision Integer toward -Infinity */ \
V(xvrdpim, XVRDPIM, 0xF00003E4) \
/* VSX Vector Round to Double-Precision Integer toward +Infinity */ \
V(xvrdpip, XVRDPIP, 0xF00003A4) \
/* VSX Vector Round to Double-Precision Integer toward Zero */ \
V(xvrdpiz, XVRDPIZ, 0xF0000364) \
/* VSX Vector Round to Single-Precision Integer */ \
V(xvrspi, XVRSPI, 0xF0000224) \
/* VSX Vector Round to Single-Precision Integer using Current rounding */ \
/* mode */ \
V(xvrspic, XVRSPIC, 0xF00002AC) \
/* VSX Vector Round to Single-Precision Integer toward -Infinity */ \
V(xvrspim, XVRSPIM, 0xF00002E4) \
/* VSX Vector Round to Single-Precision Integer toward +Infinity */ \
V(xvrspip, XVRSPIP, 0xF00002A4) \
/* VSX Vector Round to Single-Precision Integer toward Zero */ \
V(xvrspiz, XVRSPIZ, 0xF0000264) \
/* VSX Vector Reciprocal Square Root Estimate Double-Precision */ \
V(xvrsqrtedp, XVRSQRTEDP, 0xF0000328) \
/* VSX Vector Reciprocal Square Root Estimate Single-Precision */ \
V(xvrsqrtesp, XVRSQRTESP, 0xF0000228) \
/* VSX Vector Square Root Double-Precision */ \
V(xvsqrtdp, XVSQRTDP, 0xF000032C) \
/* VSX Vector Square Root Single-Precision */ \
V(xvsqrtsp, XVSQRTSP, 0xF000022C) \
/* VSX Vector Test for software Square Root Double-Precision */ \
V(xvtsqrtdp, XVTSQRTDP, 0xF00003A8) \
/* VSX Vector Test for software Square Root Single-Precision */ \
V(xvtsqrtsp, XVTSQRTSP, 0xF00002A8)
#define PPC_EVX_OPCODE_LIST(V) \
/* Vector Load Double Word into Double Word by External PID Indexed */ \
V(evlddepx, EVLDDEPX, 0x7C00063E) \
/* Vector Store Double of Double by External PID Indexed */ \
V(evstddepx, EVSTDDEPX, 0x7C00073E) \
/* Bit Reversed Increment */ \
V(brinc, BRINC, 0x1000020F) \
/* Vector Absolute Value */ \
V(evabs, EVABS, 0x10000208) \
/* Vector Add Immediate Word */ \
V(evaddiw, EVADDIW, 0x10000202) \
/* Vector Add Signed, Modulo, Integer to Accumulator Word */ \
V(evaddsmiaaw, EVADDSMIAAW, 0x100004C9) \
/* Vector Add Signed, Saturate, Integer to Accumulator Word */ \
V(evaddssiaaw, EVADDSSIAAW, 0x100004C1) \
/* Vector Add Unsigned, Modulo, Integer to Accumulator Word */ \
V(evaddumiaaw, EVADDUMIAAW, 0x100004C8) \
/* Vector Add Unsigned, Saturate, Integer to Accumulator Word */ \
V(evaddusiaaw, EVADDUSIAAW, 0x100004C0) \
/* Vector Add Word */ \
V(evaddw, EVADDW, 0x10000200) \
/* Vector AND */ \
V(evand, EVAND, 0x10000211) \
/* Vector AND with Complement */ \
V(evandc, EVANDC, 0x10000212) \
/* Vector Compare Equal */ \
V(evcmpeq, EVCMPEQ, 0x10000234) \
/* Vector Compare Greater Than Signed */ \
V(evcmpgts, EVCMPGTS, 0x10000231) \
/* Vector Compare Greater Than Unsigned */ \
V(evcmpgtu, EVCMPGTU, 0x10000230) \
/* Vector Compare Less Than Signed */ \
V(evcmplts, EVCMPLTS, 0x10000233) \
/* Vector Compare Less Than Unsigned */ \
V(evcmpltu, EVCMPLTU, 0x10000232) \
/* Vector Count Leading Signed Bits Word */ \
V(evcntlsw, EVCNTLSW, 0x1000020E) \
/* Vector Count Leading Zeros Word */ \
V(evcntlzw, EVCNTLZW, 0x1000020D) \
/* Vector Divide Word Signed */ \
V(evdivws, EVDIVWS, 0x100004C6) \
/* Vector Divide Word Unsigned */ \
V(evdivwu, EVDIVWU, 0x100004C7) \
/* Vector Equivalent */ \
V(eveqv, EVEQV, 0x10000219) \
/* Vector Extend Sign Byte */ \
V(evextsb, EVEXTSB, 0x1000020A) \
/* Vector Extend Sign Half Word */ \
V(evextsh, EVEXTSH, 0x1000020B) \
/* Vector Load Double Word into Double Word */ \
V(evldd, EVLDD, 0x10000301) \
/* Vector Load Double Word into Double Word Indexed */ \
V(evlddx, EVLDDX, 0x10000300) \
/* Vector Load Double into Four Half Words */ \
V(evldh, EVLDH, 0x10000305) \
/* Vector Load Double into Four Half Words Indexed */ \
V(evldhx, EVLDHX, 0x10000304) \
/* Vector Load Double into Two Words */ \
V(evldw, EVLDW, 0x10000303) \
/* Vector Load Double into Two Words Indexed */ \
V(evldwx, EVLDWX, 0x10000302) \
/* Vector Load Half Word into Half Words Even and Splat */ \
V(evlhhesplat, EVLHHESPLAT, 0x10000309) \
/* Vector Load Half Word into Half Words Even and Splat Indexed */ \
V(evlhhesplatx, EVLHHESPLATX, 0x10000308) \
/* Vector Load Half Word into Half Word Odd Signed and Splat */ \
V(evlhhossplat, EVLHHOSSPLAT, 0x1000030F) \
/* Vector Load Half Word into Half Word Odd Signed and Splat Indexed */ \
V(evlhhossplatx, EVLHHOSSPLATX, 0x1000030E) \
/* Vector Load Half Word into Half Word Odd Unsigned and Splat */ \
V(evlhhousplat, EVLHHOUSPLAT, 0x1000030D) \
/* Vector Load Half Word into Half Word Odd Unsigned and Splat Indexed */ \
V(evlhhousplatx, EVLHHOUSPLATX, 0x1000030C) \
/* Vector Load Word into Two Half Words Even */ \
V(evlwhe, EVLWHE, 0x10000311) \
/* Vector Load Word into Two Half Words Even Indexed */ \
V(evlwhex, EVLWHEX, 0x10000310) \
/* Vector Load Word into Two Half Words Odd Signed (with sign extension) */ \
V(evlwhos, EVLWHOS, 0x10000317) \
/* Vector Load Word into Two Half Words Odd Signed Indexed (with sign */ \
/* extension) */ \
V(evlwhosx, EVLWHOSX, 0x10000316) \
/* Vector Load Word into Two Half Words Odd Unsigned (zero-extended) */ \
V(evlwhou, EVLWHOU, 0x10000315) \
/* Vector Load Word into Two Half Words Odd Unsigned Indexed (zero- */ \
/* extended) */ \
V(evlwhoux, EVLWHOUX, 0x10000314) \
/* Vector Load Word into Two Half Words and Splat */ \
V(evlwhsplat, EVLWHSPLAT, 0x1000031D) \
/* Vector Load Word into Two Half Words and Splat Indexed */ \
V(evlwhsplatx, EVLWHSPLATX, 0x1000031C) \
/* Vector Load Word into Word and Splat */ \
V(evlwwsplat, EVLWWSPLAT, 0x10000319) \
/* Vector Load Word into Word and Splat Indexed */ \
V(evlwwsplatx, EVLWWSPLATX, 0x10000318) \
/* Vector Merge High */ \
V(evmergehi, EVMERGEHI, 0x1000022C) \
/* Vector Merge High/Low */ \
V(evmergehilo, EVMERGEHILO, 0x1000022E) \
/* Vector Merge Low */ \
V(evmergelo, EVMERGELO, 0x1000022D) \
/* Vector Merge Low/High */ \
V(evmergelohi, EVMERGELOHI, 0x1000022F) \
/* Vector Multiply Half Words, Even, Guarded, Signed, Modulo, Fractional */ \
/* and Accumulate */ \
V(evmhegsmfaa, EVMHEGSMFAA, 0x1000052B) \
/* Vector Multiply Half Words, Even, Guarded, Signed, Modulo, Fractional */ \
/* and Accumulate Negative */ \
V(evmhegsmfan, EVMHEGSMFAN, 0x100005AB) \
/* Vector Multiply Half Words, Even, Guarded, Signed, Modulo, Integer */ \
/* and Accumulate */ \
V(evmhegsmiaa, EVMHEGSMIAA, 0x10000529) \
/* Vector Multiply Half Words, Even, Guarded, Signed, Modulo, Integer */ \
/* and Accumulate Negative */ \
V(evmhegsmian, EVMHEGSMIAN, 0x100005A9) \
/* Vector Multiply Half Words, Even, Guarded, Unsigned, Modulo, Integer */ \
/* and Accumulate */ \
V(evmhegumiaa, EVMHEGUMIAA, 0x10000528) \
/* Vector Multiply Half Words, Even, Guarded, Unsigned, Modulo, Integer */ \
/* and Accumulate Negative */ \
V(evmhegumian, EVMHEGUMIAN, 0x100005A8) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Fractional */ \
V(evmhesmf, EVMHESMF, 0x1000040B) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Fractional to */ \
/* Accumulator */ \
V(evmhesmfa, EVMHESMFA, 0x1000042B) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Fractional and */ \
/* Accumulate into Words */ \
V(evmhesmfaaw, EVMHESMFAAW, 0x1000050B) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Fractional and */ \
/* Accumulate Negative into Words */ \
V(evmhesmfanw, EVMHESMFANW, 0x1000058B) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Integer */ \
V(evmhesmi, EVMHESMI, 0x10000409) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Integer to */ \
/* Accumulator */ \
V(evmhesmia, EVMHESMIA, 0x10000429) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Integer and */ \
/* Accumulate into Words */ \
V(evmhesmiaaw, EVMHESMIAAW, 0x10000509) \
/* Vector Multiply Half Words, Even, Signed, Modulo, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmhesmianw, EVMHESMIANW, 0x10000589) \
/* Vector Multiply Half Words, Even, Signed, Saturate, Fractional */ \
V(evmhessf, EVMHESSF, 0x10000403) \
/* Vector Multiply Half Words, Even, Signed, Saturate, Fractional to */ \
/* Accumulator */ \
V(evmhessfa, EVMHESSFA, 0x10000423) \
/* Vector Multiply Half Words, Even, Signed, Saturate, Fractional and */ \
/* Accumulate into Words */ \
V(evmhessfaaw, EVMHESSFAAW, 0x10000503) \
/* Vector Multiply Half Words, Even, Signed, Saturate, Fractional and */ \
/* Accumulate Negative into Words */ \
V(evmhessfanw, EVMHESSFANW, 0x10000583) \
/* Vector Multiply Half Words, Even, Signed, Saturate, Integer and */ \
/* Accumulate into Words */ \
V(evmhessiaaw, EVMHESSIAAW, 0x10000501) \
/* Vector Multiply Half Words, Even, Signed, Saturate, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmhessianw, EVMHESSIANW, 0x10000581) \
/* Vector Multiply Half Words, Even, Unsigned, Modulo, Integer */ \
V(evmheumi, EVMHEUMI, 0x10000408) \
/* Vector Multiply Half Words, Even, Unsigned, Modulo, Integer to */ \
/* Accumulator */ \
V(evmheumia, EVMHEUMIA, 0x10000428) \
/* Vector Multiply Half Words, Even, Unsigned, Modulo, Integer and */ \
/* Accumulate into Words */ \
V(evmheumiaaw, EVMHEUMIAAW, 0x10000508) \
/* Vector Multiply Half Words, Even, Unsigned, Modulo, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmheumianw, EVMHEUMIANW, 0x10000588) \
/* Vector Multiply Half Words, Even, Unsigned, Saturate, Integer and */ \
/* Accumulate into Words */ \
V(evmheusiaaw, EVMHEUSIAAW, 0x10000500) \
/* Vector Multiply Half Words, Even, Unsigned, Saturate, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmheusianw, EVMHEUSIANW, 0x10000580) \
/* Vector Multiply Half Words, Odd, Guarded, Signed, Modulo, Fractional */ \
/* and Accumulate */ \
V(evmhogsmfaa, EVMHOGSMFAA, 0x1000052F) \
/* Vector Multiply Half Words, Odd, Guarded, Signed, Modulo, Fractional */ \
/* and Accumulate Negative */ \
V(evmhogsmfan, EVMHOGSMFAN, 0x100005AF) \
/* Vector Multiply Half Words, Odd, Guarded, Signed, Modulo, Integer, */ \
/* and Accumulate */ \
V(evmhogsmiaa, EVMHOGSMIAA, 0x1000052D) \
/* Vector Multiply Half Words, Odd, Guarded, Signed, Modulo, Integer and */ \
/* Accumulate Negative */ \
V(evmhogsmian, EVMHOGSMIAN, 0x100005AD) \
/* Vector Multiply Half Words, Odd, Guarded, Unsigned, Modulo, Integer */ \
/* and Accumulate */ \
V(evmhogumiaa, EVMHOGUMIAA, 0x1000052C) \
/* Vector Multiply Half Words, Odd, Guarded, Unsigned, Modulo, Integer */ \
/* and Accumulate Negative */ \
V(evmhogumian, EVMHOGUMIAN, 0x100005AC) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Fractional */ \
V(evmhosmf, EVMHOSMF, 0x1000040F) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Fractional to */ \
/* Accumulator */ \
V(evmhosmfa, EVMHOSMFA, 0x1000042F) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Fractional and */ \
/* Accumulate into Words */ \
V(evmhosmfaaw, EVMHOSMFAAW, 0x1000050F) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Fractional and */ \
/* Accumulate Negative into Words */ \
V(evmhosmfanw, EVMHOSMFANW, 0x1000058F) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Integer */ \
V(evmhosmi, EVMHOSMI, 0x1000040D) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Integer to */ \
/* Accumulator */ \
V(evmhosmia, EVMHOSMIA, 0x1000042D) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Integer and */ \
/* Accumulate into Words */ \
V(evmhosmiaaw, EVMHOSMIAAW, 0x1000050D) \
/* Vector Multiply Half Words, Odd, Signed, Modulo, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmhosmianw, EVMHOSMIANW, 0x1000058D) \
/* Vector Multiply Half Words, Odd, Signed, Saturate, Fractional */ \
V(evmhossf, EVMHOSSF, 0x10000407) \
/* Vector Multiply Half Words, Odd, Signed, Saturate, Fractional to */ \
/* Accumulator */ \
V(evmhossfa, EVMHOSSFA, 0x10000427) \
/* Vector Multiply Half Words, Odd, Signed, Saturate, Fractional and */ \
/* Accumulate into Words */ \
V(evmhossfaaw, EVMHOSSFAAW, 0x10000507) \
/* Vector Multiply Half Words, Odd, Signed, Saturate, Fractional and */ \
/* Accumulate Negative into Words */ \
V(evmhossfanw, EVMHOSSFANW, 0x10000587) \
/* Vector Multiply Half Words, Odd, Signed, Saturate, Integer and */ \
/* Accumulate into Words */ \
V(evmhossiaaw, EVMHOSSIAAW, 0x10000505) \
/* Vector Multiply Half Words, Odd, Signed, Saturate, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmhossianw, EVMHOSSIANW, 0x10000585) \
/* Vector Multiply Half Words, Odd, Unsigned, Modulo, Integer */ \
V(evmhoumi, EVMHOUMI, 0x1000040C) \
/* Vector Multiply Half Words, Odd, Unsigned, Modulo, Integer to */ \
/* Accumulator */ \
V(evmhoumia, EVMHOUMIA, 0x1000042C) \
/* Vector Multiply Half Words, Odd, Unsigned, Modulo, Integer and */ \
/* Accumulate into Words */ \
V(evmhoumiaaw, EVMHOUMIAAW, 0x1000050C) \
/* Vector Multiply Half Words, Odd, Unsigned, Modulo, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmhoumianw, EVMHOUMIANW, 0x1000058C) \
/* Vector Multiply Half Words, Odd, Unsigned, Saturate, Integer and */ \
/* Accumulate into Words */ \
V(evmhousiaaw, EVMHOUSIAAW, 0x10000504) \
/* Vector Multiply Half Words, Odd, Unsigned, Saturate, Integer and */ \
/* Accumulate Negative into Words */ \
V(evmhousianw, EVMHOUSIANW, 0x10000584) \
/* Initialize Accumulator */ \
V(evmra, EVMRA, 0x100004C4) \
/* Vector Multiply Word High Signed, Modulo, Fractional */ \
V(evmwhsmf, EVMWHSMF, 0x1000044F) \
/* Vector Multiply Word High Signed, Modulo, Fractional to Accumulator */ \
V(evmwhsmfa, EVMWHSMFA, 0x1000046F) \
/* Vector Multiply Word High Signed, Modulo, Integer */ \
V(evmwhsmi, EVMWHSMI, 0x1000044D) \
/* Vector Multiply Word High Signed, Modulo, Integer to Accumulator */ \
V(evmwhsmia, EVMWHSMIA, 0x1000046D) \
/* Vector Multiply Word High Signed, Saturate, Fractional */ \
V(evmwhssf, EVMWHSSF, 0x10000447) \
/* Vector Multiply Word High Signed, Saturate, Fractional to Accumulator */ \
V(evmwhssfa, EVMWHSSFA, 0x10000467) \
/* Vector Multiply Word High Unsigned, Modulo, Integer */ \
V(evmwhumi, EVMWHUMI, 0x1000044C) \
/* Vector Multiply Word High Unsigned, Modulo, Integer to Accumulator */ \
V(evmwhumia, EVMWHUMIA, 0x1000046C) \
/* Vector Multiply Word Low Signed, Modulo, Integer and Accumulate in */ \
/* Words */ \
V(evmwlsmiaaw, EVMWLSMIAAW, 0x10000549) \
/* Vector Multiply Word Low Signed, Modulo, Integer and Accumulate */ \
/* Negative in Words */ \
V(evmwlsmianw, EVMWLSMIANW, 0x100005C9) \
/* Vector Multiply Word Low Signed, Saturate, Integer and Accumulate in */ \
/* Words */ \
V(evmwlssiaaw, EVMWLSSIAAW, 0x10000541) \
/* Vector Multiply Word Low Signed, Saturate, Integer and Accumulate */ \
/* Negative in Words */ \
V(evmwlssianw, EVMWLSSIANW, 0x100005C1) \
/* Vector Multiply Word Low Unsigned, Modulo, Integer */ \
V(evmwlumi, EVMWLUMI, 0x10000448) \
/* Vector Multiply Word Low Unsigned, Modulo, Integer to Accumulator */ \
V(evmwlumia, EVMWLUMIA, 0x10000468) \
/* Vector Multiply Word Low Unsigned, Modulo, Integer and Accumulate in */ \
/* Words */ \
V(evmwlumiaaw, EVMWLUMIAAW, 0x10000548) \
/* Vector Multiply Word Low Unsigned, Modulo, Integer and Accumulate */ \
/* Negative in Words */ \
V(evmwlumianw, EVMWLUMIANW, 0x100005C8) \
/* Vector Multiply Word Low Unsigned, Saturate, Integer and Accumulate */ \
/* in Words */ \
V(evmwlusiaaw, EVMWLUSIAAW, 0x10000540) \
/* Vector Multiply Word Low Unsigned, Saturate, Integer and Accumulate */ \
/* Negative in Words */ \
V(evmwlusianw, EVMWLUSIANW, 0x100005C0) \
/* Vector Multiply Word Signed, Modulo, Fractional */ \
V(evmwsmf, EVMWSMF, 0x1000045B) \
/* Vector Multiply Word Signed, Modulo, Fractional to Accumulator */ \
V(evmwsmfa, EVMWSMFA, 0x1000047B) \
/* Vector Multiply Word Signed, Modulo, Fractional and Accumulate */ \
V(evmwsmfaa, EVMWSMFAA, 0x1000055B) \
/* Vector Multiply Word Signed, Modulo, Fractional and Accumulate */ \
/* Negative */ \
V(evmwsmfan, EVMWSMFAN, 0x100005DB) \
/* Vector Multiply Word Signed, Modulo, Integer */ \
V(evmwsmi, EVMWSMI, 0x10000459) \
/* Vector Multiply Word Signed, Modulo, Integer to Accumulator */ \
V(evmwsmia, EVMWSMIA, 0x10000479) \
/* Vector Multiply Word Signed, Modulo, Integer and Accumulate */ \
V(evmwsmiaa, EVMWSMIAA, 0x10000559) \
/* Vector Multiply Word Signed, Modulo, Integer and Accumulate Negative */ \
V(evmwsmian, EVMWSMIAN, 0x100005D9) \
/* Vector Multiply Word Signed, Saturate, Fractional */ \
V(evmwssf, EVMWSSF, 0x10000453) \
/* Vector Multiply Word Signed, Saturate, Fractional to Accumulator */ \
V(evmwssfa, EVMWSSFA, 0x10000473) \
/* Vector Multiply Word Signed, Saturate, Fractional and Accumulate */ \
V(evmwssfaa, EVMWSSFAA, 0x10000553) \
/* Vector Multiply Word Signed, Saturate, Fractional and Accumulate */ \
/* Negative */ \
V(evmwssfan, EVMWSSFAN, 0x100005D3) \
/* Vector Multiply Word Unsigned, Modulo, Integer */ \
V(evmwumi, EVMWUMI, 0x10000458) \
/* Vector Multiply Word Unsigned, Modulo, Integer to Accumulator */ \
V(evmwumia, EVMWUMIA, 0x10000478) \
/* Vector Multiply Word Unsigned, Modulo, Integer and Accumulate */ \
V(evmwumiaa, EVMWUMIAA, 0x10000558) \
/* Vector Multiply Word Unsigned, Modulo, Integer and Accumulate */ \
/* Negative */ \
V(evmwumian, EVMWUMIAN, 0x100005D8) \
/* Vector NAND */ \
V(evnand, EVNAND, 0x1000021E) \
/* Vector Negate */ \
V(evneg, EVNEG, 0x10000209) \
/* Vector NOR */ \
V(evnor, EVNOR, 0x10000218) \
/* Vector OR */ \
V(evor, EVOR, 0x10000217) \
/* Vector OR with Complement */ \
V(evorc, EVORC, 0x1000021B) \
/* Vector Rotate Left Word */ \
V(evrlw, EVRLW, 0x10000228) \
/* Vector Rotate Left Word Immediate */ \
V(evrlwi, EVRLWI, 0x1000022A) \
/* Vector Round Word */ \
V(evrndw, EVRNDW, 0x1000020C) \
/* Vector Shift Left Word */ \
V(evslw, EVSLW, 0x10000224) \
/* Vector Shift Left Word Immediate */ \
V(evslwi, EVSLWI, 0x10000226) \
/* Vector Splat Fractional Immediate */ \
V(evsplatfi, EVSPLATFI, 0x1000022B) \
/* Vector Splat Immediate */ \
V(evsplati, EVSPLATI, 0x10000229) \
/* Vector Shift Right Word Immediate Signed */ \
V(evsrwis, EVSRWIS, 0x10000223) \
/* Vector Shift Right Word Immediate Unsigned */ \
V(evsrwiu, EVSRWIU, 0x10000222) \
/* Vector Shift Right Word Signed */ \
V(evsrws, EVSRWS, 0x10000221) \
/* Vector Shift Right Word Unsigned */ \
V(evsrwu, EVSRWU, 0x10000220) \
/* Vector Store Double of Double */ \
V(evstdd, EVSTDD, 0x10000321) \
/* Vector Store Double of Double Indexed */ \
V(evstddx, EVSTDDX, 0x10000320) \
/* Vector Store Double of Four Half Words */ \
V(evstdh, EVSTDH, 0x10000325) \
/* Vector Store Double of Four Half Words Indexed */ \
V(evstdhx, EVSTDHX, 0x10000324) \
/* Vector Store Double of Two Words */ \
V(evstdw, EVSTDW, 0x10000323) \
/* Vector Store Double of Two Words Indexed */ \
V(evstdwx, EVSTDWX, 0x10000322) \
/* Vector Store Word of Two Half Words from Even */ \
V(evstwhe, EVSTWHE, 0x10000331) \
/* Vector Store Word of Two Half Words from Even Indexed */ \
V(evstwhex, EVSTWHEX, 0x10000330) \
/* Vector Store Word of Two Half Words from Odd */ \
V(evstwho, EVSTWHO, 0x10000335) \
/* Vector Store Word of Two Half Words from Odd Indexed */ \
V(evstwhox, EVSTWHOX, 0x10000334) \
/* Vector Store Word of Word from Even */ \
V(evstwwe, EVSTWWE, 0x10000339) \
/* Vector Store Word of Word from Even Indexed */ \
V(evstwwex, EVSTWWEX, 0x10000338) \
/* Vector Store Word of Word from Odd */ \
V(evstwwo, EVSTWWO, 0x1000033D) \
/* Vector Store Word of Word from Odd Indexed */ \
V(evstwwox, EVSTWWOX, 0x1000033C) \
/* Vector Subtract Signed, Modulo, Integer to Accumulator Word */ \
V(evsubfsmiaaw, EVSUBFSMIAAW, 0x100004CB) \
/* Vector Subtract Signed, Saturate, Integer to Accumulator Word */ \
V(evsubfssiaaw, EVSUBFSSIAAW, 0x100004C3) \
/* Vector Subtract Unsigned, Modulo, Integer to Accumulator Word */ \
V(evsubfumiaaw, EVSUBFUMIAAW, 0x100004CA) \
/* Vector Subtract Unsigned, Saturate, Integer to Accumulator Word */ \
V(evsubfusiaaw, EVSUBFUSIAAW, 0x100004C2) \
/* Vector Subtract from Word */ \
V(evsubfw, EVSUBFW, 0x10000204) \
/* Vector Subtract Immediate from Word */ \
V(evsubifw, EVSUBIFW, 0x10000206) \
/* Vector XOR */ \
V(evxor, EVXOR, 0x10000216) \
/* Floating-Point Double-Precision Absolute Value */ \
V(efdabs, EFDABS, 0x100002E4) \
/* Floating-Point Double-Precision Add */ \
V(efdadd, EFDADD, 0x100002E0) \
/* Floating-Point Double-Precision Convert from Single-Precision */ \
V(efdcfs, EFDCFS, 0x100002EF) \
/* Convert Floating-Point Double-Precision from Signed Fraction */ \
V(efdcfsf, EFDCFSF, 0x100002F3) \
/* Convert Floating-Point Double-Precision from Signed Integer */ \
V(efdcfsi, EFDCFSI, 0x100002F1) \
/* Convert Floating-Point Double-Precision from Signed Integer */ \
/* Doubleword */ \
V(efdcfsid, EFDCFSID, 0x100002E3) \
/* Convert Floating-Point Double-Precision from Unsigned Fraction */ \
V(efdcfuf, EFDCFUF, 0x100002F2) \
/* Convert Floating-Point Double-Precision from Unsigned Integer */ \
V(efdcfui, EFDCFUI, 0x100002F0) \
/* Convert Floating-Point Double-Precision fromUnsigned Integer */ \
/* Doubleword */ \
V(efdcfuid, EFDCFUID, 0x100002E2) \
/* Floating-Point Double-Precision Compare Equal */ \
V(efdcmpeq, EFDCMPEQ, 0x100002EE) \
/* Floating-Point Double-Precision Compare Greater Than */ \
V(efdcmpgt, EFDCMPGT, 0x100002EC) \
/* Floating-Point Double-Precision Compare Less Than */ \
V(efdcmplt, EFDCMPLT, 0x100002ED) \
/* Convert Floating-Point Double-Precision to Signed Fraction */ \
V(efdctsf, EFDCTSF, 0x100002F7) \
/* Convert Floating-Point Double-Precision to Signed Integer */ \
V(efdctsi, EFDCTSI, 0x100002F5) \
/* Convert Floating-Point Double-Precision to Signed Integer Doubleword */ \
/* with Round toward Zero */ \
V(efdctsidz, EFDCTSIDZ, 0x100002EB) \
/* Convert Floating-Point Double-Precision to Signed Integer with Round */ \
/* toward Zero */ \
V(efdctsiz, EFDCTSIZ, 0x100002FA) \
/* Convert Floating-Point Double-Precision to Unsigned Fraction */ \
V(efdctuf, EFDCTUF, 0x100002F6) \
/* Convert Floating-Point Double-Precision to Unsigned Integer */ \
V(efdctui, EFDCTUI, 0x100002F4) \
/* Convert Floating-Point Double-Precision to Unsigned Integer */ \
/* Doubleword with Round toward Zero */ \
V(efdctuidz, EFDCTUIDZ, 0x100002EA) \
/* Convert Floating-Point Double-Precision to Unsigned Integer with */ \
/* Round toward Zero */ \
V(efdctuiz, EFDCTUIZ, 0x100002F8) \
/* Floating-Point Double-Precision Divide */ \
V(efddiv, EFDDIV, 0x100002E9) \
/* Floating-Point Double-Precision Multiply */ \
V(efdmul, EFDMUL, 0x100002E8) \
/* Floating-Point Double-Precision Negative Absolute Value */ \
V(efdnabs, EFDNABS, 0x100002E5) \
/* Floating-Point Double-Precision Negate */ \
V(efdneg, EFDNEG, 0x100002E6) \
/* Floating-Point Double-Precision Subtract */ \
V(efdsub, EFDSUB, 0x100002E1) \
/* Floating-Point Double-Precision Test Equal */ \
V(efdtsteq, EFDTSTEQ, 0x100002FE) \
/* Floating-Point Double-Precision Test Greater Than */ \
V(efdtstgt, EFDTSTGT, 0x100002FC) \
/* Floating-Point Double-Precision Test Less Than */ \
V(efdtstlt, EFDTSTLT, 0x100002FD) \
/* Floating-Point Single-Precision Convert from Double-Precision */ \
V(efscfd, EFSCFD, 0x100002CF) \
/* Floating-Point Absolute Value */ \
V(efsabs, EFSABS, 0x100002C4) \
/* Floating-Point Add */ \
V(efsadd, EFSADD, 0x100002C0) \
/* Convert Floating-Point from Signed Fraction */ \
V(efscfsf, EFSCFSF, 0x100002D3) \
/* Convert Floating-Point from Signed Integer */ \
V(efscfsi, EFSCFSI, 0x100002D1) \
/* Convert Floating-Point from Unsigned Fraction */ \
V(efscfuf, EFSCFUF, 0x100002D2) \
/* Convert Floating-Point from Unsigned Integer */ \
V(efscfui, EFSCFUI, 0x100002D0) \
/* Floating-Point Compare Equal */ \
V(efscmpeq, EFSCMPEQ, 0x100002CE) \
/* Floating-Point Compare Greater Than */ \
V(efscmpgt, EFSCMPGT, 0x100002CC) \
/* Floating-Point Compare Less Than */ \
V(efscmplt, EFSCMPLT, 0x100002CD) \
/* Convert Floating-Point to Signed Fraction */ \
V(efsctsf, EFSCTSF, 0x100002D7) \
/* Convert Floating-Point to Signed Integer */ \
V(efsctsi, EFSCTSI, 0x100002D5) \
/* Convert Floating-Point to Signed Integer with Round toward Zero */ \
V(efsctsiz, EFSCTSIZ, 0x100002DA) \
/* Convert Floating-Point to Unsigned Fraction */ \
V(efsctuf, EFSCTUF, 0x100002D6) \
/* Convert Floating-Point to Unsigned Integer */ \
V(efsctui, EFSCTUI, 0x100002D4) \
/* Convert Floating-Point to Unsigned Integer with Round toward Zero */ \
V(efsctuiz, EFSCTUIZ, 0x100002D8) \
/* Floating-Point Divide */ \
V(efsdiv, EFSDIV, 0x100002C9) \
/* Floating-Point Multiply */ \
V(efsmul, EFSMUL, 0x100002C8) \
/* Floating-Point Negative Absolute Value */ \
V(efsnabs, EFSNABS, 0x100002C5) \
/* Floating-Point Negate */ \
V(efsneg, EFSNEG, 0x100002C6) \
/* Floating-Point Subtract */ \
V(efssub, EFSSUB, 0x100002C1) \
/* Floating-Point Test Equal */ \
V(efststeq, EFSTSTEQ, 0x100002DE) \
/* Floating-Point Test Greater Than */ \
V(efststgt, EFSTSTGT, 0x100002DC) \
/* Floating-Point Test Less Than */ \
V(efststlt, EFSTSTLT, 0x100002DD) \
/* Vector Floating-Point Absolute Value */ \
V(evfsabs, EVFSABS, 0x10000284) \
/* Vector Floating-Point Add */ \
V(evfsadd, EVFSADD, 0x10000280) \
/* Vector Convert Floating-Point from Signed Fraction */ \
V(evfscfsf, EVFSCFSF, 0x10000293) \
/* Vector Convert Floating-Point from Signed Integer */ \
V(evfscfsi, EVFSCFSI, 0x10000291) \
/* Vector Convert Floating-Point from Unsigned Fraction */ \
V(evfscfuf, EVFSCFUF, 0x10000292) \
/* Vector Convert Floating-Point from Unsigned Integer */ \
V(evfscfui, EVFSCFUI, 0x10000290) \
/* Vector Floating-Point Compare Equal */ \
V(evfscmpeq, EVFSCMPEQ, 0x1000028E) \
/* Vector Floating-Point Compare Greater Than */ \
V(evfscmpgt, EVFSCMPGT, 0x1000028C) \
/* Vector Floating-Point Compare Less Than */ \
V(evfscmplt, EVFSCMPLT, 0x1000028D) \
/* Vector Convert Floating-Point to Signed Fraction */ \
V(evfsctsf, EVFSCTSF, 0x10000297) \
/* Vector Convert Floating-Point to Signed Integer */ \
V(evfsctsi, EVFSCTSI, 0x10000295) \
/* Vector Convert Floating-Point to Signed Integer with Round toward */ \
/* Zero */ \
V(evfsctsiz, EVFSCTSIZ, 0x1000029A) \
/* Vector Convert Floating-Point to Unsigned Fraction */ \
V(evfsctuf, EVFSCTUF, 0x10000296) \
/* Vector Convert Floating-Point to Unsigned Integer */ \
V(evfsctui, EVFSCTUI, 0x10000294) \
/* Vector Convert Floating-Point to Unsigned Integer with Round toward */ \
/* Zero */ \
V(evfsctuiz, EVFSCTUIZ, 0x10000298) \
/* Vector Floating-Point Divide */ \
V(evfsdiv, EVFSDIV, 0x10000289) \
/* Vector Floating-Point Multiply */ \
V(evfsmul, EVFSMUL, 0x10000288) \
/* Vector Floating-Point Negative Absolute Value */ \
V(evfsnabs, EVFSNABS, 0x10000285) \
/* Vector Floating-Point Negate */ \
V(evfsneg, EVFSNEG, 0x10000286) \
/* Vector Floating-Point Subtract */ \
V(evfssub, EVFSSUB, 0x10000281) \
/* Vector Floating-Point Test Equal */ \
V(evfststeq, EVFSTSTEQ, 0x1000029E) \
/* Vector Floating-Point Test Greater Than */ \
V(evfststgt, EVFSTSTGT, 0x1000029C) \
/* Vector Floating-Point Test Less Than */ \
V(evfststlt, EVFSTSTLT, 0x1000029D)
#define PPC_VC_OPCODE_LIST(V) \
/* Vector Compare Bounds Single-Precision */ \
V(vcmpbfp, VCMPBFP, 0x100003C6) \
/* Vector Compare Equal To Single-Precision */ \
V(vcmpeqfp, VCMPEQFP, 0x100000C6) \
/* Vector Compare Equal To Unsigned Byte */ \
V(vcmpequb, VCMPEQUB, 0x10000006) \
/* Vector Compare Equal To Unsigned Doubleword */ \
V(vcmpequd, VCMPEQUD, 0x100000C7) \
/* Vector Compare Equal To Unsigned Halfword */ \
V(vcmpequh, VCMPEQUH, 0x10000046) \
/* Vector Compare Equal To Unsigned Word */ \
V(vcmpequw, VCMPEQUW, 0x10000086) \
/* Vector Compare Greater Than or Equal To Single-Precision */ \
V(vcmpgefp, VCMPGEFP, 0x100001C6) \
/* Vector Compare Greater Than Single-Precision */ \
V(vcmpgtfp, VCMPGTFP, 0x100002C6) \
/* Vector Compare Greater Than Signed Byte */ \
V(vcmpgtsb, VCMPGTSB, 0x10000306) \
/* Vector Compare Greater Than Signed Doubleword */ \
V(vcmpgtsd, VCMPGTSD, 0x100003C7) \
/* Vector Compare Greater Than Signed Halfword */ \
V(vcmpgtsh, VCMPGTSH, 0x10000346) \
/* Vector Compare Greater Than Signed Word */ \
V(vcmpgtsw, VCMPGTSW, 0x10000386) \
/* Vector Compare Greater Than Unsigned Byte */ \
V(vcmpgtub, VCMPGTUB, 0x10000206) \
/* Vector Compare Greater Than Unsigned Doubleword */ \
V(vcmpgtud, VCMPGTUD, 0x100002C7) \
/* Vector Compare Greater Than Unsigned Halfword */ \
V(vcmpgtuh, VCMPGTUH, 0x10000246) \
/* Vector Compare Greater Than Unsigned Word */ \
V(vcmpgtuw, VCMPGTUW, 0x10000286)
#define PPC_X_OPCODE_LIST(V) \
/* Bit Permute Doubleword */ \
V(bpermd, BPERMD, 0x7C0001F8) \
/* Count Leading Zeros Doubleword */ \
V(cntlzd, CNTLZDX, 0x7C000074) \
/* Extend Sign Word */ \
V(extsw, EXTSW, 0x7C0007B4) \
/* Load Doubleword And Reserve Indexed */ \
V(ldarx, LDARX, 0x7C0000A8) \
/* Load Doubleword Byte-Reverse Indexed */ \
V(ldbrx, LDBRX, 0x7C000428) \
/* Load Doubleword with Update Indexed */ \
V(ldux, LDUX, 0x7C00006A) \
/* Load Doubleword Indexed */ \
V(ldx, LDX, 0x7C00002A) \
/* Load Word Algebraic with Update Indexed */ \
V(lwaux, LWAUX, 0x7C0002EA) \
/* Load Word Algebraic Indexed */ \
V(lwax, LWAX, 0x7C0002AA) \
/* Modulo Signed Dword */ \
V(modsd, MODSD, 0x7C000612) \
/* Modulo Unsigned Dword */ \
V(modud, MODUD, 0x7C000212) \
/* Population Count Doubleword */ \
V(popcntd, POPCNTD, 0x7C0003F4) \
/* Parity Doubleword */ \
V(prtyd, PRTYD, 0x7C000174) \
/* Shift Left Doubleword */ \
V(sld, SLDX, 0x7C000036) \
/* Shift Right Algebraic Doubleword */ \
V(srad, SRAD, 0x7C000634) \
/* Shift Right Doubleword */ \
V(srd, SRDX, 0x7C000436) \
/* Store Doubleword Byte-Reverse Indexed */ \
V(stdbrx, STDBRX, 0x7C000528) \
/* Store Doubleword Conditional Indexed & record CR0 */ \
V(stdcx, STDCX, 0x7C0001AD) \
/* Store Doubleword with Update Indexed */ \
V(stdux, STDUX, 0x7C00016A) \
/* Store Doubleword Indexed */ \
V(stdx, STDX, 0x7C00012A) \
/* Trap Doubleword */ \
V(td, TD, 0x7C000088) \
/* AND */ \
V(andx, ANDX, 0x7C000038) \
/* AND with Complement */ \
V(andc, ANDCX, 0x7C000078) \
/* Branch Conditional to Branch Target Address Register */ \
V(bctar, BCTAR, 0x4C000460) \
/* Compare */ \
V(cmp, CMP, 0x7C000000) \
/* Compare Byte */ \
V(cmpb, CMPB, 0x7C0003F8) \
/* Compare Logical */ \
V(cmpl, CMPL, 0x7C000040) \
/* Count Leading Zeros Word */ \
V(cntlzw, CNTLZWX, 0x7C000034) \
/* Data Cache Block Flush */ \
V(dcbf, DCBF, 0x7C0000AC) \
/* Data Cache Block Store */ \
V(dcbst, DCBST, 0x7C00006C) \
/* Data Cache Block Touch */ \
V(dcbt, DCBT, 0x7C00022C) \
/* Data Cache Block Touch for Store */ \
V(dcbtst, DCBTST, 0x7C0001EC) \
/* Data Cache Block Zero */ \
V(dcbz, DCBZ, 0x7C0007EC) \
/* Equivalent */ \
V(eqv, EQV, 0x7C000238) \
/* Extend Sign Byte */ \
V(extsb, EXTSB, 0x7C000774) \
/* Extend Sign Halfword */ \
V(extsh, EXTSH, 0x7C000734) \
/* Instruction Cache Block Invalidate */ \
V(icbi, ICBI, 0x7C0007AC) \
/* Load Byte And Reserve Indexed */ \
V(lbarx, LBARX, 0x7C000068) \
/* Load Byte and Zero with Update Indexed */ \
V(lbzux, LBZUX, 0x7C0000EE) \
/* Load Byte and Zero Indexed */ \
V(lbzx, LBZX, 0x7C0000AE) \
/* Load Halfword And Reserve Indexed Xform */ \
V(lharx, LHARX, 0x7C0000E8) \
/* Load Halfword Algebraic with Update Indexed */ \
V(lhaux, LHAUX, 0x7C0002EE) \
/* Load Halfword Algebraic Indexed */ \
V(lhax, LHAX, 0x7C0002AE) \
/* Load Halfword Byte-Reverse Indexed */ \
V(lhbrx, LHBRX, 0x7C00062C) \
/* Load Halfword and Zero with Update Indexed */ \
V(lhzux, LHZUX, 0x7C00026E) \
/* Load Halfword and Zero Indexed */ \
V(lhzx, LHZX, 0x7C00022E) \
/* Load Word and Reserve Indexed */ \
V(lwarx, LWARX, 0x7C000028) \
/* Load Word Byte-Reverse Indexed */ \
V(lwbrx, LWBRX, 0x7C00042C) \
/* Load Word and Zero with Update Indexed */ \
V(lwzux, LWZUX, 0x7C00006E) \
/* Load Word and Zero Indexed */ \
V(lwzx, LWZX, 0x7C00002E) \
/* Modulo Signed Word */ \
V(mods, MODSW, 0x7C000616) \
/* Modulo Unsigned Word */ \
V(moduw, MODUW, 0x7C000216) \
/* NAND */ \
V(nand, NAND, 0x7C0003B8) \
/* NOR */ \
V(nor, NORX, 0x7C0000F8) \
/* OR */ \
V(orx, ORX, 0x7C000378) \
/* OR with Complement */ \
V(orc, ORC, 0x7C000338) \
/* Population Count Byte-wise */ \
V(popcntb, POPCNTB, 0x7C0000F4) \
/* Population Count Words */ \
V(popcntw, POPCNTW, 0x7C0002F4) \
/* Parity Word */ \
V(prtyw, PRTYW, 0x7C000134) \
/* Shift Left Word */ \
V(slw, SLWX, 0x7C000030) \
/* Shift Right Algebraic Word */ \
V(sraw, SRAW, 0x7C000630) \
/* Shift Right Algebraic Word Immediate */ \
V(srawi, SRAWIX, 0x7C000670) \
/* Shift Right Word */ \
V(srw, SRWX, 0x7C000430) \
/* Store Byte Conditional Indexed */ \
V(stbcx, STBCX, 0x7C00056D) \
/* Store Byte with Update Indexed */ \
V(stbux, STBUX, 0x7C0001EE) \
/* Store Byte Indexed */ \
V(stbx, STBX, 0x7C0001AE) \
/* Store Halfword Byte-Reverse Indexed */ \
V(sthbrx, STHBRX, 0x7C00072C) \
/* Store Halfword Conditional Indexed Xform */ \
V(sthcx, STHCX, 0x7C0005AD) \
/* Store Halfword with Update Indexed */ \
V(sthux, STHUX, 0x7C00036E) \
/* Store Halfword Indexed */ \
V(sthx, STHX, 0x7C00032E) \
/* Store Word Byte-Reverse Indexed */ \
V(stwbrx, STWBRX, 0x7C00052C) \
/* Store Word Conditional Indexed & record CR0 */ \
V(stwcx, STWCX, 0x7C00012D) \
/* Store Word with Update Indexed */ \
V(stwux, STWUX, 0x7C00016E) \
/* Store Word Indexed */ \
V(stwx, STWX, 0x7C00012E) \
/* Synchronize */ \
V(sync, SYNC, 0x7C0004AC) \
/* Trap Word */ \
V(tw, TW, 0x7C000008) \
/* ExecuExecuted No Operation */ \
V(xnop, XNOP, 0x68000000) \
/* XOR */ \
V(xorx, XORX, 0x7C000278) \
/* Convert Binary Coded Decimal To Declets */ \
V(cbcdtd, CBCDTD, 0x7C000274) \
/* Convert Declets To Binary Coded Decimal */ \
V(cdtbcd, CDTBCD, 0x7C000234) \
/* Decimal Floating Add */ \
V(dadd, DADD, 0xEC000004) \
/* Decimal Floating Add Quad */ \
V(daddq, DADDQ, 0xFC000004) \
/* Decimal Floating Convert From Fixed */ \
V(dcffix, DCFFIX, 0xEC000644) \
/* Decimal Floating Convert From Fixed Quad */ \
V(dcffixq, DCFFIXQ, 0xFC000644) \
/* Decimal Floating Compare Ordered */ \
V(dcmpo, DCMPO, 0xEC000104) \
/* Decimal Floating Compare Ordered Quad */ \
V(dcmpoq, DCMPOQ, 0xFC000104) \
/* Decimal Floating Compare Unordered */ \
V(dcmpu, DCMPU, 0xEC000504) \
/* Decimal Floating Compare Unordered Quad */ \
V(dcmpuq, DCMPUQ, 0xFC000504) \
/* Decimal Floating Convert To DFP Long */ \
V(dctdp, DCTDP, 0xEC000204) \
/* Decimal Floating Convert To Fixed */ \
V(dctfix, DCTFIX, 0xEC000244) \
/* Decimal Floating Convert To Fixed Quad */ \
V(dctfixq, DCTFIXQ, 0xFC000244) \
/* Decimal Floating Convert To DFP Extended */ \
V(dctqpq, DCTQPQ, 0xFC000204) \
/* Decimal Floating Decode DPD To BCD */ \
V(ddedpd, DDEDPD, 0xEC000284) \
/* Decimal Floating Decode DPD To BCD Quad */ \
V(ddedpdq, DDEDPDQ, 0xFC000284) \
/* Decimal Floating Divide */ \
V(ddiv, DDIV, 0xEC000444) \
/* Decimal Floating Divide Quad */ \
V(ddivq, DDIVQ, 0xFC000444) \
/* Decimal Floating Encode BCD To DPD */ \
V(denbcd, DENBCD, 0xEC000684) \
/* Decimal Floating Encode BCD To DPD Quad */ \
V(denbcdq, DENBCDQ, 0xFC000684) \
/* Decimal Floating Insert Exponent */ \
V(diex, DIEX, 0xEC0006C4) \
/* Decimal Floating Insert Exponent Quad */ \
V(diexq, DIEXQ, 0xFC0006C4) \
/* Decimal Floating Multiply */ \
V(dmul, DMUL, 0xEC000044) \
/* Decimal Floating Multiply Quad */ \
V(dmulq, DMULQ, 0xFC000044) \
/* Decimal Floating Round To DFP Long */ \
V(drdpq, DRDPQ, 0xFC000604) \
/* Decimal Floating Round To DFP Short */ \
V(drsp, DRSP, 0xEC000604) \
/* Decimal Floating Subtract */ \
V(dsub, DSUB, 0xEC000404) \
/* Decimal Floating Subtract Quad */ \
V(dsubq, DSUBQ, 0xFC000404) \
/* Decimal Floating Test Exponent */ \
V(dtstex, DTSTEX, 0xEC000144) \
/* Decimal Floating Test Exponent Quad */ \
V(dtstexq, DTSTEXQ, 0xFC000144) \
/* Decimal Floating Test Significance */ \
V(dtstsf, DTSTSF, 0xEC000544) \
/* Decimal Floating Test Significance Quad */ \
V(dtstsfq, DTSTSFQ, 0xFC000544) \
/* Decimal Floating Extract Exponent */ \
V(dxex, DXEX, 0xEC0002C4) \
/* Decimal Floating Extract Exponent Quad */ \
V(dxexq, DXEXQ, 0xFC0002C4) \
/* Decorated Storage Notify */ \
V(dsn, DSN, 0x7C0003C6) \
/* Load Byte with Decoration Indexed */ \
V(lbdx, LBDX, 0x7C000406) \
/* Load Doubleword with Decoration Indexed */ \
V(lddx, LDDX, 0x7C0004C6) \
/* Load Floating Doubleword with Decoration Indexed */ \
V(lfddx, LFDDX, 0x7C000646) \
/* Load Halfword with Decoration Indexed */ \
V(lhdx, LHDX, 0x7C000446) \
/* Load Word with Decoration Indexed */ \
V(lwdx, LWDX, 0x7C000486) \
/* Store Byte with Decoration Indexed */ \
V(stbdx, STBDX, 0x7C000506) \
/* Store Doubleword with Decoration Indexed */ \
V(stddx, STDDX, 0x7C0005C6) \
/* Store Floating Doubleword with Decoration Indexed */ \
V(stfddx, STFDDX, 0x7C000746) \
/* Store Halfword with Decoration Indexed */ \
V(sthdx, STHDX, 0x7C000546) \
/* Store Word with Decoration Indexed */ \
V(stwdx, STWDX, 0x7C000586) \
/* Data Cache Block Allocate */ \
V(dcba, DCBA, 0x7C0005EC) \
/* Data Cache Block Invalidate */ \
V(dcbi, DCBI, 0x7C0003AC) \
/* Instruction Cache Block Touch */ \
V(icbt, ICBT, 0x7C00002C) \
/* Memory Barrier */ \
V(mbar, MBAR, 0x7C0006AC) \
/* Move to Condition Register from XER */ \
V(mcrxr, MCRXR, 0x7C000400) \
/* TLB Invalidate Local Indexed */ \
V(tlbilx, TLBILX, 0x7C000024) \
/* TLB Invalidate Virtual Address Indexed */ \
V(tlbivax, TLBIVAX, 0x7C000624) \
/* TLB Read Entry */ \
V(tlbre, TLBRE, 0x7C000764) \
/* TLB Search Indexed */ \
V(tlbsx, TLBSX, 0x7C000724) \
/* TLB Write Entry */ \
V(tlbwe, TLBWE, 0x7C0007A4) \
/* Write External Enable */ \
V(wrtee, WRTEE, 0x7C000106) \
/* Write External Enable Immediate */ \
V(wrteei, WRTEEI, 0x7C000146) \
/* Data Cache Read */ \
V(dcread, DCREAD, 0x7C00028C) \
/* Instruction Cache Read */ \
V(icread, ICREAD, 0x7C0007CC) \
/* Data Cache Invalidate */ \
V(dci, DCI, 0x7C00038C) \
/* Instruction Cache Invalidate */ \
V(ici, ICI, 0x7C00078C) \
/* Move From Device Control Register User Mode Indexed */ \
V(mfdcrux, MFDCRUX, 0x7C000246) \
/* Move From Device Control Register Indexed */ \
V(mfdcrx, MFDCRX, 0x7C000206) \
/* Move To Device Control Register User Mode Indexed */ \
V(mtdcrux, MTDCRUX, 0x7C000346) \
/* Move To Device Control Register Indexed */ \
V(mtdcrx, MTDCRX, 0x7C000306) \
/* Return From Debug Interrupt */ \
V(rfdi, RFDI, 0x4C00004E) \
/* Data Cache Block Flush by External PID */ \
V(dcbfep, DCBFEP, 0x7C0000FE) \
/* Data Cache Block Store by External PID */ \
V(dcbstep, DCBSTEP, 0x7C00007E) \
/* Data Cache Block Touch by External PID */ \
V(dcbtep, DCBTEP, 0x7C00027E) \
/* Data Cache Block Touch for Store by External PID */ \
V(dcbtstep, DCBTSTEP, 0x7C0001FE) \
/* Data Cache Block Zero by External PID */ \
V(dcbzep, DCBZEP, 0x7C0007FE) \
/* Instruction Cache Block Invalidate by External PID */ \
V(icbiep, ICBIEP, 0x7C0007BE) \
/* Load Byte and Zero by External PID Indexed */ \
V(lbepx, LBEPX, 0x7C0000BE) \
/* Load Floating-Point Double by External PID Indexed */ \
V(lfdepx, LFDEPX, 0x7C0004BE) \
/* Load Halfword and Zero by External PID Indexed */ \
V(lhepx, LHEPX, 0x7C00023E) \
/* Load Vector by External PID Indexed */ \
V(lvepx, LVEPX, 0x7C00024E) \
/* Load Vector by External PID Indexed Last */ \
V(lvepxl, LVEPXL, 0x7C00020E) \
/* Load Word and Zero by External PID Indexed */ \
V(lwepx, LWEPX, 0x7C00003E) \
/* Store Byte by External PID Indexed */ \
V(stbepx, STBEPX, 0x7C0001BE) \
/* Store Floating-Point Double by External PID Indexed */ \
V(stfdepx, STFDEPX, 0x7C0005BE) \
/* Store Halfword by External PID Indexed */ \
V(sthepx, STHEPX, 0x7C00033E) \
/* Store Vector by External PID Indexed */ \
V(stvepx, STVEPX, 0x7C00064E) \
/* Store Vector by External PID Indexed Last */ \
V(stvepxl, STVEPXL, 0x7C00060E) \
/* Store Word by External PID Indexed */ \
V(stwepx, STWEPX, 0x7C00013E) \
/* Load Doubleword by External PID Indexed */ \
V(ldepx, LDEPX, 0x7C00003A) \
/* Store Doubleword by External PID Indexed */ \
V(stdepx, STDEPX, 0x7C00013A) \
/* TLB Search and Reserve Indexed */ \
V(tlbsrx, TLBSRX, 0x7C0006A5) \
/* External Control In Word Indexed */ \
V(eciwx, ECIWX, 0x7C00026C) \
/* External Control Out Word Indexed */ \
V(ecowx, ECOWX, 0x7C00036C) \
/* Data Cache Block Lock Clear */ \
V(dcblc, DCBLC, 0x7C00030C) \
/* Data Cache Block Lock Query */ \
V(dcblq, DCBLQ, 0x7C00034D) \
/* Data Cache Block Touch and Lock Set */ \
V(dcbtls, DCBTLS, 0x7C00014C) \
/* Data Cache Block Touch for Store and Lock Set */ \
V(dcbtstls, DCBTSTLS, 0x7C00010C) \
/* Instruction Cache Block Lock Clear */ \
V(icblc, ICBLC, 0x7C0001CC) \
/* Instruction Cache Block Lock Query */ \
V(icblq, ICBLQ, 0x7C00018D) \
/* Instruction Cache Block Touch and Lock Set */ \
V(icbtls, ICBTLS, 0x7C0003CC) \
/* Floating Compare Ordered */ \
V(fcmpo, FCMPO, 0xFC000040) \
/* Floating Compare Unordered */ \
V(fcmpu, FCMPU, 0xFC000000) \
/* Floating Test for software Divide */ \
V(ftdiv, FTDIV, 0xFC000100) \
/* Floating Test for software Square Root */ \
V(ftsqrt, FTSQRT, 0xFC000140) \
/* Load Floating-Point Double with Update Indexed */ \
V(lfdux, LFDUX, 0x7C0004EE) \
/* Load Floating-Point Double Indexed */ \
V(lfdx, LFDX, 0x7C0004AE) \
/* Load Floating-Point as Integer Word Algebraic Indexed */ \
V(lfiwax, LFIWAX, 0x7C0006AE) \
/* Load Floating-Point as Integer Word and Zero Indexed */ \
V(lfiwzx, LFIWZX, 0x7C0006EE) \
/* Load Floating-Point Single with Update Indexed */ \
V(lfsux, LFSUX, 0x7C00046E) \
/* Load Floating-Point Single Indexed */ \
V(lfsx, LFSX, 0x7C00042E) \
/* Move To Condition Register from FPSCR */ \
V(mcrfs, MCRFS, 0xFC000080) \
/* Store Floating-Point Double with Update Indexed */ \
V(stfdux, STFDUX, 0x7C0005EE) \
/* Store Floating-Point Double Indexed */ \
V(stfdx, STFDX, 0x7C0005AE) \
/* Store Floating-Point as Integer Word Indexed */ \
V(stfiwx, STFIWX, 0x7C0007AE) \
/* Store Floating-Point Single with Update Indexed */ \
V(stfsux, STFSUX, 0x7C00056E) \
/* Store Floating-Point Single Indexed */ \
V(stfsx, STFSX, 0x7C00052E) \
/* Load Floating-Point Double Pair Indexed */ \
V(lfdpx, LFDPX, 0x7C00062E) \
/* Store Floating-Point Double Pair Indexed */ \
V(stfdpx, STFDPX, 0x7C00072E) \
/* Floating Absolute Value */ \
V(fabs, FABS, 0xFC000210) \
/* Floating Convert From Integer Doubleword */ \
V(fcfid, FCFID, 0xFC00069C) \
/* Floating Convert From Integer Doubleword Single */ \
V(fcfids, FCFIDS, 0xEC00069C) \
/* Floating Convert From Integer Doubleword Unsigned */ \
V(fcfidu, FCFIDU, 0xFC00079C) \
/* Floating Convert From Integer Doubleword Unsigned Single */ \
V(fcfidus, FCFIDUS, 0xEC00079C) \
/* Floating Copy Sign */ \
V(fcpsgn, FCPSGN, 0xFC000010) \
/* Floating Convert To Integer Doubleword */ \
V(fctid, FCTID, 0xFC00065C) \
/* Floating Convert To Integer Doubleword Unsigned */ \
V(fctidu, FCTIDU, 0xFC00075C) \
/* Floating Convert To Integer Doubleword Unsigned with round toward */ \
/* Zero */ \
V(fctiduz, FCTIDUZ, 0xFC00075E) \
/* Floating Convert To Integer Doubleword with round toward Zero */ \
V(fctidz, FCTIDZ, 0xFC00065E) \
/* Floating Convert To Integer Word */ \
V(fctiw, FCTIW, 0xFC00001C) \
/* Floating Convert To Integer Word Unsigned */ \
V(fctiwu, FCTIWU, 0xFC00011C) \
/* Floating Convert To Integer Word Unsigned with round toward Zero */ \
V(fctiwuz, FCTIWUZ, 0xFC00011E) \
/* Floating Convert To Integer Word with round to Zero */ \
V(fctiwz, FCTIWZ, 0xFC00001E) \
/* Floating Move Register */ \
V(fmr, FMR, 0xFC000090) \
/* Floating Negative Absolute Value */ \
V(fnabs, FNABS, 0xFC000110) \
/* Floating Negate */ \
V(fneg, FNEG, 0xFC000050) \
/* Floating Round to Single-Precision */ \
V(frsp, FRSP, 0xFC000018) \
/* Move From FPSCR */ \
V(mffs, MFFS, 0xFC00048E) \
/* Move To FPSCR Bit 0 */ \
V(mtfsb0, MTFSB0, 0xFC00008C) \
/* Move To FPSCR Bit 1 */ \
V(mtfsb1, MTFSB1, 0xFC00004C) \
/* Move To FPSCR Field Immediate */ \
V(mtfsfi, MTFSFI, 0xFC00010C) \
/* Floating Round To Integer Minus */ \
V(frim, FRIM, 0xFC0003D0) \
/* Floating Round To Integer Nearest */ \
V(frin, FRIN, 0xFC000310) \
/* Floating Round To Integer Plus */ \
V(frip, FRIP, 0xFC000390) \
/* Floating Round To Integer toward Zero */ \
V(friz, FRIZ, 0xFC000350) \
/* Multiply Cross Halfword to Word Signed */ \
V(mulchw, MULCHW, 0x10000150) \
/* Multiply Cross Halfword to Word Unsigned */ \
V(mulchwu, MULCHWU, 0x10000110) \
/* Multiply High Halfword to Word Signed */ \
V(mulhhw, MULHHW, 0x10000050) \
/* Multiply High Halfword to Word Unsigned */ \
V(mulhhwu, MULHHWU, 0x10000010) \
/* Multiply Low Halfword to Word Signed */ \
V(mullhw, MULLHW, 0x10000350) \
/* Multiply Low Halfword to Word Unsigned */ \
V(mullhwu, MULLHWU, 0x10000310) \
/* Determine Leftmost Zero Byte DQ 56 E0000000 P 58 LSQ lq Load Quadword */ \
V(dlmzb, DLMZB, 0x7C00009C) \
/* Load Quadword And Reserve Indexed */ \
V(lqarx, LQARX, 0x7C000228) \
/* Store Quadword Conditional Indexed and record CR0 */ \
V(stqcx, STQCX, 0x7C00016D) \
/* Load String Word Immediate */ \
V(lswi, LSWI, 0x7C0004AA) \
/* Load String Word Indexed */ \
V(lswx, LSWX, 0x7C00042A) \
/* Store String Word Immediate */ \
V(stswi, STSWI, 0x7C0005AA) \
/* Store String Word Indexed */ \
V(stswx, STSWX, 0x7C00052A) \
/* Clear BHRB */ \
V(clrbhrb, CLRBHRB, 0x7C00035C) \
/* Enforce In-order Execution of I/O */ \
V(eieio, EIEIO, 0x7C0006AC) \
/* Load Byte and Zero Caching Inhibited Indexed */ \
V(lbzcix, LBZCIX, 0x7C0006AA) \
/* Load Doubleword Caching Inhibited Indexed */ \
V(ldcix, LDCIX, 0x7C0006EA) \
/* Load Halfword and Zero Caching Inhibited Indexed */ \
V(lhzcix, LHZCIX, 0x7C00066A) \
/* Load Word and Zero Caching Inhibited Indexed */ \
V(lwzcix, LWZCIX, 0x7C00062A) \
/* Move From Segment Register */ \
V(mfsr, MFSR, 0x7C0004A6) \
/* Move From Segment Register Indirect */ \
V(mfsrin, MFSRIN, 0x7C000526) \
/* Move To Machine State Register Doubleword */ \
V(mtmsrd, MTMSRD, 0x7C000164) \
/* Move To Split Little Endian */ \
V(mtsle, MTSLE, 0x7C000126) \
/* Move To Segment Register */ \
V(mtsr, MTSR, 0x7C0001A4) \
/* Move To Segment Register Indirect */ \
V(mtsrin, MTSRIN, 0x7C0001E4) \
/* SLB Find Entry ESID */ \
V(slbfee, SLBFEE, 0x7C0007A7) \
/* SLB Invalidate All */ \
V(slbia, SLBIA, 0x7C0003E4) \
/* SLB Invalidate Entry */ \
V(slbie, SLBIE, 0x7C000364) \
/* SLB Move From Entry ESID */ \
V(slbmfee, SLBMFEE, 0x7C000726) \
/* SLB Move From Entry VSID */ \
V(slbmfev, SLBMFEV, 0x7C0006A6) \
/* SLB Move To Entry */ \
V(slbmte, SLBMTE, 0x7C000324) \
/* Store Byte Caching Inhibited Indexed */ \
V(stbcix, STBCIX, 0x7C0007AA) \
/* Store Doubleword Caching Inhibited Indexed */ \
V(stdcix, STDCIX, 0x7C0007EA) \
/* Store Halfword and Zero Caching Inhibited Indexed */ \
V(sthcix, STHCIX, 0x7C00076A) \
/* Store Word and Zero Caching Inhibited Indexed */ \
V(stwcix, STWCIX, 0x7C00072A) \
/* TLB Invalidate All */ \
V(tlbia, TLBIA, 0x7C0002E4) \
/* TLB Invalidate Entry */ \
V(tlbie, TLBIE, 0x7C000264) \
/* TLB Invalidate Entry Local */ \
V(tlbiel, TLBIEL, 0x7C000224) \
/* Message Clear Privileged */ \
V(msgclrp, MSGCLRP, 0x7C00015C) \
/* Message Send Privileged */ \
V(msgsndp, MSGSNDP, 0x7C00011C) \
/* Message Clear */ \
V(msgclr, MSGCLR, 0x7C0001DC) \
/* Message Send */ \
V(msgsnd, MSGSND, 0x7C00019C) \
/* Move From Machine State Register */ \
V(mfmsr, MFMSR, 0x7C0000A6) \
/* Move To Machine State Register */ \
V(mtmsr, MTMSR, 0x7C000124) \
/* TLB Synchronize */ \
V(tlbsync, TLBSYNC, 0x7C00046C) \
/* Transaction Abort */ \
V(tabort, TABORT, 0x7C00071D) \
/* Transaction Abort Doubleword Conditional */ \
V(tabortdc, TABORTDC, 0x7C00065D) \
/* Transaction Abort Doubleword Conditional Immediate */ \
V(tabortdci, TABORTDCI, 0x7C0006DD) \
/* Transaction Abort Word Conditional */ \
V(tabortwc, TABORTWC, 0x7C00061D) \
/* Transaction Abort Word Conditional Immediate */ \
V(tabortwci, TABORTWCI, 0x7C00069D) \
/* Transaction Begin */ \
V(tbegin, TBEGIN, 0x7C00051D) \
/* Transaction Check */ \
V(tcheck, TCHECK, 0x7C00059C) \
/* Transaction End */ \
V(tend, TEND, 0x7C00055C) \
/* Transaction Recheckpoint */ \
V(trechkpt, TRECHKPT, 0x7C0007DD) \
/* Transaction Reclaim */ \
V(treclaim, TRECLAIM, 0x7C00075D) \
/* Transaction Suspend or Resume */ \
V(tsr, TSR, 0x7C0005DC) \
/* Load Vector Element Byte Indexed */ \
V(lvebx, LVEBX, 0x7C00000E) \
/* Load Vector Element Halfword Indexed */ \
V(lvehx, LVEHX, 0x7C00004E) \
/* Load Vector Element Word Indexed */ \
V(lvewx, LVEWX, 0x7C00008E) \
/* Load Vector for Shift Left */ \
V(lvsl, LVSL, 0x7C00000C) \
/* Load Vector for Shift Right */ \
V(lvsr, LVSR, 0x7C00004C) \
/* Load Vector Indexed */ \
V(lvx, LVX, 0x7C0000CE) \
/* Load Vector Indexed Last */ \
V(lvxl, LVXL, 0x7C0002CE) \
/* Store Vector Element Byte Indexed */ \
V(stvebx, STVEBX, 0x7C00010E) \
/* Store Vector Element Halfword Indexed */ \
V(stvehx, STVEHX, 0x7C00014E) \
/* Store Vector Element Word Indexed */ \
V(stvewx, STVEWX, 0x7C00018E) \
/* Store Vector Indexed */ \
V(stvx, STVX, 0x7C0001CE) \
/* Store Vector Indexed Last */ \
V(stvxl, STVXL, 0x7C0003CE) \
/* Vector Minimum Signed Doubleword */ \
V(vminsd, VMINSD, 0x100003C2) \
/* Floating Merge Even Word */ \
V(fmrgew, FMRGEW, 0xFC00078C) \
/* Floating Merge Odd Word */ \
V(fmrgow, FMRGOW, 0xFC00068C) \
/* Wait for Interrupt */ \
V(wait, WAIT, 0x7C00007C)
#define PPC_EVS_OPCODE_LIST(V) \
/* Vector Select */ \
V(evsel, EVSEL, 0x10000278)
#define PPC_DS_OPCODE_LIST(V) \
/* Load Doubleword */ \
V(ld, LD, 0xE8000000) \
/* Load Doubleword with Update */ \
V(ldu, LDU, 0xE8000001) \
/* Load Word Algebraic */ \
V(lwa, LWA, 0xE8000002) \
/* Store Doubleword */ \
V(std, STD, 0xF8000000) \
/* Store Doubleword with Update */ \
V(stdu, STDU, 0xF8000001) \
/* Load Floating-Point Double Pair */ \
V(lfdp, LFDP, 0xE4000000) \
/* Store Floating-Point Double Pair */ \
V(stfdp, STFDP, 0xF4000000) \
/* Store Quadword */ \
V(stq, STQ, 0xF8000002)
#define PPC_D_OPCODE_LIST(V) \
/* Trap Doubleword Immediate */ \
V(tdi, TDI, 0x08000000) \
/* Add Immediate */ \
V(addi, ADDI, 0x38000000) \
/* Add Immediate Carrying */ \
V(addic, ADDIC, 0x30000000) \
/* Add Immediate Carrying & record CR0 */ \
V(addicx, ADDICx, 0x34000000) \
/* Add Immediate Shifted */ \
V(addis, ADDIS, 0x3C000000) \
/* AND Immediate & record CR0 */ \
V(andix, ANDIx, 0x70000000) \
/* AND Immediate Shifted & record CR0 */ \
V(andisx, ANDISx, 0x74000000) \
/* Compare Immediate */ \
V(cmpi, CMPI, 0x2C000000) \
/* Compare Logical Immediate */ \
V(cmpli, CMPLI, 0x28000000) \
/* Load Byte and Zero */ \
V(lbz, LBZ, 0x88000000) \
/* Load Byte and Zero with Update */ \
V(lbzu, LBZU, 0x8C000000) \
/* Load Halfword Algebraic */ \
V(lha, LHA, 0xA8000000) \
/* Load Halfword Algebraic with Update */ \
V(lhau, LHAU, 0xAC000000) \
/* Load Halfword and Zero */ \
V(lhz, LHZ, 0xA0000000) \
/* Load Halfword and Zero with Update */ \
V(lhzu, LHZU, 0xA4000000) \
/* Load Multiple Word */ \
V(lmw, LMW, 0xB8000000) \
/* Load Word and Zero */ \
V(lwz, LWZ, 0x80000000) \
/* Load Word and Zero with Update */ \
V(lwzu, LWZU, 0x84000000) \
/* Multiply Low Immediate */ \
V(mulli, MULLI, 0x1C000000) \
/* OR Immediate */ \
V(ori, ORI, 0x60000000) \
/* OR Immediate Shifted */ \
V(oris, ORIS, 0x64000000) \
/* Store Byte */ \
V(stb, STB, 0x98000000) \
/* Store Byte with Update */ \
V(stbu, STBU, 0x9C000000) \
/* Store Halfword */ \
V(sth, STH, 0xB0000000) \
/* Store Halfword with Update */ \
V(sthu, STHU, 0xB4000000) \
/* Store Multiple Word */ \
V(stmw, STMW, 0xBC000000) \
/* Store Word */ \
V(stw, STW, 0x90000000) \
/* Store Word with Update */ \
V(stwu, STWU, 0x94000000) \
/* Subtract From Immediate Carrying */ \
V(subfic, SUBFIC, 0x20000000) \
/* Trap Word Immediate */ \
V(twi, TWI, 0x0C000000) \
/* XOR Immediate */ \
V(xori, XORI, 0x68000000) \
/* XOR Immediate Shifted */ \
V(xoris, XORIS, 0x6C000000) \
/* Load Floating-Point Double */ \
V(lfd, LFD, 0xC8000000) \
/* Load Floating-Point Double with Update */ \
V(lfdu, LFDU, 0xCC000000) \
/* Load Floating-Point Single */ \
V(lfs, LFS, 0xC0000000) \
/* Load Floating-Point Single with Update */ \
V(lfsu, LFSU, 0xC4000000) \
/* Store Floating-Point Double */ \
V(stfd, STFD, 0xD8000000) \
/* Store Floating-Point Double with Update */ \
V(stfdu, STFDU, 0xDC000000) \
/* Store Floating-Point Single */ \
V(stfs, STFS, 0xD0000000) \
/* Store Floating-Point Single with Update */ \
V(stfsu, STFSU, 0xD4000000)
#define PPC_XFL_OPCODE_LIST(V) \
/* Move To FPSCR Fields */ \
V(mtfsf, MTFSF, 0xFC00058E)
#define PPC_XFX_OPCODE_LIST(V) \
/* Move From Condition Register */ \
V(mfcr, MFCR, 0x7C000026) \
/* Move From One Condition Register Field */ \
V(mfocrf, MFOCRF, 0x7C100026) \
/* Move From Special Purpose Register */ \
V(mfspr, MFSPR, 0x7C0002A6) \
/* Move To Condition Register Fields */ \
V(mtcrf, MTCRF, 0x7C000120) \
/* Move To One Condition Register Field */ \
V(mtocrf, MTOCRF, 0x7C100120) \
/* Move To Special Purpose Register */ \
V(mtspr, MTSPR, 0x7C0003A6) \
/* Debugger Notify Halt */ \
V(dnh, DNH, 0x4C00018C) \
/* Move From Device Control Register */ \
V(mfdcr, MFDCR, 0x7C000286) \
/* Move To Device Control Register */ \
V(mtdcr, MTDCR, 0x7C000386) \
/* Move from Performance Monitor Register */ \
V(mfpmr, MFPMR, 0x7C00029C) \
/* Move To Performance Monitor Register */ \
V(mtpmr, MTPMR, 0x7C00039C) \
/* Move From Branch History Rolling Buffer */ \
V(mfbhrbe, MFBHRBE, 0x7C00025C) \
/* Move From Time Base */ \
V(mftb, MFTB, 0x7C0002E6)
#define PPC_MDS_OPCODE_LIST(V) \
/* Rotate Left Doubleword then Clear Left */ \
V(rldcl, RLDCL, 0x78000010) \
/* Rotate Left Doubleword then Clear Right */ \
V(rldcr, RLDCR, 0x78000012)
#define PPC_A_OPCODE_LIST(V) \
/* Integer Select */ \
V(isel, ISEL, 0x7C00001E) \
/* Floating Add */ \
V(fadd, FADD, 0xFC00002A) \
/* Floating Add Single */ \
V(fadds, FADDS, 0xEC00002A) \
/* Floating Divide */ \
V(fdiv, FDIV, 0xFC000024) \
/* Floating Divide Single */ \
V(fdivs, FDIVS, 0xEC000024) \
/* Floating Multiply-Add */ \
V(fmadd, FMADD, 0xFC00003A) \
/* Floating Multiply-Add Single */ \
V(fmadds, FMADDS, 0xEC00003A) \
/* Floating Multiply-Subtract */ \
V(fmsub, FMSUB, 0xFC000038) \
/* Floating Multiply-Subtract Single */ \
V(fmsubs, FMSUBS, 0xEC000038) \
/* Floating Multiply */ \
V(fmul, FMUL, 0xFC000032) \
/* Floating Multiply Single */ \
V(fmuls, FMULS, 0xEC000032) \
/* Floating Negative Multiply-Add */ \
V(fnmadd, FNMADD, 0xFC00003E) \
/* Floating Negative Multiply-Add Single */ \
V(fnmadds, FNMADDS, 0xEC00003E) \
/* Floating Negative Multiply-Subtract */ \
V(fnmsub, FNMSUB, 0xFC00003C) \
/* Floating Negative Multiply-Subtract Single */ \
V(fnmsubs, FNMSUBS, 0xEC00003C) \
/* Floating Reciprocal Estimate Single */ \
V(fres, FRES, 0xEC000030) \
/* Floating Reciprocal Square Root Estimate */ \
V(frsqrte, FRSQRTE, 0xFC000034) \
/* Floating Select */ \
V(fsel, FSEL, 0xFC00002E) \
/* Floating Square Root */ \
V(fsqrt, FSQRT, 0xFC00002C) \
/* Floating Square Root Single */ \
V(fsqrts, FSQRTS, 0xEC00002C) \
/* Floating Subtract */ \
V(fsub, FSUB, 0xFC000028) \
/* Floating Subtract Single */ \
V(fsubs, FSUBS, 0xEC000028) \
/* Floating Reciprocal Estimate */ \
V(fre, FRE, 0xFC000030) \
/* Floating Reciprocal Square Root Estimate Single */ \
V(frsqrtes, FRSQRTES, 0xEC000034)
#define PPC_VA_OPCODE_LIST(V) \
/* Vector Add Extended & write Carry Unsigned Quadword */ \
V(vaddecuq, VADDECUQ, 0x1000003D) \
/* Vector Add Extended Unsigned Quadword Modulo */ \
V(vaddeuqm, VADDEUQM, 0x1000003C) \
/* Vector Multiply-Add Single-Precision */ \
V(vmaddfp, VMADDFP, 0x1000002E) \
/* Vector Multiply-High-Add Signed Halfword Saturate */ \
V(vmhaddshs, VMHADDSHS, 0x10000020) \
/* Vector Multiply-High-Round-Add Signed Halfword Saturate */ \
V(vmhraddshs, VMHRADDSHS, 0x10000021) \
/* Vector Multiply-Low-Add Unsigned Halfword Modulo */ \
V(vmladduhm, VMLADDUHM, 0x10000022) \
/* Vector Multiply-Sum Mixed Byte Modulo */ \
V(vmsummbm, VMSUMMBM, 0x10000025) \
/* Vector Multiply-Sum Signed Halfword Modulo */ \
V(vmsumshm, VMSUMSHM, 0x10000028) \
/* Vector Multiply-Sum Signed Halfword Saturate */ \
V(vmsumshs, VMSUMSHS, 0x10000029) \
/* Vector Multiply-Sum Unsigned Byte Modulo */ \
V(vmsumubm, VMSUMUBM, 0x10000024) \
/* Vector Multiply-Sum Unsigned Halfword Modulo */ \
V(vmsumuhm, VMSUMUHM, 0x10000026) \
/* Vector Multiply-Sum Unsigned Halfword Saturate */ \
V(vmsumuhs, VMSUMUHS, 0x10000027) \
/* Vector Negative Multiply-Subtract Single-Precision */ \
V(vnmsubfp, VNMSUBFP, 0x1000002F) \
/* Vector Permute */ \
V(vperm, VPERM, 0x1000002B) \
/* Vector Select */ \
V(vsel, VSEL, 0x1000002A) \
/* Vector Shift Left Double by Octet Immediate */ \
V(vsldoi, VSLDOI, 0x1000002C) \
/* Vector Subtract Extended & write Carry Unsigned Quadword */ \
V(vsubecuq, VSUBECUQ, 0x1000003F) \
/* Vector Subtract Extended Unsigned Quadword Modulo */ \
V(vsubeuqm, VSUBEUQM, 0x1000003E) \
/* Vector Permute and Exclusive-OR */ \
V(vpermxor, VPERMXOR, 0x1000002D)
#define PPC_XX1_OPCODE_LIST(V) \
/* Load VSR Scalar Doubleword Indexed */ \
V(lxsdx, LXSDX, 0x7C000498) \
/* Load VSX Scalar as Integer Word Algebraic Indexed */ \
V(lxsiwax, LXSIWAX, 0x7C000098) \
/* Load VSX Scalar as Integer Word and Zero Indexed */ \
V(lxsiwzx, LXSIWZX, 0x7C000018) \
/* Load VSX Scalar Single-Precision Indexed */ \
V(lxsspx, LXSSPX, 0x7C000418) \
/* Load VSR Vector Doubleword*2 Indexed */ \
V(lxvd, LXVD, 0x7C000698) \
/* Load VSR Vector Doubleword & Splat Indexed */ \
V(lxvdsx, LXVDSX, 0x7C000298) \
/* Load VSR Vector Word*4 Indexed */ \
V(lxvw, LXVW, 0x7C000618) \
/* Move From VSR Doubleword */ \
V(mfvsrd, MFVSRD, 0x7C000066) \
/* Move From VSR Word and Zero */ \
V(mfvsrwz, MFVSRWZ, 0x7C0000E6) \
/* Store VSR Scalar Doubleword Indexed */ \
V(stxsdx, STXSDX, 0x7C000598) \
/* Store VSX Scalar as Integer Word Indexed */ \
V(stxsiwx, STXSIWX, 0x7C000118) \
/* Store VSR Scalar Word Indexed */ \
V(stxsspx, STXSSPX, 0x7C000518) \
/* Store VSR Vector Doubleword*2 Indexed */ \
V(stxvd, STXVD, 0x7C000798) \
/* Store VSR Vector Word*4 Indexed */ \
V(stxvw, STXVW, 0x7C000718)
#define PPC_B_OPCODE_LIST(V) \
/* Branch Conditional */ \
V(bc, BCX, 0x40000000)
#define PPC_XO_OPCODE_LIST(V) \
/* Divide Doubleword */ \
V(divd, DIVD, 0x7C0003D2) \
/* Divide Doubleword Extended */ \
V(divde, DIVDE, 0x7C000352) \
/* Divide Doubleword Extended & record OV */ \
V(divdeo, DIVDEO, 0x7C000752) \
/* Divide Doubleword Extended Unsigned */ \
V(divdeu, DIVDEU, 0x7C000312) \
/* Divide Doubleword Extended Unsigned & record OV */ \
V(divdeuo, DIVDEUO, 0x7C000712) \
/* Divide Doubleword & record OV */ \
V(divdo, DIVDO, 0x7C0007D2) \
/* Divide Doubleword Unsigned */ \
V(divdu, DIVDU, 0x7C000392) \
/* Divide Doubleword Unsigned & record OV */ \
V(divduo, DIVDUO, 0x7C000792) \
/* Multiply High Doubleword */ \
V(mulhd, MULHD, 0x7C000092) \
/* Multiply High Doubleword Unsigned */ \
V(mulhdu, MULHDU, 0x7C000012) \
/* Multiply Low Doubleword */ \
V(mulld, MULLD, 0x7C0001D2) \
/* Multiply Low Doubleword & record OV */ \
V(mulldo, MULLDO, 0x7C0005D2) \
/* Add */ \
V(add, ADDX, 0x7C000214) \
/* Add Carrying */ \
V(addc, ADDCX, 0x7C000014) \
/* Add Carrying & record OV */ \
V(addco, ADDCO, 0x7C000414) \
/* Add Extended */ \
V(adde, ADDEX, 0x7C000114) \
/* Add Extended & record OV & record OV */ \
V(addeo, ADDEO, 0x7C000514) \
/* Add to Minus One Extended */ \
V(addme, ADDME, 0x7C0001D4) \
/* Add to Minus One Extended & record OV */ \
V(addmeo, ADDMEO, 0x7C0005D4) \
/* Add & record OV */ \
V(addo, ADDO, 0x7C000614) \
/* Add to Zero Extended */ \
V(addze, ADDZEX, 0x7C000194) \
/* Add to Zero Extended & record OV */ \
V(addzeo, ADDZEO, 0x7C000594) \
/* Divide Word Format */ \
V(divw, DIVW, 0x7C0003D6) \
/* Divide Word Extended */ \
V(divwe, DIVWE, 0x7C000356) \
/* Divide Word Extended & record OV */ \
V(divweo, DIVWEO, 0x7C000756) \
/* Divide Word Extended Unsigned */ \
V(divweu, DIVWEU, 0x7C000316) \
/* Divide Word Extended Unsigned & record OV */ \
V(divweuo, DIVWEUO, 0x7C000716) \
/* Divide Word & record OV */ \
V(divwo, DIVWO, 0x7C0007D6) \
/* Divide Word Unsigned */ \
V(divwu, DIVWU, 0x7C000396) \
/* Divide Word Unsigned & record OV */ \
V(divwuo, DIVWUO, 0x7C000796) \
/* Multiply High Word */ \
V(mulhw, MULHWX, 0x7C000096) \
/* Multiply High Word Unsigned */ \
V(mulhwu, MULHWUX, 0x7C000016) \
/* Multiply Low Word */ \
V(mullw, MULLW, 0x7C0001D6) \
/* Multiply Low Word & record OV */ \
V(mullwo, MULLWO, 0x7C0005D6) \
/* Negate */ \
V(neg, NEGX, 0x7C0000D0) \
/* Negate & record OV */ \
V(nego, NEGO, 0x7C0004D0) \
/* Subtract From */ \
V(subf, SUBFX, 0x7C000050) \
/* Subtract From Carrying */ \
V(subfc, SUBFCX, 0x7C000010) \
/* Subtract From Carrying & record OV */ \
V(subfco, SUBFCO, 0x7C000410) \
/* Subtract From Extended */ \
V(subfe, SUBFEX, 0x7C000110) \
/* Subtract From Extended & record OV */ \
V(subfeo, SUBFEO, 0x7C000510) \
/* Subtract From Minus One Extended */ \
V(subfme, SUBFME, 0x7C0001D0) \
/* Subtract From Minus One Extended & record OV */ \
V(subfmeo, SUBFMEO, 0x7C0005D0) \
/* Subtract From & record OV */ \
V(subfo, SUBFO, 0x7C000450) \
/* Subtract From Zero Extended */ \
V(subfze, SUBFZE, 0x7C000190) \
/* Subtract From Zero Extended & record OV */ \
V(subfzeo, SUBFZEO, 0x7C000590) \
/* Add and Generate Sixes */ \
V(addg, ADDG, 0x7C000094) \
/* Multiply Accumulate Cross Halfword to Word Modulo Signed */ \
V(macchw, MACCHW, 0x10000158) \
/* Multiply Accumulate Cross Halfword to Word Modulo Signed & record OV */ \
V(macchwo, MACCHWO, 0x10000158) \
/* Multiply Accumulate Cross Halfword to Word Saturate Signed */ \
V(macchws, MACCHWS, 0x100001D8) \
/* Multiply Accumulate Cross Halfword to Word Saturate Signed & record */ \
/* OV */ \
V(macchwso, MACCHWSO, 0x100001D8) \
/* Multiply Accumulate Cross Halfword to Word Saturate Unsigned */ \
V(macchwsu, MACCHWSU, 0x10000198) \
/* Multiply Accumulate Cross Halfword to Word Saturate Unsigned & record */ \
/* OV */ \
V(macchwsuo, MACCHWSUO, 0x10000198) \
/* Multiply Accumulate Cross Halfword to Word Modulo Unsigned */ \
V(macchwu, MACCHWU, 0x10000118) \
/* Multiply Accumulate Cross Halfword to Word Modulo Unsigned & record */ \
/* OV */ \
V(macchwuo, MACCHWUO, 0x10000118) \
/* Multiply Accumulate High Halfword to Word Modulo Signed */ \
V(machhw, MACHHW, 0x10000058) \
/* Multiply Accumulate High Halfword to Word Modulo Signed & record OV */ \
V(machhwo, MACHHWO, 0x10000058) \
/* Multiply Accumulate High Halfword to Word Saturate Signed */ \
V(machhws, MACHHWS, 0x100000D8) \
/* Multiply Accumulate High Halfword to Word Saturate Signed & record OV */ \
V(machhwso, MACHHWSO, 0x100000D8) \
/* Multiply Accumulate High Halfword to Word Saturate Unsigned */ \
V(machhwsu, MACHHWSU, 0x10000098) \
/* Multiply Accumulate High Halfword to Word Saturate Unsigned & record */ \
/* OV */ \
V(machhwsuo, MACHHWSUO, 0x10000098) \
/* Multiply Accumulate High Halfword to Word Modulo Unsigned */ \
V(machhwu, MACHHWU, 0x10000018) \
/* Multiply Accumulate High Halfword to Word Modulo Unsigned & record OV */ \
V(machhwuo, MACHHWUO, 0x10000018) \
/* Multiply Accumulate Low Halfword to Word Modulo Signed */ \
V(maclhw, MACLHW, 0x10000358) \
/* Multiply Accumulate Low Halfword to Word Modulo Signed & record OV */ \
V(maclhwo, MACLHWO, 0x10000358) \
/* Multiply Accumulate Low Halfword to Word Saturate Signed */ \
V(maclhws, MACLHWS, 0x100003D8) \
/* Multiply Accumulate Low Halfword to Word Saturate Signed & record OV */ \
V(maclhwso, MACLHWSO, 0x100003D8) \
/* Multiply Accumulate Low Halfword to Word Saturate Unsigned */ \
V(maclhwsu, MACLHWSU, 0x10000398) \
/* Multiply Accumulate Low Halfword to Word Saturate Unsigned & record */ \
/* OV */ \
V(maclhwsuo, MACLHWSUO, 0x10000398) \
/* Multiply Accumulate Low Halfword to Word Modulo Unsigned */ \
V(maclhwu, MACLHWU, 0x10000318) \
/* Multiply Accumulate Low Halfword to Word Modulo Unsigned & record OV */ \
V(maclhwuo, MACLHWUO, 0x10000318) \
/* Negative Multiply Accumulate Cross Halfword to Word Modulo Signed */ \
V(nmacchw, NMACCHW, 0x1000015C) \
/* Negative Multiply Accumulate Cross Halfword to Word Modulo Signed & */ \
/* record OV */ \
V(nmacchwo, NMACCHWO, 0x1000015C) \
/* Negative Multiply Accumulate Cross Halfword to Word Saturate Signed */ \
V(nmacchws, NMACCHWS, 0x100001DC) \
/* Negative Multiply Accumulate Cross Halfword to Word Saturate Signed & */ \
/* record OV */ \
V(nmacchwso, NMACCHWSO, 0x100001DC) \
/* Negative Multiply Accumulate High Halfword to Word Modulo Signed */ \
V(nmachhw, NMACHHW, 0x1000005C) \
/* Negative Multiply Accumulate High Halfword to Word Modulo Signed & */ \
/* record OV */ \
V(nmachhwo, NMACHHWO, 0x1000005C) \
/* Negative Multiply Accumulate High Halfword to Word Saturate Signed */ \
V(nmachhws, NMACHHWS, 0x100000DC) \
/* Negative Multiply Accumulate High Halfword to Word Saturate Signed & */ \
/* record OV */ \
V(nmachhwso, NMACHHWSO, 0x100000DC) \
/* Negative Multiply Accumulate Low Halfword to Word Modulo Signed */ \
V(nmaclhw, NMACLHW, 0x1000035C) \
/* Negative Multiply Accumulate Low Halfword to Word Modulo Signed & */ \
/* record OV */ \
V(nmaclhwo, NMACLHWO, 0x1000035C) \
/* Negative Multiply Accumulate Low Halfword to Word Saturate Signed */ \
V(nmaclhws, NMACLHWS, 0x100003DC) \
/* Negative Multiply Accumulate Low Halfword to Word Saturate Signed & */ \
/* record OV */ \
V(nmaclhwso, NMACLHWSO, 0x100003DC)
#define PPC_XL_OPCODE_LIST(V) \
/* Branch Conditional to Count Register */ \
V(bcctr, BCCTRX, 0x4C000420) \
/* Branch Conditional to Link Register */ \
V(bclr, BCLRX, 0x4C000020) \
/* Condition Register AND */ \
V(crand, CRAND, 0x4C000202) \
/* Condition Register AND with Complement */ \
V(crandc, CRANDC, 0x4C000102) \
/* Condition Register Equivalent */ \
V(creqv, CREQV, 0x4C000242) \
/* Condition Register NAND */ \
V(crnand, CRNAND, 0x4C0001C2) \
/* Condition Register NOR */ \
V(crnor, CRNOR, 0x4C000042) \
/* Condition Register OR */ \
V(cror, CROR, 0x4C000382) \
/* Condition Register OR with Complement */ \
V(crorc, CRORC, 0x4C000342) \
/* Condition Register XOR */ \
V(crxor, CRXOR, 0x4C000182) \
/* Instruction Synchronize */ \
V(isync, ISYNC, 0x4C00012C) \
/* Move Condition Register Field */ \
V(mcrf, MCRF, 0x4C000000) \
/* Return From Critical Interrupt */ \
V(rfci, RFCI, 0x4C000066) \
/* Return From Interrupt */ \
V(rfi, RFI, 0x4C000064) \
/* Return From Machine Check Interrupt */ \
V(rfmci, RFMCI, 0x4C00004C) \
/* Embedded Hypervisor Privilege */ \
V(ehpriv, EHPRIV, 0x7C00021C) \
/* Return From Guest Interrupt */ \
V(rfgi, RFGI, 0x4C0000CC) \
/* Doze */ \
V(doze, DOZE, 0x4C000324) \
/* Return From Interrupt Doubleword Hypervisor */ \
V(hrfid, HRFID, 0x4C000224) \
/* Nap */ \
V(nap, NAP, 0x4C000364) \
/* Return from Event Based Branch */ \
V(rfebb, RFEBB, 0x4C000124) \
/* Return from Interrupt Doubleword */ \
V(rfid, RFID, 0x4C000024) \
/* Rip Van Winkle */ \
V(rvwinkle, RVWINKLE, 0x4C0003E4) \
/* Sleep */ \
V(sleep, SLEEP, 0x4C0003A4)
#define PPC_XX4_OPCODE_LIST(V) \
/* VSX Select */ \
V(xxsel, XXSEL, 0xF0000030)
#define PPC_I_OPCODE_LIST(V) \
/* Branch */ \
V(b, BX, 0x48000000)
#define PPC_M_OPCODE_LIST(V) \
/* Rotate Left Word Immediate then Mask Insert */ \
V(rlwimi, RLWIMIX, 0x50000000) \
/* Rotate Left Word Immediate then AND with Mask */ \
V(rlwinm, RLWINMX, 0x54000000) \
/* Rotate Left Word then AND with Mask */ \
V(rlwnm, RLWNMX, 0x5C000000)
#define PPC_VX_OPCODE_LIST(V) \
/* Decimal Add Modulo */ \
V(bcdadd, BCDADD, 0xF0000400) \
/* Decimal Subtract Modulo */ \
V(bcdsub, BCDSUB, 0xF0000440) \
/* Move From Vector Status and Control Register */ \
V(mfvscr, MFVSCR, 0x10000604) \
/* Move To Vector Status and Control Register */ \
V(mtvscr, MTVSCR, 0x10000644) \
/* Vector Add & write Carry Unsigned Quadword */ \
V(vaddcuq, VADDCUQ, 0x10000140) \
/* Vector Add and Write Carry-Out Unsigned Word */ \
V(vaddcuw, VADDCUW, 0x10000180) \
/* Vector Add Single-Precision */ \
V(vaddfp, VADDFP, 0x1000000A) \
/* Vector Add Signed Byte Saturate */ \
V(vaddsbs, VADDSBS, 0x10000300) \
/* Vector Add Signed Halfword Saturate */ \
V(vaddshs, VADDSHS, 0x10000340) \
/* Vector Add Signed Word Saturate */ \
V(vaddsws, VADDSWS, 0x10000380) \
/* Vector Add Unsigned Byte Modulo */ \
V(vaddubm, VADDUBM, 0x10000000) \
/* Vector Add Unsigned Byte Saturate */ \
V(vaddubs, VADDUBS, 0x10000200) \
/* Vector Add Unsigned Doubleword Modulo */ \
V(vaddudm, VADDUDM, 0x100000C0) \
/* Vector Add Unsigned Halfword Modulo */ \
V(vadduhm, VADDUHM, 0x10000040) \
/* Vector Add Unsigned Halfword Saturate */ \
V(vadduhs, VADDUHS, 0x10000240) \
/* Vector Add Unsigned Quadword Modulo */ \
V(vadduqm, VADDUQM, 0x10000100) \
/* Vector Add Unsigned Word Modulo */ \
V(vadduwm, VADDUWM, 0x10000080) \
/* Vector Add Unsigned Word Saturate */ \
V(vadduws, VADDUWS, 0x10000280) \
/* Vector Logical AND */ \
V(vand, VAND, 0x10000404) \
/* Vector Logical AND with Complement */ \
V(vandc, VANDC, 0x10000444) \
/* Vector Average Signed Byte */ \
V(vavgsb, VAVGSB, 0x10000502) \
/* Vector Average Signed Halfword */ \
V(vavgsh, VAVGSH, 0x10000542) \
/* Vector Average Signed Word */ \
V(vavgsw, VAVGSW, 0x10000582) \
/* Vector Average Unsigned Byte */ \
V(vavgub, VAVGUB, 0x10000402) \
/* Vector Average Unsigned Halfword */ \
V(vavguh, VAVGUH, 0x10000442) \
/* Vector Average Unsigned Word */ \
V(vavguw, VAVGUW, 0x10000482) \
/* Vector Bit Permute Quadword */ \
V(vbpermq, VBPERMQ, 0x1000054C) \
/* Vector Convert From Signed Fixed-Point Word To Single-Precision */ \
V(vcfsx, VCFSX, 0x1000034A) \
/* Vector Convert From Unsigned Fixed-Point Word To Single-Precision */ \
V(vcfux, VCFUX, 0x1000030A) \
/* Vector Count Leading Zeros Byte */ \
V(vclzb, VCLZB, 0x10000702) \
/* Vector Count Leading Zeros Doubleword */ \
V(vclzd, VCLZD, 0x100007C2) \
/* Vector Count Leading Zeros Halfword */ \
V(vclzh, VCLZH, 0x10000742) \
/* Vector Count Leading Zeros Word */ \
V(vclzw, VCLZW, 0x10000782) \
/* Vector Convert From Single-Precision To Signed Fixed-Point Word */ \
/* Saturate */ \
V(vctsxs, VCTSXS, 0x100003CA) \
/* Vector Convert From Single-Precision To Unsigned Fixed-Point Word */ \
/* Saturate */ \
V(vctuxs, VCTUXS, 0x1000038A) \
/* Vector Equivalence */ \
V(veqv, VEQV, 0x10000684) \
/* Vector 2 Raised to the Exponent Estimate Single-Precision */ \
V(vexptefp, VEXPTEFP, 0x1000018A) \
/* Vector Gather Bits by Byte by Doubleword */ \
V(vgbbd, VGBBD, 0x1000050C) \
/* Vector Log Base 2 Estimate Single-Precision */ \
V(vlogefp, VLOGEFP, 0x100001CA) \
/* Vector Maximum Single-Precision */ \
V(vmaxfp, VMAXFP, 0x1000040A) \
/* Vector Maximum Signed Byte */ \
V(vmaxsb, VMAXSB, 0x10000102) \
/* Vector Maximum Signed Doubleword */ \
V(vmaxsd, VMAXSD, 0x100001C2) \
/* Vector Maximum Signed Halfword */ \
V(vmaxsh, VMAXSH, 0x10000142) \
/* Vector Maximum Signed Word */ \
V(vmaxsw, VMAXSW, 0x10000182) \
/* Vector Maximum Unsigned Byte */ \
V(vmaxub, VMAXUB, 0x10000002) \
/* Vector Maximum Unsigned Doubleword */ \
V(vmaxud, VMAXUD, 0x100000C2) \
/* Vector Maximum Unsigned Halfword */ \
V(vmaxuh, VMAXUH, 0x10000042) \
/* Vector Maximum Unsigned Word */ \
V(vmaxuw, VMAXUW, 0x10000082) \
/* Vector Minimum Single-Precision */ \
V(vminfp, VMINFP, 0x1000044A) \
/* Vector Minimum Signed Byte */ \
V(vminsb, VMINSB, 0x10000302) \
/* Vector Minimum Signed Halfword */ \
V(vminsh, VMINSH, 0x10000342) \
/* Vector Minimum Signed Word */ \
V(vminsw, VMINSW, 0x10000382) \
/* Vector Minimum Unsigned Byte */ \
V(vminub, VMINUB, 0x10000202) \
/* Vector Minimum Unsigned Doubleword */ \
V(vminud, VMINUD, 0x100002C2) \
/* Vector Minimum Unsigned Halfword */ \
V(vminuh, VMINUH, 0x10000242) \
/* Vector Minimum Unsigned Word */ \
V(vminuw, VMINUW, 0x10000282) \
/* Vector Merge High Byte */ \
V(vmrghb, VMRGHB, 0x1000000C) \
/* Vector Merge High Halfword */ \
V(vmrghh, VMRGHH, 0x1000004C) \
/* Vector Merge High Word */ \
V(vmrghw, VMRGHW, 0x1000008C) \
/* Vector Merge Low Byte */ \
V(vmrglb, VMRGLB, 0x1000010C) \
/* Vector Merge Low Halfword */ \
V(vmrglh, VMRGLH, 0x1000014C) \
/* Vector Merge Low Word */ \
V(vmrglw, VMRGLW, 0x1000018C) \
/* Vector Multiply Even Signed Byte */ \
V(vmulesb, VMULESB, 0x10000308) \
/* Vector Multiply Even Signed Halfword */ \
V(vmulesh, VMULESH, 0x10000348) \
/* Vector Multiply Even Signed Word */ \
V(vmulesw, VMULESW, 0x10000388) \
/* Vector Multiply Even Unsigned Byte */ \
V(vmuleub, VMULEUB, 0x10000208) \
/* Vector Multiply Even Unsigned Halfword */ \
V(vmuleuh, VMULEUH, 0x10000248) \
/* Vector Multiply Even Unsigned Word */ \
V(vmuleuw, VMULEUW, 0x10000288) \
/* Vector Multiply Odd Signed Byte */ \
V(vmulosb, VMULOSB, 0x10000108) \
/* Vector Multiply Odd Signed Halfword */ \
V(vmulosh, VMULOSH, 0x10000148) \
/* Vector Multiply Odd Signed Word */ \
V(vmulosw, VMULOSW, 0x10000188) \
/* Vector Multiply Odd Unsigned Byte */ \
V(vmuloub, VMULOUB, 0x10000008) \
/* Vector Multiply Odd Unsigned Halfword */ \
V(vmulouh, VMULOUH, 0x10000048) \
/* Vector Multiply Odd Unsigned Word */ \
V(vmulouw, VMULOUW, 0x10000088) \
/* Vector Multiply Unsigned Word Modulo */ \
V(vmuluwm, VMULUWM, 0x10000089) \
/* Vector NAND */ \
V(vnand, VNAND, 0x10000584) \
/* Vector Logical NOR */ \
V(vnor, VNOR, 0x10000504) \
/* Vector Logical OR */ \
V(vor, VOR, 0x10000484) \
/* Vector OR with Complement */ \
V(vorc, VORC, 0x10000544) \
/* Vector Pack Pixel */ \
V(vpkpx, VPKPX, 0x1000030E) \
/* Vector Pack Signed Doubleword Signed Saturate */ \
V(vpksdss, VPKSDSS, 0x100005CE) \
/* Vector Pack Signed Doubleword Unsigned Saturate */ \
V(vpksdus, VPKSDUS, 0x1000054E) \
/* Vector Pack Signed Halfword Signed Saturate */ \
V(vpkshss, VPKSHSS, 0x1000018E) \
/* Vector Pack Signed Halfword Unsigned Saturate */ \
V(vpkshus, VPKSHUS, 0x1000010E) \
/* Vector Pack Signed Word Signed Saturate */ \
V(vpkswss, VPKSWSS, 0x100001CE) \
/* Vector Pack Signed Word Unsigned Saturate */ \
V(vpkswus, VPKSWUS, 0x1000014E) \
/* Vector Pack Unsigned Doubleword Unsigned Modulo */ \
V(vpkudum, VPKUDUM, 0x1000044E) \
/* Vector Pack Unsigned Doubleword Unsigned Saturate */ \
V(vpkudus, VPKUDUS, 0x100004CE) \
/* Vector Pack Unsigned Halfword Unsigned Modulo */ \
V(vpkuhum, VPKUHUM, 0x1000000E) \
/* Vector Pack Unsigned Halfword Unsigned Saturate */ \
V(vpkuhus, VPKUHUS, 0x1000008E) \
/* Vector Pack Unsigned Word Unsigned Modulo */ \
V(vpkuwum, VPKUWUM, 0x1000004E) \
/* Vector Pack Unsigned Word Unsigned Saturate */ \
V(vpkuwus, VPKUWUS, 0x100000CE) \
/* Vector Polynomial Multiply-Sum Byte */ \
V(vpmsumb, VPMSUMB, 0x10000408) \
/* Vector Polynomial Multiply-Sum Doubleword */ \
V(vpmsumd, VPMSUMD, 0x100004C8) \
/* Vector Polynomial Multiply-Sum Halfword */ \
V(vpmsumh, VPMSUMH, 0x10000448) \
/* Vector Polynomial Multiply-Sum Word */ \
V(vpmsumw, VPMSUMW, 0x10000488) \
/* Vector Population Count Byte */ \
V(vpopcntb, VPOPCNTB, 0x10000703) \
/* Vector Population Count Doubleword */ \
V(vpopcntd, VPOPCNTD, 0x100007C3) \
/* Vector Population Count Halfword */ \
V(vpopcnth, VPOPCNTH, 0x10000743) \
/* Vector Population Count Word */ \
V(vpopcntw, VPOPCNTW, 0x10000783) \
/* Vector Reciprocal Estimate Single-Precision */ \
V(vrefp, VREFP, 0x1000010A) \
/* Vector Round to Single-Precision Integer toward -Infinity */ \
V(vrfim, VRFIM, 0x100002CA) \
/* Vector Round to Single-Precision Integer Nearest */ \
V(vrfin, VRFIN, 0x1000020A) \
/* Vector Round to Single-Precision Integer toward +Infinity */ \
V(vrfip, VRFIP, 0x1000028A) \
/* Vector Round to Single-Precision Integer toward Zero */ \
V(vrfiz, VRFIZ, 0x1000024A) \
/* Vector Rotate Left Byte */ \
V(vrlb, VRLB, 0x10000004) \
/* Vector Rotate Left Doubleword */ \
V(vrld, VRLD, 0x100000C4) \
/* Vector Rotate Left Halfword */ \
V(vrlh, VRLH, 0x10000044) \
/* Vector Rotate Left Word */ \
V(vrlw, VRLW, 0x10000084) \
/* Vector Reciprocal Square Root Estimate Single-Precision */ \
V(vrsqrtefp, VRSQRTEFP, 0x1000014A) \
/* Vector Shift Left */ \
V(vsl, VSL, 0x100001C4) \
/* Vector Shift Left Byte */ \
V(vslb, VSLB, 0x10000104) \
/* Vector Shift Left Doubleword */ \
V(vsld, VSLD, 0x100005C4) \
/* Vector Shift Left Halfword */ \
V(vslh, VSLH, 0x10000144) \
/* Vector Shift Left by Octet */ \
V(vslo, VSLO, 0x1000040C) \
/* Vector Shift Left Word */ \
V(vslw, VSLW, 0x10000184) \
/* Vector Splat Byte */ \
V(vspltb, VSPLTB, 0x1000020C) \
/* Vector Splat Halfword */ \
V(vsplth, VSPLTH, 0x1000024C) \
/* Vector Splat Immediate Signed Byte */ \
V(vspltisb, VSPLTISB, 0x1000030C) \
/* Vector Splat Immediate Signed Halfword */ \
V(vspltish, VSPLTISH, 0x1000034C) \
/* Vector Splat Immediate Signed Word */ \
V(vspltisw, VSPLTISW, 0x1000038C) \
/* Vector Splat Word */ \
V(vspltw, VSPLTW, 0x1000028C) \
/* Vector Shift Right */ \
V(vsr, VSR, 0x100002C4) \
/* Vector Shift Right Algebraic Byte */ \
V(vsrab, VSRAB, 0x10000304) \
/* Vector Shift Right Algebraic Doubleword */ \
V(vsrad, VSRAD, 0x100003C4) \
/* Vector Shift Right Algebraic Halfword */ \
V(vsrah, VSRAH, 0x10000344) \
/* Vector Shift Right Algebraic Word */ \
V(vsraw, VSRAW, 0x10000384) \
/* Vector Shift Right Byte */ \
V(vsrb, VSRB, 0x10000204) \
/* Vector Shift Right Doubleword */ \
V(vsrd, VSRD, 0x100006C4) \
/* Vector Shift Right Halfword */ \
V(vsrh, VSRH, 0x10000244) \
/* Vector Shift Right by Octet */ \
V(vsro, VSRO, 0x1000044C) \
/* Vector Shift Right Word */ \
V(vsrw, VSRW, 0x10000284) \
/* Vector Subtract & write Carry Unsigned Quadword */ \
V(vsubcuq, VSUBCUQ, 0x10000540) \
/* Vector Subtract and Write Carry-Out Unsigned Word */ \
V(vsubcuw, VSUBCUW, 0x10000580) \
/* Vector Subtract Single-Precision */ \
V(vsubfp, VSUBFP, 0x1000004A) \
/* Vector Subtract Signed Byte Saturate */ \
V(vsubsbs, VSUBSBS, 0x10000700) \
/* Vector Subtract Signed Halfword Saturate */ \
V(vsubshs, VSUBSHS, 0x10000740) \
/* Vector Subtract Signed Word Saturate */ \
V(vsubsws, VSUBSWS, 0x10000780) \
/* Vector Subtract Unsigned Byte Modulo */ \
V(vsububm, VSUBUBM, 0x10000400) \
/* Vector Subtract Unsigned Byte Saturate */ \
V(vsububs, VSUBUBS, 0x10000600) \
/* Vector Subtract Unsigned Doubleword Modulo */ \
V(vsubudm, VSUBUDM, 0x100004C0) \
/* Vector Subtract Unsigned Halfword Modulo */ \
V(vsubuhm, VSUBUHM, 0x10000440) \
/* Vector Subtract Unsigned Halfword Saturate */ \
V(vsubuhs, VSUBUHS, 0x10000640) \
/* Vector Subtract Unsigned Quadword Modulo */ \
V(vsubuqm, VSUBUQM, 0x10000500) \
/* Vector Subtract Unsigned Word Modulo */ \
V(vsubuwm, VSUBUWM, 0x10000480) \
/* Vector Subtract Unsigned Word Saturate */ \
V(vsubuws, VSUBUWS, 0x10000680) \
/* Vector Sum across Half Signed Word Saturate */ \
V(vsum2sws, VSUM2SWS, 0x10000688) \
/* Vector Sum across Quarter Signed Byte Saturate */ \
V(vsum4sbs, VSUM4SBS, 0x10000708) \
/* Vector Sum across Quarter Signed Halfword Saturate */ \
V(vsum4shs, VSUM4SHS, 0x10000648) \
/* Vector Sum across Quarter Unsigned Byte Saturate */ \
V(vsum4bus, VSUM4BUS, 0x10000608) \
/* Vector Sum across Signed Word Saturate */ \
V(vsumsws, VSUMSWS, 0x10000788) \
/* Vector Unpack High Pixel */ \
V(vupkhpx, VUPKHPX, 0x1000034E) \
/* Vector Unpack High Signed Byte */ \
V(vupkhsb, VUPKHSB, 0x1000020E) \
/* Vector Unpack High Signed Halfword */ \
V(vupkhsh, VUPKHSH, 0x1000024E) \
/* Vector Unpack High Signed Word */ \
V(vupkhsw, VUPKHSW, 0x1000064E) \
/* Vector Unpack Low Pixel */ \
V(vupklpx, VUPKLPX, 0x100003CE) \
/* Vector Unpack Low Signed Byte */ \
V(vupklsb, VUPKLSB, 0x1000028E) \
/* Vector Unpack Low Signed Halfword */ \
V(vupklsh, VUPKLSH, 0x100002CE) \
/* Vector Unpack Low Signed Word */ \
V(vupklsw, VUPKLSW, 0x100006CE) \
/* Vector Logical XOR */ \
V(vxor, VXOR, 0x100004C4) \
/* Vector AES Cipher */ \
V(vcipher, VCIPHER, 0x10000508) \
/* Vector AES Cipher Last */ \
V(vcipherlast, VCIPHERLAST, 0x10000509) \
/* Vector AES Inverse Cipher */ \
V(vncipher, VNCIPHER, 0x10000548) \
/* Vector AES Inverse Cipher Last */ \
V(vncipherlast, VNCIPHERLAST, 0x10000549) \
/* Vector AES S-Box */ \
V(vsbox, VSBOX, 0x100005C8) \
/* Vector SHA-512 Sigma Doubleword */ \
V(vshasigmad, VSHASIGMAD, 0x100006C2) \
/* Vector SHA-256 Sigma Word */ \
V(vshasigmaw, VSHASIGMAW, 0x10000682) \
/* Vector Merge Even Word */ \
V(vmrgew, VMRGEW, 0x1000078C) \
/* Vector Merge Odd Word */ \
V(vmrgow, VMRGOW, 0x1000068C)
#define PPC_XS_OPCODE_LIST(V) \
/* Shift Right Algebraic Doubleword Immediate */ \
V(sradi, SRADIX, 0x7C000674)
#define PPC_MD_OPCODE_LIST(V) \
/* Rotate Left Doubleword Immediate then Clear */ \
V(rldic, RLDIC, 0x78000008) \
/* Rotate Left Doubleword Immediate then Clear Left */ \
V(rldicl, RLDICL, 0x78000000) \
/* Rotate Left Doubleword Immediate then Clear Right */ \
V(rldicr, RLDICR, 0x78000004) \
/* Rotate Left Doubleword Immediate then Mask Insert */ \
V(rldimi, RLDIMI, 0x7800000C)
#define PPC_SC_OPCODE_LIST(V) \
/* System Call */ \
V(sc, SC, 0x44000002)
#define PPC_OPCODE_LIST(V) \
PPC_X_OPCODE_LIST(V) \
PPC_XO_OPCODE_LIST(V) \
PPC_DS_OPCODE_LIST(V) \
PPC_MDS_OPCODE_LIST(V) \
PPC_MD_OPCODE_LIST(V) \
PPC_XS_OPCODE_LIST(V) \
PPC_D_OPCODE_LIST(V) \
PPC_I_OPCODE_LIST(V) \
PPC_B_OPCODE_LIST(V) \
PPC_XL_OPCODE_LIST(V) \
PPC_A_OPCODE_LIST(V) \
PPC_XFX_OPCODE_LIST(V) \
PPC_M_OPCODE_LIST(V) \
PPC_SC_OPCODE_LIST(V) \
PPC_Z23_OPCODE_LIST(V) \
PPC_Z22_OPCODE_LIST(V) \
PPC_EVX_OPCODE_LIST(V) \
PPC_XFL_OPCODE_LIST(V) \
PPC_EVS_OPCODE_LIST(V) \
PPC_VX_OPCODE_LIST(V) \
PPC_VA_OPCODE_LIST(V) \
PPC_VC_OPCODE_LIST(V) \
PPC_XX1_OPCODE_LIST(V) \
PPC_XX2_OPCODE_LIST(V) \
PPC_XX3_OPCODE_LIST(V) \
PPC_XX4_OPCODE_LIST(V)
enum Opcode : uint32_t {
#define DECLARE_INSTRUCTION(name, opcode_name, opcode_value) \
opcode_name = opcode_value,
PPC_OPCODE_LIST(DECLARE_INSTRUCTION)
#undef DECLARE_INSTRUCTION
EXT1 = 0x4C000000, // Extended code set 1
EXT2 = 0x7C000000, // Extended code set 2
EXT3 = 0xEC000000, // Extended code set 3
EXT4 = 0xFC000000, // Extended code set 4
EXT5 = 0x78000000, // Extended code set 5 - 64bit only
EXT6 = 0xF0000000, // Extended code set 6
};
// Instruction encoding bits and masks.
enum {
// Instruction encoding bit
B1 = 1 << 1,
B2 = 1 << 2,
B3 = 1 << 3,
B4 = 1 << 4,
B5 = 1 << 5,
B7 = 1 << 7,
B8 = 1 << 8,
B9 = 1 << 9,
B12 = 1 << 12,
B18 = 1 << 18,
B19 = 1 << 19,
B20 = 1 << 20,
B22 = 1 << 22,
B23 = 1 << 23,
B24 = 1 << 24,
B25 = 1 << 25,
B26 = 1 << 26,
B27 = 1 << 27,
B28 = 1 << 28,
B6 = 1 << 6,
B10 = 1 << 10,
B11 = 1 << 11,
B16 = 1 << 16,
B17 = 1 << 17,
B21 = 1 << 21,
// Instruction bit masks
kCondMask = 0x1F << 21,
kOff12Mask = (1 << 12) - 1,
kImm24Mask = (1 << 24) - 1,
kOff16Mask = (1 << 16) - 1,
kImm16Mask = (1 << 16) - 1,
kImm26Mask = (1 << 26) - 1,
kBOfieldMask = 0x1f << 21,
kOpcodeMask = 0x3f << 26,
kExt1OpcodeMask = 0x3ff << 1,
kExt2OpcodeMask = 0x3ff << 1,
kExt2OpcodeVariant2Mask = 0x1ff << 2,
kExt5OpcodeMask = 0x3 << 2,
kBOMask = 0x1f << 21,
kBIMask = 0x1F << 16,
kBDMask = 0x14 << 2,
kAAMask = 0x01 << 1,
kLKMask = 0x01,
kRCMask = 0x01,
kTOMask = 0x1f << 21
};
// -----------------------------------------------------------------------------
// Addressing modes and instruction variants.
// Overflow Exception
enum OEBit {
SetOE = 1 << 10, // Set overflow exception
LeaveOE = 0 << 10 // No overflow exception
};
// Record bit
enum RCBit { // Bit 0
SetRC = 1, // LT,GT,EQ,SO
LeaveRC = 0 // None
};
// Link bit
enum LKBit { // Bit 0
SetLK = 1, // Load effective address of next instruction
LeaveLK = 0 // No action
};
enum BOfield { // Bits 25-21
DCBNZF = 0 << 21, // Decrement CTR; branch if CTR != 0 and condition false
DCBEZF = 2 << 21, // Decrement CTR; branch if CTR == 0 and condition false
BF = 4 << 21, // Branch if condition false
DCBNZT = 8 << 21, // Decrement CTR; branch if CTR != 0 and condition true
DCBEZT = 10 << 21, // Decrement CTR; branch if CTR == 0 and condition true
BT = 12 << 21, // Branch if condition true
DCBNZ = 16 << 21, // Decrement CTR; branch if CTR != 0
DCBEZ = 18 << 21, // Decrement CTR; branch if CTR == 0
BA = 20 << 21 // Branch always
};
#if V8_OS_AIX
#undef CR_LT
#undef CR_GT
#undef CR_EQ
#undef CR_SO
#endif
enum CRBit { CR_LT = 0, CR_GT = 1, CR_EQ = 2, CR_SO = 3, CR_FU = 3 };
#define CRWIDTH 4
// These are the documented bit positions biased down by 32
enum FPSCRBit {
VXSOFT = 21, // 53: Software-Defined Condition
VXSQRT = 22, // 54: Invalid Square Root
VXCVI = 23 // 55: Invalid Integer Convert
};
// -----------------------------------------------------------------------------
// Supervisor Call (svc) specific support.
// Special Software Interrupt codes when used in the presence of the PPC
// simulator.
// svc (formerly swi) provides a 24bit immediate value. Use bits 22:0 for
// standard SoftwareInterrupCode. Bit 23 is reserved for the stop feature.
enum SoftwareInterruptCodes {
// transition to C code
kCallRtRedirected = 0x10,
// break point
kBreakpoint = 0x821008, // bits23-0 of 0x7d821008 = twge r2, r2
// stop
kStopCode = 1 << 23
};
const uint32_t kStopCodeMask = kStopCode - 1;
const uint32_t kMaxStopCode = kStopCode - 1;
const int32_t kDefaultStopCode = -1;
// FP rounding modes.
enum FPRoundingMode {
RN = 0, // Round to Nearest.
RZ = 1, // Round towards zero.
RP = 2, // Round towards Plus Infinity.
RM = 3, // Round towards Minus Infinity.
// Aliases.
kRoundToNearest = RN,
kRoundToZero = RZ,
kRoundToPlusInf = RP,
kRoundToMinusInf = RM
};
const uint32_t kFPRoundingModeMask = 3;
enum CheckForInexactConversion {
kCheckForInexactConversion,
kDontCheckForInexactConversion
};
// -----------------------------------------------------------------------------
// Specific instructions, constants, and masks.
// These constants are declared in assembler-arm.cc, as they use named registers
// and other constants.
// add(sp, sp, 4) instruction (aka Pop())
extern const Instr kPopInstruction;
// str(r, MemOperand(sp, 4, NegPreIndex), al) instruction (aka push(r))
// register r is not encoded.
extern const Instr kPushRegPattern;
// ldr(r, MemOperand(sp, 4, PostIndex), al) instruction (aka pop(r))
// register r is not encoded.
extern const Instr kPopRegPattern;
// use TWI to indicate redirection call for simulation mode
const Instr rtCallRedirInstr = TWI;
// -----------------------------------------------------------------------------
// Instruction abstraction.
// The class Instruction enables access to individual fields defined in the PPC
// architecture instruction set encoding.
// Note that the Assembler uses typedef int32_t Instr.
//
// Example: Test whether the instruction at ptr does set the condition code
// bits.
//
// bool InstructionSetsConditionCodes(byte* ptr) {
// Instruction* instr = Instruction::At(ptr);
// int type = instr->TypeValue();
// return ((type == 0) || (type == 1)) && instr->HasS();
// }
//
class Instruction {
public:
enum { kInstrSize = 4, kInstrSizeLog2 = 2, kPCReadOffset = 8 };
// Helper macro to define static accessors.
// We use the cast to char* trick to bypass the strict anti-aliasing rules.
#define DECLARE_STATIC_TYPED_ACCESSOR(return_type, Name) \
static inline return_type Name(Instr instr) { \
char* temp = reinterpret_cast<char*>(&instr); \
return reinterpret_cast<Instruction*>(temp)->Name(); \
}
#define DECLARE_STATIC_ACCESSOR(Name) DECLARE_STATIC_TYPED_ACCESSOR(int, Name)
// Get the raw instruction bits.
inline Instr InstructionBits() const {
return *reinterpret_cast<const Instr*>(this);
}
// Set the raw instruction bits to value.
inline void SetInstructionBits(Instr value) {
*reinterpret_cast<Instr*>(this) = value;
}
// Read one particular bit out of the instruction bits.
inline int Bit(int nr) const { return (InstructionBits() >> nr) & 1; }
// Read a bit field's value out of the instruction bits.
inline int Bits(int hi, int lo) const {
return (InstructionBits() >> lo) & ((2 << (hi - lo)) - 1);
}
// Read a bit field out of the instruction bits.
inline uint32_t BitField(int hi, int lo) const {
return InstructionBits() & (((2 << (hi - lo)) - 1) << lo);
}
// Static support.
// Read one particular bit out of the instruction bits.
static inline int Bit(Instr instr, int nr) { return (instr >> nr) & 1; }
// Read the value of a bit field out of the instruction bits.
static inline int Bits(Instr instr, int hi, int lo) {
return (instr >> lo) & ((2 << (hi - lo)) - 1);
}
// Read a bit field out of the instruction bits.
static inline uint32_t BitField(Instr instr, int hi, int lo) {
return instr & (((2 << (hi - lo)) - 1) << lo);
}
inline int RSValue() const { return Bits(25, 21); }
inline int RTValue() const { return Bits(25, 21); }
inline int RAValue() const { return Bits(20, 16); }
DECLARE_STATIC_ACCESSOR(RAValue);
inline int RBValue() const { return Bits(15, 11); }
DECLARE_STATIC_ACCESSOR(RBValue);
inline int RCValue() const { return Bits(10, 6); }
DECLARE_STATIC_ACCESSOR(RCValue);
inline int OpcodeValue() const { return static_cast<Opcode>(Bits(31, 26)); }
inline Opcode OpcodeField() const {
return static_cast<Opcode>(BitField(31, 26));
}
// Fields used in Software interrupt instructions
inline SoftwareInterruptCodes SvcValue() const {
return static_cast<SoftwareInterruptCodes>(Bits(23, 0));
}
// Instructions are read of out a code stream. The only way to get a
// reference to an instruction is to convert a pointer. There is no way
// to allocate or create instances of class Instruction.
// Use the At(pc) function to create references to Instruction.
static Instruction* At(byte* pc) {
return reinterpret_cast<Instruction*>(pc);
}
private:
// We need to prevent the creation of instances of class Instruction.
DISALLOW_IMPLICIT_CONSTRUCTORS(Instruction);
};
// Helper functions for converting between register numbers and names.
class Registers {
public:
// Lookup the register number for the name provided.
static int Number(const char* name);
private:
static const char* names_[kNumRegisters];
};
// Helper functions for converting between FP register numbers and names.
class DoubleRegisters {
public:
// Lookup the register number for the name provided.
static int Number(const char* name);
private:
static const char* names_[kNumDoubleRegisters];
};
} // namespace internal
} // namespace v8
#endif // V8_PPC_CONSTANTS_PPC_H_