OpenCores
URL https://opencores.org/ocsvn/or1k/or1k/trunk

Subversion Repositories or1k

[/] [or1k/] [tags/] [nog_patch_66/] [or1ksim/] [cache/] [dcache_model.c] - Diff between revs 638 and 884

Go to most recent revision | Only display areas with differences | Details | Blame | View Log

Rev 638 Rev 884
/* dcache_model.c -- data cache simulation
/* dcache_model.c -- data cache simulation
   Copyright (C) 1999 Damjan Lampret, lampret@opencores.org
   Copyright (C) 1999 Damjan Lampret, lampret@opencores.org
 
 
This file is part of OpenRISC 1000 Architectural Simulator.
This file is part of OpenRISC 1000 Architectural Simulator.
 
 
This program is free software; you can redistribute it and/or modify
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
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
(at your option) any later version.
 
 
This program is distributed in the hope that it will be useful,
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.
GNU General Public License for more details.
 
 
You should have received a copy of the GNU General Public License
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
along with this program; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 
/* Cache functions.
/* Cache functions.
   At the moment this functions only simulate functionality of data
   At the moment this functions only simulate functionality of data
   caches and do not influence on fetche/decode/execute stages and timings.
   caches and do not influence on fetche/decode/execute stages and timings.
   They are here only to verify performance of various cache configurations.
   They are here only to verify performance of various cache configurations.
 */
 */
 
 
#include <stdio.h>
#include <stdio.h>
#include <string.h>
#include <string.h>
#include <errno.h>
#include <errno.h>
#include <stdarg.h>
#include <stdarg.h>
 
 
#include "dcache_model.h"
#include "dcache_model.h"
#include "abstract.h"
#include "abstract.h"
#include "stats.h"
#include "stats.h"
#include "spr_defs.h"
#include "spr_defs.h"
#include "sprs.h"
#include "sprs.h"
#include "sim-config.h"
#include "sim-config.h"
 
 
/* Data cache */
/* Data cache */
 
 
struct dc_set {
struct dc_set {
  struct {
  struct {
    unsigned long line[MAX_DC_BLOCK_SIZE];
    unsigned long line[MAX_DC_BLOCK_SIZE];
    unsigned long tagaddr;  /* tag address */
    unsigned long tagaddr;  /* tag address */
    int lru;    /* least recently used */
    int lru;    /* least recently used */
  } way[MAX_DC_WAYS];
  } way[MAX_DC_WAYS];
} dc[MAX_DC_SETS];
} dc[MAX_DC_SETS];
 
 
void dc_info()
void dc_info()
{
{
  if (!testsprbits(SPR_UPR, SPR_UPR_DCP)) {
  if (!testsprbits(SPR_UPR, SPR_UPR_DCP)) {
    printf("DCache not implemented. Set UPR[DCP].\n");
    printf("DCache not implemented. Set UPR[DCP].\n");
    return;
    return;
  }
  }
 
 
  printf("Data cache %dKB: ", config.dc.nsets * config.dc.blocksize * config.dc.nways / 1024);
  printf("Data cache %dKB: ", config.dc.nsets * config.dc.blocksize * config.dc.nways / 1024);
  printf("%d ways, %d sets, block size %d bytes\n", config.dc.nways, config.dc.nsets, config.dc.blocksize);
  printf("%d ways, %d sets, block size %d bytes\n", config.dc.nways, config.dc.nsets, config.dc.blocksize);
}
}
 
 
/* First check if data is already in the cache and if it is:
/* First check if data is already in the cache and if it is:
    - increment DC read hit stats,
    - increment DC read hit stats,
    - set 'lru' at this way to config.dc.ustates - 1 and
    - set 'lru' at this way to config.dc.ustates - 1 and
      decrement 'lru' of other ways unless they have reached 0,
      decrement 'lru' of other ways unless they have reached 0,
   and if not:
   and if not:
    - increment DC read miss stats
    - increment DC read miss stats
    - find lru way and entry and replace old tag with tag of the 'dataaddr'
    - find lru way and entry and replace old tag with tag of the 'dataaddr'
    - set 'lru' with config.dc.ustates - 1 and decrement 'lru' of other
    - set 'lru' with config.dc.ustates - 1 and decrement 'lru' of other
      ways unless they have reached 0
      ways unless they have reached 0
    - refill cache line
    - refill cache line
*/
*/
 
 
unsigned long dc_simulate_read(unsigned long dataaddr, int width)
unsigned long dc_simulate_read(unsigned long dataaddr, int width)
{
{
  int set, way = -1;
  int set, way = -1;
  int i;
  int i;
  unsigned long tagaddr;
  unsigned long tagaddr;
  extern int mem_cycles;
 
  unsigned long tmp;
  unsigned long tmp;
 
 
  if ((!testsprbits(SPR_UPR, SPR_UPR_DCP)) ||
  if ((!testsprbits(SPR_UPR, SPR_UPR_DCP)) ||
      (!testsprbits(SPR_SR, SPR_SR_DCE))   ||
      (!testsprbits(SPR_SR, SPR_SR_DCE))   ||
      data_ci) {
      data_ci) {
    if (width == 4)
    if (width == 4)
      return evalsim_mem32(dataaddr);
      return evalsim_mem32(dataaddr);
    else if (width == 2)
    else if (width == 2)
      return (unsigned long)evalsim_mem16(dataaddr);
      return (unsigned long)evalsim_mem16(dataaddr);
    else if (width == 1)
    else if (width == 1)
      return (unsigned long)evalsim_mem8(dataaddr);
      return (unsigned long)evalsim_mem8(dataaddr);
  }
  }
 
 
  /* Which set to check out? */
  /* Which set to check out? */
  set = (dataaddr / config.dc.blocksize) % config.dc.nsets;
  set = (dataaddr / config.dc.blocksize) % config.dc.nsets;
  tagaddr = (dataaddr / config.dc.blocksize) / config.dc.nsets;
  tagaddr = (dataaddr / config.dc.blocksize) / config.dc.nsets;
 
 
  /* Scan all ways and try to find a matching way. */
  /* Scan all ways and try to find a matching way. */
  for (i = 0; i < config.dc.nways; i++)
  for (i = 0; i < config.dc.nways; i++)
    if (dc[set].way[i].tagaddr == tagaddr)
    if (dc[set].way[i].tagaddr == tagaddr)
      way = i;
      way = i;
 
 
  /* Did we find our cached data? */
  /* Did we find our cached data? */
  if (way >= 0) { /* Yes, we did. */
  if (way >= 0) { /* Yes, we did. */
    dc_stats.readhit++;
    dc_stats.readhit++;
 
 
    for (i = 0; i < config.dc.nways; i++)
    for (i = 0; i < config.dc.nways; i++)
      if (dc[set].way[i].lru > dc[set].way[way].lru)
      if (dc[set].way[i].lru > dc[set].way[way].lru)
        dc[set].way[i].lru--;
        dc[set].way[i].lru--;
    dc[set].way[way].lru = config.dc.ustates - 1;
    dc[set].way[way].lru = config.dc.ustates - 1;
    mem_cycles += config.dc.load_hitdelay;
    runtime.sim.mem_cycles += config.dc.load_hitdelay;
 
 
    tmp = dc[set].way[way].line[(dataaddr & (config.dc.blocksize - 1)) >> 2];
    tmp = dc[set].way[way].line[(dataaddr & (config.dc.blocksize - 1)) >> 2];
    if (width == 4)
    if (width == 4)
      return tmp;
      return tmp;
    else if (width == 2) {
    else if (width == 2) {
      tmp = (unsigned long)((tmp >> ((dataaddr & 2) ? 0 : 16)) & 0xffff);
      tmp = (unsigned long)((tmp >> ((dataaddr & 2) ? 0 : 16)) & 0xffff);
      return tmp;
      return tmp;
    }
    }
    else if (width == 1) {
    else if (width == 1) {
      tmp = (unsigned long)((tmp  >> (8 * (3 - (dataaddr & 3)))) & 0xff);
      tmp = (unsigned long)((tmp  >> (8 * (3 - (dataaddr & 3)))) & 0xff);
      return tmp;
      return tmp;
    }
    }
  } else {  /* No, we didn't. */
  } else {  /* No, we didn't. */
    int minlru = config.dc.ustates - 1;
    int minlru = config.dc.ustates - 1;
    int minway = 0;
    int minway = 0;
 
 
    dc_stats.readmiss++;
    dc_stats.readmiss++;
 
 
    for (i = 0; i < config.dc.nways; i++)
    for (i = 0; i < config.dc.nways; i++)
      if (dc[set].way[i].lru < minlru)
      if (dc[set].way[i].lru < minlru)
        minway = i;
        minway = i;
 
 
    for (i = 0; i < (config.dc.blocksize); i += 4) {
    for (i = 0; i < (config.dc.blocksize); i += 4) {
      dc[set].way[minway].line[((dataaddr + i) & (config.dc.blocksize - 1)) >> 2] =
      dc[set].way[minway].line[((dataaddr + i) & (config.dc.blocksize - 1)) >> 2] =
        evalsim_mem32((dataaddr & ~(config.dc.blocksize - 1)) + (((dataaddr & ~3ul)+ i) & (config.dc.blocksize - 1)));
        evalsim_mem32((dataaddr & ~(config.dc.blocksize - 1)) + (((dataaddr & ~3ul)+ i) & (config.dc.blocksize - 1)));
      if(!cur_area) {
      if(!cur_area) {
        dc[set].way[minway].tagaddr = -1;
        dc[set].way[minway].tagaddr = -1;
        dc[set].way[minway].lru = 0;
        dc[set].way[minway].lru = 0;
        return 0;
        return 0;
      }
      }
    }
    }
 
 
    dc[set].way[minway].tagaddr = tagaddr;
    dc[set].way[minway].tagaddr = tagaddr;
    for (i = 0; i < config.dc.nways; i++)
    for (i = 0; i < config.dc.nways; i++)
      if (dc[set].way[i].lru)
      if (dc[set].way[i].lru)
        dc[set].way[i].lru--;
        dc[set].way[i].lru--;
    dc[set].way[minway].lru = config.dc.ustates - 1;
    dc[set].way[minway].lru = config.dc.ustates - 1;
    mem_cycles += config.dc.load_missdelay;
    runtime.sim.mem_cycles += config.dc.load_missdelay;
 
 
    tmp = dc[set].way[minway].line[(dataaddr & (config.dc.blocksize - 1)) >> 2];
    tmp = dc[set].way[minway].line[(dataaddr & (config.dc.blocksize - 1)) >> 2];
    if (width == 4)
    if (width == 4)
      return tmp;
      return tmp;
    else if (width == 2) {
    else if (width == 2) {
      tmp = (unsigned long)((tmp >> ((dataaddr & 2) ? 0 : 16)) & 0xffff);
      tmp = (unsigned long)((tmp >> ((dataaddr & 2) ? 0 : 16)) & 0xffff);
      return tmp;
      return tmp;
    }
    }
    else if (width == 1) {
    else if (width == 1) {
      tmp = (unsigned long)((tmp  >> (8 * (3 - (dataaddr & 3)))) & 0xff);
      tmp = (unsigned long)((tmp  >> (8 * (3 - (dataaddr & 3)))) & 0xff);
      return tmp;
      return tmp;
    }
    }
  }
  }
}
}
 
 
/* First check if data is already in the cache and if it is:
/* First check if data is already in the cache and if it is:
    - increment DC write hit stats,
    - increment DC write hit stats,
    - set 'lru' at this way to config.dc.ustates - 1 and
    - set 'lru' at this way to config.dc.ustates - 1 and
      decrement 'lru' of other ways unless they have reached 0,
      decrement 'lru' of other ways unless they have reached 0,
   and if not:
   and if not:
    - increment DC write miss stats
    - increment DC write miss stats
    - find lru way and entry and replace old tag with tag of the 'dataaddr'
    - find lru way and entry and replace old tag with tag of the 'dataaddr'
    - set 'lru' with config.dc.ustates - 1 and decrement 'lru' of other
    - set 'lru' with config.dc.ustates - 1 and decrement 'lru' of other
      ways unless they have reached 0
      ways unless they have reached 0
*/
*/
 
 
void dc_simulate_write(unsigned long dataaddr, unsigned long data, int width)
void dc_simulate_write(unsigned long dataaddr, unsigned long data, int width)
{
{
  int set, way = -1;
  int set, way = -1;
  int i;
  int i;
  unsigned long tagaddr;
  unsigned long tagaddr;
  extern int mem_cycles;
 
  unsigned long tmp;
  unsigned long tmp;
 
 
  if (width == 4)
  if (width == 4)
    setsim_mem32(dataaddr, data);
    setsim_mem32(dataaddr, data);
  else if (width == 2)
  else if (width == 2)
    setsim_mem16(dataaddr, (unsigned short)data);
    setsim_mem16(dataaddr, (unsigned short)data);
  else if (width == 1)
  else if (width == 1)
    setsim_mem8(dataaddr, (unsigned char)data);
    setsim_mem8(dataaddr, (unsigned char)data);
 
 
  if ((!testsprbits(SPR_UPR, SPR_UPR_DCP)) ||
  if ((!testsprbits(SPR_UPR, SPR_UPR_DCP)) ||
      (!testsprbits(SPR_SR, SPR_SR_DCE)) ||
      (!testsprbits(SPR_SR, SPR_SR_DCE)) ||
      data_ci ||
      data_ci ||
      (!cur_area))
      (!cur_area))
    return;
    return;
 
 
  /* Which set to check out? */
  /* Which set to check out? */
  set = (dataaddr / config.dc.blocksize) % config.dc.nsets;
  set = (dataaddr / config.dc.blocksize) % config.dc.nsets;
  tagaddr = (dataaddr / config.dc.blocksize) / config.dc.nsets;
  tagaddr = (dataaddr / config.dc.blocksize) / config.dc.nsets;
 
 
  /* Scan all ways and try to find a matching way. */
  /* Scan all ways and try to find a matching way. */
  for (i = 0; i < config.dc.nways; i++)
  for (i = 0; i < config.dc.nways; i++)
    if (dc[set].way[i].tagaddr == tagaddr)
    if (dc[set].way[i].tagaddr == tagaddr)
      way = i;
      way = i;
 
 
  /* Did we find our cached data? */
  /* Did we find our cached data? */
  if (way >= 0) { /* Yes, we did. */
  if (way >= 0) { /* Yes, we did. */
    dc_stats.writehit++;
    dc_stats.writehit++;
 
 
    for (i = 0; i < config.dc.nways; i++)
    for (i = 0; i < config.dc.nways; i++)
      if (dc[set].way[i].lru > dc[set].way[way].lru)
      if (dc[set].way[i].lru > dc[set].way[way].lru)
        dc[set].way[i].lru--;
        dc[set].way[i].lru--;
    dc[set].way[way].lru = config.dc.ustates - 1;
    dc[set].way[way].lru = config.dc.ustates - 1;
    mem_cycles += config.dc.store_hitdelay;
    runtime.sim.mem_cycles += config.dc.store_hitdelay;
 
 
    tmp = dc[set].way[way].line[(dataaddr & (config.dc.blocksize - 1)) >> 2];
    tmp = dc[set].way[way].line[(dataaddr & (config.dc.blocksize - 1)) >> 2];
    if (width == 4)
    if (width == 4)
      tmp = data;
      tmp = data;
    else if (width == 2) {
    else if (width == 2) {
      tmp &= 0xffff << ((dataaddr & 2) ? 16 : 0);
      tmp &= 0xffff << ((dataaddr & 2) ? 16 : 0);
      tmp |= (unsigned long)(data & 0xffff) << ((dataaddr & 2) ? 0 : 16);
      tmp |= (unsigned long)(data & 0xffff) << ((dataaddr & 2) ? 0 : 16);
    }
    }
    else if (width == 1) {
    else if (width == 1) {
      tmp &= ~(0xff << (8 * (3 - (dataaddr & 3))));
      tmp &= ~(0xff << (8 * (3 - (dataaddr & 3))));
      tmp |= (unsigned long)(data & 0xff) << (8 * (3 - (dataaddr & 3)));
      tmp |= (unsigned long)(data & 0xff) << (8 * (3 - (dataaddr & 3)));
    }
    }
    dc[set].way[way].line[(dataaddr & (config.dc.blocksize - 1)) >> 2] = tmp;
    dc[set].way[way].line[(dataaddr & (config.dc.blocksize - 1)) >> 2] = tmp;
  }
  }
  else {  /* No, we didn't. */
  else {  /* No, we didn't. */
    int minlru = config.dc.ustates - 1;
    int minlru = config.dc.ustates - 1;
    int minway = 0;
    int minway = 0;
 
 
    dc_stats.writemiss++;
    dc_stats.writemiss++;
 
 
    for (i = 0; i < config.dc.nways; i++)
    for (i = 0; i < config.dc.nways; i++)
      if (dc[set].way[i].lru < minlru)
      if (dc[set].way[i].lru < minlru)
        minway = i;
        minway = i;
 
 
    for (i = 0; i < (config.dc.blocksize); i += 4) {
    for (i = 0; i < (config.dc.blocksize); i += 4) {
      dc[set].way[minway].line[((dataaddr + i) & (config.dc.blocksize - 1)) >> 2] =
      dc[set].way[minway].line[((dataaddr + i) & (config.dc.blocksize - 1)) >> 2] =
        evalsim_mem32((dataaddr & ~(config.dc.blocksize - 1)) + (((dataaddr & ~3ul)+ i) & (config.dc.blocksize - 1)));
        evalsim_mem32((dataaddr & ~(config.dc.blocksize - 1)) + (((dataaddr & ~3ul)+ i) & (config.dc.blocksize - 1)));
      if(!cur_area) {
      if(!cur_area) {
        dc[set].way[minway].tagaddr = -1;
        dc[set].way[minway].tagaddr = -1;
        dc[set].way[minway].lru = 0;
        dc[set].way[minway].lru = 0;
        return;
        return;
      }
      }
    }
    }
 
 
    dc[set].way[minway].tagaddr = tagaddr;
    dc[set].way[minway].tagaddr = tagaddr;
    for (i = 0; i < config.dc.nways; i++)
    for (i = 0; i < config.dc.nways; i++)
      if (dc[set].way[i].lru)
      if (dc[set].way[i].lru)
        dc[set].way[i].lru--;
        dc[set].way[i].lru--;
    dc[set].way[minway].lru = config.dc.ustates - 1;
    dc[set].way[minway].lru = config.dc.ustates - 1;
    mem_cycles += config.dc.store_missdelay;
    runtime.sim.mem_cycles += config.dc.store_missdelay;
  }
  }
}
}
 
 
/* First check if data is already in the cache and if it is:
/* First check if data is already in the cache and if it is:
    - invalidate block if way isn't locked
    - invalidate block if way isn't locked
   otherwise don't do anything.
   otherwise don't do anything.
*/
*/
 
 
void dc_inv(unsigned long dataaddr)
void dc_inv(unsigned long dataaddr)
{
{
  int set, way = -1;
  int set, way = -1;
  int i;
  int i;
  unsigned long tagaddr;
  unsigned long tagaddr;
 
 
  if (!testsprbits(SPR_UPR, SPR_UPR_DCP))
  if (!testsprbits(SPR_UPR, SPR_UPR_DCP))
    return;
    return;
 
 
  /* Which set to check out? */
  /* Which set to check out? */
  set = (dataaddr / config.dc.blocksize) % config.dc.nsets;
  set = (dataaddr / config.dc.blocksize) % config.dc.nsets;
  tagaddr = (dataaddr / config.dc.blocksize) / config.dc.nsets;
  tagaddr = (dataaddr / config.dc.blocksize) / config.dc.nsets;
 
 
  if (!testsprbits(SPR_SR, SPR_SR_DCE)) {
  if (!testsprbits(SPR_SR, SPR_SR_DCE)) {
    for (i = 0; i < config.dc.nways; i++) {
    for (i = 0; i < config.dc.nways; i++) {
      dc[set].way[i].tagaddr = -1;
      dc[set].way[i].tagaddr = -1;
      dc[set].way[i].lru = 0;
      dc[set].way[i].lru = 0;
    }
    }
    return;
    return;
  }
  }
   /* Scan all ways and try to find a matching way. */
   /* Scan all ways and try to find a matching way. */
  for (i = 0; i < config.dc.nways; i++)
  for (i = 0; i < config.dc.nways; i++)
    if (dc[set].way[i].tagaddr == tagaddr)
    if (dc[set].way[i].tagaddr == tagaddr)
      way = i;
      way = i;
 
 
  /* Did we find our cached data? */
  /* Did we find our cached data? */
  if (way >= 0) { /* Yes, we did. */
  if (way >= 0) { /* Yes, we did. */
    dc[set].way[way].tagaddr = -1;
    dc[set].way[way].tagaddr = -1;
    dc[set].way[way].lru = 0;
    dc[set].way[way].lru = 0;
  }
  }
}
}
 
 
inline void dc_clock()
inline void dc_clock()
{
{
  unsigned long addr;
  unsigned long addr;
 
 
  if (addr = mfspr(SPR_DCBPR)) {
  if (addr = mfspr(SPR_DCBPR)) {
    dc_simulate_read(addr, 4);
    dc_simulate_read(addr, 4);
    mtspr(SPR_DCBPR, 0);
    mtspr(SPR_DCBPR, 0);
  }
  }
  if ((addr = mfspr(SPR_DCBFR)) != -1) {
  if ((addr = mfspr(SPR_DCBFR)) != -1) {
    dc_inv(addr);
    dc_inv(addr);
    mtspr(SPR_DCBFR, -1);
    mtspr(SPR_DCBFR, -1);
  }
  }
  if (addr = mfspr(SPR_DCBIR)) {
  if (addr = mfspr(SPR_DCBIR)) {
    dc_inv(addr);
    dc_inv(addr);
    mtspr(SPR_DCBIR, 0);
    mtspr(SPR_DCBIR, 0);
  }
  }
  if (addr = mfspr(SPR_DCBWR)) {
  if (addr = mfspr(SPR_DCBWR)) {
    mtspr(SPR_DCBWR, 0);
    mtspr(SPR_DCBWR, 0);
  }
  }
  if (addr = mfspr(SPR_DCBLR)) {
  if (addr = mfspr(SPR_DCBLR)) {
    mtspr(SPR_DCBLR, 0);
    mtspr(SPR_DCBLR, 0);
  }
  }
}
}
 
 

powered by: WebSVN 2.1.0

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.