/* Simulator for Motorola's MCore processor
Copyright (C) 1999-2017 Free Software Foundation, Inc.
Contributed by Cygnus Solutions.
This file is part of GDB, the GNU debugger.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see . */
#include "config.h"
#include
#include
#include
#include
#include
#include
#include "bfd.h"
#include "gdb/callback.h"
#include "libiberty.h"
#include "gdb/remote-sim.h"
#include "sim-main.h"
#include "sim-base.h"
#include "sim-syscall.h"
#include "sim-options.h"
#define target_big_endian (CURRENT_TARGET_BYTE_ORDER == BIG_ENDIAN)
static unsigned long
mcore_extract_unsigned_integer (unsigned char *addr, int len)
{
unsigned long retval;
unsigned char * p;
unsigned char * startaddr = (unsigned char *)addr;
unsigned char * endaddr = startaddr + len;
if (len > (int) sizeof (unsigned long))
printf ("That operation is not available on integers of more than %zu bytes.",
sizeof (unsigned long));
/* Start at the most significant end of the integer, and work towards
the least significant. */
retval = 0;
if (! target_big_endian)
{
for (p = endaddr; p > startaddr;)
retval = (retval << 8) | * -- p;
}
else
{
for (p = startaddr; p < endaddr;)
retval = (retval << 8) | * p ++;
}
return retval;
}
static void
mcore_store_unsigned_integer (unsigned char *addr, int len, unsigned long val)
{
unsigned char * p;
unsigned char * startaddr = (unsigned char *)addr;
unsigned char * endaddr = startaddr + len;
if (! target_big_endian)
{
for (p = startaddr; p < endaddr;)
{
* p ++ = val & 0xff;
val >>= 8;
}
}
else
{
for (p = endaddr; p > startaddr;)
{
* -- p = val & 0xff;
val >>= 8;
}
}
}
static int memcycles = 1;
#define gr cpu->active_gregs
#define cr cpu->regs.cregs
#define sr cr[0]
#define vbr cr[1]
#define esr cr[2]
#define fsr cr[3]
#define epc cr[4]
#define fpc cr[5]
#define ss0 cr[6]
#define ss1 cr[7]
#define ss2 cr[8]
#define ss3 cr[9]
#define ss4 cr[10]
#define gcr cr[11]
#define gsr cr[12]
/* maniuplate the carry bit */
#define C_ON() (sr & 1)
#define C_VALUE() (sr & 1)
#define C_OFF() ((sr & 1) == 0)
#define SET_C() {sr |= 1;}
#define CLR_C() {sr &= 0xfffffffe;}
#define NEW_C(v) {CLR_C(); sr |= ((v) & 1);}
#define SR_AF() ((sr >> 1) & 1)
static void set_active_regs (SIM_CPU *cpu)
{
if (SR_AF())
cpu->active_gregs = cpu->regs.alt_gregs;
else
cpu->active_gregs = cpu->regs.gregs;
}
#define TRAPCODE 1 /* r1 holds which function we want */
#define PARM1 2 /* first parameter */
#define PARM2 3
#define PARM3 4
#define PARM4 5
#define RET1 2 /* register for return values. */
/* Default to a 8 Mbyte (== 2^23) memory space. */
#define DEFAULT_MEMORY_SIZE 0x800000
static void
set_initial_gprs (SIM_CPU *cpu)
{
/* Set up machine just out of reset. */
CPU_PC_SET (cpu, 0);
sr = 0;
/* Clean out the GPRs and alternate GPRs. */
memset (&cpu->regs.gregs, 0, sizeof(cpu->regs.gregs));
memset (&cpu->regs.alt_gregs, 0, sizeof(cpu->regs.alt_gregs));
/* Make our register set point to the right place. */
set_active_regs (cpu);
/* ABI specifies initial values for these registers. */
gr[0] = DEFAULT_MEMORY_SIZE - 4;
/* dac fix, the stack address must be 8-byte aligned! */
gr[0] = gr[0] - gr[0] % 8;
gr[PARM1] = 0;
gr[PARM2] = 0;
gr[PARM3] = 0;
gr[PARM4] = gr[0];
}
/* Simulate a monitor trap. */
static void
handle_trap1 (SIM_DESC sd, SIM_CPU *cpu)
{
/* XXX: We don't pass back the actual errno value. */
gr[RET1] = sim_syscall (cpu, gr[TRAPCODE], gr[PARM1], gr[PARM2], gr[PARM3],
gr[PARM4]);
}
static void
process_stub (SIM_DESC sd, SIM_CPU *cpu, int what)
{
/* These values should match those in libgloss/mcore/syscalls.s. */
switch (what)
{
case 3: /* _read */
case 4: /* _write */
case 5: /* _open */
case 6: /* _close */
case 10: /* _unlink */
case 19: /* _lseek */
case 43: /* _times */
gr[TRAPCODE] = what;
handle_trap1 (sd, cpu);
break;
default:
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "Unhandled stub opcode: %d\n", what);
break;
}
}
static void
util (SIM_DESC sd, SIM_CPU *cpu, unsigned what)
{
switch (what)
{
case 0: /* exit */
sim_engine_halt (sd, cpu, NULL, cpu->regs.pc, sim_exited, gr[PARM1]);
break;
case 1: /* printf */
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "WARNING: printf unimplemented\n");
break;
case 2: /* scanf */
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "WARNING: scanf unimplemented\n");
break;
case 3: /* utime */
gr[RET1] = cpu->insts;
break;
case 0xFF:
process_stub (sd, cpu, gr[1]);
break;
default:
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "Unhandled util code: %x\n", what);
break;
}
}
/* For figuring out whether we carried; addc/subc use this. */
static int
iu_carry (unsigned long a, unsigned long b, int cin)
{
unsigned long x;
x = (a & 0xffff) + (b & 0xffff) + cin;
x = (x >> 16) + (a >> 16) + (b >> 16);
x >>= 16;
return (x != 0);
}
/* TODO: Convert to common watchpoints. */
#undef WATCHFUNCTIONS
#ifdef WATCHFUNCTIONS
#define MAXWL 80
word WL[MAXWL];
char * WLstr[MAXWL];
int ENDWL=0;
int WLincyc;
int WLcyc[MAXWL];
int WLcnts[MAXWL];
int WLmax[MAXWL];
int WLmin[MAXWL];
word WLendpc;
int WLbcyc;
int WLW;
#endif
#define RD (inst & 0xF)
#define RS ((inst >> 4) & 0xF)
#define RX ((inst >> 8) & 0xF)
#define IMM5 ((inst >> 4) & 0x1F)
#define IMM4 ((inst) & 0xF)
#define rbat(X) sim_core_read_1 (cpu, 0, read_map, X)
#define rhat(X) sim_core_read_2 (cpu, 0, read_map, X)
#define rlat(X) sim_core_read_4 (cpu, 0, read_map, X)
#define wbat(X, D) sim_core_write_1 (cpu, 0, write_map, X, D)
#define what(X, D) sim_core_write_2 (cpu, 0, write_map, X, D)
#define wlat(X, D) sim_core_write_4 (cpu, 0, write_map, X, D)
static int tracing = 0;
#define ILLEGAL() \
sim_engine_halt (sd, cpu, NULL, pc, sim_stopped, SIM_SIGILL)
static void
step_once (SIM_DESC sd, SIM_CPU *cpu)
{
int needfetch;
word ibuf;
word pc;
unsigned short inst;
int memops;
int bonus_cycles;
int insts;
int w;
int cycs;
#ifdef WATCHFUNCTIONS
word WLhash;
#endif
pc = CPU_PC_GET (cpu);
/* Fetch the initial instructions that we'll decode. */
ibuf = rlat (pc & 0xFFFFFFFC);
needfetch = 0;
memops = 0;
bonus_cycles = 0;
insts = 0;
/* make our register set point to the right place */
set_active_regs (cpu);
#ifdef WATCHFUNCTIONS
/* make a hash to speed exec loop, hope it's nonzero */
WLhash = 0xFFFFFFFF;
for (w = 1; w <= ENDWL; w++)
WLhash = WLhash & WL[w];
#endif
/* TODO: Unindent this block. */
{
word oldpc;
insts ++;
if (pc & 02)
{
if (! target_big_endian)
inst = ibuf >> 16;
else
inst = ibuf & 0xFFFF;
needfetch = 1;
}
else
{
if (! target_big_endian)
inst = ibuf & 0xFFFF;
else
inst = ibuf >> 16;
}
#ifdef WATCHFUNCTIONS
/* now scan list of watch addresses, if match, count it and
note return address and count cycles until pc=return address */
if ((WLincyc == 1) && (pc == WLendpc))
{
cycs = (cpu->cycles + (insts + bonus_cycles +
(memops * memcycles)) - WLbcyc);
if (WLcnts[WLW] == 1)
{
WLmax[WLW] = cycs;
WLmin[WLW] = cycs;
WLcyc[WLW] = 0;
}
if (cycs > WLmax[WLW])
{
WLmax[WLW] = cycs;
}
if (cycs < WLmin[WLW])
{
WLmin[WLW] = cycs;
}
WLcyc[WLW] += cycs;
WLincyc = 0;
WLendpc = 0;
}
/* Optimize with a hash to speed loop. */
if (WLincyc == 0)
{
if ((WLhash == 0) || ((WLhash & pc) != 0))
{
for (w=1; w <= ENDWL; w++)
{
if (pc == WL[w])
{
WLcnts[w]++;
WLbcyc = cpu->cycles + insts
+ bonus_cycles + (memops * memcycles);
WLendpc = gr[15];
WLincyc = 1;
WLW = w;
break;
}
}
}
}
#endif
if (tracing)
fprintf (stderr, "%.4lx: inst = %.4x ", pc, inst);
oldpc = pc;
pc += 2;
switch (inst >> 8)
{
case 0x00:
switch RS
{
case 0x0:
switch RD
{
case 0x0: /* bkpt */
pc -= 2;
sim_engine_halt (sd, cpu, NULL, pc - 2,
sim_stopped, SIM_SIGTRAP);
break;
case 0x1: /* sync */
break;
case 0x2: /* rte */
pc = epc;
sr = esr;
needfetch = 1;
set_active_regs (cpu);
break;
case 0x3: /* rfi */
pc = fpc;
sr = fsr;
needfetch = 1;
set_active_regs (cpu);
break;
case 0x4: /* stop */
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "WARNING: stop unimplemented\n");
break;
case 0x5: /* wait */
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "WARNING: wait unimplemented\n");
break;
case 0x6: /* doze */
if (STATE_VERBOSE_P (sd))
fprintf (stderr, "WARNING: doze unimplemented\n");
break;
case 0x7:
ILLEGAL (); /* illegal */
break;
case 0x8: /* trap 0 */
case 0xA: /* trap 2 */
case 0xB: /* trap 3 */
sim_engine_halt (sd, cpu, NULL, pc,
sim_stopped, SIM_SIGTRAP);
break;
case 0xC: /* trap 4 */
case 0xD: /* trap 5 */
case 0xE: /* trap 6 */
ILLEGAL (); /* illegal */
break;
case 0xF: /* trap 7 */
sim_engine_halt (sd, cpu, NULL, pc, /* integer div-by-0 */
sim_stopped, SIM_SIGTRAP);
break;
case 0x9: /* trap 1 */
handle_trap1 (sd, cpu);
break;
}
break;
case 0x1:
ILLEGAL (); /* illegal */
break;
case 0x2: /* mvc */
gr[RD] = C_VALUE();
break;
case 0x3: /* mvcv */
gr[RD] = C_OFF();
break;
case 0x4: /* ldq */
{
word addr = gr[RD];
int regno = 4; /* always r4-r7 */
bonus_cycles++;
memops += 4;
do
{
gr[regno] = rlat (addr);
addr += 4;
regno++;
}
while ((regno&0x3) != 0);
}
break;
case 0x5: /* stq */
{
word addr = gr[RD];
int regno = 4; /* always r4-r7 */
memops += 4;
bonus_cycles++;
do
{
wlat (addr, gr[regno]);
addr += 4;
regno++;
}
while ((regno & 0x3) != 0);
}
break;
case 0x6: /* ldm */
{
word addr = gr[0];
int regno = RD;
/* bonus cycle is really only needed if
the next insn shifts the last reg loaded.
bonus_cycles++;
*/
memops += 16-regno;
while (regno <= 0xF)
{
gr[regno] = rlat (addr);
addr += 4;
regno++;
}
}
break;
case 0x7: /* stm */
{
word addr = gr[0];
int regno = RD;
/* this should be removed! */
/* bonus_cycles ++; */
memops += 16 - regno;
while (regno <= 0xF)
{
wlat (addr, gr[regno]);
addr += 4;
regno++;
}
}
break;
case 0x8: /* dect */
gr[RD] -= C_VALUE();
break;
case 0x9: /* decf */
gr[RD] -= C_OFF();
break;
case 0xA: /* inct */
gr[RD] += C_VALUE();
break;
case 0xB: /* incf */
gr[RD] += C_OFF();
break;
case 0xC: /* jmp */
pc = gr[RD];
if (tracing && RD == 15)
fprintf (stderr, "Func return, r2 = %lxx, r3 = %lx\n",
gr[2], gr[3]);
bonus_cycles++;
needfetch = 1;
break;
case 0xD: /* jsr */
gr[15] = pc;
pc = gr[RD];
bonus_cycles++;
needfetch = 1;
break;
case 0xE: /* ff1 */
{
word tmp, i;
tmp = gr[RD];
for (i = 0; !(tmp & 0x80000000) && i < 32; i++)
tmp <<= 1;
gr[RD] = i;
}
break;
case 0xF: /* brev */
{
word tmp;
tmp = gr[RD];
tmp = ((tmp & 0xaaaaaaaa) >> 1) | ((tmp & 0x55555555) << 1);
tmp = ((tmp & 0xcccccccc) >> 2) | ((tmp & 0x33333333) << 2);
tmp = ((tmp & 0xf0f0f0f0) >> 4) | ((tmp & 0x0f0f0f0f) << 4);
tmp = ((tmp & 0xff00ff00) >> 8) | ((tmp & 0x00ff00ff) << 8);
gr[RD] = ((tmp & 0xffff0000) >> 16) | ((tmp & 0x0000ffff) << 16);
}
break;
}
break;
case 0x01:
switch RS
{
case 0x0: /* xtrb3 */
gr[1] = (gr[RD]) & 0xFF;
NEW_C (gr[RD] != 0);
break;
case 0x1: /* xtrb2 */
gr[1] = (gr[RD]>>8) & 0xFF;
NEW_C (gr[RD] != 0);
break;
case 0x2: /* xtrb1 */
gr[1] = (gr[RD]>>16) & 0xFF;
NEW_C (gr[RD] != 0);
break;
case 0x3: /* xtrb0 */
gr[1] = (gr[RD]>>24) & 0xFF;
NEW_C (gr[RD] != 0);
break;
case 0x4: /* zextb */
gr[RD] &= 0x000000FF;
break;
case 0x5: /* sextb */
{
long tmp;
tmp = gr[RD];
tmp <<= 24;
tmp >>= 24;
gr[RD] = tmp;
}
break;
case 0x6: /* zexth */
gr[RD] &= 0x0000FFFF;
break;
case 0x7: /* sexth */
{
long tmp;
tmp = gr[RD];
tmp <<= 16;
tmp >>= 16;
gr[RD] = tmp;
}
break;
case 0x8: /* declt */
--gr[RD];
NEW_C ((long)gr[RD] < 0);
break;
case 0x9: /* tstnbz */
{
word tmp = gr[RD];
NEW_C ((tmp & 0xFF000000) != 0 &&
(tmp & 0x00FF0000) != 0 && (tmp & 0x0000FF00) != 0 &&
(tmp & 0x000000FF) != 0);
}
break;
case 0xA: /* decgt */
--gr[RD];
NEW_C ((long)gr[RD] > 0);
break;
case 0xB: /* decne */
--gr[RD];
NEW_C ((long)gr[RD] != 0);
break;
case 0xC: /* clrt */
if (C_ON())
gr[RD] = 0;
break;
case 0xD: /* clrf */
if (C_OFF())
gr[RD] = 0;
break;
case 0xE: /* abs */
if (gr[RD] & 0x80000000)
gr[RD] = ~gr[RD] + 1;
break;
case 0xF: /* not */
gr[RD] = ~gr[RD];
break;
}
break;
case 0x02: /* movt */
if (C_ON())
gr[RD] = gr[RS];
break;
case 0x03: /* mult */
/* consume 2 bits per cycle from rs, until rs is 0 */
{
unsigned int t = gr[RS];
int ticks;
for (ticks = 0; t != 0 ; t >>= 2)
ticks++;
bonus_cycles += ticks;
}
bonus_cycles += 2; /* min. is 3, so add 2, plus ticks above */
if (tracing)
fprintf (stderr, " mult %lx by %lx to give %lx",
gr[RD], gr[RS], gr[RD] * gr[RS]);
gr[RD] = gr[RD] * gr[RS];
break;
case 0x04: /* loopt */
if (C_ON())
{
pc += (IMM4 << 1) - 32;
bonus_cycles ++;
needfetch = 1;
}
--gr[RS]; /* not RD! */
NEW_C (((long)gr[RS]) > 0);
break;
case 0x05: /* subu */
gr[RD] -= gr[RS];
break;
case 0x06: /* addc */
{
unsigned long tmp, a, b;
a = gr[RD];
b = gr[RS];
gr[RD] = a + b + C_VALUE ();
tmp = iu_carry (a, b, C_VALUE ());
NEW_C (tmp);
}
break;
case 0x07: /* subc */
{
unsigned long tmp, a, b;
a = gr[RD];
b = gr[RS];
gr[RD] = a - b + C_VALUE () - 1;
tmp = iu_carry (a,~b, C_VALUE ());
NEW_C (tmp);
}
break;
case 0x08: /* illegal */
case 0x09: /* illegal*/
ILLEGAL ();
break;
case 0x0A: /* movf */
if (C_OFF())
gr[RD] = gr[RS];
break;
case 0x0B: /* lsr */
{
unsigned long dst, src;
dst = gr[RD];
src = gr[RS];
/* We must not rely solely upon the native shift operations, since they
may not match the M*Core's behaviour on boundary conditions. */
dst = src > 31 ? 0 : dst >> src;
gr[RD] = dst;
}
break;
case 0x0C: /* cmphs */
NEW_C ((unsigned long )gr[RD] >=
(unsigned long)gr[RS]);
break;
case 0x0D: /* cmplt */
NEW_C ((long)gr[RD] < (long)gr[RS]);
break;
case 0x0E: /* tst */
NEW_C ((gr[RD] & gr[RS]) != 0);
break;
case 0x0F: /* cmpne */
NEW_C (gr[RD] != gr[RS]);
break;
case 0x10: case 0x11: /* mfcr */
{
unsigned r;
r = IMM5;
if (r <= LAST_VALID_CREG)
gr[RD] = cr[r];
else
ILLEGAL ();
}
break;
case 0x12: /* mov */
gr[RD] = gr[RS];
if (tracing)
fprintf (stderr, "MOV %lx into reg %d", gr[RD], RD);
break;
case 0x13: /* bgenr */
if (gr[RS] & 0x20)
gr[RD] = 0;
else
gr[RD] = 1 << (gr[RS] & 0x1F);
break;
case 0x14: /* rsub */
gr[RD] = gr[RS] - gr[RD];
break;
case 0x15: /* ixw */
gr[RD] += gr[RS]<<2;
break;
case 0x16: /* and */
gr[RD] &= gr[RS];
break;
case 0x17: /* xor */
gr[RD] ^= gr[RS];
break;
case 0x18: case 0x19: /* mtcr */
{
unsigned r;
r = IMM5;
if (r <= LAST_VALID_CREG)
cr[r] = gr[RD];
else
ILLEGAL ();
/* we might have changed register sets... */
set_active_regs (cpu);
}
break;
case 0x1A: /* asr */
/* We must not rely solely upon the native shift operations, since they
may not match the M*Core's behaviour on boundary conditions. */
if (gr[RS] > 30)
gr[RD] = ((long) gr[RD]) < 0 ? -1 : 0;
else
gr[RD] = (long) gr[RD] >> gr[RS];
break;
case 0x1B: /* lsl */
/* We must not rely solely upon the native shift operations, since they
may not match the M*Core's behaviour on boundary conditions. */
gr[RD] = gr[RS] > 31 ? 0 : gr[RD] << gr[RS];
break;
case 0x1C: /* addu */
gr[RD] += gr[RS];
break;
case 0x1D: /* ixh */
gr[RD] += gr[RS] << 1;
break;
case 0x1E: /* or */
gr[RD] |= gr[RS];
break;
case 0x1F: /* andn */
gr[RD] &= ~gr[RS];
break;
case 0x20: case 0x21: /* addi */
gr[RD] =
gr[RD] + (IMM5 + 1);
break;
case 0x22: case 0x23: /* cmplti */
{
int tmp = (IMM5 + 1);
if (gr[RD] < tmp)
{
SET_C();
}
else
{
CLR_C();
}
}
break;
case 0x24: case 0x25: /* subi */
gr[RD] =
gr[RD] - (IMM5 + 1);
break;
case 0x26: case 0x27: /* illegal */
ILLEGAL ();
break;
case 0x28: case 0x29: /* rsubi */
gr[RD] =
IMM5 - gr[RD];
break;
case 0x2A: case 0x2B: /* cmpnei */
if (gr[RD] != IMM5)
{
SET_C();
}
else
{
CLR_C();
}
break;
case 0x2C: case 0x2D: /* bmaski, divu */
{
unsigned imm = IMM5;
if (imm == 1)
{
int exe;
int rxnlz, r1nlz;
unsigned int rx, r1;
rx = gr[RD];
r1 = gr[1];
exe = 0;
/* unsigned divide */
gr[RD] = (word) ((unsigned int) gr[RD] / (unsigned int)gr[1] );
/* compute bonus_cycles for divu */
for (r1nlz = 0; ((r1 & 0x80000000) == 0) && (r1nlz < 32); r1nlz ++)
r1 = r1 << 1;
for (rxnlz = 0; ((rx & 0x80000000) == 0) && (rxnlz < 32); rxnlz ++)
rx = rx << 1;
if (r1nlz < rxnlz)
exe += 4;
else
exe += 5 + r1nlz - rxnlz;
if (exe >= (2 * memcycles - 1))
{
bonus_cycles += exe - (2 * memcycles) + 1;
}
}
else if (imm == 0 || imm >= 8)
{
/* bmaski */
if (imm == 0)
gr[RD] = -1;
else
gr[RD] = (1 << imm) - 1;
}
else
{
/* illegal */
ILLEGAL ();
}
}
break;
case 0x2E: case 0x2F: /* andi */
gr[RD] = gr[RD] & IMM5;
break;
case 0x30: case 0x31: /* bclri */
gr[RD] = gr[RD] & ~(1< 0)) || ((rx >= 0) && (r1 < 0)))
sc = 1;
else
sc = 0;
rx = abs (rx);
r1 = abs (r1);
/* signed divide, general registers are of type int, so / op is OK */
gr[RD] = gr[RD] / gr[1];
for (r1nlz = 0; ((r1 & 0x80000000) == 0) && (r1nlz < 32) ; r1nlz ++ )
r1 = r1 << 1;
for (rxnlz = 0; ((rx & 0x80000000) == 0) && (rxnlz < 32) ; rxnlz ++ )
rx = rx << 1;
if (r1nlz < rxnlz)
exe += 5;
else
exe += 6 + r1nlz - rxnlz + sc;
if (exe >= (2 * memcycles - 1))
{
bonus_cycles += exe - (2 * memcycles) + 1;
}
}
else if (imm >= 7)
{
/* bgeni */
gr[RD] = (1 << IMM5);
}
else
{
/* illegal */
ILLEGAL ();
}
break;
}
case 0x34: case 0x35: /* bseti */
gr[RD] = gr[RD] | (1 << IMM5);
break;
case 0x36: case 0x37: /* btsti */
NEW_C (gr[RD] >> IMM5);
break;
case 0x38: case 0x39: /* xsr, rotli */
{
unsigned imm = IMM5;
unsigned long tmp = gr[RD];
if (imm == 0)
{
word cbit;
cbit = C_VALUE();
NEW_C (tmp);
gr[RD] = (cbit << 31) | (tmp >> 1);
}
else
gr[RD] = (tmp << imm) | (tmp >> (32 - imm));
}
break;
case 0x3A: case 0x3B: /* asrc, asri */
{
unsigned imm = IMM5;
long tmp = gr[RD];
if (imm == 0)
{
NEW_C (tmp);
gr[RD] = tmp >> 1;
}
else
gr[RD] = tmp >> imm;
}
break;
case 0x3C: case 0x3D: /* lslc, lsli */
{
unsigned imm = IMM5;
unsigned long tmp = gr[RD];
if (imm == 0)
{
NEW_C (tmp >> 31);
gr[RD] = tmp << 1;
}
else
gr[RD] = tmp << imm;
}
break;
case 0x3E: case 0x3F: /* lsrc, lsri */
{
unsigned imm = IMM5;
unsigned long tmp = gr[RD];
if (imm == 0)
{
NEW_C (tmp);
gr[RD] = tmp >> 1;
}
else
gr[RD] = tmp >> imm;
}
break;
case 0x40: case 0x41: case 0x42: case 0x43:
case 0x44: case 0x45: case 0x46: case 0x47:
case 0x48: case 0x49: case 0x4A: case 0x4B:
case 0x4C: case 0x4D: case 0x4E: case 0x4F:
ILLEGAL ();
break;
case 0x50:
util (sd, cpu, inst & 0xFF);
break;
case 0x51: case 0x52: case 0x53:
case 0x54: case 0x55: case 0x56: case 0x57:
case 0x58: case 0x59: case 0x5A: case 0x5B:
case 0x5C: case 0x5D: case 0x5E: case 0x5F:
ILLEGAL ();
break;
case 0x60: case 0x61: case 0x62: case 0x63: /* movi */
case 0x64: case 0x65: case 0x66: case 0x67:
gr[RD] = (inst >> 4) & 0x7F;
break;
case 0x68: case 0x69: case 0x6A: case 0x6B:
case 0x6C: case 0x6D: case 0x6E: case 0x6F: /* illegal */
ILLEGAL ();
break;
case 0x71: case 0x72: case 0x73:
case 0x74: case 0x75: case 0x76: case 0x77:
case 0x78: case 0x79: case 0x7A: case 0x7B:
case 0x7C: case 0x7D: case 0x7E: /* lrw */
gr[RX] = rlat ((pc + ((inst & 0xFF) << 2)) & 0xFFFFFFFC);
if (tracing)
fprintf (stderr, "LRW of 0x%x from 0x%lx to reg %d",
rlat ((pc + ((inst & 0xFF) << 2)) & 0xFFFFFFFC),
(pc + ((inst & 0xFF) << 2)) & 0xFFFFFFFC, RX);
memops++;
break;
case 0x7F: /* jsri */
gr[15] = pc;
if (tracing)
fprintf (stderr,
"func call: r2 = %lx r3 = %lx r4 = %lx r5 = %lx r6 = %lx r7 = %lx\n",
gr[2], gr[3], gr[4], gr[5], gr[6], gr[7]);
case 0x70: /* jmpi */
pc = rlat ((pc + ((inst & 0xFF) << 2)) & 0xFFFFFFFC);
memops++;
bonus_cycles++;
needfetch = 1;
break;
case 0x80: case 0x81: case 0x82: case 0x83:
case 0x84: case 0x85: case 0x86: case 0x87:
case 0x88: case 0x89: case 0x8A: case 0x8B:
case 0x8C: case 0x8D: case 0x8E: case 0x8F: /* ld */
gr[RX] = rlat (gr[RD] + ((inst >> 2) & 0x003C));
if (tracing)
fprintf (stderr, "load reg %d from 0x%lx with 0x%lx",
RX,
gr[RD] + ((inst >> 2) & 0x003C), gr[RX]);
memops++;
break;
case 0x90: case 0x91: case 0x92: case 0x93:
case 0x94: case 0x95: case 0x96: case 0x97:
case 0x98: case 0x99: case 0x9A: case 0x9B:
case 0x9C: case 0x9D: case 0x9E: case 0x9F: /* st */
wlat (gr[RD] + ((inst >> 2) & 0x003C), gr[RX]);
if (tracing)
fprintf (stderr, "store reg %d (containing 0x%lx) to 0x%lx",
RX, gr[RX],
gr[RD] + ((inst >> 2) & 0x003C));
memops++;
break;
case 0xA0: case 0xA1: case 0xA2: case 0xA3:
case 0xA4: case 0xA5: case 0xA6: case 0xA7:
case 0xA8: case 0xA9: case 0xAA: case 0xAB:
case 0xAC: case 0xAD: case 0xAE: case 0xAF: /* ld.b */
gr[RX] = rbat (gr[RD] + RS);
memops++;
break;
case 0xB0: case 0xB1: case 0xB2: case 0xB3:
case 0xB4: case 0xB5: case 0xB6: case 0xB7:
case 0xB8: case 0xB9: case 0xBA: case 0xBB:
case 0xBC: case 0xBD: case 0xBE: case 0xBF: /* st.b */
wbat (gr[RD] + RS, gr[RX]);
memops++;
break;
case 0xC0: case 0xC1: case 0xC2: case 0xC3:
case 0xC4: case 0xC5: case 0xC6: case 0xC7:
case 0xC8: case 0xC9: case 0xCA: case 0xCB:
case 0xCC: case 0xCD: case 0xCE: case 0xCF: /* ld.h */
gr[RX] = rhat (gr[RD] + ((inst >> 3) & 0x001E));
memops++;
break;
case 0xD0: case 0xD1: case 0xD2: case 0xD3:
case 0xD4: case 0xD5: case 0xD6: case 0xD7:
case 0xD8: case 0xD9: case 0xDA: case 0xDB:
case 0xDC: case 0xDD: case 0xDE: case 0xDF: /* st.h */
what (gr[RD] + ((inst >> 3) & 0x001E), gr[RX]);
memops++;
break;
case 0xE8: case 0xE9: case 0xEA: case 0xEB:
case 0xEC: case 0xED: case 0xEE: case 0xEF: /* bf */
if (C_OFF())
{
int disp;
disp = inst & 0x03FF;
if (inst & 0x0400)
disp |= 0xFFFFFC00;
pc += disp<<1;
bonus_cycles++;
needfetch = 1;
}
break;
case 0xE0: case 0xE1: case 0xE2: case 0xE3:
case 0xE4: case 0xE5: case 0xE6: case 0xE7: /* bt */
if (C_ON())
{
int disp;
disp = inst & 0x03FF;
if (inst & 0x0400)
disp |= 0xFFFFFC00;
pc += disp<<1;
bonus_cycles++;
needfetch = 1;
}
break;
case 0xF8: case 0xF9: case 0xFA: case 0xFB:
case 0xFC: case 0xFD: case 0xFE: case 0xFF: /* bsr */
gr[15] = pc;
case 0xF0: case 0xF1: case 0xF2: case 0xF3:
case 0xF4: case 0xF5: case 0xF6: case 0xF7: /* br */
{
int disp;
disp = inst & 0x03FF;
if (inst & 0x0400)
disp |= 0xFFFFFC00;
pc += disp<<1;
bonus_cycles++;
needfetch = 1;
}
break;
}
if (tracing)
fprintf (stderr, "\n");
if (needfetch)
{
ibuf = rlat (pc & 0xFFFFFFFC);
needfetch = 0;
}
}
/* Hide away the things we've cached while executing. */
CPU_PC_SET (cpu, pc);
cpu->insts += insts; /* instructions done ... */
cpu->cycles += insts; /* and each takes a cycle */
cpu->cycles += bonus_cycles; /* and extra cycles for branches */
cpu->cycles += memops * memcycles; /* and memop cycle delays */
}
void
sim_engine_run (SIM_DESC sd,
int next_cpu_nr, /* ignore */
int nr_cpus, /* ignore */
int siggnal) /* ignore */
{
sim_cpu *cpu;
SIM_ASSERT (STATE_MAGIC (sd) == SIM_MAGIC_NUMBER);
cpu = STATE_CPU (sd, 0);
while (1)
{
step_once (sd, cpu);
if (sim_events_tick (sd))
sim_events_process (sd);
}
}
static int
mcore_reg_store (SIM_CPU *cpu, int rn, unsigned char *memory, int length)
{
if (rn < NUM_MCORE_REGS && rn >= 0)
{
if (length == 4)
{
long ival;
/* misalignment safe */
ival = mcore_extract_unsigned_integer (memory, 4);
cpu->asints[rn] = ival;
}
return 4;
}
else
return 0;
}
static int
mcore_reg_fetch (SIM_CPU *cpu, int rn, unsigned char *memory, int length)
{
if (rn < NUM_MCORE_REGS && rn >= 0)
{
if (length == 4)
{
long ival = cpu->asints[rn];
/* misalignment-safe */
mcore_store_unsigned_integer (memory, 4, ival);
}
return 4;
}
else
return 0;
}
void
sim_info (SIM_DESC sd, int verbose)
{
SIM_CPU *cpu = STATE_CPU (sd, 0);
#ifdef WATCHFUNCTIONS
int w, wcyc;
#endif
double virttime = cpu->cycles / 36.0e6;
host_callback *callback = STATE_CALLBACK (sd);
callback->printf_filtered (callback, "\n\n# instructions executed %10d\n",
cpu->insts);
callback->printf_filtered (callback, "# cycles %10d\n",
cpu->cycles);
callback->printf_filtered (callback, "# pipeline stalls %10d\n",
cpu->stalls);
callback->printf_filtered (callback, "# virtual time taken %10.4f\n",
virttime);
#ifdef WATCHFUNCTIONS
callback->printf_filtered (callback, "\nNumber of watched functions: %d\n",
ENDWL);
wcyc = 0;
for (w = 1; w <= ENDWL; w++)
{
callback->printf_filtered (callback, "WL = %s %8x\n",WLstr[w],WL[w]);
callback->printf_filtered (callback, " calls = %d, cycles = %d\n",
WLcnts[w],WLcyc[w]);
if (WLcnts[w] != 0)
callback->printf_filtered (callback,
" maxcpc = %d, mincpc = %d, avecpc = %d\n",
WLmax[w],WLmin[w],WLcyc[w]/WLcnts[w]);
wcyc += WLcyc[w];
}
callback->printf_filtered (callback,
"Total cycles for watched functions: %d\n",wcyc);
#endif
}
static sim_cia
mcore_pc_get (sim_cpu *cpu)
{
return cpu->regs.pc;
}
static void
mcore_pc_set (sim_cpu *cpu, sim_cia pc)
{
cpu->regs.pc = pc;
}
static void
free_state (SIM_DESC sd)
{
if (STATE_MODULES (sd) != NULL)
sim_module_uninstall (sd);
sim_cpu_free_all (sd);
sim_state_free (sd);
}
SIM_DESC
sim_open (SIM_OPEN_KIND kind, host_callback *cb,
struct bfd *abfd, char * const *argv)
{
int i;
SIM_DESC sd = sim_state_alloc (kind, cb);
SIM_ASSERT (STATE_MAGIC (sd) == SIM_MAGIC_NUMBER);
/* The cpu data is kept in a separately allocated chunk of memory. */
if (sim_cpu_alloc_all (sd, 1, /*cgen_cpu_max_extra_bytes ()*/0) != SIM_RC_OK)
{
free_state (sd);
return 0;
}
if (sim_pre_argv_init (sd, argv[0]) != SIM_RC_OK)
{
free_state (sd);
return 0;
}
/* The parser will print an error message for us, so we silently return. */
if (sim_parse_args (sd, argv) != SIM_RC_OK)
{
free_state (sd);
return 0;
}
/* Check for/establish the a reference program image. */
if (sim_analyze_program (sd,
(STATE_PROG_ARGV (sd) != NULL
? *STATE_PROG_ARGV (sd)
: NULL), abfd) != SIM_RC_OK)
{
free_state (sd);
return 0;
}
/* Configure/verify the target byte order and other runtime
configuration options. */
if (sim_config (sd) != SIM_RC_OK)
{
sim_module_uninstall (sd);
return 0;
}
if (sim_post_argv_init (sd) != SIM_RC_OK)
{
/* Uninstall the modules to avoid memory leaks,
file descriptor leaks, etc. */
sim_module_uninstall (sd);
return 0;
}
/* CPU specific initialization. */
for (i = 0; i < MAX_NR_PROCESSORS; ++i)
{
SIM_CPU *cpu = STATE_CPU (sd, i);
CPU_REG_FETCH (cpu) = mcore_reg_fetch;
CPU_REG_STORE (cpu) = mcore_reg_store;
CPU_PC_FETCH (cpu) = mcore_pc_get;
CPU_PC_STORE (cpu) = mcore_pc_set;
set_initial_gprs (cpu); /* Reset the GPR registers. */
}
/* Default to a 8 Mbyte (== 2^23) memory space. */
sim_do_commandf (sd, "memory-size %#x", DEFAULT_MEMORY_SIZE);
return sd;
}
SIM_RC
sim_create_inferior (SIM_DESC sd, struct bfd *prog_bfd,
char * const *argv, char * const *env)
{
SIM_CPU *cpu = STATE_CPU (sd, 0);
char ** avp;
int nargs = 0;
int nenv = 0;
int s_length;
int l;
unsigned long strings;
unsigned long pointers;
unsigned long hi_stack;
/* Set the initial register set. */
set_initial_gprs (cpu);
hi_stack = DEFAULT_MEMORY_SIZE - 4;
CPU_PC_SET (cpu, bfd_get_start_address (prog_bfd));
/* Calculate the argument and environment strings. */
s_length = 0;
nargs = 0;
avp = argv;
while (avp && *avp)
{
l = strlen (*avp) + 1; /* include the null */
s_length += (l + 3) & ~3; /* make it a 4 byte boundary */
nargs++; avp++;
}
nenv = 0;
avp = env;
while (avp && *avp)
{
l = strlen (*avp) + 1; /* include the null */
s_length += (l + 3) & ~ 3;/* make it a 4 byte boundary */
nenv++; avp++;
}
/* Claim some memory for the pointers and strings. */
pointers = hi_stack - sizeof(word) * (nenv+1+nargs+1);
pointers &= ~3; /* must be 4-byte aligned */
gr[0] = pointers;
strings = gr[0] - s_length;
strings &= ~3; /* want to make it 4-byte aligned */
gr[0] = strings;
/* dac fix, the stack address must be 8-byte aligned! */
gr[0] = gr[0] - gr[0] % 8;
/* Loop through the arguments and fill them in. */
gr[PARM1] = nargs;
if (nargs == 0)
{
/* No strings to fill in. */
gr[PARM2] = 0;
}
else
{
gr[PARM2] = pointers;
avp = argv;
while (avp && *avp)
{
/* Save where we're putting it. */
wlat (pointers, strings);
/* Copy the string. */
l = strlen (* avp) + 1;
sim_core_write_buffer (sd, cpu, write_map, *avp, strings, l);
/* Bump the pointers. */
avp++;
pointers += 4;
strings += l+1;
}
/* A null to finish the list. */
wlat (pointers, 0);
pointers += 4;
}
/* Now do the environment pointers. */
if (nenv == 0)
{
/* No strings to fill in. */
gr[PARM3] = 0;
}
else
{
gr[PARM3] = pointers;
avp = env;
while (avp && *avp)
{
/* Save where we're putting it. */
wlat (pointers, strings);
/* Copy the string. */
l = strlen (* avp) + 1;
sim_core_write_buffer (sd, cpu, write_map, *avp, strings, l);
/* Bump the pointers. */
avp++;
pointers += 4;
strings += l+1;
}
/* A null to finish the list. */
wlat (pointers, 0);
pointers += 4;
}
return SIM_RC_OK;
}