/*

 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

 *

 * This code is free software; you can redistribute it and/or modify it

 * under the terms of the GNU General Public License version 2 only, as

 * published by the Free Software Foundation.

 *

 * This code is distributed in the hope that it will be useful, but WITHOUT

 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or

 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License

 * version 2 for more details (a copy is included in the LICENSE file that

 * accompanied this code).

 *

 * You should have received a copy of the GNU General Public License version

 * 2 along with this work; if not, write to the Free Software Foundation,

 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

 *

 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

 * or visit www.oracle.com if you need additional information or have any

 * questions.

 */

package org.graalvm.compiler.lir.aarch64;

import static org.graalvm.compiler.lir.LIRInstruction.OperandFlag.REG;

import static org.graalvm.compiler.lir.aarch64.AArch64ArithmeticOp.ARMv8ConstantCategory.ARITHMETIC;

import static org.graalvm.compiler.lir.aarch64.AArch64ArithmeticOp.ARMv8ConstantCategory.LOGICAL;

import static org.graalvm.compiler.lir.aarch64.AArch64ArithmeticOp.ARMv8ConstantCategory.NONE;

import static org.graalvm.compiler.lir.aarch64.AArch64ArithmeticOp.ARMv8ConstantCategory.SHIFT;

import static jdk.vm.ci.aarch64.AArch64.zr;

import static jdk.vm.ci.code.ValueUtil.asRegister;

import org.graalvm.compiler.asm.aarch64.AArch64Assembler;

import org.graalvm.compiler.asm.aarch64.AArch64Assembler.ConditionFlag;

import org.graalvm.compiler.debug.GraalError;

import org.graalvm.compiler.asm.aarch64.AArch64MacroAssembler;

import org.graalvm.compiler.lir.LIRInstructionClass;

import org.graalvm.compiler.lir.Opcode;

import org.graalvm.compiler.lir.asm.CompilationResultBuilder;

import jdk.vm.ci.code.Register;

import jdk.vm.ci.meta.AllocatableValue;

import jdk.vm.ci.meta.JavaConstant;

public enum AArch64ArithmeticOp {

    // TODO At least add and sub *can* be used with SP, so this should be supported

    NEG,

    NOT,

    ADD(ARITHMETIC),

    ADDS(ARITHMETIC),

    SUB(ARITHMETIC),

    SUBS(ARITHMETIC),

    MUL,

    MULVS,

    MNEG,

    DIV,

    SMULH,

    UMULH,

    REM,

    UDIV,

    UREM,

    AND(LOGICAL),

    ANDS(LOGICAL),

    OR(LOGICAL),

    XOR(LOGICAL),

    SHL(SHIFT),

    LSHR(SHIFT),

    ASHR(SHIFT),

    ABS,

    FADD,

    FSUB,

    FMUL,

    FDIV,

    FREM,

    FNEG,

    FABS,

    FRINTM,

    FRINTN,

    FRINTP,

    SQRT;

    /**

     * Specifies what constants can be used directly without having to be loaded into a register

     * with the given instruction.

     */

    public enum ARMv8ConstantCategory {

        NONE,

        LOGICAL,

        ARITHMETIC,

        SHIFT

    }

    public final ARMv8ConstantCategory category;

    AArch64ArithmeticOp(ARMv8ConstantCategory category) {

        this.category = category;

    }

    AArch64ArithmeticOp() {

        this(NONE);

    }

    public static class UnaryOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<UnaryOp> TYPE = LIRInstructionClass.create(UnaryOp.class);

        @Opcode private final AArch64ArithmeticOp opcode;

        @Def({REG}) protected AllocatableValue result;

        @Use({REG}) protected AllocatableValue x;

        public UnaryOp(AArch64ArithmeticOp opcode, AllocatableValue result, AllocatableValue x) {

            super(TYPE);

            this.opcode = opcode;

            this.result = result;

            this.x = x;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            Register dst = asRegister(result);

            Register src = asRegister(x);

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            switch (opcode) {

                case NEG:

                    masm.sub(size, dst, zr, src);

                    break;

                case FNEG:

                    masm.fneg(size, dst, src);

                    break;

                case NOT:

                    masm.not(size, dst, src);

                    break;

                case ABS:

                    masm.cmp(size, src, 0);

                    masm.csneg(size, dst, src, ConditionFlag.LT);

                    break;

                case FABS:

                    masm.fabs(size, dst, src);

                    break;

                case FRINTM:

                    masm.frintm(size, dst, src);

                    break;

                case FRINTN:

                    masm.frintn(size, dst, src);

                    break;

                case FRINTP:

                    masm.frintp(size, dst, src);

                    break;

                case SQRT:

                    masm.fsqrt(size, dst, src);

                    break;

                default:

                    throw GraalError.shouldNotReachHere("op=" + opcode.name());

            }

        }

    }

    public static class BinaryConstOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<BinaryConstOp> TYPE = LIRInstructionClass.create(BinaryConstOp.class);

        @Opcode private final AArch64ArithmeticOp op;

        @Def({REG}) protected AllocatableValue result;

        @Use({REG}) protected AllocatableValue a;

        private final JavaConstant b;

        public BinaryConstOp(AArch64ArithmeticOp op, AllocatableValue result, AllocatableValue a, JavaConstant b) {

            super(TYPE);

            this.op = op;

            this.result = result;

            this.a = a;

            this.b = b;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            assert op.category != NONE;

            Register dst = asRegister(result);

            Register src = asRegister(a);

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            switch (op) {

                case ADD:

                    // Don't use asInt() here, since we can't use asInt on a long variable, even

                    // if the constant easily fits as an int.

                    assert AArch64MacroAssembler.isArithmeticImmediate(b.asLong());

                    masm.add(size, dst, src, (int) b.asLong());

                    break;

                case SUB:

                    // Don't use asInt() here, since we can't use asInt on a long variable, even

                    // if the constant easily fits as an int.

                    assert AArch64MacroAssembler.isArithmeticImmediate(b.asLong());

                    masm.sub(size, dst, src, (int) b.asLong());

                    break;

                case ADDS:

                    assert AArch64MacroAssembler.isArithmeticImmediate(b.asLong());

                    masm.adds(size, dst, src, (int) b.asLong());

                    break;

                case SUBS:

                    assert AArch64MacroAssembler.isArithmeticImmediate(b.asLong());

                    masm.subs(size, dst, src, (int) b.asLong());

                    break;

                case AND:

                    // XXX Should this be handled somewhere else?

                    if (size == 32 && b.asLong() == 0xFFFF_FFFFL) {

                        masm.mov(size, dst, src);

                    } else {

                        masm.and(size, dst, src, b.asLong());

                    }

                    break;

                case ANDS:

                    masm.ands(size, dst, src, b.asLong());

                    break;

                case OR:

                    masm.or(size, dst, src, b.asLong());

                    break;

                case XOR:

                    masm.eor(size, dst, src, b.asLong());

                    break;

                case SHL:

                    masm.shl(size, dst, src, b.asLong());

                    break;

                case LSHR:

                    masm.lshr(size, dst, src, b.asLong());

                    break;

                case ASHR:

                    masm.ashr(size, dst, src, b.asLong());

                    break;

                default:

                    throw GraalError.shouldNotReachHere("op=" + op.name());

            }

        }

    }

    public static class BinaryOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<BinaryOp> TYPE = LIRInstructionClass.create(BinaryOp.class);

        @Opcode private final AArch64ArithmeticOp op;

        @Def({REG}) protected AllocatableValue result;

        @Use({REG}) protected AllocatableValue a;

        @Use({REG}) protected AllocatableValue b;

        public BinaryOp(AArch64ArithmeticOp op, AllocatableValue result, AllocatableValue a, AllocatableValue b) {

            super(TYPE);

            this.op = op;

            this.result = result;

            this.a = a;

            this.b = b;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            Register dst = asRegister(result);

            Register src1 = asRegister(a);

            Register src2 = asRegister(b);

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            switch (op) {

                case ADD:

                    masm.add(size, dst, src1, src2);

                    break;

                case ADDS:

                    masm.adds(size, dst, src1, src2);

                    break;

                case SUB:

                    masm.sub(size, dst, src1, src2);

                    break;

                case SUBS:

                    masm.subs(size, dst, src1, src2);

                    break;

                case MUL:

                    masm.mul(size, dst, src1, src2);

                    break;

                case UMULH:

                    masm.umulh(size, dst, src1, src2);

                    break;

                case SMULH:

                    masm.smulh(size, dst, src1, src2);

                    break;

                case MNEG:

                    masm.mneg(size, dst, src1, src2);

                    break;

                case DIV:

                    masm.sdiv(size, dst, src1, src2);

                    break;

                case UDIV:

                    masm.udiv(size, dst, src1, src2);

                    break;

                case AND:

                    masm.and(size, dst, src1, src2);

                    break;

                case ANDS:

                    masm.ands(size, dst, src1, src2);

                    break;

                case OR:

                    masm.or(size, dst, src1, src2);

                    break;

                case XOR:

                    masm.eor(size, dst, src1, src2);

                    break;

                case SHL:

                    masm.shl(size, dst, src1, src2);

                    break;

                case LSHR:

                    masm.lshr(size, dst, src1, src2);

                    break;

                case ASHR:

                    masm.ashr(size, dst, src1, src2);

                    break;

                case FADD:

                    masm.fadd(size, dst, src1, src2);

                    break;

                case FSUB:

                    masm.fsub(size, dst, src1, src2);

                    break;

                case FMUL:

                    masm.fmul(size, dst, src1, src2);

                    break;

                case FDIV:

                    masm.fdiv(size, dst, src1, src2);

                    break;

                case MULVS:

                    masm.mulvs(size, dst, src1, src2);

                    break;

                default:

                    throw GraalError.shouldNotReachHere("op=" + op.name());

            }

        }

    }

    /**

     * Class used for instructions that have to reuse one of their arguments. This only applies to

     * the remainder instructions at the moment, since we have to compute n % d using rem = n -

     * TruncatingDivision(n, d) * d

     *

     * TODO (das) Replace the remainder nodes in the LIR.

     */

    public static class BinaryCompositeOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<BinaryCompositeOp> TYPE = LIRInstructionClass.create(BinaryCompositeOp.class);

        @Opcode private final AArch64ArithmeticOp op;

        @Def({REG}) protected AllocatableValue result;

        @Alive({REG}) protected AllocatableValue a;

        @Alive({REG}) protected AllocatableValue b;

        public BinaryCompositeOp(AArch64ArithmeticOp op, AllocatableValue result, AllocatableValue a, AllocatableValue b) {

            super(TYPE);

            this.op = op;

            this.result = result;

            this.a = a;

            this.b = b;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            Register dst = asRegister(result);

            Register src1 = asRegister(a);

            Register src2 = asRegister(b);

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            switch (op) {

                case REM:

                    masm.rem(size, dst, src1, src2);

                    break;

                case UREM:

                    masm.urem(size, dst, src1, src2);

                    break;

                case FREM:

                    masm.frem(size, dst, src1, src2);

                    break;

                default:

                    throw GraalError.shouldNotReachHere();

            }

        }

    }

    public static class BinaryShiftOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<BinaryShiftOp> TYPE = LIRInstructionClass.create(BinaryShiftOp.class);

        @Opcode private final AArch64ArithmeticOp op;

        @Def(REG) protected AllocatableValue result;

        @Use(REG) protected AllocatableValue src1;

        @Use(REG) protected AllocatableValue src2;

        private final AArch64MacroAssembler.ShiftType shiftType;

        private final int shiftAmt;

        private final boolean isShiftNot;

        /**

         * If shiftNot: Computes <code>result = src1 <op> ~(src2 <shiftType> <shiftAmt>)</code>

         * (Only for logic ops). else: Computes

         * <code>result = src1 <op> src2 <shiftType> <shiftAmt></code>.

         */

        public BinaryShiftOp(AArch64ArithmeticOp op, AllocatableValue result, AllocatableValue src1, AllocatableValue src2,

                        AArch64MacroAssembler.ShiftType shiftType, int shiftAmt, boolean isShiftNot) {

            super(TYPE);

            assert op == ADD || op == SUB || op == AND || op == OR || op == XOR;

            this.op = op;

            this.result = result;

            this.src1 = src1;

            this.src2 = src2;

            this.shiftType = shiftType;

            this.shiftAmt = shiftAmt;

            this.isShiftNot = isShiftNot;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            switch (op) {

                case ADD:

                    masm.add(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    break;

                case SUB:

                    masm.sub(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    break;

                case AND:

                    if (!isShiftNot) {

                        masm.and(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    } else {

                        masm.bic(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    }

                    break;

                case OR:

                    if (!isShiftNot) {

                        masm.or(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    } else {

                        masm.orn(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    }

                    break;

                case XOR:

                    if (!isShiftNot) {

                        masm.eor(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    } else {

                        masm.eon(size, asRegister(result), asRegister(src1), asRegister(src2), shiftType, shiftAmt);

                    }

                    break;

                default:

                    throw GraalError.shouldNotReachHere();

            }

        }

    }

    public static class ExtendedAddShiftOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<ExtendedAddShiftOp> TYPE = LIRInstructionClass.create(ExtendedAddShiftOp.class);

        @Def(REG) protected AllocatableValue result;

        @Use(REG) protected AllocatableValue src1;

        @Use(REG) protected AllocatableValue src2;

        private final AArch64Assembler.ExtendType extendType;

        private final int shiftAmt;

        /**

         * Computes <code>result = src1 + extendType(src2) << shiftAmt</code>.

         *

         * @param extendType defines how src2 is extended to the same size as src1.

         * @param shiftAmt must be in range 0 to 4.

         */

        public ExtendedAddShiftOp(AllocatableValue result, AllocatableValue src1, AllocatableValue src2, AArch64Assembler.ExtendType extendType, int shiftAmt) {

            super(TYPE);

            this.result = result;

            this.src1 = src1;

            this.src2 = src2;

            this.extendType = extendType;

            this.shiftAmt = shiftAmt;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            masm.add(size, asRegister(result), asRegister(src1), asRegister(src2), extendType, shiftAmt);

        }

    }

    public static class MultiplyAddSubOp extends AArch64LIRInstruction {

        private static final LIRInstructionClass<MultiplyAddSubOp> TYPE = LIRInstructionClass.create(MultiplyAddSubOp.class);

        @Opcode private final AArch64ArithmeticOp op;

        @Def(REG) protected AllocatableValue result;

        @Use(REG) protected AllocatableValue src1;

        @Use(REG) protected AllocatableValue src2;

        @Use(REG) protected AllocatableValue src3;

        /**

         * Computes <code>result = src3 <op> src1 * src2</code>.

         */

        public MultiplyAddSubOp(AArch64ArithmeticOp op, AllocatableValue result, AllocatableValue src1, AllocatableValue src2, AllocatableValue src3) {

            super(TYPE);

            this.op = op;

            this.result = result;

            this.src1 = src1;

            this.src2 = src2;

            this.src3 = src3;

        }

        @Override

        public void emitCode(CompilationResultBuilder crb, AArch64MacroAssembler masm) {

            int size = result.getPlatformKind().getSizeInBytes() * Byte.SIZE;

            switch (op) {

                case ADD:

                    masm.madd(size, asRegister(result), asRegister(src1), asRegister(src2), asRegister(src3));

                    break;

                case SUB:

                    masm.msub(size, asRegister(result), asRegister(src1), asRegister(src2), asRegister(src3));

                    break;

                default:

                    throw GraalError.shouldNotReachHere();

            }

        }

    }

}