| 1 |
19 |
jeremybenn |
/* execute.c -- OR1K architecture dependent simulation
|
| 2 |
|
|
|
| 3 |
|
|
Copyright (C) 1999 Damjan Lampret, lampret@opencores.org
|
| 4 |
|
|
Copyright (C) 2005 György `nog' Jeney, nog@sdf.lonestar.org
|
| 5 |
|
|
Copyright (C) 2008 Embecosm Limited
|
| 6 |
|
|
|
| 7 |
|
|
Contributor Jeremy Bennett <jeremy.bennett@embecosm.com>
|
| 8 |
|
|
|
| 9 |
|
|
This file is part of OpenRISC 1000 Architectural Simulator.
|
| 10 |
|
|
|
| 11 |
|
|
This program is free software; you can redistribute it and/or modify it
|
| 12 |
|
|
under the terms of the GNU General Public License as published by the Free
|
| 13 |
|
|
Software Foundation; either version 3 of the License, or (at your option)
|
| 14 |
|
|
any later version.
|
| 15 |
|
|
|
| 16 |
|
|
This program is distributed in the hope that it will be useful, but WITHOUT
|
| 17 |
|
|
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
| 18 |
|
|
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
|
| 19 |
|
|
more details.
|
| 20 |
|
|
|
| 21 |
|
|
You should have received a copy of the GNU General Public License along
|
| 22 |
|
|
with this program. If not, see <http://www.gnu.org/licenses/>. */
|
| 23 |
|
|
|
| 24 |
|
|
/* This program is commented throughout in a fashion suitable for processing
|
| 25 |
|
|
with Doxygen. */
|
| 26 |
|
|
|
| 27 |
|
|
|
| 28 |
|
|
/* Most of the OR1K simulation is done in here.
|
| 29 |
|
|
|
| 30 |
|
|
When SIMPLE_EXECUTION is defined below a file insnset.c is included!
|
| 31 |
|
|
*/
|
| 32 |
|
|
|
| 33 |
|
|
/* Autoconf and/or portability configuration */
|
| 34 |
|
|
#include "config.h"
|
| 35 |
|
|
#include "port.h"
|
| 36 |
|
|
|
| 37 |
|
|
/* System includes */
|
| 38 |
|
|
#include <stdlib.h>
|
| 39 |
|
|
|
| 40 |
|
|
/* Package includes */
|
| 41 |
|
|
#include "execute.h"
|
| 42 |
|
|
#include "toplevel-support.h"
|
| 43 |
|
|
#include "except.h"
|
| 44 |
|
|
#include "labels.h"
|
| 45 |
|
|
#include "gdbcomm.h"
|
| 46 |
|
|
#include "sched.h"
|
| 47 |
|
|
#include "stats.h"
|
| 48 |
|
|
#include "opcode/or32.h"
|
| 49 |
|
|
#include "dmmu.h"
|
| 50 |
|
|
#include "immu.h"
|
| 51 |
|
|
#include "sim-cmd.h"
|
| 52 |
|
|
#include "vapi.h"
|
| 53 |
|
|
#include "debug-unit.h"
|
| 54 |
|
|
#include "branch-predict.h"
|
| 55 |
|
|
#include "sprs.h"
|
| 56 |
|
|
#include "rsp-server.h"
|
| 57 |
226 |
julius |
#include <fenv.h> // Floating point environment for FPU instructions
|
| 58 |
|
|
#include "execute-fp.h"
|
| 59 |
19 |
jeremybenn |
|
| 60 |
|
|
/* Includes and macros for simple execution */
|
| 61 |
|
|
#if SIMPLE_EXECUTION
|
| 62 |
|
|
|
| 63 |
|
|
#define SET_PARAM0(val) set_operand(0, val, current->insn_index, current->insn)
|
| 64 |
|
|
|
| 65 |
|
|
#define PARAM0 eval_operand(0, current->insn_index, current->insn)
|
| 66 |
|
|
#define PARAM1 eval_operand(1, current->insn_index, current->insn)
|
| 67 |
|
|
#define PARAM2 eval_operand(2, current->insn_index, current->insn)
|
| 68 |
|
|
|
| 69 |
|
|
#define INSTRUCTION(name) void name (struct iqueue_entry *current)
|
| 70 |
|
|
|
| 71 |
|
|
#endif /* SIMPLE_EXECUTION */
|
| 72 |
|
|
|
| 73 |
|
|
|
| 74 |
|
|
/*! Current cpu state. Globally available. */
|
| 75 |
|
|
struct cpu_state cpu_state;
|
| 76 |
|
|
|
| 77 |
|
|
/*! Temporary program counter. Globally available */
|
| 78 |
|
|
oraddr_t pcnext;
|
| 79 |
|
|
|
| 80 |
|
|
/*! Num cycles waiting for stores to complete. Globally available */
|
| 81 |
|
|
int sbuf_wait_cyc = 0;
|
| 82 |
|
|
|
| 83 |
|
|
/*! Number of total store cycles. Globally available */
|
| 84 |
|
|
int sbuf_total_cyc = 0;
|
| 85 |
|
|
|
| 86 |
|
|
/*! Whether we are doing statistical analysis. Globally available */
|
| 87 |
|
|
int do_stats = 0;
|
| 88 |
|
|
|
| 89 |
|
|
/*! History of execution. Globally available */
|
| 90 |
|
|
struct hist_exec *hist_exec_tail = NULL;
|
| 91 |
|
|
|
| 92 |
|
|
/* Benchmark multi issue execution. This file only */
|
| 93 |
|
|
static int multissue[20];
|
| 94 |
|
|
static int issued_per_cycle = 4;
|
| 95 |
|
|
|
| 96 |
|
|
/* Store buffer analysis - stores are accumulated and commited when IO is
|
| 97 |
|
|
idle. This file only */
|
| 98 |
|
|
static int sbuf_head = 0;
|
| 99 |
|
|
static int sbuf_tail = 0;
|
| 100 |
|
|
static int sbuf_count = 0;
|
| 101 |
|
|
#if !(DYNAMIC_EXECUTION)
|
| 102 |
|
|
static int sbuf_buf[MAX_SBUF_LEN] = { 0 };
|
| 103 |
|
|
#endif
|
| 104 |
|
|
|
| 105 |
|
|
static int sbuf_prev_cycles = 0;
|
| 106 |
|
|
|
| 107 |
|
|
/* Variables used throughout this file to share information */
|
| 108 |
|
|
static int breakpoint;
|
| 109 |
|
|
static int next_delay_insn;
|
| 110 |
|
|
|
| 111 |
226 |
julius |
/* Functions to configure the host machine's FPU before doing simulated OR1k
|
| 112 |
|
|
FP ops, handle flags and restore things when it's done. One for before and
|
| 113 |
|
|
one for afterwards.
|
| 114 |
|
|
*/
|
| 115 |
|
|
void fp_set_or1k_rm(void);
|
| 116 |
|
|
void fp_set_flags_restore_host_rm(void);
|
| 117 |
|
|
|
| 118 |
|
|
/* A place to store the host's FPU rounding mode while OR1k's is used */
|
| 119 |
|
|
int host_fp_rm;
|
| 120 |
|
|
|
| 121 |
19 |
jeremybenn |
/* Forward declaration of static functions */
|
| 122 |
|
|
#if !(DYNAMIC_EXECUTION)
|
| 123 |
|
|
static void decode_execute (struct iqueue_entry *current);
|
| 124 |
|
|
#endif
|
| 125 |
|
|
|
| 126 |
|
|
/*---------------------------------------------------------------------------*/
|
| 127 |
|
|
/*!Get an actual value of a specific register
|
| 128 |
|
|
|
| 129 |
|
|
Implementation specific. Abort if we are given a duff register. Only used
|
| 130 |
82 |
jeremybenn |
externally to support simprintf(), which is now obsolete.
|
| 131 |
19 |
jeremybenn |
|
| 132 |
|
|
@param[in] regno The register of interest
|
| 133 |
|
|
|
| 134 |
|
|
@return The value of the register */
|
| 135 |
|
|
/*---------------------------------------------------------------------------*/
|
| 136 |
|
|
uorreg_t
|
| 137 |
|
|
evalsim_reg (unsigned int regno)
|
| 138 |
|
|
{
|
| 139 |
|
|
if (regno < MAX_GPRS)
|
| 140 |
|
|
{
|
| 141 |
|
|
#if RAW_RANGE_STATS
|
| 142 |
|
|
int delta = (runtime.sim.cycles - raw_stats.reg[regno]);
|
| 143 |
|
|
|
| 144 |
|
|
if ((unsigned long) delta < (unsigned long) RAW_RANGE)
|
| 145 |
|
|
{
|
| 146 |
|
|
raw_stats.range[delta]++;
|
| 147 |
|
|
}
|
| 148 |
|
|
#endif /* RAW_RANGE */
|
| 149 |
|
|
|
| 150 |
|
|
return cpu_state.reg[regno];
|
| 151 |
|
|
}
|
| 152 |
|
|
else
|
| 153 |
|
|
{
|
| 154 |
|
|
PRINTF ("\nABORT: read out of registers\n");
|
| 155 |
|
|
sim_done ();
|
| 156 |
|
|
return 0;
|
| 157 |
|
|
}
|
| 158 |
|
|
} /* evalsim_reg() */
|
| 159 |
|
|
|
| 160 |
|
|
|
| 161 |
|
|
/*---------------------------------------------------------------------------*/
|
| 162 |
|
|
/*!Set a specific register with value
|
| 163 |
|
|
|
| 164 |
|
|
Implementation specific. Abort if we are given a duff register.
|
| 165 |
|
|
|
| 166 |
|
|
@param[in] regno The register of interest
|
| 167 |
|
|
@param[in] value The value to be set */
|
| 168 |
|
|
/*---------------------------------------------------------------------------*/
|
| 169 |
|
|
void
|
| 170 |
|
|
setsim_reg (unsigned int regno,
|
| 171 |
|
|
uorreg_t value)
|
| 172 |
|
|
{
|
| 173 |
|
|
if (regno == 0) /* gpr0 is always zero */
|
| 174 |
|
|
{
|
| 175 |
|
|
value = 0;
|
| 176 |
|
|
}
|
| 177 |
|
|
|
| 178 |
|
|
if (regno < MAX_GPRS)
|
| 179 |
|
|
{
|
| 180 |
|
|
cpu_state.reg[regno] = value;
|
| 181 |
|
|
}
|
| 182 |
|
|
else
|
| 183 |
|
|
{
|
| 184 |
|
|
PRINTF ("\nABORT: write out of registers\n");
|
| 185 |
|
|
sim_done ();
|
| 186 |
|
|
}
|
| 187 |
|
|
|
| 188 |
|
|
#if RAW_RANGE_STATS
|
| 189 |
|
|
raw_stats.reg[regno] = runtime.sim.cycles;
|
| 190 |
|
|
#endif /* RAW_RANGE */
|
| 191 |
|
|
|
| 192 |
|
|
} /* setsim_reg() */
|
| 193 |
|
|
|
| 194 |
|
|
|
| 195 |
|
|
/*---------------------------------------------------------------------------*/
|
| 196 |
|
|
/*!Evaluates source operand operand
|
| 197 |
|
|
|
| 198 |
|
|
Implementation specific. Declared global, although this is only actually
|
| 199 |
|
|
required for DYNAMIC_EXECUTION,
|
| 200 |
|
|
|
| 201 |
|
|
@param[in] insn The instruction
|
| 202 |
|
|
@param[in] opd The operand
|
| 203 |
|
|
|
| 204 |
|
|
@return The value of the source operand */
|
| 205 |
|
|
/*---------------------------------------------------------------------------*/
|
| 206 |
|
|
uorreg_t
|
| 207 |
|
|
eval_operand_val (uint32_t insn,
|
| 208 |
|
|
struct insn_op_struct *opd)
|
| 209 |
|
|
{
|
| 210 |
|
|
unsigned long operand = 0;
|
| 211 |
|
|
unsigned long sbit;
|
| 212 |
|
|
unsigned int nbits = 0;
|
| 213 |
|
|
|
| 214 |
|
|
while (1)
|
| 215 |
|
|
{
|
| 216 |
|
|
operand |=
|
| 217 |
|
|
((insn >> (opd->type & OPTYPE_SHR)) & ((1 << opd->data) - 1)) <<
|
| 218 |
|
|
nbits;
|
| 219 |
|
|
nbits += opd->data;
|
| 220 |
|
|
|
| 221 |
|
|
if (opd->type & OPTYPE_OP)
|
| 222 |
|
|
{
|
| 223 |
|
|
break;
|
| 224 |
|
|
}
|
| 225 |
|
|
|
| 226 |
|
|
opd++;
|
| 227 |
|
|
}
|
| 228 |
|
|
|
| 229 |
|
|
if (opd->type & OPTYPE_SIG)
|
| 230 |
|
|
{
|
| 231 |
|
|
sbit = (opd->type & OPTYPE_SBIT) >> OPTYPE_SBIT_SHR;
|
| 232 |
|
|
|
| 233 |
|
|
if (operand & (1 << sbit))
|
| 234 |
|
|
{
|
| 235 |
|
|
operand |= ~REG_C (0) << sbit;
|
| 236 |
|
|
}
|
| 237 |
|
|
}
|
| 238 |
|
|
|
| 239 |
|
|
return operand;
|
| 240 |
|
|
|
| 241 |
|
|
} /* eval_operand_val() */
|
| 242 |
|
|
|
| 243 |
|
|
|
| 244 |
|
|
/*---------------------------------------------------------------------------*/
|
| 245 |
|
|
/*!Does source operand depend on computation of dest operand?
|
| 246 |
|
|
|
| 247 |
|
|
Cycle t Cycle t+1
|
| 248 |
|
|
dst: irrelevant src: immediate always 0
|
| 249 |
|
|
dst: reg1 direct src: reg2 direct 0 if reg1 != reg2
|
| 250 |
|
|
dst: reg1 disp src: reg2 direct always 0
|
| 251 |
|
|
dst: reg1 direct src: reg2 disp 0 if reg1 != reg2
|
| 252 |
|
|
dst: reg1 disp src: reg2 disp always 1 (store must
|
| 253 |
|
|
finish before load)
|
| 254 |
|
|
dst: flag src: flag always 1
|
| 255 |
|
|
|
| 256 |
|
|
@param[in] prev Previous instruction
|
| 257 |
|
|
@param[in] next Next instruction
|
| 258 |
|
|
|
| 259 |
|
|
@return Non-zero if yes. */
|
| 260 |
|
|
/*---------------------------------------------------------------------------*/
|
| 261 |
|
|
static int
|
| 262 |
|
|
check_depend (struct iqueue_entry *prev,
|
| 263 |
|
|
struct iqueue_entry *next)
|
| 264 |
|
|
{
|
| 265 |
|
|
/* Find destination type. */
|
| 266 |
|
|
unsigned long type = 0;
|
| 267 |
|
|
int prev_dis;
|
| 268 |
|
|
int next_dis;
|
| 269 |
|
|
orreg_t prev_reg_val = 0;
|
| 270 |
|
|
struct insn_op_struct *opd;
|
| 271 |
|
|
|
| 272 |
|
|
if (or32_opcodes[prev->insn_index].flags & OR32_W_FLAG
|
| 273 |
|
|
&& or32_opcodes[next->insn_index].flags & OR32_R_FLAG)
|
| 274 |
|
|
{
|
| 275 |
|
|
return 1;
|
| 276 |
|
|
}
|
| 277 |
|
|
|
| 278 |
|
|
opd = op_start[prev->insn_index];
|
| 279 |
|
|
prev_dis = 0;
|
| 280 |
|
|
|
| 281 |
|
|
while (1)
|
| 282 |
|
|
{
|
| 283 |
|
|
if (opd->type & OPTYPE_DIS)
|
| 284 |
|
|
{
|
| 285 |
|
|
prev_dis = 1;
|
| 286 |
|
|
}
|
| 287 |
|
|
|
| 288 |
|
|
if (opd->type & OPTYPE_DST)
|
| 289 |
|
|
{
|
| 290 |
|
|
type = opd->type;
|
| 291 |
|
|
|
| 292 |
|
|
if (prev_dis)
|
| 293 |
|
|
{
|
| 294 |
|
|
type |= OPTYPE_DIS;
|
| 295 |
|
|
}
|
| 296 |
|
|
|
| 297 |
|
|
/* Destination is always a register */
|
| 298 |
|
|
prev_reg_val = eval_operand_val (prev->insn, opd);
|
| 299 |
|
|
break;
|
| 300 |
|
|
}
|
| 301 |
|
|
|
| 302 |
|
|
if (opd->type & OPTYPE_LAST)
|
| 303 |
|
|
{
|
| 304 |
|
|
return 0; /* Doesn't have a destination operand */
|
| 305 |
|
|
}
|
| 306 |
|
|
|
| 307 |
|
|
if (opd->type & OPTYPE_OP)
|
| 308 |
|
|
{
|
| 309 |
|
|
prev_dis = 0;
|
| 310 |
|
|
}
|
| 311 |
|
|
|
| 312 |
|
|
opd++;
|
| 313 |
|
|
}
|
| 314 |
|
|
|
| 315 |
|
|
/* We search all source operands - if we find confict => return 1 */
|
| 316 |
|
|
opd = op_start[next->insn_index];
|
| 317 |
|
|
next_dis = 0;
|
| 318 |
|
|
|
| 319 |
|
|
while (1)
|
| 320 |
|
|
{
|
| 321 |
|
|
if (opd->type & OPTYPE_DIS)
|
| 322 |
|
|
{
|
| 323 |
|
|
next_dis = 1;
|
| 324 |
|
|
}
|
| 325 |
|
|
|
| 326 |
|
|
/* This instruction sequence also depends on order of execution:
|
| 327 |
|
|
l.lw r1, k(r1)
|
| 328 |
|
|
l.sw k(r1), r4
|
| 329 |
|
|
Here r1 is a destination in l.sw */
|
| 330 |
|
|
|
| 331 |
|
|
/* FIXME: This situation is not handeld here when r1 == r2:
|
| 332 |
|
|
l.sw k(r1), r4
|
| 333 |
|
|
l.lw r3, k(r2) */
|
| 334 |
|
|
if (!(opd->type & OPTYPE_DST) || (next_dis && (opd->type & OPTYPE_DST)))
|
| 335 |
|
|
{
|
| 336 |
|
|
if (opd->type & OPTYPE_REG)
|
| 337 |
|
|
{
|
| 338 |
|
|
if (eval_operand_val (next->insn, opd) == prev_reg_val)
|
| 339 |
|
|
{
|
| 340 |
|
|
return 1;
|
| 341 |
|
|
}
|
| 342 |
|
|
}
|
| 343 |
|
|
}
|
| 344 |
|
|
|
| 345 |
|
|
if (opd->type & OPTYPE_LAST)
|
| 346 |
|
|
{
|
| 347 |
|
|
break;
|
| 348 |
|
|
}
|
| 349 |
|
|
|
| 350 |
|
|
opd++;
|
| 351 |
|
|
}
|
| 352 |
|
|
|
| 353 |
|
|
return 0;
|
| 354 |
|
|
|
| 355 |
|
|
} /* check_depend() */
|
| 356 |
|
|
|
| 357 |
|
|
|
| 358 |
|
|
/*---------------------------------------------------------------------------*/
|
| 359 |
|
|
/*!Should instruction NOT be executed?
|
| 360 |
|
|
|
| 361 |
|
|
Modified by CZ 26/05/01 for new mode execution.
|
| 362 |
|
|
|
| 363 |
|
|
@return Nonzero if instruction should NOT be executed */
|
| 364 |
|
|
/*---------------------------------------------------------------------------*/
|
| 365 |
|
|
static int
|
| 366 |
|
|
fetch ()
|
| 367 |
|
|
{
|
| 368 |
|
|
static int break_just_hit = 0;
|
| 369 |
|
|
|
| 370 |
|
|
if (NULL != breakpoints)
|
| 371 |
|
|
{
|
| 372 |
|
|
/* MM: Check for breakpoint. This has to be done in fetch cycle,
|
| 373 |
|
|
because of peripheria.
|
| 374 |
|
|
MM1709: if we cannot access the memory entry, we could not set the
|
| 375 |
|
|
breakpoint earlier, so just check the breakpoint list. */
|
| 376 |
|
|
if (has_breakpoint (peek_into_itlb (cpu_state.pc)) && !break_just_hit)
|
| 377 |
|
|
{
|
| 378 |
|
|
break_just_hit = 1;
|
| 379 |
|
|
return 1; /* Breakpoint set. */
|
| 380 |
|
|
}
|
| 381 |
|
|
break_just_hit = 0;
|
| 382 |
|
|
}
|
| 383 |
|
|
|
| 384 |
|
|
breakpoint = 0;
|
| 385 |
|
|
cpu_state.iqueue.insn_addr = cpu_state.pc;
|
| 386 |
|
|
cpu_state.iqueue.insn = eval_insn (cpu_state.pc, &breakpoint);
|
| 387 |
|
|
|
| 388 |
|
|
/* Fetch instruction. */
|
| 389 |
|
|
if (!except_pending)
|
| 390 |
|
|
{
|
| 391 |
|
|
runtime.cpu.instructions++;
|
| 392 |
|
|
}
|
| 393 |
|
|
|
| 394 |
|
|
/* update_pc will be called after execution */
|
| 395 |
|
|
return 0;
|
| 396 |
|
|
|
| 397 |
|
|
} /* fetch() */
|
| 398 |
|
|
|
| 399 |
|
|
|
| 400 |
|
|
/*---------------------------------------------------------------------------*/
|
| 401 |
|
|
/*!This code actually updates the PC value */
|
| 402 |
|
|
/*---------------------------------------------------------------------------*/
|
| 403 |
|
|
static void
|
| 404 |
|
|
update_pc ()
|
| 405 |
|
|
{
|
| 406 |
|
|
cpu_state.delay_insn = next_delay_insn;
|
| 407 |
|
|
cpu_state.sprs[SPR_PPC] = cpu_state.pc; /* Store value for later */
|
| 408 |
|
|
cpu_state.pc = pcnext;
|
| 409 |
|
|
pcnext = cpu_state.delay_insn ? cpu_state.pc_delay :
|
| 410 |
|
|
pcnext + 4;
|
| 411 |
|
|
} /* update_pc() */
|
| 412 |
|
|
|
| 413 |
|
|
|
| 414 |
|
|
/*---------------------------------------------------------------------------*/
|
| 415 |
|
|
/*!Perform analysis of the instruction being executed
|
| 416 |
|
|
|
| 417 |
|
|
This could be static for SIMPLE_EXECUTION, but made global for general use.
|
| 418 |
|
|
|
| 419 |
|
|
@param[in] current The instruction being executed */
|
| 420 |
|
|
/*---------------------------------------------------------------------------*/
|
| 421 |
|
|
void
|
| 422 |
|
|
analysis (struct iqueue_entry *current)
|
| 423 |
|
|
{
|
| 424 |
|
|
if (config.cpu.dependstats)
|
| 425 |
|
|
{
|
| 426 |
|
|
/* Dynamic, dependency stats. */
|
| 427 |
|
|
adddstats (cpu_state.icomplet.insn_index, current->insn_index, 1,
|
| 428 |
|
|
check_depend (&cpu_state.icomplet, current));
|
| 429 |
|
|
|
| 430 |
|
|
/* Dynamic, functional units stats. */
|
| 431 |
|
|
addfstats (or32_opcodes[cpu_state.icomplet.insn_index].func_unit,
|
| 432 |
|
|
or32_opcodes[current->insn_index].func_unit, 1,
|
| 433 |
|
|
check_depend (&cpu_state.icomplet, current));
|
| 434 |
|
|
|
| 435 |
|
|
/* Dynamic, single stats. */
|
| 436 |
|
|
addsstats (current->insn_index, 1);
|
| 437 |
|
|
}
|
| 438 |
|
|
|
| 439 |
|
|
if (config.cpu.superscalar)
|
| 440 |
|
|
{
|
| 441 |
|
|
if ((or32_opcodes[current->insn_index].func_unit == it_branch) ||
|
| 442 |
|
|
(or32_opcodes[current->insn_index].func_unit == it_jump))
|
| 443 |
|
|
runtime.sim.storecycles += 0;
|
| 444 |
|
|
|
| 445 |
|
|
if (or32_opcodes[current->insn_index].func_unit == it_store)
|
| 446 |
|
|
runtime.sim.storecycles += 1;
|
| 447 |
|
|
|
| 448 |
|
|
if (or32_opcodes[current->insn_index].func_unit == it_load)
|
| 449 |
|
|
runtime.sim.loadcycles += 1;
|
| 450 |
|
|
|
| 451 |
|
|
/* Pseudo multiple issue benchmark */
|
| 452 |
|
|
if ((multissue[or32_opcodes[current->insn_index].func_unit] < 1) ||
|
| 453 |
|
|
(check_depend (&cpu_state.icomplet, current))
|
| 454 |
|
|
|| (issued_per_cycle < 1))
|
| 455 |
|
|
{
|
| 456 |
|
|
int i;
|
| 457 |
|
|
for (i = 0; i < 20; i++)
|
| 458 |
|
|
multissue[i] = 2;
|
| 459 |
|
|
issued_per_cycle = 2;
|
| 460 |
|
|
runtime.cpu.supercycles++;
|
| 461 |
|
|
if (check_depend (&cpu_state.icomplet, current))
|
| 462 |
|
|
runtime.cpu.hazardwait++;
|
| 463 |
|
|
multissue[it_unknown] = 2;
|
| 464 |
|
|
multissue[it_shift] = 2;
|
| 465 |
|
|
multissue[it_compare] = 1;
|
| 466 |
|
|
multissue[it_branch] = 1;
|
| 467 |
|
|
multissue[it_jump] = 1;
|
| 468 |
|
|
multissue[it_extend] = 2;
|
| 469 |
|
|
multissue[it_nop] = 2;
|
| 470 |
|
|
multissue[it_move] = 2;
|
| 471 |
|
|
multissue[it_movimm] = 2;
|
| 472 |
|
|
multissue[it_arith] = 2;
|
| 473 |
|
|
multissue[it_store] = 2;
|
| 474 |
|
|
multissue[it_load] = 2;
|
| 475 |
|
|
}
|
| 476 |
|
|
multissue[or32_opcodes[current->insn_index].func_unit]--;
|
| 477 |
|
|
issued_per_cycle--;
|
| 478 |
|
|
}
|
| 479 |
|
|
|
| 480 |
|
|
if (config.cpu.dependstats)
|
| 481 |
|
|
/* Instruction waits in completition buffer until retired. */
|
| 482 |
|
|
memcpy (&cpu_state.icomplet, current, sizeof (struct iqueue_entry));
|
| 483 |
|
|
|
| 484 |
|
|
if (config.sim.history)
|
| 485 |
|
|
{
|
| 486 |
|
|
/* History of execution */
|
| 487 |
|
|
hist_exec_tail = hist_exec_tail->next;
|
| 488 |
|
|
hist_exec_tail->addr = cpu_state.icomplet.insn_addr;
|
| 489 |
|
|
}
|
| 490 |
|
|
|
| 491 |
|
|
if (config.sim.exe_log)
|
| 492 |
|
|
dump_exe_log ();
|
| 493 |
|
|
|
| 494 |
202 |
julius |
if (config.sim.exe_bin_insn_log)
|
| 495 |
|
|
dump_exe_bin_insn_log (current);
|
| 496 |
|
|
|
| 497 |
19 |
jeremybenn |
} /* analysis() */
|
| 498 |
|
|
|
| 499 |
|
|
|
| 500 |
|
|
#if !(DYNAMIC_EXECUTION)
|
| 501 |
|
|
|
| 502 |
|
|
/*---------------------------------------------------------------------------*/
|
| 503 |
|
|
/*!Store buffer analysis for store instructions
|
| 504 |
|
|
|
| 505 |
|
|
Stores are accumulated and commited when IO is idle
|
| 506 |
|
|
|
| 507 |
|
|
@param[in] cyc Number of cycles being analysed */
|
| 508 |
|
|
/*---------------------------------------------------------------------------*/
|
| 509 |
|
|
static void
|
| 510 |
|
|
sbuf_store (int cyc)
|
| 511 |
|
|
{
|
| 512 |
|
|
int delta = runtime.sim.cycles - sbuf_prev_cycles;
|
| 513 |
|
|
|
| 514 |
|
|
sbuf_total_cyc += cyc;
|
| 515 |
|
|
sbuf_prev_cycles = runtime.sim.cycles;
|
| 516 |
|
|
|
| 517 |
|
|
/* Take stores from buffer, that occured meanwhile */
|
| 518 |
|
|
while (sbuf_count && delta >= sbuf_buf[sbuf_tail])
|
| 519 |
|
|
{
|
| 520 |
|
|
delta -= sbuf_buf[sbuf_tail];
|
| 521 |
|
|
sbuf_tail = (sbuf_tail + 1) % MAX_SBUF_LEN;
|
| 522 |
|
|
sbuf_count--;
|
| 523 |
|
|
}
|
| 524 |
|
|
|
| 525 |
|
|
if (sbuf_count)
|
| 526 |
|
|
{
|
| 527 |
|
|
sbuf_buf[sbuf_tail] -= delta;
|
| 528 |
|
|
}
|
| 529 |
|
|
|
| 530 |
|
|
/* Store buffer is full, take one out */
|
| 531 |
|
|
if (sbuf_count >= config.cpu.sbuf_len)
|
| 532 |
|
|
{
|
| 533 |
|
|
sbuf_wait_cyc += sbuf_buf[sbuf_tail];
|
| 534 |
|
|
runtime.sim.mem_cycles += sbuf_buf[sbuf_tail];
|
| 535 |
|
|
sbuf_prev_cycles += sbuf_buf[sbuf_tail];
|
| 536 |
|
|
sbuf_tail = (sbuf_tail + 1) % MAX_SBUF_LEN;
|
| 537 |
|
|
sbuf_count--;
|
| 538 |
|
|
}
|
| 539 |
|
|
|
| 540 |
|
|
/* Put newest store in the buffer */
|
| 541 |
|
|
sbuf_buf[sbuf_head] = cyc;
|
| 542 |
|
|
sbuf_head = (sbuf_head + 1) % MAX_SBUF_LEN;
|
| 543 |
|
|
sbuf_count++;
|
| 544 |
|
|
|
| 545 |
|
|
} /* sbuf_store() */
|
| 546 |
|
|
|
| 547 |
|
|
|
| 548 |
|
|
/*---------------------------------------------------------------------------*/
|
| 549 |
|
|
/*!Store buffer analysis for load instructions
|
| 550 |
|
|
|
| 551 |
|
|
Previous stores should commit, before any load */
|
| 552 |
|
|
/*---------------------------------------------------------------------------*/
|
| 553 |
|
|
static void
|
| 554 |
|
|
sbuf_load ()
|
| 555 |
|
|
{
|
| 556 |
|
|
int delta = runtime.sim.cycles - sbuf_prev_cycles;
|
| 557 |
|
|
sbuf_prev_cycles = runtime.sim.cycles;
|
| 558 |
|
|
|
| 559 |
|
|
/* Take stores from buffer, that occured meanwhile */
|
| 560 |
|
|
while (sbuf_count && delta >= sbuf_buf[sbuf_tail])
|
| 561 |
|
|
{
|
| 562 |
|
|
delta -= sbuf_buf[sbuf_tail];
|
| 563 |
|
|
sbuf_tail = (sbuf_tail + 1) % MAX_SBUF_LEN;
|
| 564 |
|
|
sbuf_count--;
|
| 565 |
|
|
}
|
| 566 |
|
|
|
| 567 |
|
|
if (sbuf_count)
|
| 568 |
|
|
{
|
| 569 |
|
|
sbuf_buf[sbuf_tail] -= delta;
|
| 570 |
|
|
}
|
| 571 |
|
|
|
| 572 |
|
|
/* Wait for all stores to complete */
|
| 573 |
|
|
while (sbuf_count > 0)
|
| 574 |
|
|
{
|
| 575 |
|
|
sbuf_wait_cyc += sbuf_buf[sbuf_tail];
|
| 576 |
|
|
runtime.sim.mem_cycles += sbuf_buf[sbuf_tail];
|
| 577 |
|
|
sbuf_prev_cycles += sbuf_buf[sbuf_tail];
|
| 578 |
|
|
sbuf_tail = (sbuf_tail + 1) % MAX_SBUF_LEN;
|
| 579 |
|
|
sbuf_count--;
|
| 580 |
|
|
}
|
| 581 |
|
|
} /* sbuf_load() */
|
| 582 |
|
|
|
| 583 |
|
|
#endif /* !DYNAMIC_EXECUTION */
|
| 584 |
|
|
|
| 585 |
|
|
|
| 586 |
|
|
/*---------------------------------------------------------------------------*/
|
| 587 |
|
|
/*!Outputs dissasembled instruction */
|
| 588 |
|
|
/*---------------------------------------------------------------------------*/
|
| 589 |
|
|
void
|
| 590 |
|
|
dump_exe_log ()
|
| 591 |
|
|
{
|
| 592 |
|
|
oraddr_t insn_addr = cpu_state.iqueue.insn_addr;
|
| 593 |
|
|
unsigned int i;
|
| 594 |
|
|
unsigned int j;
|
| 595 |
|
|
uorreg_t operand;
|
| 596 |
|
|
|
| 597 |
|
|
if (insn_addr == 0xffffffff)
|
| 598 |
|
|
{
|
| 599 |
|
|
return;
|
| 600 |
|
|
}
|
| 601 |
|
|
|
| 602 |
|
|
if ((config.sim.exe_log_start <= runtime.cpu.instructions) &&
|
| 603 |
|
|
((config.sim.exe_log_end <= 0) ||
|
| 604 |
|
|
(runtime.cpu.instructions <= config.sim.exe_log_end)))
|
| 605 |
|
|
{
|
| 606 |
|
|
struct label_entry *entry;
|
| 607 |
|
|
|
| 608 |
|
|
if (config.sim.exe_log_marker &&
|
| 609 |
|
|
!(runtime.cpu.instructions % config.sim.exe_log_marker))
|
| 610 |
|
|
{
|
| 611 |
|
|
fprintf (runtime.sim.fexe_log,
|
| 612 |
|
|
"--------------------- %8lli instruction "
|
| 613 |
|
|
"---------------------\n",
|
| 614 |
|
|
runtime.cpu.instructions);
|
| 615 |
|
|
}
|
| 616 |
|
|
|
| 617 |
|
|
switch (config.sim.exe_log_type)
|
| 618 |
|
|
{
|
| 619 |
|
|
case EXE_LOG_HARDWARE:
|
| 620 |
|
|
fprintf (runtime.sim.fexe_log,
|
| 621 |
|
|
"\nEXECUTED(%11llu): %" PRIxADDR ": ",
|
| 622 |
|
|
runtime.cpu.instructions, insn_addr);
|
| 623 |
|
|
fprintf (runtime.sim.fexe_log, "%.2x%.2x",
|
| 624 |
|
|
eval_direct8 (insn_addr, 0, 0),
|
| 625 |
|
|
eval_direct8 (insn_addr + 1, 0, 0));
|
| 626 |
|
|
fprintf (runtime.sim.fexe_log, "%.2x%.2x",
|
| 627 |
|
|
eval_direct8 (insn_addr + 2, 0, 0),
|
| 628 |
|
|
eval_direct8 (insn_addr + 3, 0, 0));
|
| 629 |
|
|
|
| 630 |
|
|
for (i = 0; i < MAX_GPRS; i++)
|
| 631 |
|
|
{
|
| 632 |
|
|
if (i % 4 == 0)
|
| 633 |
|
|
{
|
| 634 |
|
|
fprintf (runtime.sim.fexe_log, "\n");
|
| 635 |
|
|
}
|
| 636 |
|
|
|
| 637 |
|
|
fprintf (runtime.sim.fexe_log, "GPR%2u: %" PRIxREG " ", i,
|
| 638 |
|
|
cpu_state.reg[i]);
|
| 639 |
|
|
}
|
| 640 |
|
|
|
| 641 |
|
|
fprintf (runtime.sim.fexe_log, "\n");
|
| 642 |
|
|
fprintf (runtime.sim.fexe_log, "SR : %.8" PRIx32 " ",
|
| 643 |
|
|
cpu_state.sprs[SPR_SR]);
|
| 644 |
|
|
fprintf (runtime.sim.fexe_log, "EPCR0: %" PRIxADDR " ",
|
| 645 |
|
|
cpu_state.sprs[SPR_EPCR_BASE]);
|
| 646 |
|
|
fprintf (runtime.sim.fexe_log, "EEAR0: %" PRIxADDR " ",
|
| 647 |
|
|
cpu_state.sprs[SPR_EEAR_BASE]);
|
| 648 |
|
|
fprintf (runtime.sim.fexe_log, "ESR0 : %.8" PRIx32 "\n",
|
| 649 |
|
|
cpu_state.sprs[SPR_ESR_BASE]);
|
| 650 |
|
|
break;
|
| 651 |
|
|
|
| 652 |
|
|
case EXE_LOG_SIMPLE:
|
| 653 |
|
|
case EXE_LOG_SOFTWARE:
|
| 654 |
|
|
disassemble_index (cpu_state.iqueue.insn,
|
| 655 |
|
|
cpu_state.iqueue.insn_index);
|
| 656 |
|
|
|
| 657 |
|
|
entry = get_label (insn_addr);
|
| 658 |
|
|
if (entry)
|
| 659 |
|
|
{
|
| 660 |
|
|
fprintf (runtime.sim.fexe_log, "%s:\n", entry->name);
|
| 661 |
|
|
}
|
| 662 |
|
|
|
| 663 |
|
|
if (config.sim.exe_log_type == EXE_LOG_SOFTWARE)
|
| 664 |
|
|
{
|
| 665 |
|
|
struct insn_op_struct *opd =
|
| 666 |
|
|
op_start[cpu_state.iqueue.insn_index];
|
| 667 |
|
|
|
| 668 |
|
|
j = 0;
|
| 669 |
|
|
while (1)
|
| 670 |
|
|
{
|
| 671 |
|
|
operand = eval_operand_val (cpu_state.iqueue.insn, opd);
|
| 672 |
|
|
while (!(opd->type & OPTYPE_OP))
|
| 673 |
|
|
{
|
| 674 |
|
|
opd++;
|
| 675 |
|
|
}
|
| 676 |
|
|
if (opd->type & OPTYPE_DIS)
|
| 677 |
|
|
{
|
| 678 |
|
|
fprintf (runtime.sim.fexe_log,
|
| 679 |
|
|
"EA =%" PRIxADDR " PA =%" PRIxADDR " ",
|
| 680 |
|
|
cpu_state.insn_ea,
|
| 681 |
|
|
peek_into_dtlb (cpu_state.insn_ea, 0, 0));
|
| 682 |
|
|
opd++; /* Skip of register operand */
|
| 683 |
|
|
j++;
|
| 684 |
|
|
}
|
| 685 |
|
|
else if ((opd->type & OPTYPE_REG) && operand)
|
| 686 |
|
|
{
|
| 687 |
|
|
fprintf (runtime.sim.fexe_log, "r%-2i=%" PRIxREG " ",
|
| 688 |
|
|
(int) operand, evalsim_reg (operand));
|
| 689 |
|
|
}
|
| 690 |
|
|
else
|
| 691 |
|
|
{
|
| 692 |
|
|
fprintf (runtime.sim.fexe_log, " ");
|
| 693 |
|
|
}
|
| 694 |
|
|
j++;
|
| 695 |
|
|
if (opd->type & OPTYPE_LAST)
|
| 696 |
|
|
{
|
| 697 |
|
|
break;
|
| 698 |
|
|
}
|
| 699 |
|
|
opd++;
|
| 700 |
|
|
}
|
| 701 |
|
|
if (or32_opcodes[cpu_state.iqueue.insn_index].flags & OR32_R_FLAG)
|
| 702 |
|
|
{
|
| 703 |
|
|
fprintf (runtime.sim.fexe_log, "SR =%" PRIxREG " ",
|
| 704 |
|
|
cpu_state.sprs[SPR_SR]);
|
| 705 |
|
|
j++;
|
| 706 |
|
|
}
|
| 707 |
|
|
while (j < 3)
|
| 708 |
|
|
{
|
| 709 |
|
|
fprintf (runtime.sim.fexe_log, " ");
|
| 710 |
|
|
j++;
|
| 711 |
|
|
}
|
| 712 |
|
|
}
|
| 713 |
|
|
fprintf (runtime.sim.fexe_log, "%" PRIxADDR " ", insn_addr);
|
| 714 |
|
|
fprintf (runtime.sim.fexe_log, "%s\n", disassembled);
|
| 715 |
|
|
}
|
| 716 |
|
|
}
|
| 717 |
|
|
} /* dump_exe_log() */
|
| 718 |
|
|
|
| 719 |
|
|
|
| 720 |
202 |
julius |
|
| 721 |
19 |
jeremybenn |
/*---------------------------------------------------------------------------*/
|
| 722 |
202 |
julius |
/*!Outputs binary copy of instruction to a file */
|
| 723 |
|
|
/*---------------------------------------------------------------------------*/
|
| 724 |
|
|
void
|
| 725 |
|
|
dump_exe_bin_insn_log (struct iqueue_entry *current)
|
| 726 |
|
|
{
|
| 727 |
|
|
// Do endian swap before spitting out (will be kept in LE on a LE machine)
|
| 728 |
|
|
// but more useful to see it in big endian format.
|
| 729 |
|
|
// Should probably host htonl().
|
| 730 |
|
|
uint32_t insn = (((current->insn & 0xff)<<24) |
|
| 731 |
|
|
((current->insn & 0xff00)<<8) |
|
| 732 |
|
|
((current->insn & 0xff0000)>>8) |
|
| 733 |
|
|
((current->insn & 0xff000000)>>24));
|
| 734 |
|
|
|
| 735 |
|
|
//for(i=0;i<4;i++) tmp_insn[i] = eval_direct8 (insn_addr+i, 0, 0);
|
| 736 |
|
|
|
| 737 |
|
|
// Dump it into binary log file
|
| 738 |
|
|
fwrite((void*)&insn, 4, 1, runtime.sim.fexe_bin_insn_log);
|
| 739 |
|
|
|
| 740 |
|
|
|
| 741 |
|
|
} /* dump_exe_bin_insn_log() */
|
| 742 |
|
|
|
| 743 |
|
|
|
| 744 |
|
|
/*---------------------------------------------------------------------------*/
|
| 745 |
19 |
jeremybenn |
/*!Dump registers
|
| 746 |
|
|
|
| 747 |
|
|
Supports the CLI 'r' and 't' commands */
|
| 748 |
|
|
/*---------------------------------------------------------------------------*/
|
| 749 |
|
|
void
|
| 750 |
|
|
dumpreg ()
|
| 751 |
|
|
{
|
| 752 |
|
|
int i;
|
| 753 |
|
|
oraddr_t physical_pc;
|
| 754 |
|
|
|
| 755 |
|
|
if ((physical_pc = peek_into_itlb (cpu_state.iqueue.insn_addr)))
|
| 756 |
|
|
{
|
| 757 |
|
|
disassemble_memory (physical_pc, physical_pc + 4, 0);
|
| 758 |
|
|
}
|
| 759 |
|
|
else
|
| 760 |
|
|
{
|
| 761 |
|
|
PRINTF ("INTERNAL SIMULATOR ERROR:\n");
|
| 762 |
|
|
PRINTF ("no translation for currently executed instruction\n");
|
| 763 |
|
|
}
|
| 764 |
|
|
|
| 765 |
|
|
// generate_time_pretty (temp, runtime.sim.cycles * config.sim.clkcycle_ps);
|
| 766 |
|
|
PRINTF (" (executed) [cycle %lld, #%lld]\n", runtime.sim.cycles,
|
| 767 |
|
|
runtime.cpu.instructions);
|
| 768 |
|
|
if (config.cpu.superscalar)
|
| 769 |
|
|
{
|
| 770 |
|
|
PRINTF ("Superscalar CYCLES: %u", runtime.cpu.supercycles);
|
| 771 |
|
|
}
|
| 772 |
|
|
if (config.cpu.hazards)
|
| 773 |
|
|
{
|
| 774 |
|
|
PRINTF (" HAZARDWAIT: %u\n", runtime.cpu.hazardwait);
|
| 775 |
|
|
}
|
| 776 |
|
|
else if (config.cpu.superscalar)
|
| 777 |
|
|
{
|
| 778 |
|
|
PRINTF ("\n");
|
| 779 |
|
|
}
|
| 780 |
|
|
|
| 781 |
|
|
if ((physical_pc = peek_into_itlb (cpu_state.pc)))
|
| 782 |
|
|
{
|
| 783 |
|
|
disassemble_memory (physical_pc, physical_pc + 4, 0);
|
| 784 |
|
|
}
|
| 785 |
|
|
else
|
| 786 |
|
|
{
|
| 787 |
|
|
PRINTF ("%" PRIxADDR ": : xxxxxxxx ITLB miss follows", cpu_state.pc);
|
| 788 |
|
|
}
|
| 789 |
|
|
|
| 790 |
|
|
PRINTF (" (next insn) %s", (cpu_state.delay_insn ? "(delay insn)" : ""));
|
| 791 |
|
|
|
| 792 |
|
|
for (i = 0; i < MAX_GPRS; i++)
|
| 793 |
|
|
{
|
| 794 |
|
|
if (i % 4 == 0)
|
| 795 |
|
|
{
|
| 796 |
|
|
PRINTF ("\n");
|
| 797 |
|
|
}
|
| 798 |
|
|
|
| 799 |
|
|
PRINTF ("GPR%.2u: %" PRIxREG " ", i, evalsim_reg (i));
|
| 800 |
|
|
}
|
| 801 |
|
|
|
| 802 |
|
|
PRINTF ("flag: %u\n", cpu_state.sprs[SPR_SR] & SPR_SR_F ? 1 : 0);
|
| 803 |
|
|
|
| 804 |
|
|
} /* dumpreg() */
|
| 805 |
|
|
|
| 806 |
|
|
|
| 807 |
|
|
/*---------------------------------------------------------------------------*/
|
| 808 |
|
|
/*!Wrapper around real decode_execute function
|
| 809 |
|
|
|
| 810 |
|
|
Some statistics here only
|
| 811 |
|
|
|
| 812 |
|
|
@param[in] current Instruction being executed */
|
| 813 |
|
|
/*---------------------------------------------------------------------------*/
|
| 814 |
|
|
static void
|
| 815 |
|
|
decode_execute_wrapper (struct iqueue_entry *current)
|
| 816 |
|
|
{
|
| 817 |
|
|
breakpoint = 0;
|
| 818 |
|
|
|
| 819 |
|
|
#ifndef HAVE_EXECUTION
|
| 820 |
|
|
#error HAVE_EXECUTION has to be defined in order to execute programs.
|
| 821 |
|
|
#endif
|
| 822 |
|
|
|
| 823 |
|
|
/* FIXME: Most of this file is not needed with DYNAMIC_EXECUTION */
|
| 824 |
|
|
#if !(DYNAMIC_EXECUTION)
|
| 825 |
|
|
decode_execute (current);
|
| 826 |
|
|
#endif
|
| 827 |
|
|
|
| 828 |
|
|
if (breakpoint)
|
| 829 |
|
|
{
|
| 830 |
|
|
except_handle (EXCEPT_TRAP, cpu_state.sprs[SPR_EEAR_BASE]);
|
| 831 |
|
|
}
|
| 832 |
|
|
} /* decode_execute_wrapper() */
|
| 833 |
|
|
|
| 834 |
|
|
/*---------------------------------------------------------------------------*/
|
| 835 |
|
|
/*!Reset the CPU */
|
| 836 |
|
|
/*---------------------------------------------------------------------------*/
|
| 837 |
|
|
void
|
| 838 |
|
|
cpu_reset ()
|
| 839 |
|
|
{
|
| 840 |
|
|
int i;
|
| 841 |
|
|
struct hist_exec *hist_exec_head = NULL;
|
| 842 |
|
|
struct hist_exec *hist_exec_new;
|
| 843 |
|
|
|
| 844 |
|
|
runtime.sim.cycles = 0;
|
| 845 |
|
|
runtime.sim.loadcycles = 0;
|
| 846 |
|
|
runtime.sim.storecycles = 0;
|
| 847 |
|
|
runtime.cpu.instructions = 0;
|
| 848 |
|
|
runtime.cpu.supercycles = 0;
|
| 849 |
|
|
runtime.cpu.hazardwait = 0;
|
| 850 |
|
|
|
| 851 |
|
|
for (i = 0; i < MAX_GPRS; i++)
|
| 852 |
|
|
{
|
| 853 |
|
|
setsim_reg (i, 0);
|
| 854 |
|
|
}
|
| 855 |
|
|
|
| 856 |
|
|
memset (&cpu_state.iqueue, 0, sizeof (cpu_state.iqueue));
|
| 857 |
|
|
memset (&cpu_state.icomplet, 0, sizeof (cpu_state.icomplet));
|
| 858 |
|
|
|
| 859 |
|
|
sbuf_head = 0;
|
| 860 |
|
|
sbuf_tail = 0;
|
| 861 |
|
|
sbuf_count = 0;
|
| 862 |
|
|
sbuf_prev_cycles = 0;
|
| 863 |
|
|
|
| 864 |
|
|
/* Initialise execution history circular buffer */
|
| 865 |
|
|
for (i = 0; i < HISTEXEC_LEN; i++)
|
| 866 |
|
|
{
|
| 867 |
|
|
hist_exec_new = malloc (sizeof (struct hist_exec));
|
| 868 |
|
|
|
| 869 |
|
|
if (!hist_exec_new)
|
| 870 |
|
|
{
|
| 871 |
|
|
fprintf (stderr, "Out-of-memory\n");
|
| 872 |
|
|
exit (1);
|
| 873 |
|
|
}
|
| 874 |
|
|
|
| 875 |
|
|
if (!hist_exec_head)
|
| 876 |
|
|
{
|
| 877 |
|
|
hist_exec_head = hist_exec_new;
|
| 878 |
|
|
}
|
| 879 |
|
|
else
|
| 880 |
|
|
{
|
| 881 |
|
|
hist_exec_tail->next = hist_exec_new;
|
| 882 |
|
|
}
|
| 883 |
|
|
|
| 884 |
|
|
hist_exec_new->prev = hist_exec_tail;
|
| 885 |
|
|
hist_exec_tail = hist_exec_new;
|
| 886 |
|
|
}
|
| 887 |
|
|
|
| 888 |
|
|
/* Make hist_exec_tail->next point to hist_exec_head */
|
| 889 |
|
|
hist_exec_tail->next = hist_exec_head;
|
| 890 |
|
|
hist_exec_head->prev = hist_exec_tail;
|
| 891 |
|
|
|
| 892 |
|
|
/* MM1409: All progs should start at reset vector entry! This sorted out by
|
| 893 |
|
|
setting the cpu_state.pc field below. Not clear this is very good code! */
|
| 894 |
|
|
|
| 895 |
|
|
/* Patches suggested by Shinji Wakatsuki, so that the vector address takes
|
| 896 |
|
|
notice of the Exception Prefix High bit of the Supervision register */
|
| 897 |
|
|
pcnext = (cpu_state.sprs[SPR_SR] & SPR_SR_EPH ? 0xf0000000 : 0x00000000);
|
| 898 |
|
|
|
| 899 |
|
|
if (config.sim.verbose)
|
| 900 |
|
|
{
|
| 901 |
|
|
PRINTF ("Starting at 0x%" PRIxADDR "\n", pcnext);
|
| 902 |
|
|
}
|
| 903 |
|
|
|
| 904 |
|
|
cpu_state.pc = pcnext;
|
| 905 |
|
|
pcnext += 4;
|
| 906 |
|
|
|
| 907 |
|
|
/* MM1409: All programs should set their stack pointer! */
|
| 908 |
|
|
#if !(DYNAMIC_EXECUTION)
|
| 909 |
|
|
except_handle (EXCEPT_RESET, 0);
|
| 910 |
|
|
update_pc ();
|
| 911 |
|
|
#endif
|
| 912 |
|
|
|
| 913 |
|
|
except_pending = 0;
|
| 914 |
|
|
cpu_state.pc = cpu_state.sprs[SPR_SR] & SPR_SR_EPH ?
|
| 915 |
|
|
0xf0000000 + EXCEPT_RESET : EXCEPT_RESET;
|
| 916 |
|
|
|
| 917 |
|
|
} /* cpu_reset() */
|
| 918 |
|
|
|
| 919 |
|
|
|
| 920 |
|
|
/*---------------------------------------------------------------------------*/
|
| 921 |
|
|
/*!Simulates one CPU clock cycle
|
| 922 |
|
|
|
| 923 |
|
|
@return non-zero if a breakpoint is hit, zero otherwise. */
|
| 924 |
|
|
/*---------------------------------------------------------------------------*/
|
| 925 |
|
|
int
|
| 926 |
|
|
cpu_clock ()
|
| 927 |
|
|
{
|
| 928 |
|
|
except_pending = 0;
|
| 929 |
|
|
next_delay_insn = 0;
|
| 930 |
|
|
|
| 931 |
|
|
if (fetch ())
|
| 932 |
|
|
{
|
| 933 |
|
|
PRINTF ("Breakpoint hit.\n");
|
| 934 |
|
|
return 1;
|
| 935 |
|
|
}
|
| 936 |
|
|
|
| 937 |
|
|
if (except_pending)
|
| 938 |
|
|
{
|
| 939 |
|
|
update_pc ();
|
| 940 |
|
|
except_pending = 0;
|
| 941 |
|
|
return 0;
|
| 942 |
|
|
}
|
| 943 |
|
|
|
| 944 |
|
|
if (breakpoint)
|
| 945 |
|
|
{
|
| 946 |
|
|
except_handle (EXCEPT_TRAP, cpu_state.sprs[SPR_EEAR_BASE]);
|
| 947 |
|
|
update_pc ();
|
| 948 |
|
|
except_pending = 0;
|
| 949 |
|
|
return 0;
|
| 950 |
|
|
}
|
| 951 |
|
|
|
| 952 |
|
|
decode_execute_wrapper (&cpu_state.iqueue);
|
| 953 |
|
|
update_pc ();
|
| 954 |
|
|
return 0;
|
| 955 |
|
|
|
| 956 |
|
|
} /* cpu_clock() */
|
| 957 |
|
|
|
| 958 |
|
|
|
| 959 |
|
|
/*---------------------------------------------------------------------------*/
|
| 960 |
|
|
/*!If decoding cannot be found, call this function */
|
| 961 |
|
|
/*---------------------------------------------------------------------------*/
|
| 962 |
|
|
#if SIMPLE_EXECUTION
|
| 963 |
|
|
void
|
| 964 |
|
|
l_invalid (struct iqueue_entry *current)
|
| 965 |
|
|
{
|
| 966 |
|
|
#else
|
| 967 |
|
|
void
|
| 968 |
|
|
l_invalid ()
|
| 969 |
|
|
{
|
| 970 |
|
|
#endif
|
| 971 |
|
|
except_handle (EXCEPT_ILLEGAL, cpu_state.iqueue.insn_addr);
|
| 972 |
|
|
|
| 973 |
|
|
} /* l_invalid() */
|
| 974 |
|
|
|
| 975 |
|
|
|
| 976 |
|
|
/*---------------------------------------------------------------------------*/
|
| 977 |
|
|
/*!The main execution loop */
|
| 978 |
|
|
/*---------------------------------------------------------------------------*/
|
| 979 |
|
|
void
|
| 980 |
|
|
exec_main ()
|
| 981 |
|
|
{
|
| 982 |
|
|
long long time_start;
|
| 983 |
|
|
|
| 984 |
|
|
while (1)
|
| 985 |
|
|
{
|
| 986 |
|
|
time_start = runtime.sim.cycles;
|
| 987 |
|
|
if (config.debug.enabled)
|
| 988 |
|
|
{
|
| 989 |
|
|
while (runtime.cpu.stalled)
|
| 990 |
|
|
{
|
| 991 |
|
|
if (config.debug.rsp_enabled)
|
| 992 |
|
|
{
|
| 993 |
|
|
handle_rsp ();
|
| 994 |
|
|
}
|
| 995 |
|
|
else if (config.debug.gdb_enabled)
|
| 996 |
|
|
{
|
| 997 |
|
|
block_jtag ();
|
| 998 |
|
|
handle_server_socket (FALSE);
|
| 999 |
|
|
}
|
| 1000 |
|
|
else
|
| 1001 |
|
|
{
|
| 1002 |
|
|
fprintf (stderr, "ERROR: CPU stalled and GDB connection not "
|
| 1003 |
|
|
"enabled: Invoking CLI and terminating.\n");
|
| 1004 |
|
|
/* Dump the user into interactive mode. From there he or
|
| 1005 |
|
|
she can decide what to do. */
|
| 1006 |
|
|
handle_sim_command ();
|
| 1007 |
|
|
sim_done ();
|
| 1008 |
|
|
}
|
| 1009 |
|
|
if (runtime.sim.iprompt)
|
| 1010 |
|
|
handle_sim_command ();
|
| 1011 |
|
|
}
|
| 1012 |
|
|
}
|
| 1013 |
|
|
|
| 1014 |
|
|
/* Each cycle has counter of mem_cycles; this value is joined with cycles
|
| 1015 |
|
|
at the end of the cycle; no sim originated memory accesses should be
|
| 1016 |
|
|
performed inbetween. */
|
| 1017 |
|
|
runtime.sim.mem_cycles = 0;
|
| 1018 |
|
|
|
| 1019 |
|
|
if (!config.pm.enabled ||
|
| 1020 |
|
|
!(config.pm.enabled &
|
| 1021 |
|
|
(cpu_state.sprs[SPR_PMR] & (SPR_PMR_DME | SPR_PMR_SME))))
|
| 1022 |
|
|
{
|
| 1023 |
|
|
if (cpu_clock ())
|
| 1024 |
|
|
{
|
| 1025 |
|
|
/* A breakpoint has been hit, drop to interactive mode */
|
| 1026 |
|
|
handle_sim_command ();
|
| 1027 |
|
|
}
|
| 1028 |
|
|
}
|
| 1029 |
|
|
|
| 1030 |
|
|
if (config.vapi.enabled && runtime.vapi.enabled)
|
| 1031 |
|
|
{
|
| 1032 |
|
|
vapi_check ();
|
| 1033 |
|
|
}
|
| 1034 |
|
|
|
| 1035 |
|
|
if (config.debug.gdb_enabled)
|
| 1036 |
|
|
{
|
| 1037 |
|
|
handle_server_socket (FALSE); /* block & check_stdin = false */
|
| 1038 |
|
|
}
|
| 1039 |
|
|
|
| 1040 |
|
|
if (config.debug.enabled)
|
| 1041 |
|
|
{
|
| 1042 |
|
|
if (cpu_state.sprs[SPR_DMR1] & SPR_DMR1_ST)
|
| 1043 |
|
|
{
|
| 1044 |
|
|
set_stall_state (1);
|
| 1045 |
|
|
|
| 1046 |
|
|
if (config.debug.rsp_enabled)
|
| 1047 |
|
|
{
|
| 1048 |
|
|
rsp_exception (EXCEPT_TRAP);
|
| 1049 |
|
|
}
|
| 1050 |
|
|
}
|
| 1051 |
|
|
}
|
| 1052 |
|
|
|
| 1053 |
|
|
runtime.sim.cycles += runtime.sim.mem_cycles;
|
| 1054 |
|
|
scheduler.job_queue->time -= runtime.sim.cycles - time_start;
|
| 1055 |
|
|
|
| 1056 |
|
|
if (scheduler.job_queue->time <= 0)
|
| 1057 |
|
|
{
|
| 1058 |
|
|
do_scheduler ();
|
| 1059 |
|
|
}
|
| 1060 |
|
|
}
|
| 1061 |
|
|
} /* exec_main() */
|
| 1062 |
|
|
|
| 1063 |
226 |
julius |
|
| 1064 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1065 |
|
|
/*!Floating point operation setup function
|
| 1066 |
|
|
|
| 1067 |
|
|
Save host rounding mode, set it to OR1K's according to FPCSR
|
| 1068 |
|
|
This function should be called before performing any FP instructions to
|
| 1069 |
|
|
ensure the OR1K's rounding mode is installed in the host
|
| 1070 |
|
|
*/
|
| 1071 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1072 |
|
|
void fp_set_or1k_rm(void)
|
| 1073 |
|
|
{
|
| 1074 |
|
|
// Set OR1K's RM in host machine
|
| 1075 |
|
|
// First save host RM to restore it later
|
| 1076 |
|
|
host_fp_rm = fegetround();
|
| 1077 |
|
|
// Now map OR1K RM to host RM
|
| 1078 |
|
|
int or1k_rm = 0;
|
| 1079 |
|
|
switch(cpu_state.sprs[SPR_FPCSR] & SPR_FPCSR_RM)
|
| 1080 |
|
|
{
|
| 1081 |
|
|
case FPCSR_RM_RN:
|
| 1082 |
|
|
or1k_rm = FE_TONEAREST;
|
| 1083 |
|
|
break;
|
| 1084 |
|
|
case FPCSR_RM_RZ:
|
| 1085 |
|
|
or1k_rm = FE_TOWARDZERO;
|
| 1086 |
|
|
break;
|
| 1087 |
|
|
case FPCSR_RM_RIP:
|
| 1088 |
|
|
or1k_rm = FE_UPWARD;
|
| 1089 |
|
|
break;
|
| 1090 |
|
|
case FPCSR_RM_RIN:
|
| 1091 |
|
|
or1k_rm = FE_TOWARDZERO;
|
| 1092 |
|
|
break;
|
| 1093 |
|
|
}
|
| 1094 |
|
|
// Now set this RM for the host
|
| 1095 |
|
|
fesetround(or1k_rm); // TODO - check for nonzero return here, if RM not
|
| 1096 |
|
|
// able to be set, maybe warn user
|
| 1097 |
|
|
}
|
| 1098 |
|
|
|
| 1099 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1100 |
|
|
/*!Floating point operation flag set and host restore function
|
| 1101 |
|
|
|
| 1102 |
|
|
Copy flags from floating point op into OR1K's FPCSR, and restore the host's
|
| 1103 |
|
|
rounding mode.
|
| 1104 |
|
|
*/
|
| 1105 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1106 |
|
|
void fp_set_flags_restore_host_rm(void)
|
| 1107 |
|
|
{
|
| 1108 |
|
|
// Check FP flags on host, convert to OR1K FPCSR bits
|
| 1109 |
|
|
// First clear all flags in OR1K FPCSR
|
| 1110 |
|
|
cpu_state.sprs[SPR_FPCSR] &= ~SPR_FPCSR_ALLF;
|
| 1111 |
|
|
|
| 1112 |
|
|
// Test host flags, set appropriate OR1K flags
|
| 1113 |
|
|
if (fetestexcept(FE_DIVBYZERO)) cpu_state.sprs[SPR_FPCSR] |= SPR_FPCSR_DZF;
|
| 1114 |
|
|
if (fetestexcept(FE_INEXACT)) cpu_state.sprs[SPR_FPCSR] |= SPR_FPCSR_IXF;
|
| 1115 |
|
|
if (fetestexcept(FE_INVALID)) cpu_state.sprs[SPR_FPCSR] |= SPR_FPCSR_IVF;
|
| 1116 |
|
|
if (fetestexcept(FE_OVERFLOW)) cpu_state.sprs[SPR_FPCSR] |= SPR_FPCSR_OVF;
|
| 1117 |
|
|
if (fetestexcept(FE_UNDERFLOW)) cpu_state.sprs[SPR_FPCSR] |= SPR_FPCSR_UNF;
|
| 1118 |
|
|
|
| 1119 |
|
|
// Restore the hosts's rounding mode
|
| 1120 |
|
|
fesetround(host_fp_rm);
|
| 1121 |
|
|
|
| 1122 |
|
|
// TODO: Call FP exception is FPEE set and any of the flags were set
|
| 1123 |
|
|
/*
|
| 1124 |
|
|
if ((cpu_state.sprs[SPR_FPCSR] & SPR_FPCSR_FPEE) &
|
| 1125 |
|
|
(|(cpu_state.sprs[SPR_FPCSR] & SPR_FPCSR_ALLF)))
|
| 1126 |
|
|
except_handle (EXCEPT_FPE, cpu_state.iqueue.insn_addr);
|
| 1127 |
|
|
*/
|
| 1128 |
|
|
}
|
| 1129 |
|
|
|
| 1130 |
19 |
jeremybenn |
#if COMPLEX_EXECUTION
|
| 1131 |
|
|
|
| 1132 |
|
|
/* Include generated/built in decode_execute function */
|
| 1133 |
|
|
#include "execgen.c"
|
| 1134 |
|
|
|
| 1135 |
|
|
#elif SIMPLE_EXECUTION
|
| 1136 |
|
|
|
| 1137 |
|
|
|
| 1138 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1139 |
|
|
/*!Evaluates source operand
|
| 1140 |
|
|
|
| 1141 |
|
|
Implementation specific.
|
| 1142 |
|
|
|
| 1143 |
|
|
@param[in] op_no The operand
|
| 1144 |
|
|
@param[in] insn_index Address of the instruction
|
| 1145 |
|
|
@param[in] insn The instruction
|
| 1146 |
|
|
|
| 1147 |
|
|
@return The value of the operand */
|
| 1148 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1149 |
|
|
static uorreg_t
|
| 1150 |
|
|
eval_operand (int op_no,
|
| 1151 |
|
|
unsigned long insn_index,
|
| 1152 |
|
|
uint32_t insn)
|
| 1153 |
|
|
{
|
| 1154 |
|
|
struct insn_op_struct *opd = op_start[insn_index];
|
| 1155 |
|
|
uorreg_t ret;
|
| 1156 |
|
|
|
| 1157 |
|
|
while (op_no)
|
| 1158 |
|
|
{
|
| 1159 |
|
|
if (opd->type & OPTYPE_LAST)
|
| 1160 |
|
|
{
|
| 1161 |
|
|
fprintf (stderr,
|
| 1162 |
|
|
"Instruction requested more operands than it has\n");
|
| 1163 |
|
|
exit (1);
|
| 1164 |
|
|
}
|
| 1165 |
|
|
|
| 1166 |
|
|
if ((opd->type & OPTYPE_OP) && !(opd->type & OPTYPE_DIS))
|
| 1167 |
|
|
{
|
| 1168 |
|
|
op_no--;
|
| 1169 |
|
|
}
|
| 1170 |
|
|
|
| 1171 |
|
|
opd++;
|
| 1172 |
|
|
}
|
| 1173 |
|
|
|
| 1174 |
|
|
if (opd->type & OPTYPE_DIS)
|
| 1175 |
|
|
{
|
| 1176 |
|
|
ret = eval_operand_val (insn, opd);
|
| 1177 |
|
|
|
| 1178 |
|
|
while (!(opd->type & OPTYPE_OP))
|
| 1179 |
|
|
{
|
| 1180 |
|
|
opd++;
|
| 1181 |
|
|
}
|
| 1182 |
|
|
|
| 1183 |
|
|
opd++;
|
| 1184 |
|
|
ret += evalsim_reg (eval_operand_val (insn, opd));
|
| 1185 |
|
|
cpu_state.insn_ea = ret;
|
| 1186 |
|
|
|
| 1187 |
|
|
return ret;
|
| 1188 |
|
|
}
|
| 1189 |
|
|
|
| 1190 |
|
|
if (opd->type & OPTYPE_REG)
|
| 1191 |
|
|
{
|
| 1192 |
|
|
return evalsim_reg (eval_operand_val (insn, opd));
|
| 1193 |
|
|
}
|
| 1194 |
|
|
|
| 1195 |
|
|
return eval_operand_val (insn, opd);
|
| 1196 |
|
|
|
| 1197 |
|
|
} /* eval_operand() */
|
| 1198 |
|
|
|
| 1199 |
|
|
|
| 1200 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1201 |
|
|
/*!Set destination operand (register direct) with value.
|
| 1202 |
|
|
|
| 1203 |
|
|
Implementation specific.
|
| 1204 |
|
|
|
| 1205 |
|
|
@param[in] op_no The operand
|
| 1206 |
|
|
@param[in] value The value to set
|
| 1207 |
|
|
@param[in] insn_index Address of the instruction
|
| 1208 |
|
|
@param[in] insn The instruction */
|
| 1209 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1210 |
|
|
static void
|
| 1211 |
|
|
set_operand (int op_no,
|
| 1212 |
|
|
orreg_t value,
|
| 1213 |
|
|
unsigned long insn_index,
|
| 1214 |
|
|
uint32_t insn)
|
| 1215 |
|
|
{
|
| 1216 |
|
|
struct insn_op_struct *opd = op_start[insn_index];
|
| 1217 |
|
|
|
| 1218 |
|
|
while (op_no)
|
| 1219 |
|
|
{
|
| 1220 |
|
|
if (opd->type & OPTYPE_LAST)
|
| 1221 |
|
|
{
|
| 1222 |
|
|
fprintf (stderr,
|
| 1223 |
|
|
"Instruction requested more operands than it has\n");
|
| 1224 |
|
|
exit (1);
|
| 1225 |
|
|
}
|
| 1226 |
|
|
|
| 1227 |
|
|
if ((opd->type & OPTYPE_OP) && !(opd->type & OPTYPE_DIS))
|
| 1228 |
|
|
{
|
| 1229 |
|
|
op_no--;
|
| 1230 |
|
|
}
|
| 1231 |
|
|
|
| 1232 |
|
|
opd++;
|
| 1233 |
|
|
}
|
| 1234 |
|
|
|
| 1235 |
|
|
if (!(opd->type & OPTYPE_REG))
|
| 1236 |
|
|
{
|
| 1237 |
|
|
fprintf (stderr, "Trying to set a non-register operand\n");
|
| 1238 |
|
|
exit (1);
|
| 1239 |
|
|
}
|
| 1240 |
|
|
|
| 1241 |
|
|
setsim_reg (eval_operand_val (insn, opd), value);
|
| 1242 |
|
|
|
| 1243 |
|
|
} /* set_operand() */
|
| 1244 |
|
|
|
| 1245 |
|
|
|
| 1246 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1247 |
|
|
/*!Simple and rather slow decoding function
|
| 1248 |
|
|
|
| 1249 |
|
|
Based on built automata.
|
| 1250 |
|
|
|
| 1251 |
|
|
@param[in] current The current instruction to execute */
|
| 1252 |
|
|
/*---------------------------------------------------------------------------*/
|
| 1253 |
|
|
static void
|
| 1254 |
|
|
decode_execute (struct iqueue_entry *current)
|
| 1255 |
|
|
{
|
| 1256 |
|
|
int insn_index;
|
| 1257 |
|
|
|
| 1258 |
|
|
current->insn_index = insn_index = insn_decode (current->insn);
|
| 1259 |
|
|
|
| 1260 |
|
|
if (insn_index < 0)
|
| 1261 |
|
|
{
|
| 1262 |
|
|
l_invalid (current);
|
| 1263 |
|
|
}
|
| 1264 |
|
|
else
|
| 1265 |
|
|
{
|
| 1266 |
|
|
or32_opcodes[insn_index].exec (current);
|
| 1267 |
|
|
}
|
| 1268 |
|
|
|
| 1269 |
|
|
if (do_stats)
|
| 1270 |
|
|
analysis (&cpu_state.iqueue);
|
| 1271 |
|
|
}
|
| 1272 |
|
|
|
| 1273 |
|
|
#include "insnset.c"
|
| 1274 |
|
|
|
| 1275 |
|
|
#elif defined(DYNAMIC_EXECUTION)
|
| 1276 |
|
|
|
| 1277 |
|
|
#else
|
| 1278 |
|
|
# error "Must define SIMPLE_EXECUTION, COMPLEX_EXECUTION or DYNAMIC_EXECUTION"
|
| 1279 |
|
|
#endif
|