提交 852d481f 编写于 作者: E edgar_igl

SH: Improve movca.l/ocbi emulation.

Author: Vladimir Prus <vladimir@codesourcery.com>

    Fix movcal.l/ocbi emulation.

        * target-sh4/cpu.h (memory_content): New.
        (CPUSH4State): New fields movcal_backup and movcal_backup_tail.
        * target-sh4/helper.h (helper_movcal)
        (helper_discard_movcal_backup, helper_ocbi): New.
        * target-sh4/op_helper.c (helper_movcal)
        (helper_discard_movcal_backup, helper_ocbi): New.
        * target-sh4/translate.c (DisasContext): New field has_movcal.
        (sh4_defs): Update CVS for SH7785.
        (cpu_sh4_init): Initialize env->movcal_backup_tail.
        (_decode_opc): Discard movca.l-backup.
        Make use of helper_movcal and helper_ocbi.
        (gen_intermediate_code_internal): Initialize has_movcal to 1.

Thanks to Shin-ichiro KAWASAKI and Paul Mundt for valuable feedback.
Acked-by: NEdgar E. Iglesias <edgar.iglesias@gmail.com>


git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@6966 c046a42c-6fe2-441c-8c8c-71466251a162
上级 714fa308
......@@ -100,6 +100,12 @@ enum sh_features {
SH_FEATURE_BCR3_AND_BCR4 = 2,
};
typedef struct memory_content {
uint32_t address;
uint32_t value;
struct memory_content *next;
} memory_content;
typedef struct CPUSH4State {
int id; /* CPU model */
......@@ -148,6 +154,8 @@ typedef struct CPUSH4State {
tlb_t itlb[ITLB_SIZE]; /* instruction translation table */
void *intc_handle;
int intr_at_halt; /* SR_BL ignored during sleep */
memory_content *movcal_backup;
memory_content **movcal_backup_tail;
} CPUSH4State;
CPUSH4State *cpu_sh4_init(const char *cpu_model);
......@@ -162,6 +170,8 @@ void sh4_cpu_list(FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...));
void cpu_sh4_write_mmaped_utlb_addr(CPUSH4State *s, target_phys_addr_t addr,
uint32_t mem_value);
int cpu_sh4_is_cached(CPUSH4State * env, target_ulong addr);
static inline void cpu_set_tls(CPUSH4State *env, target_ulong newtls)
{
env->gbr = newtls;
......@@ -293,6 +303,8 @@ static inline void cpu_pc_from_tb(CPUState *env, TranslationBlock *tb)
env->flags = tb->flags;
}
#define TB_FLAG_PENDING_MOVCA (1 << 4)
static inline void cpu_get_tb_cpu_state(CPUState *env, target_ulong *pc,
target_ulong *cs_base, int *flags)
{
......@@ -302,7 +314,8 @@ static inline void cpu_get_tb_cpu_state(CPUState *env, target_ulong *pc,
| DELAY_SLOT_TRUE | DELAY_SLOT_CLEARME)) /* Bits 0- 3 */
| (env->fpscr & (FPSCR_FR | FPSCR_SZ | FPSCR_PR)) /* Bits 19-21 */
| (env->sr & (SR_MD | SR_RB)) /* Bits 29-30 */
| (env->sr & SR_FD); /* Bit 15 */
| (env->sr & SR_FD) /* Bit 15 */
| (env->movcal_backup ? TB_FLAG_PENDING_MOVCA : 0); /* Bit 4 */
}
#endif /* _CPU_SH4_H */
......@@ -644,4 +644,48 @@ void cpu_sh4_write_mmaped_utlb_addr(CPUSH4State *s, target_phys_addr_t addr,
}
}
int cpu_sh4_is_cached(CPUSH4State * env, target_ulong addr)
{
int n;
int use_asid = (env->mmucr & MMUCR_SV) == 0 || (env->sr & SR_MD) == 0;
/* check area */
if (env->sr & SR_MD) {
/* For previledged mode, P2 and P4 area is not cachable. */
if ((0xA0000000 <= addr && addr < 0xC0000000) || 0xE0000000 <= addr)
return 0;
} else {
/* For user mode, only U0 area is cachable. */
if (0x80000000 <= addr)
return 0;
}
/*
* TODO : Evaluate CCR and check if the cache is on or off.
* Now CCR is not in CPUSH4State, but in SH7750State.
* When you move the ccr inot CPUSH4State, the code will be
* as follows.
*/
#if 0
/* check if operand cache is enabled or not. */
if (!(env->ccr & 1))
return 0;
#endif
/* if MMU is off, no check for TLB. */
if (env->mmucr & MMUCR_AT)
return 1;
/* check TLB */
n = find_tlb_entry(env, addr, env->itlb, ITLB_SIZE, use_asid);
if (n >= 0)
return env->itlb[n].c;
n = find_tlb_entry(env, addr, env->utlb, UTLB_SIZE, use_asid);
if (n >= 0)
return env->utlb[n].c;
return 0;
}
#endif
......@@ -9,6 +9,10 @@ DEF_HELPER_0(debug, void)
DEF_HELPER_1(sleep, void, i32)
DEF_HELPER_1(trapa, void, i32)
DEF_HELPER_2(movcal, void, i32, i32)
DEF_HELPER_0(discard_movcal_backup, void)
DEF_HELPER_1(ocbi, void, i32)
DEF_HELPER_2(addv, i32, i32, i32)
DEF_HELPER_2(addc, i32, i32, i32)
DEF_HELPER_2(subv, i32, i32, i32)
......
......@@ -18,6 +18,7 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301 USA
*/
#include <assert.h>
#include <stdlib.h>
#include "exec.h"
#include "helper.h"
......@@ -122,6 +123,57 @@ void helper_trapa(uint32_t tra)
cpu_loop_exit();
}
void helper_movcal(uint32_t address, uint32_t value)
{
if (cpu_sh4_is_cached (env, address))
{
memory_content *r = malloc (sizeof(memory_content));
r->address = address;
r->value = value;
r->next = NULL;
*(env->movcal_backup_tail) = r;
env->movcal_backup_tail = &(r->next);
}
}
void helper_discard_movcal_backup(void)
{
memory_content *current = env->movcal_backup;
while(current)
{
memory_content *next = current->next;
free (current);
env->movcal_backup = current = next;
if (current == 0)
env->movcal_backup_tail = &(env->movcal_backup);
}
}
void helper_ocbi(uint32_t address)
{
memory_content **current = &(env->movcal_backup);
while (*current)
{
uint32_t a = (*current)->address;
if ((a & ~0x1F) == (address & ~0x1F))
{
memory_content *next = (*current)->next;
stl(a, (*current)->value);
if (next == 0)
{
env->movcal_backup_tail = current;
}
free (*current);
*current = next;
break;
}
}
}
uint32_t helper_addc(uint32_t arg0, uint32_t arg1)
{
uint32_t tmp0, tmp1;
......
......@@ -50,6 +50,7 @@ typedef struct DisasContext {
uint32_t delayed_pc;
int singlestep_enabled;
uint32_t features;
int has_movcal;
} DisasContext;
#if defined(CONFIG_USER_ONLY)
......@@ -283,6 +284,7 @@ CPUSH4State *cpu_sh4_init(const char *cpu_model)
env = qemu_mallocz(sizeof(CPUSH4State));
env->features = def->features;
cpu_exec_init(env);
env->movcal_backup_tail = &(env->movcal_backup);
sh4_translate_init();
env->cpu_model_str = cpu_model;
cpu_sh4_reset(env);
......@@ -495,6 +497,37 @@ static inline void gen_store_fpr64 (TCGv_i64 t, int reg)
static void _decode_opc(DisasContext * ctx)
{
/* This code tries to make movcal emulation sufficiently
accurate for Linux purposes. This instruction writes
memory, and prior to that, always allocates a cache line.
It is used in two contexts:
- in memcpy, where data is copied in blocks, the first write
of to a block uses movca.l for performance.
- in arch/sh/mm/cache-sh4.c, movcal.l + ocbi combination is used
to flush the cache. Here, the data written by movcal.l is never
written to memory, and the data written is just bogus.
To simulate this, we simulate movcal.l, we store the value to memory,
but we also remember the previous content. If we see ocbi, we check
if movcal.l for that address was done previously. If so, the write should
not have hit the memory, so we restore the previous content.
When we see an instruction that is neither movca.l
nor ocbi, the previous content is discarded.
To optimize, we only try to flush stores when we're at the start of
TB, or if we already saw movca.l in this TB and did not flush stores
yet. */
if (ctx->has_movcal)
{
int opcode = ctx->opcode & 0xf0ff;
if (opcode != 0x0093 /* ocbi */
&& opcode != 0x00c3 /* movca.l */)
{
gen_helper_discard_movcal_backup ();
ctx->has_movcal = 0;
}
}
#if 0
fprintf(stderr, "Translating opcode 0x%04x\n", ctx->opcode);
#endif
......@@ -1545,7 +1578,13 @@ static void _decode_opc(DisasContext * ctx)
}
return;
case 0x00c3: /* movca.l R0,@Rm */
tcg_gen_qemu_st32(REG(0), REG(B11_8), ctx->memidx);
{
TCGv val = tcg_temp_new();
tcg_gen_qemu_ld32u(val, REG(B11_8), ctx->memidx);
gen_helper_movcal (REG(B11_8), val);
tcg_gen_qemu_st32(REG(0), REG(B11_8), ctx->memidx);
}
ctx->has_movcal = 1;
return;
case 0x40a9:
/* MOVUA.L @Rm,R0 (Rm) -> R0
......@@ -1594,9 +1633,7 @@ static void _decode_opc(DisasContext * ctx)
break;
case 0x0093: /* ocbi @Rn */
{
TCGv dummy = tcg_temp_new();
tcg_gen_qemu_ld32s(dummy, REG(B11_8), ctx->memidx);
tcg_temp_free(dummy);
gen_helper_ocbi (REG(B11_8));
}
return;
case 0x00a3: /* ocbp @Rn */
......@@ -1876,6 +1913,7 @@ gen_intermediate_code_internal(CPUState * env, TranslationBlock * tb,
ctx.tb = tb;
ctx.singlestep_enabled = env->singlestep_enabled;
ctx.features = env->features;
ctx.has_movcal = (tb->flags & TB_FLAG_PENDING_MOVCA);
#ifdef DEBUG_DISAS
qemu_log_mask(CPU_LOG_TB_CPU,
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册