Add MIPS64r6 support for libpixelflinger

Added ARMtoMips64Assembler class that translates ARM instructions to MIPS64r6
instructions.
Added MIPS64r6 assembly implementations for col32cb16_blend and t32cb16_blend
functions.
Added MIPS32r2 assembly implementation for col32cb16_blend function.
Added tests for MIPS64r6 (assembler, disassembler and assembly implementation
functions).
Added MIPS32 tests for assembly implementation functions.
Minor bug fixes for MIPS32r2 branch.

Change-Id: I69e49622117be5b8167628e9702db6aafb1849d7
diff --git a/libpixelflinger/codeflinger/ARMAssemblerInterface.h b/libpixelflinger/codeflinger/ARMAssemblerInterface.h
index 40cbfcf..72935ac 100644
--- a/libpixelflinger/codeflinger/ARMAssemblerInterface.h
+++ b/libpixelflinger/codeflinger/ARMAssemblerInterface.h
@@ -63,7 +63,7 @@
     };
 
     enum {
-        CODEGEN_ARCH_ARM = 1, CODEGEN_ARCH_MIPS, CODEGEN_ARCH_ARM64
+        CODEGEN_ARCH_ARM = 1, CODEGEN_ARCH_MIPS, CODEGEN_ARCH_ARM64, CODEGEN_ARCH_MIPS64
     };
 
     // -----------------------------------------------------------------------
@@ -115,7 +115,8 @@
     // data processing...
     enum {
         opAND, opEOR, opSUB, opRSB, opADD, opADC, opSBC, opRSC, 
-        opTST, opTEQ, opCMP, opCMN, opORR, opMOV, opBIC, opMVN
+        opTST, opTEQ, opCMP, opCMN, opORR, opMOV, opBIC, opMVN,
+        opADD64, opSUB64
     };
 
     virtual void
diff --git a/libpixelflinger/codeflinger/GGLAssembler.cpp b/libpixelflinger/codeflinger/GGLAssembler.cpp
index 325caba..346779f 100644
--- a/libpixelflinger/codeflinger/GGLAssembler.cpp
+++ b/libpixelflinger/codeflinger/GGLAssembler.cpp
@@ -893,7 +893,8 @@
         return;
     }
     
-    if (getCodegenArch() == CODEGEN_ARCH_MIPS) {
+    if ((getCodegenArch() == CODEGEN_ARCH_MIPS) ||
+        (getCodegenArch() == CODEGEN_ARCH_MIPS64)) {
         // MIPS can do 16-bit imm in 1 instr, 32-bit in 3 instr
         // the below ' while (mask)' code is buggy on mips
         // since mips returns true on isValidImmediate()
@@ -1057,7 +1058,8 @@
 RegisterAllocator::RegisterFile::RegisterFile(int codegen_arch)
     : mRegs(0), mTouched(0), mStatus(0), mArch(codegen_arch), mRegisterOffset(0)
 {
-    if (mArch == ARMAssemblerInterface::CODEGEN_ARCH_MIPS) {
+    if ((mArch == ARMAssemblerInterface::CODEGEN_ARCH_MIPS) ||
+        (mArch == ARMAssemblerInterface::CODEGEN_ARCH_MIPS64)) {
         mRegisterOffset = 2;    // ARM has regs 0..15, MIPS offset to 2..17
     }
     reserve(ARMAssemblerInterface::SP);
@@ -1067,7 +1069,8 @@
 RegisterAllocator::RegisterFile::RegisterFile(const RegisterFile& rhs, int codegen_arch)
     : mRegs(rhs.mRegs), mTouched(rhs.mTouched), mArch(codegen_arch), mRegisterOffset(0)
 {
-    if (mArch == ARMAssemblerInterface::CODEGEN_ARCH_MIPS) {
+    if ((mArch == ARMAssemblerInterface::CODEGEN_ARCH_MIPS) ||
+        (mArch == ARMAssemblerInterface::CODEGEN_ARCH_MIPS64)) {
         mRegisterOffset = 2;    // ARM has regs 0..15, MIPS offset to 2..17
     }
 }
diff --git a/libpixelflinger/codeflinger/MIPS64Assembler.cpp b/libpixelflinger/codeflinger/MIPS64Assembler.cpp
new file mode 100644
index 0000000..a5305cc
--- /dev/null
+++ b/libpixelflinger/codeflinger/MIPS64Assembler.cpp
@@ -0,0 +1,1452 @@
+/* libs/pixelflinger/codeflinger/MIPS64Assembler.cpp
+**
+** Copyright 2015, The Android Open Source Project
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+
+/* MIPS64 assembler and ARM->MIPS64 assembly translator
+**
+** The approach is utilize MIPSAssembler generator, using inherited MIPS64Assembler
+** that overrides just the specific MIPS64r6 instructions.
+** For now ArmToMips64Assembler is copied over from ArmToMipsAssembler class,
+** changing some MIPS64r6 related stuff.
+**
+*/
+
+
+#define LOG_TAG "MIPS64Assembler"
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <cutils/log.h>
+#include <cutils/properties.h>
+
+#if defined(WITH_LIB_HARDWARE)
+#include <hardware_legacy/qemu_tracing.h>
+#endif
+
+#include <private/pixelflinger/ggl_context.h>
+
+#include "MIPS64Assembler.h"
+#include "CodeCache.h"
+#include "mips64_disassem.h"
+
+
+#define NOT_IMPLEMENTED()  LOG_ALWAYS_FATAL("Arm instruction %s not yet implemented\n", __func__)
+
+
+// ----------------------------------------------------------------------------
+
+namespace android {
+
+// ----------------------------------------------------------------------------
+#if 0
+#pragma mark -
+#pragma mark ArmToMips64Assembler...
+#endif
+
+ArmToMips64Assembler::ArmToMips64Assembler(const sp<Assembly>& assembly,
+                                           char *abuf, int linesz, int instr_count)
+    :   ARMAssemblerInterface(),
+        mArmDisassemblyBuffer(abuf),
+        mArmLineLength(linesz),
+        mArmInstrCount(instr_count),
+        mInum(0),
+        mAssembly(assembly)
+{
+    mMips = new MIPS64Assembler(assembly, this);
+    mArmPC = (uint32_t **) malloc(ARM_MAX_INSTUCTIONS * sizeof(uint32_t *));
+    init_conditional_labels();
+}
+
+ArmToMips64Assembler::ArmToMips64Assembler(void* assembly)
+    :   ARMAssemblerInterface(),
+        mArmDisassemblyBuffer(NULL),
+        mInum(0),
+        mAssembly(NULL)
+{
+    mMips = new MIPS64Assembler(assembly, this);
+    mArmPC = (uint32_t **) malloc(ARM_MAX_INSTUCTIONS * sizeof(uint32_t *));
+    init_conditional_labels();
+}
+
+ArmToMips64Assembler::~ArmToMips64Assembler()
+{
+    delete mMips;
+    free((void *) mArmPC);
+}
+
+uint32_t* ArmToMips64Assembler::pc() const
+{
+    return mMips->pc();
+}
+
+uint32_t* ArmToMips64Assembler::base() const
+{
+    return mMips->base();
+}
+
+void ArmToMips64Assembler::reset()
+{
+    cond.labelnum = 0;
+    mInum = 0;
+    mMips->reset();
+}
+
+int ArmToMips64Assembler::getCodegenArch()
+{
+    return CODEGEN_ARCH_MIPS64;
+}
+
+void ArmToMips64Assembler::comment(const char* string)
+{
+    mMips->comment(string);
+}
+
+void ArmToMips64Assembler::label(const char* theLabel)
+{
+    mMips->label(theLabel);
+}
+
+void ArmToMips64Assembler::disassemble(const char* name)
+{
+    mMips->disassemble(name);
+}
+
+void ArmToMips64Assembler::init_conditional_labels()
+{
+    int i;
+    for (i=0;i<99; ++i) {
+        sprintf(cond.label[i], "cond_%d", i);
+    }
+}
+
+
+
+#if 0
+#pragma mark -
+#pragma mark Prolog/Epilog & Generate...
+#endif
+
+void ArmToMips64Assembler::prolog()
+{
+    mArmPC[mInum++] = pc();  // save starting PC for this instr
+
+    mMips->DADDIU(R_sp, R_sp, -(5 * 8));
+    mMips->SD(R_s0, R_sp, 0);
+    mMips->SD(R_s1, R_sp, 8);
+    mMips->SD(R_s2, R_sp, 16);
+    mMips->SD(R_s3, R_sp, 24);
+    mMips->SD(R_s4, R_sp, 32);
+    mMips->MOVE(R_v0, R_a0);    // move context * passed in a0 to v0 (arm r0)
+}
+
+void ArmToMips64Assembler::epilog(uint32_t touched)
+{
+    mArmPC[mInum++] = pc();  // save starting PC for this instr
+
+    mMips->LD(R_s0, R_sp, 0);
+    mMips->LD(R_s1, R_sp, 8);
+    mMips->LD(R_s2, R_sp, 16);
+    mMips->LD(R_s3, R_sp, 24);
+    mMips->LD(R_s4, R_sp, 32);
+    mMips->DADDIU(R_sp, R_sp, (5 * 8));
+    mMips->JR(R_ra);
+
+}
+
+int ArmToMips64Assembler::generate(const char* name)
+{
+    return mMips->generate(name);
+}
+
+void ArmToMips64Assembler::fix_branches()
+{
+    mMips->fix_branches();
+}
+
+uint32_t* ArmToMips64Assembler::pcForLabel(const char* label)
+{
+    return mMips->pcForLabel(label);
+}
+
+void ArmToMips64Assembler::set_condition(int mode, int R1, int R2) {
+    if (mode == 2) {
+        cond.type = SBIT_COND;
+    } else {
+        cond.type = CMP_COND;
+    }
+    cond.r1 = R1;
+    cond.r2 = R2;
+}
+
+//----------------------------------------------------------
+
+#if 0
+#pragma mark -
+#pragma mark Addressing modes & shifters...
+#endif
+
+
+// do not need this for MIPS, but it is in the Interface (virtual)
+int ArmToMips64Assembler::buildImmediate(
+        uint32_t immediate, uint32_t& rot, uint32_t& imm)
+{
+    // for MIPS, any 32-bit immediate is OK
+    rot = 0;
+    imm = immediate;
+    return 0;
+}
+
+// shifters...
+
+bool ArmToMips64Assembler::isValidImmediate(uint32_t immediate)
+{
+    // for MIPS, any 32-bit immediate is OK
+    return true;
+}
+
+uint32_t ArmToMips64Assembler::imm(uint32_t immediate)
+{
+    amode.value = immediate;
+    return AMODE_IMM;
+}
+
+uint32_t ArmToMips64Assembler::reg_imm(int Rm, int type, uint32_t shift)
+{
+    amode.reg = Rm;
+    amode.stype = type;
+    amode.value = shift;
+    return AMODE_REG_IMM;
+}
+
+uint32_t ArmToMips64Assembler::reg_rrx(int Rm)
+{
+    // reg_rrx mode is not used in the GLLAssember code at this time
+    return AMODE_UNSUPPORTED;
+}
+
+uint32_t ArmToMips64Assembler::reg_reg(int Rm, int type, int Rs)
+{
+    // reg_reg mode is not used in the GLLAssember code at this time
+    return AMODE_UNSUPPORTED;
+}
+
+
+// addressing modes...
+// LDR(B)/STR(B)/PLD (immediate and Rm can be negative, which indicate U=0)
+uint32_t ArmToMips64Assembler::immed12_pre(int32_t immed12, int W)
+{
+    LOG_ALWAYS_FATAL_IF(abs(immed12) >= 0x800,
+                        "LDR(B)/STR(B)/PLD immediate too big (%08x)",
+                        immed12);
+    amode.value = immed12;
+    amode.writeback = W;
+    return AMODE_IMM_12_PRE;
+}
+
+uint32_t ArmToMips64Assembler::immed12_post(int32_t immed12)
+{
+    LOG_ALWAYS_FATAL_IF(abs(immed12) >= 0x800,
+                        "LDR(B)/STR(B)/PLD immediate too big (%08x)",
+                        immed12);
+
+    amode.value = immed12;
+    return AMODE_IMM_12_POST;
+}
+
+uint32_t ArmToMips64Assembler::reg_scale_pre(int Rm, int type,
+        uint32_t shift, int W)
+{
+    LOG_ALWAYS_FATAL_IF(W | type | shift, "reg_scale_pre adv modes not yet implemented");
+
+    amode.reg = Rm;
+    // amode.stype = type;      // more advanced modes not used in GGLAssembler yet
+    // amode.value = shift;
+    // amode.writeback = W;
+    return AMODE_REG_SCALE_PRE;
+}
+
+uint32_t ArmToMips64Assembler::reg_scale_post(int Rm, int type, uint32_t shift)
+{
+    LOG_ALWAYS_FATAL("adr mode reg_scale_post not yet implemented\n");
+    return AMODE_UNSUPPORTED;
+}
+
+// LDRH/LDRSB/LDRSH/STRH (immediate and Rm can be negative, which indicate U=0)
+uint32_t ArmToMips64Assembler::immed8_pre(int32_t immed8, int W)
+{
+    LOG_ALWAYS_FATAL("adr mode immed8_pre not yet implemented\n");
+
+    LOG_ALWAYS_FATAL_IF(abs(immed8) >= 0x100,
+                        "LDRH/LDRSB/LDRSH/STRH immediate too big (%08x)",
+                        immed8);
+    return AMODE_IMM_8_PRE;
+}
+
+uint32_t ArmToMips64Assembler::immed8_post(int32_t immed8)
+{
+    LOG_ALWAYS_FATAL_IF(abs(immed8) >= 0x100,
+                        "LDRH/LDRSB/LDRSH/STRH immediate too big (%08x)",
+                        immed8);
+    amode.value = immed8;
+    return AMODE_IMM_8_POST;
+}
+
+uint32_t ArmToMips64Assembler::reg_pre(int Rm, int W)
+{
+    LOG_ALWAYS_FATAL_IF(W, "reg_pre writeback not yet implemented");
+    amode.reg = Rm;
+    return AMODE_REG_PRE;
+}
+
+uint32_t ArmToMips64Assembler::reg_post(int Rm)
+{
+    LOG_ALWAYS_FATAL("adr mode reg_post not yet implemented\n");
+    return AMODE_UNSUPPORTED;
+}
+
+
+
+// ----------------------------------------------------------------------------
+
+#if 0
+#pragma mark -
+#pragma mark Data Processing...
+#endif
+
+
+static const char * const dpOpNames[] = {
+    "AND", "EOR", "SUB", "RSB", "ADD", "ADC", "SBC", "RSC",
+    "TST", "TEQ", "CMP", "CMN", "ORR", "MOV", "BIC", "MVN"
+};
+
+// check if the operand registers from a previous CMP or S-bit instruction
+// would be overwritten by this instruction. If so, move the value to a
+// safe register.
+// Note that we cannot tell at _this_ instruction time if a future (conditional)
+// instruction will _also_ use this value (a defect of the simple 1-pass, one-
+// instruction-at-a-time translation). Therefore we must be conservative and
+// save the value before it is overwritten. This costs an extra MOVE instr.
+
+void ArmToMips64Assembler::protectConditionalOperands(int Rd)
+{
+    if (Rd == cond.r1) {
+        mMips->MOVE(R_cmp, cond.r1);
+        cond.r1 = R_cmp;
+    }
+    if (cond.type == CMP_COND && Rd == cond.r2) {
+        mMips->MOVE(R_cmp2, cond.r2);
+        cond.r2 = R_cmp2;
+    }
+}
+
+
+// interprets the addressing mode, and generates the common code
+// used by the majority of data-processing ops. Many MIPS instructions
+// have a register-based form and a different immediate form. See
+// opAND below for an example. (this could be inlined)
+//
+// this works with the imm(), reg_imm() methods above, which are directly
+// called by the GLLAssembler.
+// note: _signed parameter defaults to false (un-signed)
+// note: tmpReg parameter defaults to 1, MIPS register AT
+int ArmToMips64Assembler::dataProcAdrModes(int op, int& source, bool _signed, int tmpReg)
+{
+    if (op < AMODE_REG) {
+        source = op;
+        return SRC_REG;
+    } else if (op == AMODE_IMM) {
+        if ((!_signed && amode.value > 0xffff)
+                || (_signed && ((int)amode.value < -32768 || (int)amode.value > 32767) )) {
+            mMips->LUI(tmpReg, (amode.value >> 16));
+            if (amode.value & 0x0000ffff) {
+                mMips->ORI(tmpReg, tmpReg, (amode.value & 0x0000ffff));
+            }
+            source = tmpReg;
+            return SRC_REG;
+        } else {
+            source = amode.value;
+            return SRC_IMM;
+        }
+    } else if (op == AMODE_REG_IMM) {
+        switch (amode.stype) {
+            case LSL: mMips->SLL(tmpReg, amode.reg, amode.value); break;
+            case LSR: mMips->SRL(tmpReg, amode.reg, amode.value); break;
+            case ASR: mMips->SRA(tmpReg, amode.reg, amode.value); break;
+            case ROR: mMips->ROTR(tmpReg, amode.reg, amode.value); break;
+        }
+        source = tmpReg;
+        return SRC_REG;
+    } else {  // adr mode RRX is not used in GGL Assembler at this time
+        // we are screwed, this should be exception, assert-fail or something
+        LOG_ALWAYS_FATAL("adr mode reg_rrx not yet implemented\n");
+        return SRC_ERROR;
+    }
+}
+
+
+void ArmToMips64Assembler::dataProcessing(int opcode, int cc,
+        int s, int Rd, int Rn, uint32_t Op2)
+{
+    int src;    // src is modified by dataProcAdrModes() - passed as int&
+
+    if (cc != AL) {
+        protectConditionalOperands(Rd);
+        // the branch tests register(s) set by prev CMP or instr with 'S' bit set
+        // inverse the condition to jump past this conditional instruction
+        ArmToMips64Assembler::B(cc^1, cond.label[++cond.labelnum]);
+    } else {
+        mArmPC[mInum++] = pc();  // save starting PC for this instr
+    }
+
+    switch (opcode) {
+    case opAND:
+        if (dataProcAdrModes(Op2, src) == SRC_REG) {
+            mMips->AND(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->ANDI(Rd, Rn, src);
+        }
+        break;
+
+    case opADD:
+        // set "signed" to true for adr modes
+        if (dataProcAdrModes(Op2, src, true) == SRC_REG) {
+            mMips->ADDU(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->ADDIU(Rd, Rn, src);
+        }
+        break;
+
+    case opSUB:
+        // set "signed" to true for adr modes
+        if (dataProcAdrModes(Op2, src, true) == SRC_REG) {
+            mMips->SUBU(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->SUBIU(Rd, Rn, src);
+        }
+        break;
+
+    case opADD64:
+        // set "signed" to true for adr modes
+        if (dataProcAdrModes(Op2, src, true) == SRC_REG) {
+            mMips->DADDU(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->DADDIU(Rd, Rn, src);
+        }
+        break;
+
+    case opSUB64:
+        // set "signed" to true for adr modes
+        if (dataProcAdrModes(Op2, src, true) == SRC_REG) {
+            mMips->DSUBU(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->DSUBIU(Rd, Rn, src);
+        }
+        break;
+
+    case opEOR:
+        if (dataProcAdrModes(Op2, src) == SRC_REG) {
+            mMips->XOR(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->XORI(Rd, Rn, src);
+        }
+        break;
+
+    case opORR:
+        if (dataProcAdrModes(Op2, src) == SRC_REG) {
+            mMips->OR(Rd, Rn, src);
+        } else {                        // adr mode was SRC_IMM
+            mMips->ORI(Rd, Rn, src);
+        }
+        break;
+
+    case opBIC:
+        if (dataProcAdrModes(Op2, src) == SRC_IMM) {
+            // if we are 16-bit imnmediate, load to AT reg
+            mMips->ORI(R_at, 0, src);
+            src = R_at;
+        }
+        mMips->NOT(R_at, src);
+        mMips->AND(Rd, Rn, R_at);
+        break;
+
+    case opRSB:
+        if (dataProcAdrModes(Op2, src) == SRC_IMM) {
+            // if we are 16-bit imnmediate, load to AT reg
+            mMips->ORI(R_at, 0, src);
+            src = R_at;
+        }
+        mMips->SUBU(Rd, src, Rn);   // subu with the parameters reversed
+        break;
+
+    case opMOV:
+        if (Op2 < AMODE_REG) {  // op2 is reg # in this case
+            mMips->MOVE(Rd, Op2);
+        } else if (Op2 == AMODE_IMM) {
+            if (amode.value > 0xffff) {
+                mMips->LUI(Rd, (amode.value >> 16));
+                if (amode.value & 0x0000ffff) {
+                    mMips->ORI(Rd, Rd, (amode.value & 0x0000ffff));
+                }
+             } else {
+                mMips->ORI(Rd, 0, amode.value);
+            }
+        } else if (Op2 == AMODE_REG_IMM) {
+            switch (amode.stype) {
+            case LSL: mMips->SLL(Rd, amode.reg, amode.value); break;
+            case LSR: mMips->SRL(Rd, amode.reg, amode.value); break;
+            case ASR: mMips->SRA(Rd, amode.reg, amode.value); break;
+            case ROR: mMips->ROTR(Rd, amode.reg, amode.value); break;
+            }
+        }
+        else {
+            // adr mode RRX is not used in GGL Assembler at this time
+            mMips->UNIMPL();
+        }
+        break;
+
+    case opMVN:     // this is a 1's complement: NOT
+        if (Op2 < AMODE_REG) {  // op2 is reg # in this case
+            mMips->NOR(Rd, Op2, 0);     // NOT is NOR with 0
+            break;
+        } else if (Op2 == AMODE_IMM) {
+            if (amode.value > 0xffff) {
+                mMips->LUI(Rd, (amode.value >> 16));
+                if (amode.value & 0x0000ffff) {
+                    mMips->ORI(Rd, Rd, (amode.value & 0x0000ffff));
+                }
+             } else {
+                mMips->ORI(Rd, 0, amode.value);
+             }
+        } else if (Op2 == AMODE_REG_IMM) {
+            switch (amode.stype) {
+            case LSL: mMips->SLL(Rd, amode.reg, amode.value); break;
+            case LSR: mMips->SRL(Rd, amode.reg, amode.value); break;
+            case ASR: mMips->SRA(Rd, amode.reg, amode.value); break;
+            case ROR: mMips->ROTR(Rd, amode.reg, amode.value); break;
+            }
+        }
+        else {
+            // adr mode RRX is not used in GGL Assembler at this time
+            mMips->UNIMPL();
+        }
+        mMips->NOR(Rd, Rd, 0);     // NOT is NOR with 0
+        break;
+
+    case opCMP:
+        // Either operand of a CMP instr could get overwritten by a subsequent
+        // conditional instruction, which is ok, _UNLESS_ there is a _second_
+        // conditional instruction. Under MIPS, this requires doing the comparison
+        // again (SLT), and the original operands must be available. (and this
+        // pattern of multiple conditional instructions from same CMP _is_ used
+        // in GGL-Assembler)
+        //
+        // For now, if a conditional instr overwrites the operands, we will
+        // move them to dedicated temp regs. This is ugly, and inefficient,
+        // and should be optimized.
+        //
+        // WARNING: making an _Assumption_ that CMP operand regs will NOT be
+        // trashed by intervening NON-conditional instructions. In the general
+        // case this is legal, but it is NOT currently done in GGL-Assembler.
+
+        cond.type = CMP_COND;
+        cond.r1 = Rn;
+        if (dataProcAdrModes(Op2, src, false, R_cmp2) == SRC_REG) {
+            cond.r2 = src;
+        } else {                        // adr mode was SRC_IMM
+            mMips->ORI(R_cmp2, R_zero, src);
+            cond.r2 = R_cmp2;
+        }
+
+        break;
+
+
+    case opTST:
+    case opTEQ:
+    case opCMN:
+    case opADC:
+    case opSBC:
+    case opRSC:
+        mMips->UNIMPL(); // currently unused in GGL Assembler code
+        break;
+    }
+
+    if (cc != AL) {
+        mMips->label(cond.label[cond.labelnum]);
+    }
+    if (s && opcode != opCMP) {
+        cond.type = SBIT_COND;
+        cond.r1 = Rd;
+    }
+}
+
+
+
+#if 0
+#pragma mark -
+#pragma mark Multiply...
+#endif
+
+// multiply, accumulate
+void ArmToMips64Assembler::MLA(int cc, int s,
+        int Rd, int Rm, int Rs, int Rn) {
+
+    //ALOGW("MLA");
+    mArmPC[mInum++] = pc();  // save starting PC for this instr
+
+    mMips->MUL(R_at, Rm, Rs);
+    mMips->ADDU(Rd, R_at, Rn);
+    if (s) {
+        cond.type = SBIT_COND;
+        cond.r1 = Rd;
+    }
+}
+
+void ArmToMips64Assembler::MUL(int cc, int s,
+        int Rd, int Rm, int Rs) {
+    mArmPC[mInum++] = pc();
+    mMips->MUL(Rd, Rm, Rs);
+    if (s) {
+        cond.type = SBIT_COND;
+        cond.r1 = Rd;
+    }
+}
+
+void ArmToMips64Assembler::UMULL(int cc, int s,
+        int RdLo, int RdHi, int Rm, int Rs) {
+    mArmPC[mInum++] = pc();
+    mMips->MUH(RdHi, Rm, Rs);
+    mMips->MUL(RdLo, Rm, Rs);
+
+    if (s) {
+        cond.type = SBIT_COND;
+        cond.r1 = RdHi;     // BUG...
+        LOG_ALWAYS_FATAL("Condition on UMULL must be on 64-bit result\n");
+    }
+}
+
+void ArmToMips64Assembler::UMUAL(int cc, int s,
+        int RdLo, int RdHi, int Rm, int Rs) {
+    LOG_FATAL_IF(RdLo==Rm || RdHi==Rm || RdLo==RdHi,
+                        "UMUAL(r%u,r%u,r%u,r%u)", RdLo,RdHi,Rm,Rs);
+    // *mPC++ =    (cc<<28) | (1<<23) | (1<<21) | (s<<20) |
+    //             (RdHi<<16) | (RdLo<<12) | (Rs<<8) | 0x90 | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+    if (s) {
+        cond.type = SBIT_COND;
+        cond.r1 = RdHi;     // BUG...
+        LOG_ALWAYS_FATAL("Condition on UMULL must be on 64-bit result\n");
+    }
+}
+
+void ArmToMips64Assembler::SMULL(int cc, int s,
+        int RdLo, int RdHi, int Rm, int Rs) {
+    LOG_FATAL_IF(RdLo==Rm || RdHi==Rm || RdLo==RdHi,
+                        "SMULL(r%u,r%u,r%u,r%u)", RdLo,RdHi,Rm,Rs);
+    // *mPC++ =    (cc<<28) | (1<<23) | (1<<22) | (s<<20) |
+    //             (RdHi<<16) | (RdLo<<12) | (Rs<<8) | 0x90 | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+    if (s) {
+        cond.type = SBIT_COND;
+        cond.r1 = RdHi;     // BUG...
+        LOG_ALWAYS_FATAL("Condition on SMULL must be on 64-bit result\n");
+    }
+}
+void ArmToMips64Assembler::SMUAL(int cc, int s,
+        int RdLo, int RdHi, int Rm, int Rs) {
+    LOG_FATAL_IF(RdLo==Rm || RdHi==Rm || RdLo==RdHi,
+                        "SMUAL(r%u,r%u,r%u,r%u)", RdLo,RdHi,Rm,Rs);
+    // *mPC++ =    (cc<<28) | (1<<23) | (1<<22) | (1<<21) | (s<<20) |
+    //             (RdHi<<16) | (RdLo<<12) | (Rs<<8) | 0x90 | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+    if (s) {
+        cond.type = SBIT_COND;
+        cond.r1 = RdHi;     // BUG...
+        LOG_ALWAYS_FATAL("Condition on SMUAL must be on 64-bit result\n");
+    }
+}
+
+
+
+#if 0
+#pragma mark -
+#pragma mark Branches...
+#endif
+
+// branches...
+
+void ArmToMips64Assembler::B(int cc, const char* label)
+{
+    mArmPC[mInum++] = pc();
+    if (cond.type == SBIT_COND) { cond.r2 = R_zero; }
+
+    switch(cc) {
+        case EQ: mMips->BEQ(cond.r1, cond.r2, label); break;
+        case NE: mMips->BNE(cond.r1, cond.r2, label); break;
+        case HS: mMips->BGEU(cond.r1, cond.r2, label); break;
+        case LO: mMips->BLTU(cond.r1, cond.r2, label); break;
+        case MI: mMips->BLT(cond.r1, cond.r2, label); break;
+        case PL: mMips->BGE(cond.r1, cond.r2, label); break;
+
+        case HI: mMips->BGTU(cond.r1, cond.r2, label); break;
+        case LS: mMips->BLEU(cond.r1, cond.r2, label); break;
+        case GE: mMips->BGE(cond.r1, cond.r2, label); break;
+        case LT: mMips->BLT(cond.r1, cond.r2, label); break;
+        case GT: mMips->BGT(cond.r1, cond.r2, label); break;
+        case LE: mMips->BLE(cond.r1, cond.r2, label); break;
+        case AL: mMips->B(label); break;
+        case NV: /* B Never - no instruction */ break;
+
+        case VS:
+        case VC:
+        default:
+            LOG_ALWAYS_FATAL("Unsupported cc: %02x\n", cc);
+            break;
+    }
+}
+
+void ArmToMips64Assembler::BL(int cc, const char* label)
+{
+    LOG_ALWAYS_FATAL("branch-and-link not supported yet\n");
+    mArmPC[mInum++] = pc();
+}
+
+// no use for Branches with integer PC, but they're in the Interface class ....
+void ArmToMips64Assembler::B(int cc, uint32_t* to_pc)
+{
+    LOG_ALWAYS_FATAL("branch to absolute PC not supported, use Label\n");
+    mArmPC[mInum++] = pc();
+}
+
+void ArmToMips64Assembler::BL(int cc, uint32_t* to_pc)
+{
+    LOG_ALWAYS_FATAL("branch to absolute PC not supported, use Label\n");
+    mArmPC[mInum++] = pc();
+}
+
+void ArmToMips64Assembler::BX(int cc, int Rn)
+{
+    LOG_ALWAYS_FATAL("branch to absolute PC not supported, use Label\n");
+    mArmPC[mInum++] = pc();
+}
+
+
+
+#if 0
+#pragma mark -
+#pragma mark Data Transfer...
+#endif
+
+// data transfer...
+void ArmToMips64Assembler::LDR(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed12_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            amode.writeback = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_12_PRE:
+            if (Rn == ARMAssemblerInterface::SP) {
+                Rn = R_sp;      // convert LDR via Arm SP to LW via Mips SP
+            }
+            mMips->LW(Rd, Rn, amode.value);
+            if (amode.writeback) {      // OPTIONAL writeback on pre-index mode
+                mMips->DADDIU(Rn, Rn, amode.value);
+            }
+            break;
+        case AMODE_IMM_12_POST:
+            if (Rn == ARMAssemblerInterface::SP) {
+                Rn = R_sp;      // convert STR thru Arm SP to STR thru Mips SP
+            }
+            mMips->LW(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);
+            break;
+        case AMODE_REG_SCALE_PRE:
+            // we only support simple base + index, no advanced modes for this one yet
+            mMips->DADDU(R_at, Rn, amode.reg);
+            mMips->LW(Rd, R_at, 0);
+            break;
+    }
+}
+
+void ArmToMips64Assembler::LDRB(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed12_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            amode.writeback = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_12_PRE:
+            mMips->LBU(Rd, Rn, amode.value);
+            if (amode.writeback) {      // OPTIONAL writeback on pre-index mode
+                mMips->DADDIU(Rn, Rn, amode.value);
+            }
+            break;
+        case AMODE_IMM_12_POST:
+            mMips->LBU(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);
+            break;
+        case AMODE_REG_SCALE_PRE:
+            // we only support simple base + index, no advanced modes for this one yet
+            mMips->DADDU(R_at, Rn, amode.reg);
+            mMips->LBU(Rd, R_at, 0);
+            break;
+    }
+
+}
+
+void ArmToMips64Assembler::STR(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed12_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            amode.writeback = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_12_PRE:
+            if (Rn == ARMAssemblerInterface::SP) {
+                Rn = R_sp;  // convert STR thru Arm SP to SW thru Mips SP
+            }
+            if (amode.writeback) {      // OPTIONAL writeback on pre-index mode
+                // If we will writeback, then update the index reg, then store.
+                // This correctly handles stack-push case.
+                mMips->DADDIU(Rn, Rn, amode.value);
+                mMips->SW(Rd, Rn, 0);
+            } else {
+                // No writeback so store offset by value
+                mMips->SW(Rd, Rn, amode.value);
+            }
+            break;
+        case AMODE_IMM_12_POST:
+            mMips->SW(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);  // post index always writes back
+            break;
+        case AMODE_REG_SCALE_PRE:
+            // we only support simple base + index, no advanced modes for this one yet
+            mMips->DADDU(R_at, Rn, amode.reg);
+            mMips->SW(Rd, R_at, 0);
+            break;
+    }
+}
+
+void ArmToMips64Assembler::STRB(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed12_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            amode.writeback = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_12_PRE:
+            mMips->SB(Rd, Rn, amode.value);
+            if (amode.writeback) {      // OPTIONAL writeback on pre-index mode
+                mMips->DADDIU(Rn, Rn, amode.value);
+            }
+            break;
+        case AMODE_IMM_12_POST:
+            mMips->SB(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);
+            break;
+        case AMODE_REG_SCALE_PRE:
+            // we only support simple base + index, no advanced modes for this one yet
+            mMips->DADDU(R_at, Rn, amode.reg);
+            mMips->SB(Rd, R_at, 0);
+            break;
+    }
+}
+
+void ArmToMips64Assembler::LDRH(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed8_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_8_PRE:      // no support yet for writeback
+            mMips->LHU(Rd, Rn, amode.value);
+            break;
+        case AMODE_IMM_8_POST:
+            mMips->LHU(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);
+            break;
+        case AMODE_REG_PRE:
+            // we only support simple base +/- index
+            if (amode.reg >= 0) {
+                mMips->DADDU(R_at, Rn, amode.reg);
+            } else {
+                mMips->DSUBU(R_at, Rn, abs(amode.reg));
+            }
+            mMips->LHU(Rd, R_at, 0);
+            break;
+    }
+}
+
+void ArmToMips64Assembler::LDRSB(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::LDRSH(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::STRH(int cc, int Rd, int Rn, uint32_t offset)
+{
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed8_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_8_PRE:      // no support yet for writeback
+            mMips->SH(Rd, Rn, amode.value);
+            break;
+        case AMODE_IMM_8_POST:
+            mMips->SH(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);
+            break;
+        case AMODE_REG_PRE:
+            // we only support simple base +/- index
+            if (amode.reg >= 0) {
+                mMips->DADDU(R_at, Rn, amode.reg);
+            } else {
+                mMips->DSUBU(R_at, Rn, abs(amode.reg));
+            }
+            mMips->SH(Rd, R_at, 0);
+            break;
+    }
+}
+
+
+
+#if 0
+#pragma mark -
+#pragma mark Block Data Transfer...
+#endif
+
+// block data transfer...
+void ArmToMips64Assembler::LDM(int cc, int dir,
+        int Rn, int W, uint32_t reg_list)
+{   //                        ED FD EA FA      IB IA DB DA
+    // const uint8_t P[8] = { 1, 0, 1, 0,      1, 0, 1, 0 };
+    // const uint8_t U[8] = { 1, 1, 0, 0,      1, 1, 0, 0 };
+    // *mPC++ = (cc<<28) | (4<<25) | (uint32_t(P[dir])<<24) |
+    //         (uint32_t(U[dir])<<23) | (1<<20) | (W<<21) | (Rn<<16) | reg_list;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::STM(int cc, int dir,
+        int Rn, int W, uint32_t reg_list)
+{   //                        FA EA FD ED      IB IA DB DA
+    // const uint8_t P[8] = { 0, 1, 0, 1,      1, 0, 1, 0 };
+    // const uint8_t U[8] = { 0, 0, 1, 1,      1, 1, 0, 0 };
+    // *mPC++ = (cc<<28) | (4<<25) | (uint32_t(P[dir])<<24) |
+    //         (uint32_t(U[dir])<<23) | (0<<20) | (W<<21) | (Rn<<16) | reg_list;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+
+
+#if 0
+#pragma mark -
+#pragma mark Special...
+#endif
+
+// special...
+void ArmToMips64Assembler::SWP(int cc, int Rn, int Rd, int Rm) {
+    // *mPC++ = (cc<<28) | (2<<23) | (Rn<<16) | (Rd << 12) | 0x90 | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::SWPB(int cc, int Rn, int Rd, int Rm) {
+    // *mPC++ = (cc<<28) | (2<<23) | (1<<22) | (Rn<<16) | (Rd << 12) | 0x90 | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::SWI(int cc, uint32_t comment) {
+    // *mPC++ = (cc<<28) | (0xF<<24) | comment;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+
+#if 0
+#pragma mark -
+#pragma mark DSP instructions...
+#endif
+
+// DSP instructions...
+void ArmToMips64Assembler::PLD(int Rn, uint32_t offset) {
+    LOG_ALWAYS_FATAL_IF(!((offset&(1<<24)) && !(offset&(1<<21))),
+                        "PLD only P=1, W=0");
+    // *mPC++ = 0xF550F000 | (Rn<<16) | offset;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::CLZ(int cc, int Rd, int Rm)
+{
+    mArmPC[mInum++] = pc();
+    mMips->CLZ(Rd, Rm);
+}
+
+void ArmToMips64Assembler::QADD(int cc,  int Rd, int Rm, int Rn)
+{
+    // *mPC++ = (cc<<28) | 0x1000050 | (Rn<<16) | (Rd<<12) | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::QDADD(int cc,  int Rd, int Rm, int Rn)
+{
+    // *mPC++ = (cc<<28) | 0x1400050 | (Rn<<16) | (Rd<<12) | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::QSUB(int cc,  int Rd, int Rm, int Rn)
+{
+    // *mPC++ = (cc<<28) | 0x1200050 | (Rn<<16) | (Rd<<12) | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::QDSUB(int cc,  int Rd, int Rm, int Rn)
+{
+    // *mPC++ = (cc<<28) | 0x1600050 | (Rn<<16) | (Rd<<12) | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+// 16 x 16 signed multiply (like SMLAxx without the accumulate)
+void ArmToMips64Assembler::SMUL(int cc, int xy,
+                int Rd, int Rm, int Rs)
+{
+    mArmPC[mInum++] = pc();
+
+    // the 16 bits may be in the top or bottom half of 32-bit source reg,
+    // as defined by the codes BB, BT, TB, TT (compressed param xy)
+    // where x corresponds to Rm and y to Rs
+
+    // select half-reg for Rm
+    if (xy & xyTB) {
+        // use top 16-bits
+        mMips->SRA(R_at, Rm, 16);
+    } else {
+        // use bottom 16, but sign-extend to 32
+        mMips->SEH(R_at, Rm);
+    }
+    // select half-reg for Rs
+    if (xy & xyBT) {
+        // use top 16-bits
+        mMips->SRA(R_at2, Rs, 16);
+    } else {
+        // use bottom 16, but sign-extend to 32
+        mMips->SEH(R_at2, Rs);
+    }
+    mMips->MUL(Rd, R_at, R_at2);
+}
+
+// signed 32b x 16b multiple, save top 32-bits of 48-bit result
+void ArmToMips64Assembler::SMULW(int cc, int y,
+                int Rd, int Rm, int Rs)
+{
+    mArmPC[mInum++] = pc();
+
+    // the selector yT or yB refers to reg Rs
+    if (y & yT) {
+        // zero the bottom 16-bits, with 2 shifts, it can affect result
+        mMips->SRL(R_at, Rs, 16);
+        mMips->SLL(R_at, R_at, 16);
+
+    } else {
+        // move low 16-bit half, to high half
+        mMips->SLL(R_at, Rs, 16);
+    }
+    mMips->MUH(Rd, Rm, R_at);
+}
+
+// 16 x 16 signed multiply, accumulate: Rd = Rm{16} * Rs{16} + Rn
+void ArmToMips64Assembler::SMLA(int cc, int xy,
+                int Rd, int Rm, int Rs, int Rn)
+{
+    mArmPC[mInum++] = pc();
+
+    // the 16 bits may be in the top or bottom half of 32-bit source reg,
+    // as defined by the codes BB, BT, TB, TT (compressed param xy)
+    // where x corresponds to Rm and y to Rs
+
+    // select half-reg for Rm
+    if (xy & xyTB) {
+        // use top 16-bits
+        mMips->SRA(R_at, Rm, 16);
+    } else {
+        // use bottom 16, but sign-extend to 32
+        mMips->SEH(R_at, Rm);
+    }
+    // select half-reg for Rs
+    if (xy & xyBT) {
+        // use top 16-bits
+        mMips->SRA(R_at2, Rs, 16);
+    } else {
+        // use bottom 16, but sign-extend to 32
+        mMips->SEH(R_at2, Rs);
+    }
+
+    mMips->MUL(R_at, R_at, R_at2);
+    mMips->ADDU(Rd, R_at, Rn);
+}
+
+void ArmToMips64Assembler::SMLAL(int cc, int xy,
+                int RdHi, int RdLo, int Rs, int Rm)
+{
+    // *mPC++ = (cc<<28) | 0x1400080 | (RdHi<<16) | (RdLo<<12) | (Rs<<8) | (xy<<4) | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+void ArmToMips64Assembler::SMLAW(int cc, int y,
+                int Rd, int Rm, int Rs, int Rn)
+{
+    // *mPC++ = (cc<<28) | 0x1200080 | (Rd<<16) | (Rn<<12) | (Rs<<8) | (y<<4) | Rm;
+    mArmPC[mInum++] = pc();
+    mMips->NOP2();
+    NOT_IMPLEMENTED();
+}
+
+// used by ARMv6 version of GGLAssembler::filter32
+void ArmToMips64Assembler::UXTB16(int cc, int Rd, int Rm, int rotate)
+{
+    mArmPC[mInum++] = pc();
+
+    //Rd[31:16] := ZeroExtend((Rm ROR (8 * sh))[23:16]),
+    //Rd[15:0] := ZeroExtend((Rm ROR (8 * sh))[7:0]). sh 0-3.
+
+    mMips->ROTR(R_at2, Rm, rotate * 8);
+    mMips->LUI(R_at, 0xFF);
+    mMips->ORI(R_at, R_at, 0xFF);
+    mMips->AND(Rd, R_at2, R_at);
+}
+
+void ArmToMips64Assembler::UBFX(int cc, int Rd, int Rn, int lsb, int width)
+{
+     /* Placeholder for UBFX */
+     mArmPC[mInum++] = pc();
+
+     mMips->NOP2();
+     NOT_IMPLEMENTED();
+}
+
+// ----------------------------------------------------------------------------
+// Address Processing...
+// ----------------------------------------------------------------------------
+
+void ArmToMips64Assembler::ADDR_ADD(int cc,
+        int s, int Rd, int Rn, uint32_t Op2)
+{
+//    if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
+//    if(s  != 0) { NOT_IMPLEMENTED(); return;} //Not required
+    dataProcessing(opADD64, cc, s, Rd, Rn, Op2);
+}
+
+void ArmToMips64Assembler::ADDR_SUB(int cc,
+        int s, int Rd, int Rn, uint32_t Op2)
+{
+//    if(cc != AL){ NOT_IMPLEMENTED(); return;} //Not required
+//    if(s  != 0) { NOT_IMPLEMENTED(); return;} //Not required
+    dataProcessing(opSUB64, cc, s, Rd, Rn, Op2);
+}
+
+void ArmToMips64Assembler::ADDR_LDR(int cc, int Rd, int Rn, uint32_t offset) {
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed12_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            amode.writeback = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_12_PRE:
+            if (Rn == ARMAssemblerInterface::SP) {
+                Rn = R_sp;      // convert LDR via Arm SP to LW via Mips SP
+            }
+            mMips->LD(Rd, Rn, amode.value);
+            if (amode.writeback) {      // OPTIONAL writeback on pre-index mode
+                mMips->DADDIU(Rn, Rn, amode.value);
+            }
+            break;
+        case AMODE_IMM_12_POST:
+            if (Rn == ARMAssemblerInterface::SP) {
+                Rn = R_sp;      // convert STR thru Arm SP to STR thru Mips SP
+            }
+            mMips->LD(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);
+            break;
+        case AMODE_REG_SCALE_PRE:
+            // we only support simple base + index, no advanced modes for this one yet
+            mMips->DADDU(R_at, Rn, amode.reg);
+            mMips->LD(Rd, R_at, 0);
+            break;
+    }
+}
+
+void ArmToMips64Assembler::ADDR_STR(int cc, int Rd, int Rn, uint32_t offset) {
+    mArmPC[mInum++] = pc();
+    // work-around for ARM default address mode of immed12_pre(0)
+    if (offset > AMODE_UNSUPPORTED) offset = 0;
+    switch (offset) {
+        case 0:
+            amode.value = 0;
+            amode.writeback = 0;
+            // fall thru to next case ....
+        case AMODE_IMM_12_PRE:
+            if (Rn == ARMAssemblerInterface::SP) {
+                Rn = R_sp;  // convert STR thru Arm SP to SW thru Mips SP
+            }
+            if (amode.writeback) {      // OPTIONAL writeback on pre-index mode
+                // If we will writeback, then update the index reg, then store.
+                // This correctly handles stack-push case.
+                mMips->DADDIU(Rn, Rn, amode.value);
+                mMips->SD(Rd, Rn, 0);
+            } else {
+                // No writeback so store offset by value
+                mMips->SD(Rd, Rn, amode.value);
+            }
+            break;
+        case AMODE_IMM_12_POST:
+            mMips->SD(Rd, Rn, 0);
+            mMips->DADDIU(Rn, Rn, amode.value);  // post index always writes back
+            break;
+        case AMODE_REG_SCALE_PRE:
+            // we only support simple base + index, no advanced modes for this one yet
+            mMips->DADDU(R_at, Rn, amode.reg);
+            mMips->SD(Rd, R_at, 0);
+            break;
+    }
+}
+
+#if 0
+#pragma mark -
+#pragma mark MIPS Assembler...
+#endif
+
+
+//**************************************************************************
+//**************************************************************************
+//**************************************************************************
+
+
+/* MIPS64 assembler
+** this is a subset of mips64r6, targeted specifically at ARM instruction
+** replacement in the pixelflinger/codeflinger code.
+**
+** This class is extended from MIPSAssembler class and overrides only
+** MIPS64r6 specific stuff.
+*/
+
+MIPS64Assembler::MIPS64Assembler(const sp<Assembly>& assembly, ArmToMips64Assembler *parent)
+    : mParent(parent),
+    MIPSAssembler::MIPSAssembler(assembly, NULL)
+{
+}
+
+MIPS64Assembler::MIPS64Assembler(void* assembly, ArmToMips64Assembler *parent)
+    : mParent(parent),
+    MIPSAssembler::MIPSAssembler(NULL, NULL)
+{
+    mBase = mPC = (uint32_t *)assembly;
+}
+
+MIPS64Assembler::~MIPS64Assembler()
+{
+}
+
+void MIPS64Assembler::reset()
+{
+    if (mAssembly != NULL) {
+        mBase = mPC = (uint32_t *)mAssembly->base();
+    } else {
+        mPC = mBase = base();
+    }
+    mBranchTargets.clear();
+    mLabels.clear();
+    mLabelsInverseMapping.clear();
+    mComments.clear();
+}
+
+
+void MIPS64Assembler::disassemble(const char* name)
+{
+    char di_buf[140];
+
+    bool arm_disasm_fmt = (mParent->mArmDisassemblyBuffer == NULL) ? false : true;
+
+    typedef char dstr[40];
+    dstr *lines = (dstr *)mParent->mArmDisassemblyBuffer;
+
+    if (mParent->mArmDisassemblyBuffer != NULL) {
+        for (int i=0; i<mParent->mArmInstrCount; ++i) {
+            string_detab(lines[i]);
+        }
+    }
+
+    // iArm is an index to Arm instructions 1...n for this assembly sequence
+    // mArmPC[iArm] holds the value of the Mips-PC for the first MIPS
+    // instruction corresponding to that Arm instruction number
+
+    int iArm = 0;
+    size_t count = pc()-base();
+    uint32_t* mipsPC = base();
+
+    while (count--) {
+        ssize_t label = mLabelsInverseMapping.indexOfKey(mipsPC);
+        if (label >= 0) {
+            ALOGW("%s:\n", mLabelsInverseMapping.valueAt(label));
+        }
+        ssize_t comment = mComments.indexOfKey(mipsPC);
+        if (comment >= 0) {
+            ALOGW("; %s\n", mComments.valueAt(comment));
+        }
+        ::mips_disassem(mipsPC, di_buf, arm_disasm_fmt);
+        string_detab(di_buf);
+        string_pad(di_buf, 30);
+        ALOGW("%08lx:    %08x    %s", uintptr_t(mipsPC), uint32_t(*mipsPC), di_buf);
+        mipsPC++;
+    }
+}
+
+void MIPS64Assembler::fix_branches()
+{
+    // fixup all the branches
+    size_t count = mBranchTargets.size();
+    while (count--) {
+        const branch_target_t& bt = mBranchTargets[count];
+        uint32_t* target_pc = mLabels.valueFor(bt.label);
+        LOG_ALWAYS_FATAL_IF(!target_pc,
+                "error resolving branch targets, target_pc is null");
+        int32_t offset = int32_t(target_pc - (bt.pc+1));
+        *bt.pc |= offset & 0x00FFFF;
+    }
+}
+
+void MIPS64Assembler::DADDU(int Rd, int Rs, int Rt)
+{
+    *mPC++ = (spec_op<<OP_SHF) | (daddu_fn<<FUNC_SHF)
+                    | (Rs<<RS_SHF) | (Rt<<RT_SHF) | (Rd<<RD_SHF);
+}
+
+void MIPS64Assembler::DADDIU(int Rt, int Rs, int16_t imm)
+{
+    *mPC++ = (daddiu_op<<OP_SHF) | (Rt<<RT_SHF) | (Rs<<RS_SHF) | (imm & MSK_16);
+}
+
+void MIPS64Assembler::DSUBU(int Rd, int Rs, int Rt)
+{
+    *mPC++ = (spec_op<<OP_SHF) | (dsubu_fn<<FUNC_SHF) |
+                        (Rs<<RS_SHF) | (Rt<<RT_SHF) | (Rd<<RD_SHF) ;
+}
+
+void MIPS64Assembler::DSUBIU(int Rt, int Rs, int16_t imm)   // really addiu(d, s, -j)
+{
+    *mPC++ = (daddiu_op<<OP_SHF) | (Rt<<RT_SHF) | (Rs<<RS_SHF) | ((-imm) & MSK_16);
+}
+
+void MIPS64Assembler::MUL(int Rd, int Rs, int Rt)
+{
+    *mPC++ = (spec_op<<OP_SHF) | (mul_fn<<RE_SHF) | (sop30_fn<<FUNC_SHF) |
+                        (Rs<<RS_SHF) | (Rt<<RT_SHF) | (Rd<<RD_SHF) ;
+}
+
+void MIPS64Assembler::MUH(int Rd, int Rs, int Rt)
+{
+    *mPC++ = (spec_op<<OP_SHF) | (muh_fn<<RE_SHF) | (sop30_fn<<FUNC_SHF) |
+                        (Rs<<RS_SHF) | (Rt<<RT_SHF) | (Rd<<RD_SHF) ;
+}
+
+void MIPS64Assembler::CLO(int Rd, int Rs)
+{
+    *mPC++ = (spec_op<<OP_SHF) | (17<<FUNC_SHF) |
+                        (Rd<<RD_SHF) | (Rs<<RS_SHF) | (1<<RE_SHF);
+}
+
+void MIPS64Assembler::CLZ(int Rd, int Rs)
+{
+    *mPC++ = (spec_op<<OP_SHF) | (16<<FUNC_SHF) |
+                        (Rd<<RD_SHF) | (Rs<<RS_SHF) | (1<<RE_SHF);
+}
+
+void MIPS64Assembler::LD(int Rt, int Rbase, int16_t offset)
+{
+    *mPC++ = (ld_op<<OP_SHF) | (Rbase<<RS_SHF) | (Rt<<RT_SHF) | (offset & MSK_16);
+}
+
+void MIPS64Assembler::SD(int Rt, int Rbase, int16_t offset)
+{
+    *mPC++ = (sd_op<<OP_SHF) | (Rbase<<RS_SHF) | (Rt<<RT_SHF) | (offset & MSK_16);
+}
+
+void MIPS64Assembler::LUI(int Rt, int16_t offset)
+{
+    *mPC++ = (aui_op<<OP_SHF) | (Rt<<RT_SHF) | (offset & MSK_16);
+}
+
+
+void MIPS64Assembler::JR(int Rs)
+{
+        *mPC++ = (spec_op<<OP_SHF) | (Rs<<RS_SHF) | (jalr_fn << FUNC_SHF);
+        MIPS64Assembler::NOP();
+}
+
+}; // namespace android:
diff --git a/libpixelflinger/codeflinger/MIPS64Assembler.h b/libpixelflinger/codeflinger/MIPS64Assembler.h
new file mode 100644
index 0000000..6c760f4
--- /dev/null
+++ b/libpixelflinger/codeflinger/MIPS64Assembler.h
@@ -0,0 +1,429 @@
+/* libs/pixelflinger/codeflinger/MIPS64Assembler.h
+**
+** Copyright 2015, The Android Open Source Project
+**
+** Licensed under the Apache License, Version 2.0 (the "License");
+** you may not use this file except in compliance with the License.
+** You may obtain a copy of the License at
+**
+**     http://www.apache.org/licenses/LICENSE-2.0
+**
+** Unless required by applicable law or agreed to in writing, software
+** distributed under the License is distributed on an "AS IS" BASIS,
+** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+** See the License for the specific language governing permissions and
+** limitations under the License.
+*/
+
+#ifndef ANDROID_MIPS64ASSEMBLER_H
+#define ANDROID_MIPS64ASSEMBLER_H
+
+#include <stdint.h>
+#include <sys/types.h>
+
+#include "tinyutils/KeyedVector.h"
+#include "tinyutils/Vector.h"
+#include "tinyutils/smartpointer.h"
+
+#include "ARMAssemblerInterface.h"
+#include "MIPSAssembler.h"
+#include "CodeCache.h"
+
+namespace android {
+
+class MIPS64Assembler;    // forward reference
+
+// this class mimics ARMAssembler interface
+// intent is to translate each ARM instruction to 1 or more MIPS instr
+// implementation calls MIPS64Assembler class to generate mips code
+class ArmToMips64Assembler : public ARMAssemblerInterface
+{
+public:
+                ArmToMips64Assembler(const sp<Assembly>& assembly,
+                        char *abuf = 0, int linesz = 0, int instr_count = 0);
+                ArmToMips64Assembler(void* assembly);
+    virtual     ~ArmToMips64Assembler();
+
+    uint32_t*   base() const;
+    uint32_t*   pc() const;
+    void        disassemble(const char* name);
+
+    virtual void    reset();
+
+    virtual int     generate(const char* name);
+    virtual int     getCodegenArch();
+
+    virtual void    prolog();
+    virtual void    epilog(uint32_t touched);
+    virtual void    comment(const char* string);
+    // for testing purposes
+    void        fix_branches();
+    void        set_condition(int mode, int R1, int R2);
+
+
+    // -----------------------------------------------------------------------
+    // shifters and addressing modes
+    // -----------------------------------------------------------------------
+
+    // shifters...
+    virtual bool        isValidImmediate(uint32_t immed);
+    virtual int         buildImmediate(uint32_t i, uint32_t& rot, uint32_t& imm);
+
+    virtual uint32_t    imm(uint32_t immediate);
+    virtual uint32_t    reg_imm(int Rm, int type, uint32_t shift);
+    virtual uint32_t    reg_rrx(int Rm);
+    virtual uint32_t    reg_reg(int Rm, int type, int Rs);
+
+    // addressing modes...
+    // LDR(B)/STR(B)/PLD
+    // (immediate and Rm can be negative, which indicates U=0)
+    virtual uint32_t    immed12_pre(int32_t immed12, int W=0);
+    virtual uint32_t    immed12_post(int32_t immed12);
+    virtual uint32_t    reg_scale_pre(int Rm, int type=0, uint32_t shift=0, int W=0);
+    virtual uint32_t    reg_scale_post(int Rm, int type=0, uint32_t shift=0);
+
+    // LDRH/LDRSB/LDRSH/STRH
+    // (immediate and Rm can be negative, which indicates U=0)
+    virtual uint32_t    immed8_pre(int32_t immed8, int W=0);
+    virtual uint32_t    immed8_post(int32_t immed8);
+    virtual uint32_t    reg_pre(int Rm, int W=0);
+    virtual uint32_t    reg_post(int Rm);
+
+
+
+
+    virtual void    dataProcessing(int opcode, int cc, int s,
+                                int Rd, int Rn,
+                                uint32_t Op2);
+    virtual void MLA(int cc, int s,
+                int Rd, int Rm, int Rs, int Rn);
+    virtual void MUL(int cc, int s,
+                int Rd, int Rm, int Rs);
+    virtual void UMULL(int cc, int s,
+                int RdLo, int RdHi, int Rm, int Rs);
+    virtual void UMUAL(int cc, int s,
+                int RdLo, int RdHi, int Rm, int Rs);
+    virtual void SMULL(int cc, int s,
+                int RdLo, int RdHi, int Rm, int Rs);
+    virtual void SMUAL(int cc, int s,
+                int RdLo, int RdHi, int Rm, int Rs);
+
+    virtual void B(int cc, uint32_t* pc);
+    virtual void BL(int cc, uint32_t* pc);
+    virtual void BX(int cc, int Rn);
+    virtual void label(const char* theLabel);
+    virtual void B(int cc, const char* label);
+    virtual void BL(int cc, const char* label);
+
+    virtual uint32_t* pcForLabel(const char* label);
+
+    virtual void LDR (int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void LDRB(int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void STR (int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void STRB(int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void LDRH (int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void LDRSB(int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void LDRSH(int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+    virtual void STRH (int cc, int Rd,
+                int Rn, uint32_t offset = 0);
+
+    virtual void LDM(int cc, int dir,
+                int Rn, int W, uint32_t reg_list);
+    virtual void STM(int cc, int dir,
+                int Rn, int W, uint32_t reg_list);
+
+    virtual void SWP(int cc, int Rn, int Rd, int Rm);
+    virtual void SWPB(int cc, int Rn, int Rd, int Rm);
+    virtual void SWI(int cc, uint32_t comment);
+
+    virtual void PLD(int Rn, uint32_t offset);
+    virtual void CLZ(int cc, int Rd, int Rm);
+    virtual void QADD(int cc, int Rd, int Rm, int Rn);
+    virtual void QDADD(int cc, int Rd, int Rm, int Rn);
+    virtual void QSUB(int cc, int Rd, int Rm, int Rn);
+    virtual void QDSUB(int cc, int Rd, int Rm, int Rn);
+    virtual void SMUL(int cc, int xy,
+                int Rd, int Rm, int Rs);
+    virtual void SMULW(int cc, int y,
+                int Rd, int Rm, int Rs);
+    virtual void SMLA(int cc, int xy,
+                int Rd, int Rm, int Rs, int Rn);
+    virtual void SMLAL(int cc, int xy,
+                int RdHi, int RdLo, int Rs, int Rm);
+    virtual void SMLAW(int cc, int y,
+                int Rd, int Rm, int Rs, int Rn);
+
+    // byte/half word extract...
+    virtual void UXTB16(int cc, int Rd, int Rm, int rotate);
+
+    // bit manipulation...
+    virtual void UBFX(int cc, int Rd, int Rn, int lsb, int width);
+
+    // Address loading/storing/manipulation
+    virtual void ADDR_LDR(int cc, int Rd, int Rn, uint32_t offset = __immed12_pre(0));
+    virtual void ADDR_STR(int cc, int Rd, int Rn, uint32_t offset = __immed12_pre(0));
+    virtual void ADDR_ADD(int cc, int s, int Rd, int Rn, uint32_t Op2);
+    virtual void ADDR_SUB(int cc, int s, int Rd, int Rn, uint32_t Op2);
+
+    // this is some crap to share is MIPS64Assembler class for debug
+    char *      mArmDisassemblyBuffer;
+    int         mArmLineLength;
+    int         mArmInstrCount;
+
+    int         mInum;      // current arm instuction number (0..n)
+    uint32_t**  mArmPC;     // array: PC for 1st mips instr of
+                            //      each translated ARM instr
+
+
+private:
+    ArmToMips64Assembler(const ArmToMips64Assembler& rhs);
+    ArmToMips64Assembler& operator = (const ArmToMips64Assembler& rhs);
+
+    void init_conditional_labels(void);
+
+    void protectConditionalOperands(int Rd);
+
+    // reg__tmp set to MIPS AT, reg 1
+    int dataProcAdrModes(int op, int& source, bool sign = false, int reg_tmp = 1);
+
+    sp<Assembly>        mAssembly;
+    MIPS64Assembler*    mMips;
+
+
+    enum misc_constants_t {
+        ARM_MAX_INSTUCTIONS = 512  // based on ASSEMBLY_SCRATCH_SIZE
+    };
+
+    enum {
+        SRC_REG = 0,
+        SRC_IMM,
+        SRC_ERROR = -1
+    };
+
+    enum addr_modes {
+        // start above the range of legal mips reg #'s (0-31)
+        AMODE_REG = 0x20,
+        AMODE_IMM, AMODE_REG_IMM,               // for data processing
+        AMODE_IMM_12_PRE, AMODE_IMM_12_POST,    // for load/store
+        AMODE_REG_SCALE_PRE, AMODE_IMM_8_PRE,
+        AMODE_IMM_8_POST, AMODE_REG_PRE,
+        AMODE_UNSUPPORTED
+    };
+
+    struct addr_mode_t {    // address modes for current ARM instruction
+        int         reg;
+        int         stype;
+        uint32_t    value;
+        bool        writeback;  // writeback the adr reg after modification
+    } amode;
+
+    enum cond_types {
+        CMP_COND = 1,
+        SBIT_COND
+    };
+
+    struct cond_mode_t {    // conditional-execution info for current ARM instruction
+        cond_types  type;
+        int         r1;
+        int         r2;
+        int         labelnum;
+        char        label[100][10];
+    } cond;
+};
+
+
+
+
+// ----------------------------------------------------------------------------
+// ----------------------------------------------------------------------------
+// ----------------------------------------------------------------------------
+
+// This is the basic MIPS64 assembler, which just creates the opcodes in memory.
+// All the more complicated work is done in ArmToMips64Assember above.
+// Inherits MIPSAssembler class, and overrides only MIPS64r6 specific stuff
+
+class MIPS64Assembler : public MIPSAssembler
+{
+public:
+                MIPS64Assembler(const sp<Assembly>& assembly, ArmToMips64Assembler *parent);
+                MIPS64Assembler(void* assembly, ArmToMips64Assembler *parent);
+    virtual     ~MIPS64Assembler();
+
+    virtual void        reset();
+    virtual void        disassemble(const char* name);
+
+    void        fix_branches();
+
+    // ------------------------------------------------------------------------
+    // MIPS64AssemblerInterface...
+    // ------------------------------------------------------------------------
+
+#if 0
+#pragma mark -
+#pragma mark Arithmetic...
+#endif
+
+    void DADDU(int Rd, int Rs, int Rt);
+    void DADDIU(int Rt, int Rs, int16_t imm);
+    void DSUBU(int Rd, int Rs, int Rt);
+    void DSUBIU(int Rt, int Rs, int16_t imm);
+    virtual void MUL(int Rd, int Rs, int Rt);
+    void MUH(int Rd, int Rs, int Rt);
+
+#if 0
+#pragma mark -
+#pragma mark Logical...
+#endif
+
+    virtual void CLO(int Rd, int Rs);
+    virtual void CLZ(int Rd, int Rs);
+
+#if 0
+#pragma mark -
+#pragma mark Load/store...
+#endif
+
+    void LD(int Rt, int Rbase, int16_t offset);
+    void SD(int Rt, int Rbase, int16_t offset);
+    virtual void LUI(int Rt, int16_t offset);
+
+#if 0
+#pragma mark -
+#pragma mark Branch...
+#endif
+
+    void JR(int Rs);
+
+
+protected:
+    // void string_detab(char *s);
+    // void string_pad(char *s, int padded_len);
+
+    ArmToMips64Assembler *mParent;
+    sp<Assembly>    mAssembly;
+    uint32_t*       mBase;
+    uint32_t*       mPC;
+    uint32_t*       mPrologPC;
+    int64_t         mDuration;
+#if defined(WITH_LIB_HARDWARE)
+    bool            mQemuTracing;
+#endif
+
+    struct branch_target_t {
+        inline branch_target_t() : label(0), pc(0) { }
+        inline branch_target_t(const char* l, uint32_t* p)
+            : label(l), pc(p) { }
+        const char* label;
+        uint32_t*   pc;
+    };
+
+    Vector<branch_target_t>                 mBranchTargets;
+    KeyedVector< const char*, uint32_t* >   mLabels;
+    KeyedVector< uint32_t*, const char* >   mLabelsInverseMapping;
+    KeyedVector< uint32_t*, const char* >   mComments;
+
+
+    // opcode field of all instructions
+    enum opcode_field {
+        spec_op, regimm_op, j_op, jal_op,                  // 0x00 - 0x03
+        beq_op, bne_op, pop06_op, pop07_op,                // 0x04 - 0x07
+        pop10_op, addiu_op, slti_op, sltiu_op,             // 0x08 - 0x0b
+        andi_op, ori_op, xori_op, aui_op,                  // 0x0c - 0x0f
+        cop0_op, cop1_op, cop2_op, rsrv_opc_0,             // 0x10 - 0x13
+        rsrv_opc_1, rsrv_opc_2, pop26_op, pop27_op,        // 0x14 - 0x17
+        pop30_op, daddiu_op, rsrv_opc_3, rsrv_opc_4,       // 0x18 - 0x1b
+        rsrv_opc_5, daui_op, msa_op, spec3_op,             // 0x1c - 0x1f
+        lb_op, lh_op, rsrv_opc_6, lw_op,                   // 0x20 - 0x23
+        lbu_op, lhu_op, rsrv_opc_7, lwu_op,                // 0x24 - 0x27
+        sb_op, sh_op, rsrv_opc_8, sw_op,                   // 0x28 - 0x2b
+        rsrv_opc_9, rsrv_opc_10, rsrv_opc_11, rsrv_opc_12, // 0x2c - 0x2f
+        rsrv_opc_13, lwc1_op, bc_op, rsrv_opc_14,          // 0x2c - 0x2f
+        rsrv_opc_15, ldc1_op, pop66_op, ld_op,             // 0x30 - 0x33
+        rsrv_opc_16, swc1_op, balc_op, pcrel_op,           // 0x34 - 0x37
+        rsrv_opc_17, sdc1_op, pop76_op, sd_op              // 0x38 - 0x3b
+    };
+
+
+    // func field for special opcode
+    enum func_spec_op {
+        sll_fn, rsrv_spec_0, srl_fn, sra_fn,
+        sllv_fn, lsa_fn, srlv_fn, srav_fn,
+        rsrv_spec_1, jalr_fn, rsrv_spec_2, rsrv_spec_3,
+        syscall_fn, break_fn, sdbbp_fn, sync_fn,
+        clz_fn, clo_fn, dclz_fn, dclo_fn,
+        dsllv_fn, dlsa_fn, dsrlv_fn, dsrav_fn,
+        sop30_fn, sop31_fn, sop32_fn, sop33_fn,
+        sop34_fn, sop35_fn, sop36_fn, sop37_fn,
+        add_fn, addu_fn, sub_fn, subu_fn,
+        and_fn, or_fn, xor_fn, nor_fn,
+        rsrv_spec_4, rsrv_spec_5, slt_fn, sltu_fn,
+        dadd_fn, daddu_fn, dsub_fn, dsubu_fn,
+        tge_fn, tgeu_fn, tlt_fn, tltu_fn,
+        teq_fn, seleqz_fn, tne_fn, selnez_fn,
+        dsll_fn, rsrv_spec_6, dsrl_fn, dsra_fn,
+        dsll32_fn, rsrv_spec_7, dsrl32_fn, dsra32_fn
+    };
+
+    // func field for spec3 opcode
+    enum func_spec3_op {
+        ext_fn, dextm_fn, dextu_fn, dext_fn,
+        ins_fn, dinsm_fn, dinsu_fn, dins_fn,
+        cachee_fn = 0x1b, sbe_fn, she_fn, sce_fn, swe_fn,
+        bshfl_fn, prefe_fn = 0x23, dbshfl_fn, cache_fn, sc_fn, scd_fn,
+        lbue_fn, lhue_fn, lbe_fn = 0x2c, lhe_fn, lle_fn, lwe_fn,
+        pref_fn = 0x35, ll_fn, lld_fn, rdhwr_fn = 0x3b
+    };
+
+    // sa field for spec3 opcodes, with BSHFL function
+    enum func_spec3_bshfl {
+        bitswap_fn,
+        wsbh_fn = 0x02,
+        dshd_fn = 0x05,
+        seb_fn = 0x10,
+        seh_fn = 0x18
+    };
+
+    // rt field of regimm opcodes.
+    enum regimm_fn {
+        bltz_fn, bgez_fn,
+        dahi_fn = 0x6,
+        nal_fn = 0x10, bal_fn, bltzall_fn, bgezall_fn,
+        sigrie_fn = 0x17,
+        dati_fn = 0x1e, synci_fn
+    };
+
+    enum muldiv_fn {
+        mul_fn = 0x02, muh_fn
+    };
+
+    enum mips_inst_shifts {
+        OP_SHF       = 26,
+        JTARGET_SHF  = 0,
+        RS_SHF       = 21,
+        RT_SHF       = 16,
+        RD_SHF       = 11,
+        RE_SHF       = 6,
+        SA_SHF       = RE_SHF,  // synonym
+        IMM_SHF      = 0,
+        FUNC_SHF     = 0,
+
+        // mask values
+        MSK_16       = 0xffff,
+
+
+        CACHEOP_SHF  = 18,
+        CACHESEL_SHF = 16,
+    };
+};
+
+
+}; // namespace android
+
+#endif //ANDROID_MIPS64ASSEMBLER_H
diff --git a/libpixelflinger/codeflinger/MIPSAssembler.cpp b/libpixelflinger/codeflinger/MIPSAssembler.cpp
index a88d2fe..daa231f 100644
--- a/libpixelflinger/codeflinger/MIPSAssembler.cpp
+++ b/libpixelflinger/codeflinger/MIPSAssembler.cpp
@@ -1358,7 +1358,7 @@
         ::mips_disassem(mipsPC, di_buf, arm_disasm_fmt);
         string_detab(di_buf);
         string_pad(di_buf, 30);
-        ALOGW("%08x:    %08x    %s", uint32_t(mipsPC), uint32_t(*mipsPC), di_buf);
+        ALOGW("%08x:    %08x    %s", uintptr_t(mipsPC), uint32_t(*mipsPC), di_buf);
         mipsPC++;
     }
 }
@@ -1407,7 +1407,7 @@
 
 #if defined(WITH_LIB_HARDWARE)
     if (__builtin_expect(mQemuTracing, 0)) {
-        int err = qemu_add_mapping(int(base()), name);
+        int err = qemu_add_mapping(uintptr_t(base()), name);
         mQemuTracing = (err >= 0);
     }
 #endif
diff --git a/libpixelflinger/codeflinger/MIPSAssembler.h b/libpixelflinger/codeflinger/MIPSAssembler.h
index 430ab06..68a2b4f 100644
--- a/libpixelflinger/codeflinger/MIPSAssembler.h
+++ b/libpixelflinger/codeflinger/MIPSAssembler.h
@@ -244,20 +244,20 @@
                 MIPSAssembler(const sp<Assembly>& assembly, ArmToMipsAssembler *parent);
     virtual     ~MIPSAssembler();
 
-    uint32_t*   base() const;
-    uint32_t*   pc() const;
-    void        reset();
+    virtual uint32_t*   base() const;
+    virtual uint32_t*   pc() const;
+    virtual void        reset();
 
-    void        disassemble(const char* name);
+    virtual void        disassemble(const char* name);
 
-    void        prolog();
-    void        epilog(uint32_t touched);
-    int         generate(const char* name);
-    void        comment(const char* string);
-    void        label(const char* string);
+    virtual void        prolog();
+    virtual void        epilog(uint32_t touched);
+    virtual int         generate(const char* name);
+    virtual void        comment(const char* string);
+    virtual void        label(const char* string);
 
     // valid only after generate() has been called
-    uint32_t*   pcForLabel(const char* label);
+    virtual uint32_t*   pcForLabel(const char* label);
 
 
     // ------------------------------------------------------------------------
@@ -399,9 +399,9 @@
 
 
 
-private:
-    void string_detab(char *s);
-    void string_pad(char *s, int padded_len);
+protected:
+    virtual void string_detab(char *s);
+    virtual void string_pad(char *s, int padded_len);
 
     ArmToMipsAssembler *mParent;
     sp<Assembly>    mAssembly;
@@ -537,7 +537,11 @@
 enum mips_regnames {
     R_zero = 0,
             R_at,   R_v0,   R_v1,   R_a0,   R_a1,   R_a2,   R_a3,
+#if __mips_isa_rev < 6
     R_t0,   R_t1,   R_t2,   R_t3,   R_t4,   R_t5,   R_t6,   R_t7,
+#else
+    R_a4,   R_a5,   R_a6,   R_a7,   R_t0,   R_t1,   R_t2,   R_t3,
+#endif
     R_s0,   R_s1,   R_s2,   R_s3,   R_s4,   R_s5,   R_s6,   R_s7,
     R_t8,   R_t9,   R_k0,   R_k1,   R_gp,   R_sp,   R_s8,   R_ra,
     R_lr = R_s8,
diff --git a/libpixelflinger/codeflinger/mips64_disassem.c b/libpixelflinger/codeflinger/mips64_disassem.c
new file mode 100644
index 0000000..44b7fe7
--- /dev/null
+++ b/libpixelflinger/codeflinger/mips64_disassem.c
@@ -0,0 +1,582 @@
+/*  $NetBSD: db_disasm.c,v 1.19 2007/02/28 04:21:53 thorpej Exp $   */
+
+/*-
+ * Copyright (c) 1991, 1993
+ *  The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Ralph Campbell.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of the University nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ *  from: @(#)kadb.c    8.1 (Berkeley) 6/10/93
+ */
+
+#include <stdio.h>
+#include <stdint.h>
+#include <stdarg.h>
+#include <stdbool.h>
+#include <sys/cdefs.h>
+
+#include <sys/types.h>
+#include "mips_opcode.h"
+
+#include <cutils/log.h>
+
+static char *sprintf_buffer;
+static int sprintf_buf_len;
+
+
+typedef uint64_t db_addr_t;
+static void db_printf(const char* fmt, ...);
+
+static const char * const op_name[64] = {
+/* 0 */ "spec", "bcond", "j", "jal", "beq", "bne", "blez", "bgtz",
+/* 8 */ "pop10", "addiu", "slti", "sltiu", "andi", "ori", "xori", "aui",
+/*16 */ "cop0", "cop1", "cop2", "?", "?", "?", "pop26", "pop27",
+/*24 */ "pop30", "daddiu", "?", "?", "?", "daui", "msa", "op37",
+/*32 */ "lb", "lh", "?",  "lw", "lbu", "lhu", "?", "lwu",
+/*40 */ "sb", "sh", "?", "sw", "?", "?", "?", "?",
+/*48 */ "?", "lwc1", "bc", "?", "?",  "ldc1", "pop66", "ld",
+/*56 */ "?", "swc1", "balc", "pcrel", "?", "sdc1", "pop76", "sd"
+};
+
+static const char * const spec_name[64] = {
+/* 0 */ "sll", "?", "srl", "sra", "sllv", "?", "srlv", "srav",
+/* 8 */ "?", "jalr", "?", "?", "syscall", "break", "sdbpp", "sync",
+/*16 */ "clz", "clo", "dclz", "dclo", "dsllv", "dlsa", "dsrlv", "dsrav",
+/*24 */ "sop30", "sop31", "sop32", "sop33", "sop34", "sop35", "sop36", "sop37",
+/*32 */ "add", "addu", "sub", "subu", "and", "or", "xor", "nor",
+/*40 */ "?", "?", "slt", "sltu", "dadd", "daddu", "dsub", "dsubu",
+/*48 */ "tge", "tgeu", "tlt", "tltu", "teq", "seleqz", "tne", "selnez",
+/*56 */ "dsll", "?", "dsrl", "dsra", "dsll32", "?", "dsrl32", "dsra32"
+};
+
+static const char * const bcond_name[32] = {
+/* 0 */ "bltz", "bgez", "?", "?", "?", "?", "dahi", "?",
+/* 8 */ "?", "?", "?", "?", "?", "?", "?", "?",
+/*16 */ "nal", "bal", "?", "?", "?", "?", "?", "sigrie",
+/*24 */ "?", "?", "?", "?", "?", "?", "dati", "synci",
+};
+
+static const char * const cop1_name[64] = {
+/* 0 */ "fadd",  "fsub", "fmpy", "fdiv", "fsqrt","fabs", "fmov", "fneg",
+/* 8 */ "fop08","fop09","fop0a","fop0b","fop0c","fop0d","fop0e","fop0f",
+/*16 */ "fop10","fop11","fop12","fop13","fop14","fop15","fop16","fop17",
+/*24 */ "fop18","fop19","fop1a","fop1b","fop1c","fop1d","fop1e","fop1f",
+/*32 */ "fcvts","fcvtd","fcvte","fop23","fcvtw","fop25","fop26","fop27",
+/*40 */ "fop28","fop29","fop2a","fop2b","fop2c","fop2d","fop2e","fop2f",
+/*48 */ "fcmp.f","fcmp.un","fcmp.eq","fcmp.ueq","fcmp.olt","fcmp.ult",
+    "fcmp.ole","fcmp.ule",
+/*56 */ "fcmp.sf","fcmp.ngle","fcmp.seq","fcmp.ngl","fcmp.lt","fcmp.nge",
+    "fcmp.le","fcmp.ngt"
+};
+
+static const char * const fmt_name[16] = {
+    "s",    "d",    "e",    "fmt3",
+    "w",    "fmt5", "fmt6", "fmt7",
+    "fmt8", "fmt9", "fmta", "fmtb",
+    "fmtc", "fmtd", "fmte", "fmtf"
+};
+
+static char * const mips_reg_name[32] = {
+    "zero", "at",   "v0",   "v1",   "a0",   "a1",   "a2",   "a3",
+    "a4",   "a5",   "a6",   "a7",   "t0",   "t1",   "t2",   "t3",
+    "s0",   "s1",   "s2",   "s3",   "s4",   "s5",   "s6",   "s7",
+    "t8",   "t9",   "k0",   "k1",   "gp",   "sp",   "s8",   "ra"
+};
+
+static char * alt_arm_reg_name[32] = {  // hacked names for comparison with ARM code
+    "zero", "at",   "r0",   "r1",   "r2",   "r3",   "r4",   "r5",
+    "r6",   "r7",   "r8",   "r9",   "r10",  "r11",  "r12",  "r13",
+    "r14",  "r15",  "at2",  "cmp",  "s4",   "s5",   "s6",   "s7",
+    "t8",   "t9",   "k0",   "k1",   "gp",   "sp",   "s8",   "ra"
+};
+
+static char ** reg_name =  &mips_reg_name[0];
+
+static const char * const c0_opname[64] = {
+    "c0op00","tlbr",  "tlbwi", "c0op03","c0op04","c0op05","tlbwr", "c0op07",
+    "tlbp",  "c0op11","c0op12","c0op13","c0op14","c0op15","c0op16","c0op17",
+    "rfe",   "c0op21","c0op22","c0op23","c0op24","c0op25","c0op26","c0op27",
+    "eret",  "c0op31","c0op32","c0op33","c0op34","c0op35","c0op36","c0op37",
+    "c0op40","c0op41","c0op42","c0op43","c0op44","c0op45","c0op46","c0op47",
+    "c0op50","c0op51","c0op52","c0op53","c0op54","c0op55","c0op56","c0op57",
+    "c0op60","c0op61","c0op62","c0op63","c0op64","c0op65","c0op66","c0op67",
+    "c0op70","c0op71","c0op72","c0op73","c0op74","c0op75","c0op77","c0op77",
+};
+
+static const char * const c0_reg[32] = {
+    "index",    "random",   "tlblo0",  "tlblo1",
+    "context",  "pagemask", "wired",   "cp0r7",
+    "badvaddr", "count",    "tlbhi",   "compare",
+    "status",   "cause",    "epc",     "prid",
+    "config",   "lladdr",   "watchlo", "watchhi",
+    "xcontext", "cp0r21",   "cp0r22",  "debug",
+    "depc",     "perfcnt",  "ecc",     "cacheerr",
+    "taglo",    "taghi",    "errepc",  "desave"
+};
+
+static void print_addr(db_addr_t);
+db_addr_t mips_disassem(db_addr_t loc, char *di_buffer, int alt_dis_format);
+
+
+/*
+ * Disassemble instruction 'insn' nominally at 'loc'.
+ * 'loc' may in fact contain a breakpoint instruction.
+ */
+static db_addr_t
+db_disasm_insn(int insn, db_addr_t loc, bool altfmt)
+{
+    bool bdslot = false;
+    InstFmt i;
+
+    i.word = insn;
+
+    switch (i.JType.op) {
+    case OP_SPECIAL:
+        if (i.word == 0) {
+            db_printf("nop");
+            break;
+        }
+        if (i.word == 0x0080) {
+            db_printf("NIY");
+            break;
+        }
+        if (i.word == 0x00c0) {
+            db_printf("NOT IMPL");
+            break;
+        }
+        /* Special cases --------------------------------------------------
+         * "addu" is a "move" only in 32-bit mode.  What's the correct
+         * answer - never decode addu/daddu as "move"?
+         */
+        if ( (i.RType.func == OP_ADDU && i.RType.rt == 0)  ||
+             (i.RType.func == OP_OR   && i.RType.rt == 0) ) {
+            db_printf("move\t%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rs]);
+            break;
+        }
+
+        if (i.RType.func == OP_SRL && (i.RType.rs & 1) == 1) {
+            db_printf("rotr\t%s,%s,%d", reg_name[i.RType.rd],
+                reg_name[i.RType.rt], i.RType.shamt);
+            break;
+        }
+        if (i.RType.func == OP_SRLV && (i.RType.shamt & 1) == 1) {
+            db_printf("rotrv\t%s,%s,%s", reg_name[i.RType.rd],
+                reg_name[i.RType.rt], reg_name[i.RType.rs]);
+            break;
+        }
+
+        if (i.RType.func == OP_SOP30) {
+            if (i.RType.shamt == OP_MUL) {
+                db_printf("mul");
+            } else if (i.RType.shamt == OP_MUH) {
+                db_printf("muh");
+            }
+            db_printf("\t%s,%s,%s", reg_name[i.RType.rd],
+                reg_name[i.RType.rs], reg_name[i.RType.rt]);
+            break;
+        }
+        if (i.RType.func == OP_SOP31) {
+            if (i.RType.shamt == OP_MUL) {
+                db_printf("mulu");
+            } else if (i.RType.shamt == OP_MUH) {
+                db_printf("muhu");
+            }
+            db_printf("\t%s,%s,%s", reg_name[i.RType.rd],
+                reg_name[i.RType.rs], reg_name[i.RType.rt]);
+            break;
+        }
+
+        if (i.RType.func == OP_JALR && i.RType.rd == 0) {
+            db_printf("jr\t%s", reg_name[i.RType.rs]);
+            bdslot = true;
+            break;
+        }
+
+        db_printf("%s", spec_name[i.RType.func]);
+        switch (i.RType.func) {
+        case OP_SLL:
+        case OP_SRL:
+        case OP_SRA:
+        case OP_DSLL:
+
+        case OP_DSRL:
+        case OP_DSRA:
+        case OP_DSLL32:
+        case OP_DSRL32:
+        case OP_DSRA32:
+            db_printf("\t%s,%s,%d",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rt],
+                i.RType.shamt);
+            break;
+
+        case OP_SLLV:
+        case OP_SRLV:
+        case OP_SRAV:
+        case OP_DSLLV:
+        case OP_DSRLV:
+        case OP_DSRAV:
+            db_printf("\t%s,%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rt],
+                reg_name[i.RType.rs]);
+            break;
+
+        case OP_CLZ:
+        case OP_CLO:
+        case OP_DCLZ:
+        case OP_DCLO:
+            db_printf("\t%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rs]);
+            break;
+
+        case OP_JALR:
+            db_printf("\t");
+            if (i.RType.rd != 31) {
+                db_printf("%s,", reg_name[i.RType.rd]);
+            }
+            db_printf("%s", reg_name[i.RType.rs]);
+            bdslot = true;
+            break;
+
+        case OP_SYSCALL:
+        case OP_SYNC:
+            break;
+
+        case OP_BREAK:
+            db_printf("\t%d", (i.RType.rs << 5) | i.RType.rt);
+            break;
+
+        default:
+            db_printf("\t%s,%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rs],
+                reg_name[i.RType.rt]);
+        }
+        break;
+
+    case OP_SPECIAL3:
+        if (i.RType.func == OP_EXT)
+            db_printf("ext\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt,
+                    i.RType.rd+1);
+        else if (i.RType.func == OP_DEXT)
+            db_printf("dext\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt,
+                    i.RType.rd+1);
+        else if (i.RType.func == OP_DEXTM)
+            db_printf("dextm\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt,
+                    i.RType.rd+33);
+        else if (i.RType.func == OP_DEXTU)
+            db_printf("dextu\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt+32,
+                    i.RType.rd+1);
+        else if (i.RType.func == OP_INS)
+            db_printf("ins\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt,
+                    i.RType.rd-i.RType.shamt+1);
+        else if (i.RType.func == OP_DINS)
+            db_printf("dins\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt,
+                    i.RType.rd-i.RType.shamt+1);
+        else if (i.RType.func == OP_DINSM)
+            db_printf("dinsm\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt,
+                    i.RType.rd-i.RType.shamt+33);
+        else if (i.RType.func == OP_DINSU)
+            db_printf("dinsu\t%s,%s,%d,%d",
+                    reg_name[i.RType.rt],
+                    reg_name[i.RType.rs],
+                    i.RType.shamt+32,
+                    i.RType.rd-i.RType.shamt+1);
+        else if (i.RType.func == OP_BSHFL && i.RType.shamt == OP_WSBH)
+            db_printf("wsbh\t%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rt]);
+        else if (i.RType.func == OP_BSHFL && i.RType.shamt == OP_SEB)
+            db_printf("seb\t%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rt]);
+        else if (i.RType.func == OP_BSHFL && i.RType.shamt == OP_SEH)
+            db_printf("seh\t%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rt]);
+        else if (i.RType.func == OP_RDHWR)
+            db_printf("rdhwr\t%s,%s",
+                reg_name[i.RType.rd],
+                reg_name[i.RType.rt]);
+        else
+            db_printf("Unknown");
+        break;
+
+    case OP_BCOND:
+        db_printf("%s\t%s,", bcond_name[i.IType.rt],
+            reg_name[i.IType.rs]);
+        goto pr_displ;
+
+    case OP_BLEZ:
+    case OP_BGTZ:
+        db_printf("%s\t%s,", op_name[i.IType.op],
+            reg_name[i.IType.rs]);
+        goto pr_displ;
+
+    case OP_BEQ:
+        if (i.IType.rs == 0 && i.IType.rt == 0) {
+            db_printf("b\t");
+            goto pr_displ;
+        }
+        /* FALLTHROUGH */
+    case OP_BNE:
+        db_printf("%s\t%s,%s,", op_name[i.IType.op],
+            reg_name[i.IType.rs],
+            reg_name[i.IType.rt]);
+    pr_displ:
+        print_addr(loc + 4 + ((short)i.IType.imm << 2));
+        bdslot = true;
+        break;
+
+    case OP_COP0:
+        switch (i.RType.rs) {
+        case OP_BCx:
+        case OP_BCy:
+
+            db_printf("bc0%c\t",
+                "ft"[i.RType.rt & COPz_BC_TF_MASK]);
+            goto pr_displ;
+
+        case OP_MT:
+            db_printf("mtc0\t%s,%s",
+                reg_name[i.RType.rt],
+                c0_reg[i.RType.rd]);
+            break;
+
+        case OP_DMT:
+            db_printf("dmtc0\t%s,%s",
+                reg_name[i.RType.rt],
+                c0_reg[i.RType.rd]);
+            break;
+
+        case OP_MF:
+            db_printf("mfc0\t%s,%s",
+                reg_name[i.RType.rt],
+                c0_reg[i.RType.rd]);
+            break;
+
+        case OP_DMF:
+            db_printf("dmfc0\t%s,%s",
+                reg_name[i.RType.rt],
+                c0_reg[i.RType.rd]);
+            break;
+
+        default:
+            db_printf("%s", c0_opname[i.FRType.func]);
+        }
+        break;
+
+    case OP_COP1:
+        switch (i.RType.rs) {
+        case OP_BCx:
+        case OP_BCy:
+            db_printf("bc1%c\t",
+                "ft"[i.RType.rt & COPz_BC_TF_MASK]);
+            goto pr_displ;
+
+        case OP_MT:
+            db_printf("mtc1\t%s,f%d",
+                reg_name[i.RType.rt],
+                i.RType.rd);
+            break;
+
+        case OP_MF:
+            db_printf("mfc1\t%s,f%d",
+                reg_name[i.RType.rt],
+                i.RType.rd);
+            break;
+
+        case OP_CT:
+            db_printf("ctc1\t%s,f%d",
+                reg_name[i.RType.rt],
+                i.RType.rd);
+            break;
+
+        case OP_CF:
+            db_printf("cfc1\t%s,f%d",
+                reg_name[i.RType.rt],
+                i.RType.rd);
+            break;
+
+        default:
+            db_printf("%s.%s\tf%d,f%d,f%d",
+                cop1_name[i.FRType.func],
+                fmt_name[i.FRType.fmt],
+                i.FRType.fd, i.FRType.fs, i.FRType.ft);
+        }
+        break;
+
+    case OP_J:
+    case OP_JAL:
+        db_printf("%s\t", op_name[i.JType.op]);
+        print_addr((loc & 0xFFFFFFFFF0000000) | (i.JType.target << 2));
+        bdslot = true;
+        break;
+
+    case OP_LWC1:
+    case OP_SWC1:
+        db_printf("%s\tf%d,", op_name[i.IType.op],
+            i.IType.rt);
+        goto loadstore;
+
+    case OP_LB:
+    case OP_LH:
+    case OP_LW:
+    case OP_LD:
+    case OP_LBU:
+    case OP_LHU:
+    case OP_LWU:
+    case OP_SB:
+    case OP_SH:
+    case OP_SW:
+    case OP_SD:
+        db_printf("%s\t%s,", op_name[i.IType.op],
+            reg_name[i.IType.rt]);
+    loadstore:
+        db_printf("%d(%s)", (short)i.IType.imm,
+            reg_name[i.IType.rs]);
+        break;
+
+    case OP_ORI:
+    case OP_XORI:
+        if (i.IType.rs == 0) {
+            db_printf("li\t%s,0x%x",
+                reg_name[i.IType.rt],
+                i.IType.imm);
+            break;
+        }
+        /* FALLTHROUGH */
+    case OP_ANDI:
+        db_printf("%s\t%s,%s,0x%x", op_name[i.IType.op],
+            reg_name[i.IType.rt],
+            reg_name[i.IType.rs],
+            i.IType.imm);
+        break;
+
+    case OP_AUI:
+        if (i.IType.rs == 0) {
+            db_printf("lui\t%s,0x%x", reg_name[i.IType.rt],
+                i.IType.imm);
+        } else {
+            db_printf("%s\t%s,%s,%d", op_name[i.IType.op],
+            reg_name[i.IType.rt], reg_name[i.IType.rs],
+            (short)i.IType.imm);
+        }
+        break;
+
+    case OP_ADDIU:
+    case OP_DADDIU:
+        if (i.IType.rs == 0) {
+            db_printf("li\t%s,%d",
+                reg_name[i.IType.rt],
+                (short)i.IType.imm);
+            break;
+        }
+        /* FALLTHROUGH */
+    default:
+        db_printf("%s\t%s,%s,%d", op_name[i.IType.op],
+            reg_name[i.IType.rt],
+            reg_name[i.IType.rs],
+            (short)i.IType.imm);
+    }
+    // db_printf("\n");
+    // if (bdslot) {
+    //     db_printf("   bd: ");
+    //     mips_disassem(loc+4);
+    //     return (loc + 8);
+    // }
+    return (loc + 4);
+}
+
+static void
+print_addr(db_addr_t loc)
+{
+    db_printf("0x%08lx", loc);
+}
+
+static void db_printf(const char* fmt, ...)
+{
+    int cnt;
+    va_list argp;
+    va_start(argp, fmt);
+    if (sprintf_buffer) {
+        cnt = vsnprintf(sprintf_buffer, sprintf_buf_len, fmt, argp);
+        sprintf_buffer += cnt;
+        sprintf_buf_len -= cnt;
+    } else {
+        vprintf(fmt, argp);
+    }
+}
+
+/*
+ * Disassemble instruction at 'loc'.
+ * Return address of start of next instruction.
+ * Since this function is used by 'examine' and by 'step'
+ * "next instruction" does NOT mean the next instruction to
+ * be executed but the 'linear' next instruction.
+ */
+db_addr_t
+mips_disassem(db_addr_t loc, char *di_buffer, int alt_dis_format)
+{
+    u_int32_t instr;
+
+    if (alt_dis_format) {   // use ARM register names for disassembly
+        reg_name = &alt_arm_reg_name[0];
+    }
+
+    sprintf_buffer = di_buffer;     // quick 'n' dirty printf() vs sprintf()
+    sprintf_buf_len = 39;           // should be passed in
+
+    instr =  *(u_int32_t *)loc;
+    return (db_disasm_insn(instr, loc, false));
+}
diff --git a/libpixelflinger/codeflinger/mips64_disassem.h b/libpixelflinger/codeflinger/mips64_disassem.h
new file mode 100644
index 0000000..c94f04f
--- /dev/null
+++ b/libpixelflinger/codeflinger/mips64_disassem.h
@@ -0,0 +1,56 @@
+/*  $NetBSD: db_disasm.c,v 1.19 2007/02/28 04:21:53 thorpej Exp $   */
+
+/*-
+ * Copyright (c) 1991, 1993
+ *  The Regents of the University of California.  All rights reserved.
+ *
+ * This code is derived from software contributed to Berkeley by
+ * Ralph Campbell.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of the University nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ *  from: @(#)kadb.c    8.1 (Berkeley) 6/10/93
+ */
+
+
+
+#ifndef ANDROID_MIPS_DISASSEM_H
+#define ANDROID_MIPS_DISASSEM_H
+
+#include <sys/types.h>
+
+#if __cplusplus
+extern "C" {
+#endif
+
+/* Prototypes for callable functions */
+
+void mips_disassem(uint32_t *location, char *di_buffer, int alt_fmt);
+
+#if __cplusplus
+}
+#endif
+
+#endif /* !ANDROID_MIPS_DISASSEM_H */
diff --git a/libpixelflinger/codeflinger/mips_disassem.c b/libpixelflinger/codeflinger/mips_disassem.c
index 4ab9bd3..3007b15 100644
--- a/libpixelflinger/codeflinger/mips_disassem.c
+++ b/libpixelflinger/codeflinger/mips_disassem.c
@@ -323,14 +323,14 @@
             db_printf("ext\t%s,%s,%d,%d",
                     reg_name[i.RType.rt],
                     reg_name[i.RType.rs],
-                    i.RType.rd+1,
-                    i.RType.shamt);
+                    i.RType.shamt,
+                    i.RType.rd+1);
         else if (i.RType.func == OP_INS)
             db_printf("ins\t%s,%s,%d,%d",
                     reg_name[i.RType.rt],
                     reg_name[i.RType.rs],
-                    i.RType.rd+1,
-                    i.RType.shamt);
+                    i.RType.shamt,
+                    i.RType.rd-i.RType.shamt+1);
         else if (i.RType.func == OP_BSHFL && i.RType.shamt == OP_WSBH)
             db_printf("wsbh\t%s,%s",
                 reg_name[i.RType.rd],
diff --git a/libpixelflinger/codeflinger/mips_opcode.h b/libpixelflinger/codeflinger/mips_opcode.h
index 7ed5ef5..45bb19e 100644
--- a/libpixelflinger/codeflinger/mips_opcode.h
+++ b/libpixelflinger/codeflinger/mips_opcode.h
@@ -125,69 +125,118 @@
 #define OP_BLEZ     006
 #define OP_BGTZ     007
 
+#if __mips_isa_rev < 6
 #define OP_ADDI     010
+#else
+#define OP_POP10    010
+#endif
+
 #define OP_ADDIU    011
 #define OP_SLTI     012
 #define OP_SLTIU    013
 #define OP_ANDI     014
 #define OP_ORI      015
 #define OP_XORI     016
+
+#if __mips_isa_rev < 6
 #define OP_LUI      017
+#else
+#define OP_AUI      017
+#endif
 
 #define OP_COP0     020
 #define OP_COP1     021
 #define OP_COP2     022
+
+#if __mips_isa_rev < 6
 #define OP_COP3     023
-#define OP_BEQL     024     /* MIPS-II, for r4000 port */
-#define OP_BNEL     025     /* MIPS-II, for r4000 port */
-#define OP_BLEZL    026     /* MIPS-II, for r4000 port */
-#define OP_BGTZL    027     /* MIPS-II, for r4000 port */
+#define OP_BEQL     024
+#define OP_BNEL     025
+#define OP_BLEZL    026
+#define OP_BGTZL    027
+#define OP_DADDI    030
+#else
+#define OP_POP26    026
+#define OP_POP27    027
+#define OP_POP30    030
+#endif
 
-#define OP_DADDI    030     /* MIPS-II, for r4000 port */
-#define OP_DADDIU   031     /* MIPS-II, for r4000 port */
-#define OP_LDL      032     /* MIPS-II, for r4000 port */
-#define OP_LDR      033     /* MIPS-II, for r4000 port */
+#define OP_DADDIU   031
 
-#define OP_SPECIAL2 034     /* QED opcodes */
-#define OP_SPECIAL3 037     /* mips32r2 opcodes */
+#if __mips_isa_rev < 6
+#define OP_LDL      032
+#define OP_LDR      033
+#define OP_SPECIAL2 034
+#else
+#define OP_DAUI     035
+#endif
+
+#define OP_SPECIAL3 037
 
 #define OP_LB       040
 #define OP_LH       041
+
+#if __mips_isa_rev < 6
 #define OP_LWL      042
+#endif
+
 #define OP_LW       043
 #define OP_LBU      044
 #define OP_LHU      045
 #define OP_LWR      046
 #define OP_LHU      045
+
+#if __mips_isa_rev < 6
 #define OP_LWR      046
-#define OP_LWU      047     /* MIPS-II, for r4000 port */
+#endif
+
+#define OP_LWU      047
 
 #define OP_SB       050
 #define OP_SH       051
-#define OP_SWL      052
-#define OP_SW       053
-#define OP_SDL      054     /* MIPS-II, for r4000 port */
-#define OP_SDR      055     /* MIPS-II, for r4000 port */
-#define OP_SWR      056
-#define OP_CACHE    057     /* MIPS-II, for r4000 port */
 
+#if __mips_isa_rev < 6
+#define OP_SWL      052
+#endif
+
+#define OP_SW       053
+
+#if __mips_isa_rev < 6
+#define OP_SDL      054
+#define OP_SDR      055
+#define OP_SWR      056
+#define OP_CACHE    057
 #define OP_LL       060
-#define OP_LWC0     OP_LL   /* backwards source compatibility */
+#define OP_LWC0     OP_LL
 #define OP_LWC1     061
 #define OP_LWC2     062
 #define OP_LWC3     063
-#define OP_LLD      064     /* MIPS-II, for r4000 port */
-#define OP_LDC1     065
-#define OP_LD       067     /* MIPS-II, for r4000 port */
+#define OP_LLD      064
+#else
+#define OP_LWC1     061
+#define OP_BC       062
+#endif
 
+#define OP_LDC1     065
+#define OP_LD       067
+
+#if __mips_isa_rev < 6
 #define OP_SC       070
-#define OP_SWC0     OP_SC   /* backwards source compatibility */
+#define OP_SWC0     OP_SC
+#endif
+
 #define OP_SWC1     071
+
+#if __mips_isa_rev < 6
 #define OP_SWC2     072
 #define OP_SWC3     073
-#define OP_SCD      074     /* MIPS-II, for r4000 port */
+#define OP_SCD      074
+#else
+#define OP_BALC     072
+#endif
+
 #define OP_SDC1     075
-#define OP_SD       077     /* MIPS-II, for r4000 port */
+#define OP_SD       077
 
 /*
  * Values for the 'func' field when 'op' == OP_SPECIAL.
@@ -199,28 +248,50 @@
 #define OP_SRLV     006
 #define OP_SRAV     007
 
+#if __mips_isa_rev < 6
 #define OP_JR       010
+#endif
+
 #define OP_JALR     011
 #define OP_SYSCALL  014
 #define OP_BREAK    015
-#define OP_SYNC     017     /* MIPS-II, for r4000 port */
+#define OP_SYNC     017
 
+#if __mips_isa_rev < 6
 #define OP_MFHI     020
 #define OP_MTHI     021
 #define OP_MFLO     022
 #define OP_MTLO     023
-#define OP_DSLLV    024     /* MIPS-II, for r4000 port */
-#define OP_DSRLV    026     /* MIPS-II, for r4000 port */
-#define OP_DSRAV    027     /* MIPS-II, for r4000 port */
+#else
+#define OP_CLZ      020
+#define OP_CLO      021
+#define OP_DCLZ     022
+#define OP_DCLO     023
+#endif
 
+#define OP_DSLLV    024
+#define OP_DSRLV    026
+#define OP_DSRAV    027
+
+#if __mips_isa_rev < 6
 #define OP_MULT     030
 #define OP_MULTU    031
 #define OP_DIV      032
 #define OP_DIVU     033
-#define OP_DMULT    034     /* MIPS-II, for r4000 port */
-#define OP_DMULTU   035     /* MIPS-II, for r4000 port */
-#define OP_DDIV     036     /* MIPS-II, for r4000 port */
-#define OP_DDIVU    037     /* MIPS-II, for r4000 port */
+#define OP_DMULT    034
+#define OP_DMULTU   035
+#define OP_DDIV     036
+#define OP_DDIVU    037
+#else
+#define OP_SOP30    030
+#define OP_SOP31    031
+#define OP_SOP32    032
+#define OP_SOP33    033
+#define OP_SOP34    034
+#define OP_SOP35    035
+#define OP_SOP36    036
+#define OP_SOP37    037
+#endif
 
 #define OP_ADD      040
 #define OP_ADDU     041
@@ -233,73 +304,96 @@
 
 #define OP_SLT      052
 #define OP_SLTU     053
-#define OP_DADD     054     /* MIPS-II, for r4000 port */
-#define OP_DADDU    055     /* MIPS-II, for r4000 port */
-#define OP_DSUB     056     /* MIPS-II, for r4000 port */
-#define OP_DSUBU    057     /* MIPS-II, for r4000 port */
+#define OP_DADD     054
+#define OP_DADDU    055
+#define OP_DSUB     056
+#define OP_DSUBU    057
 
-#define OP_TGE      060     /* MIPS-II, for r4000 port */
-#define OP_TGEU     061     /* MIPS-II, for r4000 port */
-#define OP_TLT      062     /* MIPS-II, for r4000 port */
-#define OP_TLTU     063     /* MIPS-II, for r4000 port */
-#define OP_TEQ      064     /* MIPS-II, for r4000 port */
-#define OP_TNE      066     /* MIPS-II, for r4000 port */
+#define OP_TGE      060
+#define OP_TGEU     061
+#define OP_TLT      062
+#define OP_TLTU     063
+#define OP_TEQ      064
+#define OP_TNE      066
 
-#define OP_DSLL     070     /* MIPS-II, for r4000 port */
-#define OP_DSRL     072     /* MIPS-II, for r4000 port */
-#define OP_DSRA     073     /* MIPS-II, for r4000 port */
-#define OP_DSLL32   074     /* MIPS-II, for r4000 port */
-#define OP_DSRL32   076     /* MIPS-II, for r4000 port */
-#define OP_DSRA32   077     /* MIPS-II, for r4000 port */
+#define OP_DSLL     070
+#define OP_DSRL     072
+#define OP_DSRA     073
+#define OP_DSLL32   074
+#define OP_DSRL32   076
+#define OP_DSRA32   077
 
+#if __mips_isa_rev < 6
 /*
  * Values for the 'func' field when 'op' == OP_SPECIAL2.
+ * OP_SPECIAL2 opcodes are removed in mips32r6
  */
 #define OP_MAD      000     /* QED */
 #define OP_MADU     001     /* QED */
 #define OP_MUL      002     /* QED */
+#endif
 
 /*
  * Values for the 'func' field when 'op' == OP_SPECIAL3.
  */
 #define OP_EXT      000
+#define OP_DEXTM    001
+#define OP_DEXTU    002
+#define OP_DEXT     003
 #define OP_INS      004
+#define OP_DINSM    005
+#define OP_DINSU    006
+#define OP_DINS     007
 #define OP_BSHFL    040
+#define OP_RDHWR    073
 
 /*
  * Values for the 'shamt' field when OP_SPECIAL3 && func OP_BSHFL.
  */
+
 #define OP_WSBH     002
 #define OP_SEB      020
 #define OP_SEH      030
 
+#if __mips_isa_rev == 6
+/*
+ * Values for the 'shamt' field when OP_SOP30.
+ */
+#define OP_MUL      002
+#define OP_MUH      003
+#endif
+
 /*
  * Values for the 'func' field when 'op' == OP_BCOND.
  */
 #define OP_BLTZ     000
 #define OP_BGEZ     001
-#define OP_BLTZL    002     /* MIPS-II, for r4000 port */
-#define OP_BGEZL    003     /* MIPS-II, for r4000 port */
 
-#define OP_TGEI     010     /* MIPS-II, for r4000 port */
-#define OP_TGEIU    011     /* MIPS-II, for r4000 port */
-#define OP_TLTI     012     /* MIPS-II, for r4000 port */
-#define OP_TLTIU    013     /* MIPS-II, for r4000 port */
-#define OP_TEQI     014     /* MIPS-II, for r4000 port */
-#define OP_TNEI     016     /* MIPS-II, for r4000 port */
-
-#define OP_BLTZAL   020     /* MIPS-II, for r4000 port */
+#if __mips_isa_rev < 6
+#define OP_BLTZL    002
+#define OP_BGEZL    003
+#define OP_TGEI     010
+#define OP_TGEIU    011
+#define OP_TLTI     012
+#define OP_TLTIU    013
+#define OP_TEQI     014
+#define OP_TNEI     016
+#define OP_BLTZAL   020
 #define OP_BGEZAL   021
 #define OP_BLTZALL  022
 #define OP_BGEZALL  023
+#else
+#define OP_NAL      020
+#define OP_BAL      021
+#endif
 
 /*
  * Values for the 'rs' field when 'op' == OP_COPz.
  */
 #define OP_MF       000
-#define OP_DMF      001     /* MIPS-II, for r4000 port */
+#define OP_DMF      001
 #define OP_MT       004
-#define OP_DMT      005     /* MIPS-II, for r4000 port */
+#define OP_DMT      005
 #define OP_BCx      010
 #define OP_BCy      014
 #define OP_CF       002
@@ -311,6 +405,6 @@
 #define COPz_BC_TF_MASK 0x01
 #define COPz_BC_TRUE    0x01
 #define COPz_BC_FALSE   0x00
-#define COPz_BCL_TF_MASK    0x02        /* MIPS-II, for r4000 port */
-#define COPz_BCL_TRUE   0x02        /* MIPS-II, for r4000 port */
-#define COPz_BCL_FALSE  0x00        /* MIPS-II, for r4000 port */
+#define COPz_BCL_TF_MASK    0x02
+#define COPz_BCL_TRUE   0x02
+#define COPz_BCL_FALSE  0x00