* Removed all Id tags.
[cacao.git] / src / vm / jit / x86_64 / emit.c
index 5114ea1b6cb1b055ffe83292801553b6ca19bf2c..44f3742a704166127f9e5f0a97401c5cfbe85e59 100644 (file)
@@ -22,8 +22,6 @@
    Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
    02110-1301, USA.
 
-   $Id: emit.c 7248 2007-01-29 19:28:12Z twisti $
-
 */
 
 #include "config.h"
 
 #include "mm/memory.h"
 
-#if defined(ENABLE_THREADS)
-# include "threads/native/lock.h"
-#endif
+#include "threads/lock-common.h"
 
 #include "vm/builtin.h"
+#include "vm/exceptions.h"
 
+#include "vm/jit/abi.h"
 #include "vm/jit/abi-asm.h"
 #include "vm/jit/asmpart.h"
 #include "vm/jit/codegen-common.h"
@@ -74,19 +72,24 @@ s4 emit_load(jitdata *jd, instruction *iptr, varinfo *src, s4 tempreg)
        if (IS_INMEMORY(src->flags)) {
                COUNT_SPILLS;
 
-               disp = src->vv.regoff * 8;
-
-               if (IS_FLT_DBL_TYPE(src->type)) {
-                       if (IS_2_WORD_TYPE(src->type))
-                               M_DLD(tempreg, REG_SP, disp);
-                       else
-                               M_FLD(tempreg, REG_SP, disp);
-               }
-               else {
-                       if (IS_INT_TYPE(src->type))
-                               M_ILD(tempreg, REG_SP, disp);
-                       else
-                               M_LLD(tempreg, REG_SP, disp);
+               disp = src->vv.regoff;
+
+               switch (src->type) {
+               case TYPE_INT:
+                       M_ILD(tempreg, REG_SP, disp);
+                       break;
+               case TYPE_LNG:
+               case TYPE_ADR:
+                       M_LLD(tempreg, REG_SP, disp);
+                       break;
+               case TYPE_FLT:
+                       M_FLD(tempreg, REG_SP, disp);
+                       break;
+               case TYPE_DBL:
+                       M_DLD(tempreg, REG_SP, disp);
+                       break;
+               default:
+                       vm_abort("emit_load: unknown type %d", src->type);
                }
 
                reg = tempreg;
@@ -146,16 +149,23 @@ inline void emit_store(jitdata *jd, instruction *iptr, varinfo *dst, s4 d)
        if (IS_INMEMORY(dst->flags)) {
                COUNT_SPILLS;
 
-               disp = dst->vv.regoff * 8;
+               disp = dst->vv.regoff;
 
-               if (IS_FLT_DBL_TYPE(dst->type)) {
-                       if (IS_2_WORD_TYPE(dst->type))
-                               M_DST(d, REG_SP, disp);
-                       else
-                               M_FST(d, REG_SP, disp);
-               }
-               else
+               switch (dst->type) {
+               case TYPE_INT:
+               case TYPE_LNG:
+               case TYPE_ADR:
                        M_LST(d, REG_SP, disp);
+                       break;
+               case TYPE_FLT:
+                       M_FST(d, REG_SP, disp);
+                       break;
+               case TYPE_DBL:
+                       M_DST(d, REG_SP, disp);
+                       break;
+               default:
+                       vm_abort("emit_store: unknown type %d", dst->type);
+               }
        }
 }
 
@@ -166,36 +176,57 @@ inline void emit_store(jitdata *jd, instruction *iptr, varinfo *dst, s4 d)
 
 *******************************************************************************/
 
-void emit_copy(jitdata *jd, instruction *iptr, varinfo *src, varinfo *dst)
+void emit_copy(jitdata *jd, instruction *iptr)
 {
-       codegendata  *cd;
-       s4            s1, d;
+       codegendata *cd;
+       varinfo     *src;
+       varinfo     *dst;
+       s4           s1, d;
 
        /* get required compiler data */
 
        cd = jd->cd;
 
+       /* get source and destination variables */
+
+       src = VAROP(iptr->s1);
+       dst = VAROP(iptr->dst);
+
        if ((src->vv.regoff != dst->vv.regoff) ||
                ((src->flags ^ dst->flags) & INMEMORY)) {
 
+               if ((src->type == TYPE_RET) || (dst->type == TYPE_RET)) {
+                       /* emit nothing, as the value won't be used anyway */
+                       return;
+               }
+
                /* If one of the variables resides in memory, we can eliminate
                   the register move from/to the temporary register with the
                   order of getting the destination register and the load. */
 
                if (IS_INMEMORY(src->flags)) {
-                       d = codegen_reg_of_var(iptr->opc, dst, REG_IFTMP);
+                       d  = codegen_reg_of_var(iptr->opc, dst, REG_IFTMP);
                        s1 = emit_load(jd, iptr, src, d);
                }
                else {
                        s1 = emit_load(jd, iptr, src, REG_IFTMP);
-                       d = codegen_reg_of_var(iptr->opc, dst, s1);
+                       d  = codegen_reg_of_var(iptr->opc, dst, s1);
                }
 
                if (s1 != d) {
-                       if (IS_FLT_DBL_TYPE(src->type))
-                               M_FMOV(s1, d);
-                       else
+                       switch (src->type) {
+                       case TYPE_INT:
+                       case TYPE_LNG:
+                       case TYPE_ADR:
                                M_MOV(s1, d);
+                               break;
+                       case TYPE_FLT:
+                       case TYPE_DBL:
+                               M_FMOV(s1, d);
+                               break;
+                       default:
+                               vm_abort("emit_copy: unknown type %d", src->type);
+                       }
                }
 
                emit_store(jd, iptr, dst, d);
@@ -230,6 +261,71 @@ void emit_cmovxx(codegendata *cd, instruction *iptr, s4 s, s4 d)
 }
 
 
+/* emit_branch *****************************************************************
+
+   Emits the code for conditional and unconditional branchs.
+
+*******************************************************************************/
+
+void emit_branch(codegendata *cd, s4 disp, s4 condition, s4 reg, u4 options)
+{
+       s4 branchdisp;
+
+       /* NOTE: A displacement overflow cannot happen. */
+
+       /* check which branch to generate */
+
+       if (condition == BRANCH_UNCONDITIONAL) {
+
+               /* calculate the different displacements */
+
+               branchdisp = disp - BRANCH_UNCONDITIONAL_SIZE;
+
+               M_JMP_IMM(branchdisp);
+       }
+       else {
+               /* calculate the different displacements */
+
+               branchdisp = disp - BRANCH_CONDITIONAL_SIZE;
+
+               switch (condition) {
+               case BRANCH_EQ:
+                       M_BEQ(branchdisp);
+                       break;
+               case BRANCH_NE:
+                       M_BNE(branchdisp);
+                       break;
+               case BRANCH_LT:
+                       M_BLT(branchdisp);
+                       break;
+               case BRANCH_GE:
+                       M_BGE(branchdisp);
+                       break;
+               case BRANCH_GT:
+                       M_BGT(branchdisp);
+                       break;
+               case BRANCH_LE:
+                       M_BLE(branchdisp);
+                       break;
+               case BRANCH_ULT:
+                       M_BULT(branchdisp);
+                       break;
+               case BRANCH_ULE:
+                       M_BULE(branchdisp);
+                       break;
+               case BRANCH_UGE:
+                       M_BUGE(branchdisp);
+                       break;
+               case BRANCH_UGT:
+                       M_BUGT(branchdisp);
+                       break;
+               default:
+                       vm_abort("emit_branch: unknown condition %d", condition);
+               }
+       }
+}
+
+
 /* emit_arithmetic_check *******************************************************
 
    Emit an ArithmeticException check.
@@ -240,8 +336,8 @@ void emit_arithmetic_check(codegendata *cd, instruction *iptr, s4 reg)
 {
        if (INSTRUCTION_MUST_CHECK(iptr)) {
                M_TEST(reg);
-               M_BEQ(0);
-               codegen_add_arithmeticexception_ref(cd);
+               M_BNE(8);
+               M_ALD_MEM(reg, EXCEPTION_HARDWARE_ARITHMETIC);
        }
 }
 
@@ -255,10 +351,10 @@ void emit_arithmetic_check(codegendata *cd, instruction *iptr, s4 reg)
 void emit_arrayindexoutofbounds_check(codegendata *cd, instruction *iptr, s4 s1, s4 s2)
 {
        if (INSTRUCTION_MUST_CHECK(iptr)) {
-        M_ILD(REG_ITMP3, s1, OFFSET(java_arrayheader, size));
+        M_ILD(REG_ITMP3, s1, OFFSET(java_array_t, size));
         M_ICMP(REG_ITMP3, s2);
-        M_BAE(0);
-        codegen_add_arrayindexoutofboundsexception_ref(cd, s2);
+               M_BULT(8);
+               M_ALD_MEM(s2, EXCEPTION_HARDWARE_ARRAYINDEXOUTOFBOUNDS);
        }
 }
 
@@ -271,7 +367,22 @@ void emit_arrayindexoutofbounds_check(codegendata *cd, instruction *iptr, s4 s1,
 
 void emit_classcast_check(codegendata *cd, instruction *iptr, s4 condition, s4 reg, s4 s1)
 {
-       vm_abort("IMPLEMENT ME!");
+       if (INSTRUCTION_MUST_CHECK(iptr)) {
+               switch (condition) {
+               case BRANCH_LE:
+                       M_BGT(8);
+                       break;
+               case BRANCH_EQ:
+                       M_BNE(8);
+                       break;
+               case BRANCH_UGT:
+                       M_BULE(8);
+                       break;
+               default:
+                       vm_abort("emit_classcast_check: unknown condition %d", condition);
+               }
+               M_ALD_MEM(s1, EXCEPTION_HARDWARE_CLASSCAST);
+       }
 }
 
 
@@ -285,85 +396,24 @@ void emit_nullpointer_check(codegendata *cd, instruction *iptr, s4 reg)
 {
        if (INSTRUCTION_MUST_CHECK(iptr)) {
                M_TEST(reg);
-               M_BEQ(0);
-               codegen_add_nullpointerexception_ref(cd);
+               M_BNE(8);
+               M_ALD_MEM(reg, EXCEPTION_HARDWARE_NULLPOINTER);
        }
 }
 
 
-/* emit_exception_stubs ********************************************************
+/* emit_exception_check ********************************************************
 
-   Generates the code for the exception stubs.
+   Emit an Exception check.
 
 *******************************************************************************/
 
-void emit_exception_stubs(jitdata *jd)
+void emit_exception_check(codegendata *cd, instruction *iptr)
 {
-       codegendata  *cd;
-       registerdata *rd;
-       exceptionref *er;
-       s4            branchmpc;
-       s4            targetmpc;
-       s4            targetdisp;
-
-       /* get required compiler data */
-
-       cd = jd->cd;
-       rd = jd->rd;
-
-       /* generate exception stubs */
-
-       targetdisp = 0;
-
-       for (er = cd->exceptionrefs; er != NULL; er = er->next) {
-               /* back-patch the branch to this exception code */
-
-               branchmpc = er->branchpos;
-               targetmpc = cd->mcodeptr - cd->mcodebase;
-
-               md_codegen_patch_branch(cd, branchmpc, targetmpc);
-
-               MCODECHECK(512);
-
-               /* Check if the exception is an
-                  ArrayIndexOutOfBoundsException.  If so, move index register
-                  into a4. */
-
-               if (er->reg != -1)
-                       M_MOV(er->reg, rd->argintregs[4]);
-
-               /* calcuate exception address */
-
-               M_MOV_IMM(0, rd->argintregs[3]);
-               dseg_adddata(cd);
-               M_AADD_IMM32(er->branchpos - 6, rd->argintregs[3]);
-
-               /* move function to call into REG_ITMP3 */
-
-               M_MOV_IMM(er->function, REG_ITMP3);
-
-               if (targetdisp == 0) {
-                       targetdisp = cd->mcodeptr - cd->mcodebase;
-
-                       emit_lea_membase_reg(cd, RIP, -((cd->mcodeptr + 7) - cd->mcodebase), rd->argintregs[0]);
-                       M_MOV(REG_SP, rd->argintregs[1]);
-                       M_ALD(rd->argintregs[2], REG_SP, cd->stackframesize * 8);
-
-                       M_ASUB_IMM(2 * 8, REG_SP);
-                       M_AST(rd->argintregs[3], REG_SP, 0 * 8);             /* store XPC */
-
-                       M_CALL(REG_ITMP3);
-
-                       M_ALD(REG_ITMP2_XPC, REG_SP, 0 * 8);
-                       M_AADD_IMM(2 * 8, REG_SP);
-
-                       M_MOV_IMM(asm_handle_exception, REG_ITMP3);
-                       M_JMP(REG_ITMP3);
-               }
-               else {
-                       M_JMP_IMM((cd->mcodebase + targetdisp) -
-                                         (cd->mcodeptr + PATCHER_CALL_SIZE));
-               }
+       if (INSTRUCTION_MUST_CHECK(iptr)) {
+               M_TEST(REG_RESULT);
+               M_BNE(8);
+               M_ALD_MEM(REG_RESULT, EXCEPTION_HARDWARE_EXCEPTION);
        }
 }
 
@@ -455,66 +505,25 @@ void emit_patcher_stubs(jitdata *jd)
 }
 
 
-/* emit_replacement_stubs ******************************************************
+/* emit_trap *******************************************************************
 
-   Generates the code for the replacement stubs.
+   Emit a trap instruction and return the original machine code.
 
 *******************************************************************************/
 
-#if defined(ENABLE_REPLACEMENT)
-void emit_replacement_stubs(jitdata *jd)
+uint32_t emit_trap(codegendata *cd)
 {
-       codegendata *cd;
-       codeinfo    *code;
-       rplpoint    *rplp;
-       s4           disp;
-       s4           i;
-#if !defined(NDEBUG)
-       u1          *savedmcodeptr;
-#endif
-
-       /* get required compiler data */
-
-       cd   = jd->cd;
-       code = jd->code;
+       uint32_t mcode;
 
-       rplp = code->rplpoints;
-
-       /* store beginning of replacement stubs */
-
-       code->replacementstubs = (u1*) (cd->mcodeptr - cd->mcodebase);
-
-       for (i = 0; i < code->rplpointcount; ++i, ++rplp) {
-               /* do not generate stubs for non-trappable points */
-
-               if (rplp->flags & RPLPOINT_FLAG_NOTRAP)
-                       continue;
-
-               /* check code segment size */
-
-               MCODECHECK(512);
+       /* Get machine code which is patched back in later. The
+          trap is 1 instruction word long. */
 
-               /* note start of stub code */
+       mcode = *((uint32_t *) cd->mcodeptr);
 
-#if !defined(NDEBUG)
-               savedmcodeptr = cd->mcodeptr;
-#endif
-
-               /* push address of `rplpoint` struct */
-                       
-               M_MOV_IMM(rplp, REG_ITMP3);
-               M_PUSH(REG_ITMP3);
-
-               /* jump to replacement function */
-
-               M_MOV_IMM(asm_replacement_out, REG_ITMP3);
-               M_PUSH(REG_ITMP3);
-               M_RET;
+       M_NOP;
 
-               assert((cd->mcodeptr - savedmcodeptr) == REPLACEMENT_STUB_SIZE);
-       }
+       return mcode;
 }
-#endif /* defined(ENABLE_REPLACEMENT) */
 
 
 /* emit_verbosecall_enter ******************************************************
@@ -551,10 +560,10 @@ void emit_verbosecall_enter(jitdata *jd)
        /* save argument registers */
 
        for (i = 0; i < INT_ARG_CNT; i++)
-               M_LST(rd->argintregs[i], REG_SP, (1 + i) * 8);
+               M_LST(abi_registers_integer_argument[i], REG_SP, (1 + i) * 8);
 
        for (i = 0; i < FLT_ARG_CNT; i++)
-               M_DST(rd->argfltregs[i], REG_SP, (1 + INT_ARG_CNT + i) * 8);
+               M_DST(abi_registers_float_argument[i], REG_SP, (1 + INT_ARG_CNT + i) * 8);
 
        /* save temporary registers for leaf methods */
 
@@ -574,25 +583,27 @@ void emit_verbosecall_enter(jitdata *jd)
        
                if (IS_FLT_DBL_TYPE(md->paramtypes[i].type)) {
                        for (k = INT_ARG_CNT - 2; k >= i; k--)
-                               M_MOV(rd->argintregs[k], rd->argintregs[k + 1]);
+                               M_MOV(abi_registers_integer_argument[k],
+                                         abi_registers_integer_argument[k + 1]);
 
-                       emit_movd_freg_reg(cd, rd->argfltregs[j], rd->argintregs[i]);
+                       emit_movd_freg_reg(cd, abi_registers_float_argument[j],
+                                                          abi_registers_integer_argument[i]);
                        j++;
                }
        }
 
        M_MOV_IMM(m, REG_ITMP2);
        M_AST(REG_ITMP2, REG_SP, 0 * 8);
-       M_MOV_IMM(builtin_trace_args, REG_ITMP1);
+       M_MOV_IMM(builtin_verbosecall_enter, REG_ITMP1);
        M_CALL(REG_ITMP1);
 
        /* restore argument registers */
 
        for (i = 0; i < INT_ARG_CNT; i++)
-               M_LLD(rd->argintregs[i], REG_SP, (1 + i) * 8);
+               M_LLD(abi_registers_integer_argument[i], REG_SP, (1 + i) * 8);
 
        for (i = 0; i < FLT_ARG_CNT; i++)
-               M_DLD(rd->argfltregs[i], REG_SP, (1 + INT_ARG_CNT + i) * 8);
+               M_DLD(abi_registers_float_argument[i], REG_SP, (1 + INT_ARG_CNT + i) * 8);
 
        /* restore temporary registers for leaf methods */
 
@@ -641,12 +652,12 @@ void emit_verbosecall_exit(jitdata *jd)
        M_LST(REG_RESULT, REG_SP, 0 * 8);
        M_DST(REG_FRESULT, REG_SP, 1 * 8);
 
-       M_MOV_IMM(m, rd->argintregs[0]);
-       M_MOV(REG_RESULT, rd->argintregs[1]);
-       M_FLTMOVE(REG_FRESULT, rd->argfltregs[0]);
-       M_FLTMOVE(REG_FRESULT, rd->argfltregs[1]);
+       M_INTMOVE(REG_RESULT, REG_A0);
+       M_FLTMOVE(REG_FRESULT, REG_FA0);
+       M_FLTMOVE(REG_FRESULT, REG_FA1);
+       M_MOV_IMM(m, REG_A1);
 
-       M_MOV_IMM(builtin_displaymethodstop, REG_ITMP1);
+       M_MOV_IMM(builtin_verbosecall_exit, REG_ITMP1);
        M_CALL(REG_ITMP1);
 
        M_LLD(REG_RESULT, REG_SP, 0 * 8);
@@ -762,53 +773,53 @@ void emit_ishift(jitdata *jd, s4 shift_op, instruction *iptr)
        if (IS_INMEMORY(v_dst->flags)) {
                if (IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
                        if (s1 == d) {
-                               M_ILD(RCX, REG_SP, s2 * 8);
-                               emit_shiftl_membase(cd, shift_op, REG_SP, d * 8);
+                               M_ILD(RCX, REG_SP, s2);
+                               emit_shiftl_membase(cd, shift_op, REG_SP, d);
 
                        } else {
-                               M_ILD(RCX, REG_SP, s2 * 8);
-                               M_ILD(REG_ITMP2, REG_SP, s1 * 8);
+                               M_ILD(RCX, REG_SP, s2);
+                               M_ILD(REG_ITMP2, REG_SP, s1);
                                emit_shiftl_reg(cd, shift_op, REG_ITMP2);
-                               M_IST(REG_ITMP2, REG_SP, d * 8);
+                               M_IST(REG_ITMP2, REG_SP, d);
                        }
 
                } else if (IS_INMEMORY(v_s2->flags) && !IS_INMEMORY(v_s1->flags)) {
                        /* s1 may be equal to RCX */
                        if (s1 == RCX) {
                                if (s2 == d) {
-                                       M_ILD(REG_ITMP1, REG_SP, s2 * 8);
-                                       M_IST(s1, REG_SP, d * 8);
+                                       M_ILD(REG_ITMP1, REG_SP, s2);
+                                       M_IST(s1, REG_SP, d);
                                        M_INTMOVE(REG_ITMP1, RCX);
 
                                } else {
-                                       M_IST(s1, REG_SP, d * 8);
-                                       M_ILD(RCX, REG_SP, s2 * 8);
+                                       M_IST(s1, REG_SP, d);
+                                       M_ILD(RCX, REG_SP, s2);
                                }
 
                        } else {
-                               M_ILD(RCX, REG_SP, s2 * 8);
-                               M_IST(s1, REG_SP, d * 8);
+                               M_ILD(RCX, REG_SP, s2);
+                               M_IST(s1, REG_SP, d);
                        }
 
-                       emit_shiftl_membase(cd, shift_op, REG_SP, d * 8);
+                       emit_shiftl_membase(cd, shift_op, REG_SP, d);
 
                } else if (!IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
                        if (s1 == d) {
                                M_INTMOVE(s2, RCX);
-                               emit_shiftl_membase(cd, shift_op, REG_SP, d * 8);
+                               emit_shiftl_membase(cd, shift_op, REG_SP, d);
 
                        } else {
                                M_INTMOVE(s2, RCX);
-                               M_ILD(REG_ITMP2, REG_SP, s1 * 8);
+                               M_ILD(REG_ITMP2, REG_SP, s1);
                                emit_shiftl_reg(cd, shift_op, REG_ITMP2);
-                               M_IST(REG_ITMP2, REG_SP, d * 8);
+                               M_IST(REG_ITMP2, REG_SP, d);
                        }
 
                } else {
                        /* s1 may be equal to RCX */
-                       M_IST(s1, REG_SP, d * 8);
+                       M_IST(s1, REG_SP, d);
                        M_INTMOVE(s2, RCX);
-                       emit_shiftl_membase(cd, shift_op, REG_SP, d * 8);
+                       emit_shiftl_membase(cd, shift_op, REG_SP, d);
                }
 
                M_INTMOVE(REG_ITMP1, RCX);                             /* restore RCX */
@@ -820,19 +831,19 @@ void emit_ishift(jitdata *jd, s4 shift_op, instruction *iptr)
                }
                                        
                if (IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
-                       M_ILD(RCX, REG_SP, s2 * 8);
-                       M_ILD(d, REG_SP, s1 * 8);
+                       M_ILD(RCX, REG_SP, s2);
+                       M_ILD(d, REG_SP, s1);
                        emit_shiftl_reg(cd, shift_op, d);
 
                } else if (IS_INMEMORY(v_s2->flags) && !IS_INMEMORY(v_s1->flags)) {
                        /* s1 may be equal to RCX */
                        M_INTMOVE(s1, d);
-                       M_ILD(RCX, REG_SP, s2 * 8);
+                       M_ILD(RCX, REG_SP, s2);
                        emit_shiftl_reg(cd, shift_op, d);
 
                } else if (!IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
                        M_INTMOVE(s2, RCX);
-                       M_ILD(d, REG_SP, s1 * 8);
+                       M_ILD(d, REG_SP, s1);
                        emit_shiftl_reg(cd, shift_op, d);
 
                } else {
@@ -889,53 +900,53 @@ void emit_lshift(jitdata *jd, s4 shift_op, instruction *iptr)
        if (IS_INMEMORY(v_dst->flags)) {
                if (IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
                        if (s1 == d) {
-                               M_ILD(RCX, REG_SP, s2 * 8);
-                               emit_shift_membase(cd, shift_op, REG_SP, d * 8);
+                               M_ILD(RCX, REG_SP, s2);
+                               emit_shift_membase(cd, shift_op, REG_SP, d);
 
                        } else {
-                               M_ILD(RCX, REG_SP, s2 * 8);
-                               M_LLD(REG_ITMP2, REG_SP, s1 * 8);
+                               M_ILD(RCX, REG_SP, s2);
+                               M_LLD(REG_ITMP2, REG_SP, s1);
                                emit_shift_reg(cd, shift_op, REG_ITMP2);
-                               M_LST(REG_ITMP2, REG_SP, d * 8);
+                               M_LST(REG_ITMP2, REG_SP, d);
                        }
 
                } else if (IS_INMEMORY(v_s2->flags) && !IS_INMEMORY(v_s1->flags)) {
                        /* s1 may be equal to RCX */
                        if (s1 == RCX) {
                                if (s2 == d) {
-                                       M_ILD(REG_ITMP1, REG_SP, s2 * 8);
-                                       M_LST(s1, REG_SP, d * 8);
+                                       M_ILD(REG_ITMP1, REG_SP, s2);
+                                       M_LST(s1, REG_SP, d);
                                        M_INTMOVE(REG_ITMP1, RCX);
 
                                } else {
-                                       M_LST(s1, REG_SP, d * 8);
-                                       M_ILD(RCX, REG_SP, s2 * 8);
+                                       M_LST(s1, REG_SP, d);
+                                       M_ILD(RCX, REG_SP, s2);
                                }
 
                        } else {
-                               M_ILD(RCX, REG_SP, s2 * 8);
-                               M_LST(s1, REG_SP, d * 8);
+                               M_ILD(RCX, REG_SP, s2);
+                               M_LST(s1, REG_SP, d);
                        }
 
-                       emit_shift_membase(cd, shift_op, REG_SP, d * 8);
+                       emit_shift_membase(cd, shift_op, REG_SP, d);
 
                } else if (!IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
                        if (s1 == d) {
                                M_INTMOVE(s2, RCX);
-                               emit_shift_membase(cd, shift_op, REG_SP, d * 8);
+                               emit_shift_membase(cd, shift_op, REG_SP, d);
 
                        } else {
                                M_INTMOVE(s2, RCX);
-                               M_LLD(REG_ITMP2, REG_SP, s1 * 8);
+                               M_LLD(REG_ITMP2, REG_SP, s1);
                                emit_shift_reg(cd, shift_op, REG_ITMP2);
-                               M_LST(REG_ITMP2, REG_SP, d * 8);
+                               M_LST(REG_ITMP2, REG_SP, d);
                        }
 
                } else {
                        /* s1 may be equal to RCX */
-                       M_LST(s1, REG_SP, d * 8);
+                       M_LST(s1, REG_SP, d);
                        M_INTMOVE(s2, RCX);
-                       emit_shift_membase(cd, shift_op, REG_SP, d * 8);
+                       emit_shift_membase(cd, shift_op, REG_SP, d);
                }
 
                M_INTMOVE(REG_ITMP1, RCX);                             /* restore RCX */
@@ -947,19 +958,19 @@ void emit_lshift(jitdata *jd, s4 shift_op, instruction *iptr)
                }
 
                if (IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
-                       M_ILD(RCX, REG_SP, s2 * 8);
-                       M_LLD(d, REG_SP, s1 * 8);
+                       M_ILD(RCX, REG_SP, s2);
+                       M_LLD(d, REG_SP, s1);
                        emit_shift_reg(cd, shift_op, d);
 
                } else if (IS_INMEMORY(v_s2->flags) && !IS_INMEMORY(v_s1->flags)) {
                        /* s1 may be equal to RCX */
                        M_INTMOVE(s1, d);
-                       M_ILD(RCX, REG_SP, s2 * 8);
+                       M_ILD(RCX, REG_SP, s2);
                        emit_shift_reg(cd, shift_op, d);
 
                } else if (!IS_INMEMORY(v_s2->flags) && IS_INMEMORY(v_s1->flags)) {
                        M_INTMOVE(s2, RCX);
-                       M_LLD(d, REG_SP, s1 * 8);
+                       M_LLD(d, REG_SP, s1);
                        emit_shift_reg(cd, shift_op, d);
 
                } else {
@@ -1269,6 +1280,15 @@ void emit_movb_imm_memindex(codegendata *cd, s4 imm, s4 disp, s4 basereg, s4 ind
 }
 
 
+void emit_mov_mem_reg(codegendata *cd, s4 disp, s4 dreg)
+{
+       emit_rex(1, dreg, 0, 0);
+       *(cd->mcodeptr++) = 0x8b;
+       emit_address_byte(0, dreg, 4);
+       emit_mem(4, disp);
+}
+
+
 /*
  * alu operations
  */
@@ -1335,7 +1355,8 @@ void emit_alu_imm_reg(codegendata *cd, s8 opc, s8 imm, s8 dreg) {
 }
 
 
-void emit_alu_imm32_reg(codegendata *cd, s8 opc, s8 imm, s8 dreg) {
+void emit_alu_imm32_reg(codegendata *cd, s4 opc, s4 imm, s4 dreg)
+{
        emit_rex(1,0,0,(dreg));
        *(cd->mcodeptr++) = 0x81;
        emit_reg((opc),(dreg));
@@ -1343,6 +1364,15 @@ void emit_alu_imm32_reg(codegendata *cd, s8 opc, s8 imm, s8 dreg) {
 }
 
 
+void emit_alul_imm32_reg(codegendata *cd, s4 opc, s4 imm, s4 dreg)
+{
+       emit_rex(0,0,0,(dreg));
+       *(cd->mcodeptr++) = 0x81;
+       emit_reg((opc),(dreg));
+       emit_imm32((imm));
+}
+
+
 void emit_alul_imm_reg(codegendata *cd, s8 opc, s8 imm, s8 dreg) {
        if (IS_IMM8(imm)) {
                emit_rex(0,0,0,(dreg));
@@ -1705,7 +1735,8 @@ void emit_jcc(codegendata *cd, s8 opc, s8 imm) {
  */
 
 /* we need the rex byte to get all low bytes */
-void emit_setcc_reg(codegendata *cd, s8 opc, s8 reg) {
+void emit_setcc_reg(codegendata *cd, s4 opc, s4 reg)
+{
        *(cd->mcodeptr++) = (0x40 | (((reg) >> 3) & 0x01));
        *(cd->mcodeptr++) = 0x0f;
        *(cd->mcodeptr++) = (0x90 + (opc));
@@ -1714,7 +1745,8 @@ void emit_setcc_reg(codegendata *cd, s8 opc, s8 reg) {
 
 
 /* we need the rex byte to get all low bytes */
-void emit_setcc_membase(codegendata *cd, s8 opc, s8 basereg, s8 disp) {
+void emit_setcc_membase(codegendata *cd, s4 opc, s4 basereg, s4 disp)
+{
        *(cd->mcodeptr++) = (0x40 | (((basereg) >> 3) & 0x01));
        *(cd->mcodeptr++) = 0x0f;
        *(cd->mcodeptr++) = (0x90 + (opc));
@@ -1722,7 +1754,7 @@ void emit_setcc_membase(codegendata *cd, s8 opc, s8 basereg, s8 disp) {
 }
 
 
-void emit_cmovcc_reg_reg(codegendata *cd, s8 opc, s8 reg, s8 dreg)
+void emit_cmovcc_reg_reg(codegendata *cd, s4 opc, s4 reg, s4 dreg)
 {
        emit_rex(1,(dreg),0,(reg));
        *(cd->mcodeptr++) = 0x0f;
@@ -1731,7 +1763,7 @@ void emit_cmovcc_reg_reg(codegendata *cd, s8 opc, s8 reg, s8 dreg)
 }
 
 
-void emit_cmovccl_reg_reg(codegendata *cd, s8 opc, s8 reg, s8 dreg)
+void emit_cmovccl_reg_reg(codegendata *cd, s4 opc, s4 reg, s4 dreg)
 {
        emit_rex(0,(dreg),0,(reg));
        *(cd->mcodeptr++) = 0x0f;
@@ -1740,7 +1772,6 @@ void emit_cmovccl_reg_reg(codegendata *cd, s8 opc, s8 reg, s8 dreg)
 }
 
 
-
 void emit_neg_reg(codegendata *cd, s8 reg)
 {
        emit_rex(1,0,0,(reg));
@@ -1791,14 +1822,16 @@ void emit_nop(codegendata *cd) {
 /*
  * call instructions
  */
-void emit_call_reg(codegendata *cd, s8 reg) {
-       emit_rex(1,0,0,(reg));
+void emit_call_reg(codegendata *cd, s8 reg)
+{
+       emit_rex(0,0,0,(reg));
        *(cd->mcodeptr++) = 0xff;
        emit_reg(2,(reg));
 }
 
 
-void emit_call_imm(codegendata *cd, s8 imm) {
+void emit_call_imm(codegendata *cd, s8 imm)
+{
        *(cd->mcodeptr++) = 0xe8;
        emit_imm32((imm));
 }