From 1747b6712da0238f7a9c9b6a4bc08337e3e4aa71 Mon Sep 17 00:00:00 2001 From: kub Date: Tue, 25 Jun 2019 20:24:11 +0200 Subject: [PATCH] SH2 drc: register cache overhaul (bugfixing, speed, readability) --- cpu/sh2/compiler.c | 1178 ++++++++++++++++++++++++-------------------- 1 file changed, 657 insertions(+), 521 deletions(-) diff --git a/cpu/sh2/compiler.c b/cpu/sh2/compiler.c index c1ba3f32..3b03d0c2 100644 --- a/cpu/sh2/compiler.c +++ b/cpu/sh2/compiler.c @@ -60,6 +60,7 @@ // 08 - runtime block entry log // 10 - smc self-check // 20 - runtime block entry counter +// 40 - rcache checking // 80 - branch cache statistics // 100 - write trace // 200 - compare trace @@ -67,7 +68,7 @@ // 800 - state dump on exit // { #ifndef DRC_DEBUG -#define DRC_DEBUG 0x0 +#define DRC_DEBUG 0//x8e7 #endif #if DRC_DEBUG @@ -152,13 +153,17 @@ enum op_types { OP_UNDEFINED, }; -#define OP_ISBRANCH(op) (BITRANGE(OP_BRANCH, OP_BRANCH_RF) & BITMASK1(op)) +// XXX consider trap insns: OP_TRAPA, OP_UNDEFINED? +#define OP_ISBRANCH(op) ((BITRANGE(OP_BRANCH, OP_BRANCH_RF)| BITMASK1(OP_RTE)) \ + & BITMASK1(op)) #define OP_ISBRAUC(op) (BITMASK4(OP_BRANCH, OP_BRANCH_R, OP_BRANCH_RF, OP_RTE) \ & BITMASK1(op)) -#define OP_ISBRACND(op) (BITMASK2(OP_BRANCH_CT, OP_BRANCH_CF) & BITMASK1(op)) +#define OP_ISBRACND(op) (BITMASK3(OP_BRANCH_CT, OP_BRANCH_CF, OP_BRANCH_N) \ + & BITMASK1(op)) #define OP_ISBRAIMM(op) (BITMASK3(OP_BRANCH, OP_BRANCH_CT, OP_BRANCH_CF) \ - & BITMASK1(op)) -#define OP_ISBRAIND(op) (BITMASK2(OP_BRANCH_R, OP_BRANCH_RF) & BITMASK1(op)) + & BITMASK1(op)) +#define OP_ISBRAIND(op) (BITMASK3(OP_BRANCH_R, OP_BRANCH_RF, OP_RTE) \ + & BITMASK1(op)) #ifdef DRC_SH2 @@ -192,7 +197,9 @@ static char sh2dasm_buff[64]; } #if (DRC_DEBUG & (8|256|512|1024)) || defined(PDB) +#if (DRC_DEBUG & (256|512|1024)) static SH2 csh2[2][8]; +#endif static void REGPARM(3) *sh2_drc_log_entry(void *block, SH2 *sh2, u32 sr) { if (block != NULL) { @@ -386,7 +393,8 @@ enum { typedef struct { u8 hreg; // "host" reg u8 flags:4; // TEMP or REG? - u8 type:4; + u8 type:2; // CACHED or TEMP? + u8 ref:2; // ref counter u16 stamp; // kind of a timestamp u32 gregs; // "guest" reg mask } cache_reg_t; @@ -415,8 +423,9 @@ static void rcache_free_tmp(int hr); // used by handlers in worst case (currently 4). // Register assignment goes by ABI convention. Caller save registers are TEMP, // the others are either static or REG. SR must be static, R0 very recommended. +// VBR, PC, PR must not be static (read from context in utils). // TEMP registers first, REG last. alloc/evict algorithm depends on this. -// The 1st TEMP must not be RET_REG on x86 (it uses temps for some insns). +// The 1st TEMP must not be RET_REG on platforms using temps in insns (eg. x86). // XXX shouldn't this be somehow defined in the code emitters? #ifdef __arm__ #include "../drc/emit_arm.c" @@ -438,7 +447,7 @@ static guest_reg_t guest_regs[] = { }; // OABI/EABI: params: r0-r3, return: r0-r1, temp: r12,r14, saved: r4-r8,r10,r11 -// SP,PC: r13,r15 must not be used. saved: r9 (for platform use, e.g. on OSx) +// SP,PC: r13,r15 must not be used. saved: r9 (for platform use, e.g. on ios) static cache_reg_t cache_regs[] = { { 12, HRF_TEMP }, // temps { 14, HRF_TEMP }, @@ -1216,43 +1225,93 @@ typedef struct { gconst_t gconsts[ARRAY_SIZE(guest_regs)]; static int rcache_get_reg_(sh2_reg_e r, rc_gr_mode mode, int do_locking, int *hr); +static inline int rcache_is_cached(sh2_reg_e r); static void rcache_add_vreg_alias(int x, sh2_reg_e r); static void rcache_remove_vreg_alias(int x, sh2_reg_e r); +static void rcache_evict_vreg(int x); +static void rcache_remap_vreg(int x); #define RCACHE_DUMP(msg) { \ cache_reg_t *cp; \ guest_reg_t *gp; \ int i; \ printf("cache dump %s:\n",msg); \ - printf("cache_regs:\n"); \ + printf(" cache_regs:\n"); \ for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { \ cp = &cache_regs[i]; \ - if (cp->type != HR_FREE || cp->gregs) \ - printf("%d: hr=%d t=%d f=%x m=%x\n", i, cp->hreg, cp->type, cp->flags, cp->gregs); \ + if (cp->type != HR_FREE || cp->gregs || (cp->flags & ~(HRF_REG|HRF_TEMP))) \ + printf(" %d: hr=%d t=%d f=%x c=%d m=%x\n", i, cp->hreg, cp->type, cp->flags, cp->ref, cp->gregs); \ } \ - printf("guest_regs:\n"); \ + printf(" guest_regs:\n"); \ for (i = 0; i < ARRAY_SIZE(guest_regs); i++) { \ gp = &guest_regs[i]; \ - if (gp->vreg != -1 || gp->sreg >= 0) \ - printf("%d: v=%d f=%x s=%d\n", i, gp->vreg, gp->flags, gp->sreg); \ + if (gp->vreg != -1 || gp->sreg >= 0 || gp->flags) \ + printf(" %d: v=%d f=%x s=%d c=%d\n", i, gp->vreg, gp->flags, gp->sreg, gp->cnst); \ + } \ + printf(" gconsts:\n"); \ + for (i = 0; i < ARRAY_SIZE(gconsts); i++) { \ + if (gconsts[i].gregs) \ + printf(" %d: m=%x v=%x\n", i, gconsts[i].gregs, gconsts[i].val); \ } \ } +#define RCACHE_CHECK(msg) { \ + cache_reg_t *cp; \ + guest_reg_t *gp; \ + int i, x, d = 0; \ + for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { \ + cp = &cache_regs[i]; \ + if (cp->type == HR_FREE || cp->type == HR_TEMP) continue; \ + /* check connectivity greg->vreg */ \ + FOR_ALL_BITS_SET_DO(cp->gregs, x, \ + if (guest_regs[x].vreg != i) \ + { d = 1; printf("cache check v=%d r=%d not connected?\n",i,x); } \ + ) \ + } \ + for (i = 0; i < ARRAY_SIZE(guest_regs); i++) { \ + gp = &guest_regs[i]; \ + if (gp->vreg != -1 && !(cache_regs[gp->vreg].gregs & (1 << i))) \ + { d = 1; printf("cache check r=%d v=%d not connected?\n", i, gp->vreg); }\ + if (gp->vreg != -1 && cache_regs[gp->vreg].type != HR_STATIC && cache_regs[gp->vreg].type != HR_CACHED) \ + { d = 1; printf("cache check r=%d v=%d wrong type?\n", i, gp->vreg); }\ + if ((gp->flags & GRF_CONST) && !(gconsts[gp->cnst].gregs & (1 << i))) \ + { d = 1; printf("cache check r=%d c=%d not connected?\n", i, gp->cnst); }\ + if ((gp->flags & GRF_CDIRTY) && (gp->vreg != -1 || !(gp->flags & GRF_CONST)) )\ + { d = 1; printf("cache check r=%d CDIRTY?\n", i); } \ + } \ + for (i = 0; i < ARRAY_SIZE(gconsts); i++) { \ + FOR_ALL_BITS_SET_DO(gconsts[i].gregs, x, \ + if (guest_regs[x].cnst != i || !(guest_regs[x].flags & GRF_CONST)) \ + { d = 1; printf("cache check c=%d v=%d not connected?\n",i,x); } \ + ) \ + } \ + if (d) RCACHE_DUMP(msg) \ +/* else { \ + printf("locked regs %s:\n",msg); \ + for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { \ + cp = &cache_regs[i]; \ + if (cp->flags & HRF_LOCKED) \ + printf(" %d: hr=%d t=%d f=%x c=%d m=%x\n", i, cp->hreg, cp->type, cp->flags, cp->ref, cp->gregs); \ + } \ + } */ \ +} + #if PROPAGATE_CONSTANTS static inline int gconst_alloc(sh2_reg_e r) { int i, n = -1; for (i = 0; i < ARRAY_SIZE(gconsts); i++) { - if (gconsts[i].gregs & (1 << r)) - gconsts[i].gregs &= ~(1 << r); + gconsts[i].gregs &= ~(1 << r); if (gconsts[i].gregs == 0 && n < 0) n = i; } if (n >= 0) gconsts[n].gregs = (1 << r); - else + else { + printf("all gconst buffers in use, aborting\n"); exit(1); // cannot happen - more constants than guest regs? + } return n; } @@ -1274,19 +1333,6 @@ static void gconst_new(sh2_reg_e r, u32 val) if (guest_regs[r].vreg >= 0) rcache_remove_vreg_alias(guest_regs[r].vreg, r); } - -static void gconst_copy(sh2_reg_e rd, sh2_reg_e rs) -{ - if (guest_regs[rd].flags & GRF_CONST) { - guest_regs[rd].flags &= ~(GRF_CONST|GRF_CDIRTY); - gconsts[guest_regs[rd].cnst].gregs &= ~(1 << rd); - } - if (guest_regs[rs].flags & GRF_CONST) { - guest_regs[rd].flags |= GRF_CONST; - guest_regs[rd].cnst = guest_regs[rs].cnst; - gconsts[guest_regs[rd].cnst].gregs |= (1 << rd); - } -} #endif static int gconst_get(sh2_reg_e r, u32 *val) @@ -1310,17 +1356,22 @@ static int gconst_check(sh2_reg_e r) static int gconst_try_read(int vreg, sh2_reg_e r) { int i, x; + if (guest_regs[r].flags & GRF_CDIRTY) { x = guest_regs[r].cnst; emith_move_r_imm(cache_regs[vreg].hreg, gconsts[x].val); FOR_ALL_BITS_SET_DO(gconsts[x].gregs, i, { - if (guest_regs[i].vreg >= 0 && i != r) + if (guest_regs[i].vreg >= 0 && guest_regs[i].vreg != vreg) rcache_remove_vreg_alias(guest_regs[i].vreg, i); - rcache_add_vreg_alias(vreg, i); + if (guest_regs[i].vreg < 0) + rcache_add_vreg_alias(vreg, i); guest_regs[i].flags &= ~GRF_CDIRTY; guest_regs[i].flags |= GRF_DIRTY; }); + if (cache_regs[vreg].type != HR_STATIC) + cache_regs[vreg].type = HR_CACHED; + cache_regs[vreg].flags |= HRF_DIRTY; return 1; } return 0; @@ -1339,11 +1390,23 @@ static u32 gconst_dirty_mask(void) static void gconst_kill(sh2_reg_e r) { - if (guest_regs[r].flags &= ~(GRF_CONST|GRF_CDIRTY)) + if (guest_regs[r].flags & (GRF_CONST|GRF_CDIRTY)) gconsts[guest_regs[r].cnst].gregs &= ~(1 << r); guest_regs[r].flags &= ~(GRF_CONST|GRF_CDIRTY); } +static void gconst_copy(sh2_reg_e rd, sh2_reg_e rs) +{ + gconst_kill(rd); + if (guest_regs[rs].flags & GRF_CONST) { + guest_regs[rd].flags |= GRF_CONST; + if (guest_regs[rd].vreg < 0) + guest_regs[rd].flags |= GRF_CDIRTY; + guest_regs[rd].cnst = guest_regs[rs].cnst; + gconsts[guest_regs[rd].cnst].gregs |= (1 << rd); + } +} + static void gconst_clean(void) { int i; @@ -1367,25 +1430,76 @@ static void gconst_invalidate(void) } } + static u16 rcache_counter; -static u32 rcache_static; -static u32 rcache_locked; -static u32 rcache_hint_soon; -static u32 rcache_hint_late; -static u32 rcache_hint_write; -static u32 rcache_hint_clean; -#define rcache_hint (rcache_hint_soon|rcache_hint_late) +// SH2 register usage bitmasks +static u32 rcache_regs_static; // statically allocated regs +static u32 rcache_regs_now; // regs used in current insn +static u32 rcache_regs_soon; // regs used in the next few insns +static u32 rcache_regs_late; // regs used in later insns +static u32 rcache_regs_discard; // regs overwritten without being used +static u32 rcache_regs_clean; // regs needing cleaning +// combination masks XXX this seems obscure +#define rcache_regs_used (rcache_regs_soon|rcache_regs_late|rcache_regs_clean) +#define rcache_regs_nowused (rcache_regs_now|rcache_regs_used) +#define rcache_regs_nowsoon (rcache_regs_now|rcache_regs_soon) +#define rcache_regs_soonclean (rcache_regs_soon|rcache_regs_clean) + +static void rcache_ref_vreg(int x) +{ + if (x >= 0) { + cache_regs[x].ref ++; + cache_regs[x].flags |= HRF_LOCKED; + } +} + +static void rcache_unref_vreg(int x) +{ + if (x >= 0 && -- cache_regs[x].ref == 0) { + cache_regs[x].flags &= ~HRF_LOCKED; + } +} + +static void rcache_free_vreg(int x) +{ + if (cache_regs[x].type != HR_STATIC) + cache_regs[x].type = HR_FREE; + cache_regs[x].flags &= (HRF_REG|HRF_TEMP); + cache_regs[x].gregs = 0; + cache_regs[x].ref = 0; +} static void rcache_unmap_vreg(int x) { int i; FOR_ALL_BITS_SET_DO(cache_regs[x].gregs, i, + if (guest_regs[i].flags & GRF_DIRTY) { + // if a dirty reg is unmapped save its value to context + if (~rcache_regs_discard & (1 << i)) + emith_ctx_write(cache_regs[x].hreg, i * 4); + guest_regs[i].flags &= ~GRF_DIRTY; + } guest_regs[i].vreg = -1); - if (cache_regs[x].type != HR_STATIC) - cache_regs[x].type = HR_FREE; - cache_regs[x].gregs = 0; - cache_regs[x].flags &= (HRF_REG|HRF_TEMP); + rcache_free_vreg(x); +} + +static void rcache_move_vreg(int d, int x) +{ + int i; + + if (cache_regs[d].type != HR_STATIC) + cache_regs[d].type = HR_CACHED; + cache_regs[d].gregs = cache_regs[x].gregs; + cache_regs[d].flags &= (HRF_TEMP|HRF_REG); + cache_regs[d].flags |= cache_regs[x].flags & ~(HRF_TEMP|HRF_REG); + cache_regs[d].ref = 0; + cache_regs[d].stamp = cache_regs[x].stamp; + emith_move_r_r(cache_regs[d].hreg, cache_regs[x].hreg); + for (i = 0; i < ARRAY_SIZE(guest_regs); i++) + if (guest_regs[i].vreg == x) + guest_regs[i].vreg = d; + rcache_free_vreg(x); } static void rcache_clean_vreg(int x) @@ -1394,99 +1508,112 @@ static void rcache_clean_vreg(int x) if (cache_regs[x].flags & HRF_DIRTY) { // writeback cache_regs[x].flags &= ~HRF_DIRTY; + rcache_ref_vreg(x); FOR_ALL_BITS_SET_DO(cache_regs[x].gregs, r, if (guest_regs[r].flags & GRF_DIRTY) { if (guest_regs[r].flags & GRF_STATIC) { if (guest_regs[r].vreg != guest_regs[r].sreg) { if (!(cache_regs[guest_regs[r].sreg].flags & HRF_LOCKED)) { // statically mapped reg not in its sreg. move back to sreg - rcache_clean_vreg(guest_regs[r].sreg); - rcache_unmap_vreg(guest_regs[r].sreg); - emith_move_r_r(cache_regs[guest_regs[r].sreg].hreg, cache_regs[guest_regs[r].vreg].hreg); + rcache_evict_vreg(guest_regs[r].sreg); + emith_move_r_r(cache_regs[guest_regs[r].sreg].hreg, + cache_regs[guest_regs[r].vreg].hreg); rcache_remove_vreg_alias(x, r); rcache_add_vreg_alias(guest_regs[r].sreg, r); cache_regs[guest_regs[r].sreg].flags |= HRF_DIRTY; } else { // must evict since sreg is locked - emith_ctx_write(cache_regs[x].hreg, r * 4); + if (~rcache_regs_discard & (1 << r)) + emith_ctx_write(cache_regs[x].hreg, r * 4); guest_regs[r].flags &= ~GRF_DIRTY; - guest_regs[r].vreg = -1; + rcache_remove_vreg_alias(x, r); } - } - } else if (~rcache_hint_write & (1 << r)) { - emith_ctx_write(cache_regs[x].hreg, r * 4); + } else + cache_regs[x].flags |= HRF_DIRTY; + } else { + if (~rcache_regs_discard & (1 << r)) + emith_ctx_write(cache_regs[x].hreg, r * 4); guest_regs[r].flags &= ~GRF_DIRTY; } + rcache_regs_clean &= ~(1 << r); }) + rcache_unref_vreg(x); } +#if DRC_DEBUG & 64 + RCACHE_CHECK("after clean"); +#endif } static void rcache_add_vreg_alias(int x, sh2_reg_e r) { cache_regs[x].gregs |= (1 << r); guest_regs[r].vreg = x; + if (cache_regs[x].type != HR_STATIC) + cache_regs[x].type = HR_CACHED; } static void rcache_remove_vreg_alias(int x, sh2_reg_e r) { cache_regs[x].gregs &= ~(1 << r); - if (!cache_regs[x].gregs) { + if (!cache_regs[x].gregs) // no reg mapped -> free vreg - if (cache_regs[x].type != HR_STATIC) - cache_regs[x].type = HR_FREE; - cache_regs[x].flags &= (HRF_REG|HRF_TEMP); - } + rcache_free_vreg(x); guest_regs[r].vreg = -1; } static void rcache_evict_vreg(int x) { +#if REMAP_REGISTER + rcache_remap_vreg(x); +#else rcache_clean_vreg(x); +#endif rcache_unmap_vreg(x); } static void rcache_evict_vreg_aliases(int x, sh2_reg_e r) { - cache_regs[x].gregs &= ~(1 << r); + rcache_remove_vreg_alias(x, r); rcache_evict_vreg(x); - cache_regs[x].gregs = (1 << r); - if (cache_regs[x].type != HR_STATIC) - cache_regs[x].type = HR_CACHED; - if (guest_regs[r].flags & GRF_DIRTY) - cache_regs[x].flags |= HRF_DIRTY; + rcache_add_vreg_alias(x, r); } -static cache_reg_t *rcache_evict(void) +static int rcache_allocate(int what, int minprio) { // evict reg with oldest stamp (only for HRF_REG, no temps) int i, i_prio, oldest = -1, prio = 0; u16 min_stamp = (u16)-1; for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { - // consider only unlocked REG - if (!(cache_regs[i].flags & HRF_REG) || (cache_regs[i].flags & HRF_LOCKED)) + // consider only unlocked REG or non-TEMP + if (cache_regs[i].flags == 0 || (cache_regs[i].flags & HRF_LOCKED)) + continue; + if ((what > 0 && !(cache_regs[i].flags & HRF_REG)) || + (what == 0 && (cache_regs[i].flags & HRF_TEMP)) || + (what < 0 && !(cache_regs[i].flags & HRF_TEMP))) continue; - if (cache_regs[i].type == HR_FREE || (cache_regs[i].type == HR_TEMP)) { + if (cache_regs[i].type == HR_FREE || cache_regs[i].type == HR_TEMP) { + // REG is free + prio = 6; oldest = i; break; } if (cache_regs[i].type == HR_CACHED) { - if (rcache_locked & cache_regs[i].gregs) + if (rcache_regs_now & cache_regs[i].gregs) // REGs needed for the current insn i_prio = 1; - else if (rcache_hint_soon & cache_regs[i].gregs) - // REGs needed in some future insn + else if (rcache_regs_soon & cache_regs[i].gregs) + // REGs needed in the next insns i_prio = 2; - else if (rcache_hint_late & cache_regs[i].gregs) + else if (rcache_regs_late & cache_regs[i].gregs) // REGs needed in some future insn i_prio = 3; - else if ((rcache_hint_write & cache_regs[i].gregs) != cache_regs[i].gregs) - // REGs not needed soon + else if (!(~rcache_regs_discard & cache_regs[i].gregs)) + // REGs not needed in the foreseeable future i_prio = 4; else // REGs soon overwritten anyway i_prio = 5; - if (prio < i_prio || (prio == i_prio && cache_regs[i].stamp < min_stamp)) { min_stamp = cache_regs[i].stamp; oldest = i; @@ -1495,25 +1622,66 @@ static cache_reg_t *rcache_evict(void) } } - if (oldest == -1) { - printf("no registers to evict, aborting\n"); - exit(1); - } + + if (prio < minprio || oldest == -1) + return -1; if (cache_regs[oldest].type == HR_CACHED) rcache_evict_vreg(oldest); - cache_regs[oldest].type = HR_FREE; - cache_regs[oldest].flags &= (HRF_TEMP|HRF_REG); - cache_regs[oldest].gregs = 0; + else + rcache_free_vreg(oldest); + + return oldest; +} + +static int rcache_allocate_vreg(int needed) +{ + int x; + + // get a free reg, but use temps only if r is not needed soon + for (x = ARRAY_SIZE(cache_regs) - 1; x >= 0; x--) { + if (cache_regs[x].flags && (cache_regs[x].type == HR_FREE || + (cache_regs[x].type == HR_TEMP && !(cache_regs[x].flags & HRF_LOCKED))) && + (!needed || (cache_regs[x].flags & HRF_REG))) + break; + } + + if (x < 0) + x = rcache_allocate(1, 0); + return x; +} - return &cache_regs[oldest]; +static int rcache_allocate_nontemp(void) +{ + int x = rcache_allocate(0, 3); + return x; +} + +static int rcache_allocate_temp(void) +{ + int x; + + // use any free reg, but prefer TEMP regs + for (x = 0; x < ARRAY_SIZE(cache_regs); x++) { + if (cache_regs[x].flags && (cache_regs[x].type == HR_FREE || + (cache_regs[x].type == HR_TEMP && !(cache_regs[x].flags & HRF_LOCKED)))) + break; + } + + if (x >= ARRAY_SIZE(cache_regs)) + x = rcache_allocate(-1, 1); + if (x < 0) { + printf("no temp register available, aborting\n"); + exit(1); + } + return x; } #if REMAP_REGISTER // maps a host register to a REG static int rcache_map_reg(sh2_reg_e r, int hr, int mode) { - int i; + int x, i; gconst_kill(r); @@ -1527,11 +1695,13 @@ static int rcache_map_reg(sh2_reg_e r, int hr, int mode) // deal with statically mapped regs if (mode == RC_GR_RMW && (guest_regs[r].flags & GRF_STATIC)) { - if (guest_regs[r].vreg == guest_regs[r].sreg) { + x = guest_regs[r].sreg; + if (guest_regs[r].vreg == x) { // STATIC in its sreg with no aliases, and some processing pending - if (cache_regs[guest_regs[r].vreg].gregs == 1 << r) - return cache_regs[guest_regs[r].vreg].hreg; - } else if (!cache_regs[guest_regs[r].sreg].gregs) + if (cache_regs[x].gregs == 1 << r) + return cache_regs[x].hreg; + } else if (cache_regs[x].type == HR_FREE || + (cache_regs[x].type == HR_TEMP && !(cache_regs[x].flags & HRF_LOCKED))) // STATIC not in its sreg, with sreg available -> move it i = guest_regs[r].sreg; } @@ -1540,187 +1710,184 @@ static int rcache_map_reg(sh2_reg_e r, int hr, int mode) if (guest_regs[r].vreg >= 0) rcache_remove_vreg_alias(guest_regs[r].vreg, r); if (cache_regs[i].type == HR_CACHED) - rcache_unmap_vreg(i); + rcache_evict_vreg(i); // set new mappping if (cache_regs[i].type != HR_STATIC) cache_regs[i].type = HR_CACHED; cache_regs[i].gregs = 1 << r; cache_regs[i].flags &= (HRF_TEMP|HRF_REG); + cache_regs[i].ref = 0; cache_regs[i].stamp = ++rcache_counter; - cache_regs[i].flags |= HRF_DIRTY|HRF_LOCKED; + cache_regs[i].flags |= HRF_DIRTY; + rcache_ref_vreg(i); guest_regs[r].flags |= GRF_DIRTY; guest_regs[r].vreg = i; +#if DRC_DEBUG & 64 + RCACHE_CHECK("after map"); +#endif return cache_regs[i].hreg; } -// remap vreg from a TEMP to a REG if it is hinted (upcoming TEMP invalidation) -static void rcache_remap_vreg(int r) +// remap vreg from a TEMP to a REG if it will be used (upcoming TEMP invalidation) +static void rcache_remap_vreg(int x) { - int i, j, free = -1, cached = -1, hinted = -1; - u16 min_stamp_cached = (u16)-1, min_stamp_hinted = -1; + int d; - // r must be a vreg - if (cache_regs[r].type != HR_CACHED) + // x must be a cached vreg + if (cache_regs[x].type != HR_CACHED && cache_regs[x].type != HR_STATIC) return; - // if r is already a REG or isn't used, clean here to avoid data loss on inval - if ((cache_regs[r].flags & HRF_REG) || !(rcache_hint & cache_regs[r].gregs)) { - rcache_clean_vreg(r); + // don't do it if x is already a REG or isn't used or to be cleaned anyway + if ((cache_regs[x].flags & HRF_REG) || + !(rcache_regs_used & ~rcache_regs_clean & cache_regs[x].gregs)) { + // clean here to avoid data loss on invalidation + rcache_clean_vreg(x); return; } - // find REG, either free or unused temp or oldest cached - for (i = 0; i < ARRAY_SIZE(cache_regs) && free < 0; i++) { - if ((cache_regs[i].flags & HRF_TEMP) || (cache_regs[i].flags & HRF_LOCKED)) - continue; - if (cache_regs[i].type == HR_FREE || cache_regs[i].type == HR_TEMP) - free = i; - if (cache_regs[i].type == HR_CACHED && !(rcache_hint & cache_regs[i].gregs)) { - if (cache_regs[i].stamp < min_stamp_cached) { - min_stamp_cached = cache_regs[i].stamp; - cached = i; - } - } - if (cache_regs[i].type == HR_CACHED && !(rcache_hint_soon & cache_regs[i].gregs) - && (rcache_hint_soon & cache_regs[r].gregs)) - if (cache_regs[i].stamp < min_stamp_hinted) { - min_stamp_hinted = cache_regs[i].stamp; - hinted = i; - } + if (cache_regs[x].flags & HRF_LOCKED) { + printf("remap vreg %d is locked\n", x); + exit(1); } - if (free >= 0) { - i = free; - } else if (cached >= 0 && cached != r) { - i = cached; - rcache_evict_vreg(i); - } else if (hinted >= 0 && hinted != r) { - i = hinted; - rcache_evict_vreg(i); - } else { - rcache_clean_vreg(r); + // allocate a non-TEMP vreg + rcache_ref_vreg(x); // lock to avoid evicting x + d = rcache_allocate_nontemp(); + rcache_unref_vreg(x); + if (d < 0) { + rcache_clean_vreg(x); return; } - // set new mapping and remove old one - cache_regs[i].type = HR_CACHED; - cache_regs[i].gregs = cache_regs[r].gregs; - cache_regs[i].flags &= (HRF_TEMP|HRF_REG); - cache_regs[i].flags |= cache_regs[r].flags & ~(HRF_TEMP|HRF_REG); - cache_regs[i].stamp = cache_regs[r].stamp; - emith_move_r_r(cache_regs[i].hreg, cache_regs[r].hreg); - for (j = 0; j < ARRAY_SIZE(guest_regs); j++) - if (guest_regs[j].vreg == r) - guest_regs[j].vreg = i; - cache_regs[r].type = HR_FREE; - cache_regs[r].flags &= (HRF_TEMP|HRF_REG); - cache_regs[r].gregs = 0; + // move vreg to new location + rcache_move_vreg(d, x); +#if DRC_DEBUG & 64 + RCACHE_CHECK("after remap"); +#endif +} +#endif + +#if ALIAS_REGISTERS +static void rcache_alias_vreg(sh2_reg_e rd, sh2_reg_e rs) +{ + int x; + + // if s isn't constant, it must be in cache for aliasing + if (!gconst_check(rs)) + rcache_get_reg_(rs, RC_GR_READ, 0, NULL); + + // if d and s are not already aliased + x = guest_regs[rs].vreg; + if (guest_regs[rd].vreg != x) { + // remove possible old mapping of dst + if (guest_regs[rd].vreg >= 0) + rcache_remove_vreg_alias(guest_regs[rd].vreg, rd); + // make dst an alias of src + if (x >= 0) + rcache_add_vreg_alias(x, rd); + // if d is now in cache, it must be dirty + if (guest_regs[rd].vreg >= 0) { + x = guest_regs[rd].vreg; + cache_regs[x].flags |= HRF_DIRTY; + guest_regs[rd].flags |= GRF_DIRTY; + } + } + + gconst_copy(rd, rs); +#if DRC_DEBUG & 64 + RCACHE_CHECK("after alias"); +#endif } #endif // note: must not be called when doing conditional code static int rcache_get_reg_(sh2_reg_e r, rc_gr_mode mode, int do_locking, int *hr) { - cache_reg_t *tr = NULL; - int i, h, split = -1; + int src, dst, ali; + cache_reg_t *tr; - rcache_counter++; + dst = src = guest_regs[r].vreg; - // maybe already cached? - // if so, prefer against gconst (they must be in sync) - i = guest_regs[r].vreg; - if ((guest_regs[r].flags & GRF_STATIC) && i != guest_regs[r].sreg && + rcache_ref_vreg(src); // lock to avoid evicting src + // good opportunity to relocate a remapped STATIC? + if ((guest_regs[r].flags & GRF_STATIC) && src != guest_regs[r].sreg && !(cache_regs[guest_regs[r].sreg].flags & HRF_LOCKED) && - (i < 0 || mode != RC_GR_READ) && - !((rcache_hint_soon|rcache_locked) & cache_regs[guest_regs[r].sreg].gregs)) { - // good opportunity to relocate a remapped STATIC - h = guest_regs[r].sreg; - rcache_evict_vreg(h); - tr = &cache_regs[h]; - tr->gregs = 1 << r; - if (i >= 0) { - if (mode != RC_GR_WRITE) { - if (hr) - *hr = cache_regs[i].hreg; - else - emith_move_r_r(cache_regs[h].hreg, cache_regs[i].hreg); - hr = NULL; - } - rcache_remove_vreg_alias(guest_regs[r].vreg, r); - } else if (mode != RC_GR_WRITE) { - if (gconst_try_read(h, r)) { - tr->flags |= HRF_DIRTY; - guest_regs[r].flags |= GRF_DIRTY; - } else - emith_ctx_read(tr->hreg, r * 4); - } - guest_regs[r].vreg = guest_regs[r].sreg; - goto end; - } else if (i >= 0) { - if (mode == RC_GR_READ || !(cache_regs[i].gregs & ~(1 << r))) { - // either only reading, or no multiple mapping - tr = &cache_regs[i]; - goto end; - } - // split if aliases needed rsn, or already locked, or r is STATIC in sreg - if (((rcache_hint|rcache_locked) & cache_regs[i].gregs & ~(1 << r)) || - (cache_regs[i].flags & HRF_LOCKED) || - (cache_regs[i].type == HR_STATIC && !(guest_regs[r].flags & GRF_STATIC))) { - // need to split up. take reg out here to avoid unnecessary writebacks - rcache_remove_vreg_alias(i, r); - split = i; - } else { - // aliases not needed anytime soon, remove them - // XXX split aliases away if writing and static and not locked and hinted? - rcache_evict_vreg_aliases(i, r); - tr = &cache_regs[i]; - goto end; + (src < 0 || mode != RC_GR_READ) && + !(rcache_regs_nowsoon & cache_regs[guest_regs[r].sreg].gregs)) { + dst = guest_regs[r].sreg; + rcache_evict_vreg(dst); + } else if (dst < 0) { + // allocate a cache register + if ((dst = rcache_allocate_vreg(rcache_regs_nowsoon & (1 << r))) < 0) { + printf("no registers to evict, aborting\n"); + exit(1); } } - - // get a free reg, but use temps only if r is not needed soon - for (i = ARRAY_SIZE(cache_regs) - 1; i >= 0; i--) { - if ((cache_regs[i].type == HR_FREE || - (cache_regs[i].type == HR_TEMP && !(cache_regs[i].flags & HRF_LOCKED))) && - (!(rcache_hint & (1 << r)) || (cache_regs[i].flags & HRF_REG))) { - tr = &cache_regs[i]; - break; + tr = &cache_regs[dst]; + tr->stamp = rcache_counter; + rcache_unref_vreg(src); + // remove r from src + if (src >= 0 && src != dst) + rcache_remove_vreg_alias(src, r); + + // if r has a constant it may have aliases + if (mode != RC_GR_WRITE && gconst_try_read(dst, r)) + src = dst; + + // if r will be modified, check for aliases being needed rsn + ali = tr->gregs & ~(1 << r); + if (mode != RC_GR_READ && src == dst && ali) { + int x = -1; + if (rcache_regs_nowsoon & ali) { + if (tr->type == HR_STATIC && guest_regs[r].sreg == dst && + !(tr->flags & HRF_LOCKED)) { + // split aliases if r is STATIC in sreg and dst isn't already locked + rcache_ref_vreg(dst); // lock to avoid evicting dst + if ((x = rcache_allocate_vreg(rcache_regs_nowsoon & ali)) >= 0) { + src = x; + rcache_move_vreg(src, dst); + } + rcache_unref_vreg(dst); + } else { + // split r + rcache_ref_vreg(src); // lock to avoid evicting src + if ((x = rcache_allocate_vreg(rcache_regs_nowsoon & (1 << r))) >= 0) { + dst = x; + tr = &cache_regs[dst]; + tr->stamp = rcache_counter; + } + rcache_unref_vreg(src); + } } + if (x < 0) + // aliases not needed or no vreg available, remove them + rcache_evict_vreg_aliases(dst, r); + else if (src != dst) + rcache_remove_vreg_alias(src, r); } - if (!tr) - tr = rcache_evict(); - - tr->type = HR_CACHED; - tr->gregs = 1 << r; - guest_regs[r].vreg = tr - cache_regs; - - if (mode != RC_GR_WRITE) { - if (gconst_try_read(guest_regs[r].vreg, r)) { - tr->flags |= HRF_DIRTY; - guest_regs[r].flags |= GRF_DIRTY; - } else if (split >= 0) { - if (hr) { - cache_regs[split].flags |= HRF_LOCKED; - *hr = cache_regs[split].hreg; - hr = NULL; - } else if (tr->hreg != cache_regs[split].hreg) - emith_move_r_r(tr->hreg, cache_regs[split].hreg); - } else - emith_ctx_read(tr->hreg, r * 4); - } + // assign r to dst + rcache_add_vreg_alias(dst, r); -end: - if (hr) - *hr = tr->hreg; + // handle dst register transfer + if (src < 0 && mode != RC_GR_WRITE) + emith_ctx_read(tr->hreg, r * 4); + if (hr) { + *hr = (src >= 0 ? cache_regs[src].hreg : tr->hreg); + rcache_ref_vreg(reg_map_host[*hr]); + } else if (src >= 0 && cache_regs[src].hreg != tr->hreg) + emith_move_r_r(tr->hreg, cache_regs[src].hreg); + + // housekeeping if (do_locking) - tr->flags |= HRF_LOCKED; - tr->stamp = rcache_counter; + rcache_ref_vreg(dst); if (mode != RC_GR_READ) { tr->flags |= HRF_DIRTY; guest_regs[r].flags |= GRF_DIRTY; gconst_kill(r); } - +#if DRC_DEBUG & 64 + RCACHE_CHECK("after getreg"); +#endif return tr->hreg; } @@ -1731,38 +1898,25 @@ static int rcache_get_reg(sh2_reg_e r, rc_gr_mode mode, int *hr) static int rcache_get_tmp(void) { - cache_reg_t *tr = NULL; int i; - // use any free reg, but prefer TEMP regs - for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { - if (cache_regs[i].type == HR_FREE || - (cache_regs[i].type == HR_TEMP && !(cache_regs[i].flags & HRF_LOCKED))) { - tr = &cache_regs[i]; - break; - } - } - - if (!tr) - tr = rcache_evict(); + i = rcache_allocate_temp(); + rcache_ref_vreg(i); - tr->type = HR_TEMP; - tr->flags |= HRF_LOCKED; - return tr->hreg; + cache_regs[i].type = HR_TEMP; + return cache_regs[i].hreg; } -static int rcache_get_hr_id(int hr) +static int rcache_get_vreg_hr(int hr) { int i; i = reg_map_host[hr]; - if (i < 0) // can't happen + if (i < 0 || (cache_regs[i].flags & HRF_LOCKED)) { + printf("host register %d is locked\n", hr); exit(1); + } -#if REMAP_REGISTER - if (cache_regs[i].type == HR_CACHED) - rcache_remap_vreg(i); -#endif if (cache_regs[i].type == HR_CACHED) rcache_evict_vreg(i); else if (cache_regs[i].type == HR_TEMP && (cache_regs[i].flags & HRF_LOCKED)) { @@ -1773,167 +1927,110 @@ static int rcache_get_hr_id(int hr) return i; } -static int rcache_get_arg_id(int arg) +static int rcache_get_vreg_arg(int arg) { int hr = 0; host_arg2reg(hr, arg); - return rcache_get_hr_id(hr); + return rcache_get_vreg_hr(hr); } // get a reg to be used as function arg static int rcache_get_tmp_arg(int arg) { - int id = rcache_get_arg_id(arg); - cache_regs[id].type = HR_TEMP; - cache_regs[id].flags |= HRF_LOCKED; + int x = rcache_get_vreg_arg(arg); + cache_regs[x].type = HR_TEMP; + rcache_ref_vreg(x); - return cache_regs[id].hreg; + return cache_regs[x].hreg; } // ... as return value after a call static int rcache_get_tmp_ret(void) { - int id = rcache_get_hr_id(RET_REG); - cache_regs[id].type = HR_TEMP; - cache_regs[id].flags |= HRF_LOCKED; + int x = rcache_get_vreg_hr(RET_REG); + cache_regs[x].type = HR_TEMP; + rcache_ref_vreg(x); - return cache_regs[id].hreg; + return cache_regs[x].hreg; } // same but caches a reg if access is readonly (announced by hr being NULL) static int rcache_get_reg_arg(int arg, sh2_reg_e r, int *hr) { - int i, srcr, dstr, dstid; - int dirty = 0, src_dirty = 0, is_const = 0, is_cached = 0; + int i, srcr, dstr, dstid, keep; u32 val; host_arg2reg(dstr, arg); i = guest_regs[r].vreg; if (i >= 0 && cache_regs[i].type == HR_CACHED && cache_regs[i].hreg == dstr) - // r is already in arg + // r is already in arg, avoid evicting dstid = i; else - dstid = rcache_get_arg_id(arg); + dstid = rcache_get_vreg_arg(arg); dstr = cache_regs[dstid].hreg; - if (rcache_hint & (1 << r)) { + if (rcache_is_cached(r)) { // r is needed later on anyway srcr = rcache_get_reg_(r, RC_GR_READ, 0, NULL); - is_cached = (cache_regs[reg_map_host[srcr]].type == HR_CACHED); - } else if (!(rcache_hint_clean & (1 << r)) && - (guest_regs[r].flags & GRF_CDIRTY) && gconst_get(r, &val)) { + keep = 1; + } else if ((guest_regs[r].flags & GRF_CDIRTY) && gconst_get(r, &val)) { // r has an uncomitted const - load into arg, but keep constant uncomitted srcr = dstr; - is_const = 1; - } else if ((i = guest_regs[r].vreg) >= 0) { - // maybe already cached? - srcr = cache_regs[i].hreg; - is_cached = (cache_regs[reg_map_host[srcr]].type == HR_CACHED); + emith_move_r_imm(srcr, val); + keep = 0; } else { - // must read either const or from ctx + // must read from ctx srcr = dstr; - if (rcache_static & (1 << r)) - srcr = rcache_get_reg_(r, RC_GR_READ, 0, NULL); - else if (gconst_try_read(dstid, r)) - dirty = 1; - else - emith_ctx_read(srcr, r * 4); + emith_ctx_read(srcr, r * 4); + keep = 1; } - if (is_cached) { - i = reg_map_host[srcr]; - if (srcr == dstr) { // evict aliases here since it is reallocated below - if (guest_regs[r].flags & GRF_STATIC) // move STATIC back to its sreg - rcache_clean_vreg(guest_regs[r].vreg); -#if REMAP_REGISTER - rcache_remap_vreg(i); -#endif - if (cache_regs[i].type == HR_CACHED) - rcache_evict_vreg(i); - } - else if (hr != NULL) // must lock srcr if not copied here - cache_regs[i].flags |= HRF_LOCKED; - if (guest_regs[r].flags & GRF_DIRTY) - src_dirty = 1; - } + if (cache_regs[dstid].type == HR_CACHED) + rcache_evict_vreg(dstid); cache_regs[dstid].type = HR_TEMP; - if (is_const) { - // uncomitted constant - emith_move_r_imm(srcr, val); - } else if (dstr != srcr) { - // arg is a copy of cached r - if (hr == NULL) + if (hr == NULL) { + if (dstr != srcr) + // arg is a copy of cached r emith_move_r_r(dstr, srcr); - } else if (hr != NULL) { - // caller will modify arg, so it will soon be out of sync with r - if (dirty || src_dirty) { - if (~rcache_hint_write & (1 << r)) { - emith_ctx_write(dstr, r * 4); // must clean since arg will be modified - guest_regs[r].flags &= ~GRF_DIRTY; - } - } + else if (keep && guest_regs[r].vreg < 0) + // keep arg as vreg for r + rcache_add_vreg_alias(dstid, r); } else { - // keep arg as vreg for r - cache_regs[dstid].type = HR_CACHED; - if (guest_regs[r].vreg < 0) { - cache_regs[dstid].gregs = 1 << r; - guest_regs[r].vreg = dstid; - } - if (dirty || src_dirty) { // mark as modifed for cleaning later on - cache_regs[dstid].flags |= HRF_DIRTY; - guest_regs[r].flags |= GRF_DIRTY; - } - } - - if (hr) *hr = srcr; + if (dstr != srcr) // must lock srcr if not copied here + rcache_ref_vreg(reg_map_host[srcr]); + } cache_regs[dstid].stamp = ++rcache_counter; - cache_regs[dstid].flags |= HRF_LOCKED; + rcache_ref_vreg(dstid); +#if DRC_DEBUG & 64 + RCACHE_CHECK("after getarg"); +#endif return dstr; } static void rcache_free_tmp(int hr) { int i = reg_map_host[hr]; + if (i < 0 || cache_regs[i].type != HR_TEMP) { printf("rcache_free_tmp fail: #%i hr %d, type %d\n", i, hr, cache_regs[i].type); - return; + exit(1); } - cache_regs[i].type = HR_FREE; - cache_regs[i].flags &= (HRF_REG|HRF_TEMP); + rcache_free_vreg(i); } // saves temporary result either in REG or in drctmp static int rcache_save_tmp(int hr) { - int i, free = -1, cached = -1; - u16 min_stamp = (u16)-1; + int i; // find REG, either free or unlocked temp or oldest non-hinted cached - for (i = 0; i < ARRAY_SIZE(cache_regs) && free < 0; i++) { - if ((cache_regs[i].flags & HRF_TEMP) || (cache_regs[i].flags & HRF_LOCKED)) - continue; - if (cache_regs[i].type == HR_FREE || cache_regs[i].type == HR_TEMP) - free = i; - if (cache_regs[i].type == HR_CACHED && - !((rcache_hint | rcache_locked) & cache_regs[i].gregs)) { - if (cache_regs[i].stamp < min_stamp) { - min_stamp = cache_regs[i].stamp; - cached = i; - } - } - } - - if (free >= 0) - i = free; - else if (cached >= 0) { - i = cached; - rcache_evict_vreg(i); - } else { + i = rcache_allocate_nontemp(); + if (i < 0) { // if none is available, store in drctmp emith_ctx_write(hr, offsetof(SH2, drc_tmp)); rcache_free_tmp(hr); @@ -1943,27 +2040,27 @@ static int rcache_save_tmp(int hr) cache_regs[i].type = HR_CACHED; cache_regs[i].gregs = 0; // not storing any guest register cache_regs[i].flags &= (HRF_TEMP|HRF_REG); - cache_regs[i].flags |= HRF_LOCKED; + cache_regs[i].ref = 0; cache_regs[i].stamp = ++rcache_counter; + rcache_ref_vreg(i); emith_move_r_r(cache_regs[i].hreg, hr); rcache_free_tmp(hr); return i; } -static int rcache_restore_tmp(int r) +static int rcache_restore_tmp(int x) { int hr; // find REG with tmp store: cached but with no gregs - if (r >= 0) { - if (cache_regs[r].type != HR_CACHED || cache_regs[r].gregs) { - printf("invalid tmp storage %d\n", r); + if (x >= 0) { + if (cache_regs[x].type != HR_CACHED || cache_regs[x].gregs) { + printf("invalid tmp storage %d\n", x); exit(1); } // found, transform to a TEMP - cache_regs[r].type = HR_TEMP; - cache_regs[r].flags |= HRF_LOCKED; - return cache_regs[r].hreg; + cache_regs[x].type = HR_TEMP; + return cache_regs[x].hreg; } // if not available, create a TEMP store and fetch from drctmp @@ -1973,51 +2070,57 @@ static int rcache_restore_tmp(int r) return hr; } -static void rcache_unlock(int hr) +static void rcache_free(int hr) { - if (hr >= 0) { - cache_regs[hr].flags &= ~HRF_LOCKED; - rcache_locked &= ~cache_regs[hr].gregs; + int x = reg_map_host[hr]; + if (cache_regs[x].type == HR_TEMP) + rcache_free_tmp(hr); + else + rcache_unref_vreg(x); +} + +static void rcache_unlock(int x) +{ + if (x >= 0) { + cache_regs[x].flags &= ~HRF_LOCKED; + cache_regs[x].ref = 0; +// rcache_regs_now &= ~cache_regs[x].gregs; } } static void rcache_unlock_all(void) { int i; - for (i = 0; i < ARRAY_SIZE(cache_regs); i++) + for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { cache_regs[i].flags &= ~HRF_LOCKED; + cache_regs[i].ref = 0; + } } -static inline void rcache_set_locked(u32 mask) -{ - rcache_locked = mask & ~rcache_static; -} - -static inline void rcache_set_hint_soon(u32 mask) +static inline void rcache_set_usage_now(u32 mask) { - rcache_hint_soon = mask & ~rcache_static; + rcache_regs_now = mask; } -static inline void rcache_set_hint_late(u32 mask) +static inline void rcache_set_usage_soon(u32 mask) { - rcache_hint_late = mask & ~rcache_static; + rcache_regs_soon = mask; } -static inline void rcache_set_hint_write(u32 mask) +static inline void rcache_set_usage_late(u32 mask) { - rcache_hint_write = mask & ~rcache_static; + rcache_regs_late = mask; } -static inline int rcache_is_hinted(sh2_reg_e r) +static inline void rcache_set_usage_discard(u32 mask) { - // consider static REGs as always hinted, since they are always there - return ((rcache_hint | rcache_static) & (1 << r)); + rcache_regs_discard = mask; } static inline int rcache_is_cached(sh2_reg_e r) { - // consider static REGs as always hinted, since they are always there - return (guest_regs[r].vreg >= 0); + // is r in cache or needed RSN? + return (guest_regs[r].vreg >= 0 || (rcache_regs_soonclean & (1 << r))); } static inline int rcache_is_hreg_used(int hr) @@ -2028,7 +2131,7 @@ static inline int rcache_is_hreg_used(int hr) (cache_regs[x].type != HR_TEMP || (cache_regs[x].flags & HRF_LOCKED)); } -static inline u32 rcache_used_hreg_mask(void) +static inline u32 rcache_used_hregs_mask(void) { u32 mask = 0; int i; @@ -2038,7 +2141,7 @@ static inline u32 rcache_used_hreg_mask(void) (cache_regs[i].type != HR_TEMP || (cache_regs[i].flags & HRF_LOCKED))) mask |= 1 << cache_regs[i].hreg; - return mask & ~rcache_static; + return mask; } static inline u32 rcache_dirty_mask(void) @@ -2054,13 +2157,13 @@ static inline u32 rcache_dirty_mask(void) return mask; } -static inline u32 rcache_reg_mask(void) +static inline u32 rcache_cached_mask(void) { u32 mask = 0; int i; for (i = 0; i < ARRAY_SIZE(cache_regs); i++) - if (cache_regs[i].type == HR_CACHED) + if (cache_regs[i].type == HR_CACHED || cache_regs[i].type == HR_STATIC) mask |= cache_regs[i].gregs; return mask; @@ -2070,26 +2173,40 @@ static void rcache_clean_tmp(void) { int i; + rcache_regs_clean = (1 << ARRAY_SIZE(guest_regs)) - 1; for (i = 0; i < ARRAY_SIZE(cache_regs); i++) - if (cache_regs[i].type == HR_CACHED && (cache_regs[i].flags & HRF_TEMP)) + if (cache_regs[i].type == HR_CACHED && (cache_regs[i].flags & HRF_TEMP)) { + rcache_unlock(i); #if REMAP_REGISTER rcache_remap_vreg(i); #else rcache_clean_vreg(i); #endif + } + rcache_regs_clean = 0; } -static void rcache_clean_mask(u32 mask) +static void rcache_clean_masked(u32 mask) { - int i; + int i, r, hr; - if (!(mask &= ~rcache_static)) + if (!(mask &= ~rcache_regs_static)) return; - rcache_hint_clean |= mask; - - // clean only vregs where all aliases are covered by the mask + rcache_regs_clean |= mask; + + // clean constants where all aliases are covered by the mask + for (i = 0; i < ARRAY_SIZE(gconsts); i++) + if ((gconsts[i].gregs & mask) && !(gconsts[i].gregs & ~mask)) { + FOR_ALL_BITS_SET_DO(gconsts[i].gregs, r, + if (guest_regs[r].flags & GRF_CDIRTY) { + hr = rcache_get_reg_(r, RC_GR_READ, 0, NULL); + rcache_clean_vreg(reg_map_host[hr]); + break; + }); + } + // clean vregs where all aliases are covered by the mask for (i = 0; i < ARRAY_SIZE(cache_regs); i++) - if (cache_regs[i].type == HR_CACHED && + if ((cache_regs[i].type == HR_CACHED || cache_regs[i].type == HR_STATIC) && (cache_regs[i].gregs & mask) && !(cache_regs[i].gregs & ~mask)) rcache_clean_vreg(i); } @@ -2099,9 +2216,30 @@ static void rcache_clean(void) int i; gconst_clean(); + rcache_regs_clean = (1 << ARRAY_SIZE(guest_regs)) - 1; for (i = ARRAY_SIZE(cache_regs)-1; i >= 0; i--) if (cache_regs[i].type == HR_CACHED || cache_regs[i].type == HR_STATIC) rcache_clean_vreg(i); + + // relocate statics to their sregs (necessary before conditional jumps) + for (i = 0; i < ARRAY_SIZE(guest_regs); i++) { + if ((guest_regs[i].flags & GRF_STATIC) && + guest_regs[i].vreg != guest_regs[i].sreg) { + rcache_ref_vreg(guest_regs[i].vreg); + rcache_evict_vreg(guest_regs[i].sreg); + rcache_unref_vreg(guest_regs[i].vreg); + if (guest_regs[i].vreg < 0) + emith_ctx_read(cache_regs[guest_regs[i].sreg].hreg, i*4); + else + emith_move_r_r(cache_regs[guest_regs[i].sreg].hreg, + cache_regs[guest_regs[i].vreg].hreg); + cache_regs[guest_regs[i].sreg].gregs = 1 << i; + cache_regs[guest_regs[i].sreg].flags |= HRF_DIRTY; + guest_regs[i].flags |= GRF_DIRTY; + guest_regs[i].vreg = guest_regs[i].sreg; + } + } + rcache_regs_clean = 0; } static void rcache_invalidate_tmp(void) @@ -2110,11 +2248,11 @@ static void rcache_invalidate_tmp(void) for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { if (cache_regs[i].flags & HRF_TEMP) { + rcache_unlock(i); if (cache_regs[i].type == HR_CACHED) - rcache_unmap_vreg(i); - cache_regs[i].type = HR_FREE; - cache_regs[i].flags &= (HRF_TEMP|HRF_REG); - cache_regs[i].gregs = 0; + rcache_evict_vreg(i); + else + rcache_free_vreg(i); } } } @@ -2122,33 +2260,26 @@ static void rcache_invalidate_tmp(void) static void rcache_invalidate(void) { int i; - gconst_invalidate(); - for (i = 0; i < ARRAY_SIZE(cache_regs); i++) { - cache_regs[i].flags &= (HRF_TEMP|HRF_REG); - if (cache_regs[i].type != HR_STATIC) - cache_regs[i].type = HR_FREE; - cache_regs[i].gregs = 0; - } + for (i = 0; i < ARRAY_SIZE(cache_regs); i++) + rcache_free_vreg(i); for (i = 0; i < ARRAY_SIZE(guest_regs); i++) { guest_regs[i].flags &= GRF_STATIC; if (!(guest_regs[i].flags & GRF_STATIC)) guest_regs[i].vreg = -1; else { - if (guest_regs[i].vreg < 0) - emith_ctx_read(cache_regs[guest_regs[i].sreg].hreg, i*4); - else if (guest_regs[i].vreg != guest_regs[i].sreg) - emith_move_r_r(cache_regs[guest_regs[i].sreg].hreg, - cache_regs[guest_regs[i].vreg].hreg); cache_regs[guest_regs[i].sreg].gregs = 1 << i; + cache_regs[guest_regs[i].sreg].flags |= HRF_DIRTY; + guest_regs[i].flags |= GRF_DIRTY; guest_regs[i].vreg = guest_regs[i].sreg; } } rcache_counter = 0; - rcache_hint_soon = rcache_hint_late = rcache_hint_write = rcache_hint_clean = 0; + rcache_regs_now = rcache_regs_soon = rcache_regs_late = 0; + rcache_regs_discard = rcache_regs_clean = 0; } static void rcache_flush(void) @@ -2171,7 +2302,7 @@ static void rcache_init(void) for (i = 0; i < ARRAY_SIZE(guest_regs); i++) if (guest_regs[i].flags & GRF_STATIC) { - rcache_static |= (1 << i); + rcache_regs_static |= (1 << i); guest_regs[i].sreg = reg_map_host[guest_regs[i].sreg]; cache_regs[guest_regs[i].sreg].type = HR_STATIC; } else @@ -2191,7 +2322,7 @@ static void rcache_init(void) // --------------------------------------------------------------- // NB may return either REG or TEMP -static int emit_get_rbase_and_offs(SH2 *sh2, sh2_reg_e r, int rmod, u32 *offs) +static int emit_get_rbase_and_offs(SH2 *sh2, sh2_reg_e r, int rmode, u32 *offs) { uptr omask = 0xff; // offset mask, XXX: ARM oriented.. u32 mask = 0; @@ -2225,21 +2356,19 @@ static int emit_get_rbase_and_offs(SH2 *sh2, sh2_reg_e r, int rmod, u32 *offs) return hr; } + // ROM, SDRAM. Host address should be mmapped to be equal to SH2 address. la = (uptr)*(void **)((char *)sh2 + poffs); - // accessing ROM or SDRAM, code location doesn't matter. The host address - // for these should be mmapped to be equal to the SH2 address. - // if r is in rcache or needed soon anyway, and offs is relative to region - // use rcached const to avoid loading a literal on ARM - if ((guest_regs[r].vreg >= 0 || ((guest_regs[r].flags & GRF_CDIRTY) && - ((rcache_hint_soon|rcache_hint_clean) & (1 << r)))) && !(*offs & ~mask)) { + + // if r is in rcache or needed soon anyway, and offs is relative to region, + // and address translation fits in add_ptr_imm (s32), then use rcached const + if (la == (s32)la && !(*offs & ~mask) && rcache_is_cached(r)) { u32 odd = a & 1; // need to fix odd address for correct byte addressing la -= (s32)((a & ~mask) - *offs - odd); // diff between reg and memory - // if reg is modified later on, allocate it RMW to remove aliases here - // else the aliases vreg stays locked and a vreg shortage may occur. - hr = hr2 = rcache_get_reg(r, rmod ? RC_GR_RMW : RC_GR_READ, NULL); + hr = hr2 = rcache_get_reg(r, rmode, NULL); if ((la & ~omask) - odd) { hr = rcache_get_tmp(); emith_add_r_r_ptr_imm(hr, hr2, (la & ~omask) - odd); + rcache_free(hr2); } *offs = (la & omask); } else { @@ -2285,39 +2414,55 @@ static void emit_move_r_imm32(sh2_reg_e dst, u32 imm) static void emit_move_r_r(sh2_reg_e dst, sh2_reg_e src) { - int hr_d, hr_s; - - if (guest_regs[src].vreg >= 0 || gconst_check(src) || rcache_is_hinted(src)) { - hr_s = rcache_get_reg(src, RC_GR_READ, NULL); + if (gconst_check(src) || rcache_is_cached(src)) { #if ALIAS_REGISTERS - // check for aliasing - int i = guest_regs[src].vreg; - if (guest_regs[dst].vreg != i) { - // remove possible old mapping of dst - if (guest_regs[dst].vreg >= 0) - rcache_remove_vreg_alias(guest_regs[dst].vreg, dst); - // make dst an alias of src - rcache_add_vreg_alias(i, dst); - cache_regs[i].flags |= HRF_DIRTY; - guest_regs[dst].flags |= GRF_DIRTY; - gconst_kill(dst); -#if PROPAGATE_CONSTANTS - gconst_copy(dst, src); -#endif - return; - } -#endif - hr_d = rcache_get_reg(dst, RC_GR_WRITE, NULL); + rcache_alias_vreg(dst, src); +#else + int hr_s = rcache_get_reg(src, RC_GR_READ, NULL); + int hr_d = rcache_get_reg(dst, RC_GR_WRITE, NULL); emith_move_r_r(hr_d, hr_s); -#if PROPAGATE_CONSTANTS gconst_copy(dst, src); #endif } else { - hr_d = rcache_get_reg(dst, RC_GR_WRITE, NULL); + int hr_d = rcache_get_reg(dst, RC_GR_WRITE, NULL); emith_ctx_read(hr_d, src * 4); } } +static void emit_add_r_imm(sh2_reg_e r, u32 imm) +{ + u32 val; + int isgc = gconst_get(r, &val); + int hr, hr2; + + if (!isgc || rcache_is_cached(r)) { + // not constant, or r is already in cache + hr = rcache_get_reg(r, RC_GR_RMW, &hr2); + emith_add_r_r_imm(hr, hr2, imm); + rcache_free(hr2); + if (isgc) + gconst_set(r, val + imm); + } else + gconst_new(r, val + imm); +} + +static void emit_sub_r_imm(sh2_reg_e r, u32 imm) +{ + u32 val; + int isgc = gconst_get(r, &val); + int hr, hr2; + + if (!isgc || rcache_is_cached(r)) { + // not constant, or r is already in cache + hr = rcache_get_reg(r, RC_GR_RMW, &hr2); + emith_sub_r_r_imm(hr, hr2, imm); + rcache_free(hr2); + if (isgc) + gconst_set(r, val - imm); + } else + gconst_new(r, val - imm); +} + static void emit_sync_t_to_sr(void) { // avoid reloading SR from context if there's nothing to do @@ -2335,8 +2480,9 @@ static int emit_memhandler_read(int size) #ifndef DRC_SR_REG // must writeback cycles for poll detection stuff if (guest_regs[SHR_SR].vreg != -1) - rcache_evict_vreg(guest_regs[SHR_SR].vreg); + rcache_unmap_vreg(guest_regs[SHR_SR].vreg); #endif + rcache_invalidate_tmp(); if (size & MF_POLLING) switch (size & MF_SIZEMASK) { @@ -2351,7 +2497,6 @@ static int emit_memhandler_read(int size) case 2: emith_call(sh2_drc_read32); break; // 32 } - rcache_invalidate_tmp(); return rcache_get_tmp_ret(); } @@ -2362,16 +2507,15 @@ static void emit_memhandler_write(int size) rcache_clean_tmp(); #ifndef DRC_SR_REG if (guest_regs[SHR_SR].vreg != -1) - rcache_evict_vreg(guest_regs[SHR_SR].vreg); + rcache_unmap_vreg(guest_regs[SHR_SR].vreg); #endif + rcache_invalidate_tmp(); switch (size & MF_SIZEMASK) { case 0: emith_call(sh2_drc_write8); break; // 8 case 1: emith_call(sh2_drc_write16); break; // 16 case 2: emith_call(sh2_drc_write32); break; // 32 } - - rcache_invalidate_tmp(); } // rd = @(Rs,#offs); rd < 0 -> return a temp @@ -2389,12 +2533,13 @@ static int emit_memhandler_read_rr(SH2 *sh2, sh2_reg_e rd, sh2_reg_e rs, u32 off emit_move_r_imm32(rd, val); hr2 = rcache_get_reg(rd, RC_GR_RMW, NULL); } - if ((size & MF_POSTINCR) && gconst_get(rs, &val)) - gconst_new(rs, val + (1 << (size & MF_SIZEMASK))); + if (size & MF_POSTINCR) + emit_add_r_imm(rs, 1 << (size & MF_SIZEMASK)); return hr2; } - hr = emit_get_rbase_and_offs(sh2, rs, size & MF_POSTINCR, &offs); + val = size & MF_POSTINCR; + hr = emit_get_rbase_and_offs(sh2, rs, val ? RC_GR_RMW : RC_GR_READ, &offs); if (hr != -1) { if (rd == SHR_TMP) hr2 = rcache_get_tmp(); @@ -2405,24 +2550,14 @@ static int emit_memhandler_read_rr(SH2 *sh2, sh2_reg_e rd, sh2_reg_e rs, u32 off case 1: emith_read16s_r_r_offs(hr2, hr, offs); break; // 16 case 2: emith_read_r_r_offs(hr2, hr, offs); emith_ror(hr2, hr2, 16); break; } - if (cache_regs[reg_map_host[hr]].type == HR_TEMP) // may also return REG - rcache_free_tmp(hr); - if (size & MF_POSTINCR) { - int isgc = gconst_get(rs, &val); - if (!isgc || guest_regs[rs].vreg >= 0) { - // already loaded - hr = rcache_get_reg(rs, RC_GR_RMW, NULL); - emith_add_r_r_imm(hr, hr, 1 << (size & MF_SIZEMASK)); - if (isgc) - gconst_set(rs, val + (1 << (size & MF_SIZEMASK))); - } else - gconst_new(rs, val + (1 << (size & MF_SIZEMASK))); - } + rcache_free(hr); + if (size & MF_POSTINCR) + emit_add_r_imm(rs, 1 << (size & MF_SIZEMASK)); return hr2; } #endif - if (gconst_get(rs, &val) && guest_regs[rs].vreg < 0 && !(rcache_hint_soon & (1 << rs))) { + if (gconst_get(rs, &val) && !rcache_is_cached(rs)) { hr = rcache_get_tmp_arg(0); emith_move_r_imm(hr, val + offs); if (size & MF_POSTINCR) @@ -2432,6 +2567,8 @@ static int emit_memhandler_read_rr(SH2 *sh2, sh2_reg_e rd, sh2_reg_e rs, u32 off hr2 = rcache_get_reg(rs, RC_GR_RMW, NULL); emith_add_r_r_imm(hr, hr2, offs); emith_add_r_imm(hr2, 1 << (size & MF_SIZEMASK)); + if (gconst_get(rs, &val)) + gconst_set(rs, val + (1 << (size & MF_SIZEMASK))); } else { hr = rcache_get_reg_arg(0, rs, &hr2); if (offs || hr != hr2) @@ -2463,30 +2600,34 @@ static void emit_memhandler_write_rr(SH2 *sh2, sh2_reg_e rd, sh2_reg_e rs, u32 o u32 val; if (rd == SHR_TMP) { - host_arg2reg(hr2, 1); + host_arg2reg(hr2, 1); // already locked and prepared by caller } else if ((size & MF_PREDECR) && rd == rs) { // must avoid caching rd in arg1 hr2 = rcache_get_reg_arg(1, rd, &hr); - if (hr != hr2) emith_move_r_r(hr2, hr); + if (hr != hr2) { + emith_move_r_r(hr2, hr); + rcache_free(hr2); + } } else hr2 = rcache_get_reg_arg(1, rd, NULL); + if (rd != SHR_TMP) + rcache_unlock(guest_regs[rd].vreg); // unlock in case rd is in arg0 - if (gconst_get(rs, &val) && guest_regs[rs].vreg < 0 && !(rcache_hint_soon & (1 << rs))) { + if (gconst_get(rs, &val) && !rcache_is_cached(rs)) { + hr = rcache_get_tmp_arg(0); if (size & MF_PREDECR) { val -= 1 << (size & MF_SIZEMASK); gconst_new(rs, val); } - hr = rcache_get_tmp_arg(0); emith_move_r_imm(hr, val + offs); } else if (offs || (size & MF_PREDECR)) { - if (size & MF_PREDECR) { - hr = rcache_get_reg(rs, RC_GR_RMW, &hr2); - emith_sub_r_r_imm(hr, hr2, 1 << (size & MF_SIZEMASK)); - } + if (size & MF_PREDECR) + emit_sub_r_imm(rs, 1 << (size & MF_SIZEMASK)); + rcache_unlock(guest_regs[rs].vreg); // unlock in case rs is in arg0 hr = rcache_get_reg_arg(0, rs, &hr2); if (offs || hr != hr2) emith_add_r_r_imm(hr, hr2, offs); } else - rcache_get_reg_arg(0, rs, NULL); + hr = rcache_get_reg_arg(0, rs, NULL); emit_memhandler_write(size); } @@ -2696,7 +2837,7 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) if (op_flags[i] & OF_BTARGET) ADD_TO_ARRAY(branch_target_pc, branch_target_count, pc, ); if (ops[i].op == OP_LDC && (ops[i].dest & BITMASK1(SHR_SR)) && pc+2 < end_pc) - op_flags[i+1] |= OF_BTARGET; // RTE entrypoint in case of SR(IMASK) change + op_flags[i+1] |= OF_BTARGET; // RTE entrypoint in case of SR.IMASK change #if LOOP_DETECTION // loop types detected: // 1. target: ... BRA target -> idle loop @@ -2855,10 +2996,6 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) drcf.polling = (drcf.loop_type == OF_POLL_LOOP ? MF_POLLING : 0); #endif -#if (DRC_DEBUG & ~7) - // must update PC - emit_move_r_imm32(SHR_PC, pc); -#endif rcache_clean(); #if (DRC_DEBUG & 0x10) @@ -2902,17 +3039,20 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) sr = rcache_get_reg(SHR_SR, RC_GR_RMW, NULL); emith_sync_t(sr); rcache_clean(); - tmp = rcache_used_hreg_mask(); + tmp = rcache_used_hregs_mask(); emith_save_caller_regs(tmp); emit_do_static_regs(1, 0); rcache_get_reg_arg(2, SHR_SR, NULL); tmp2 = rcache_get_tmp_arg(0); tmp3 = rcache_get_tmp_arg(1); + tmp4 = rcache_get_tmp_arg(3); emith_move_r_ptr_imm(tmp2, tcache_ptr); - emith_move_r_r_ptr(tmp3,CONTEXT_REG); + emith_move_r_r_ptr(tmp3, CONTEXT_REG); + emith_move_r_imm(tmp4, pc); + emith_ctx_write(tmp4, SHR_PC * 4); + rcache_invalidate_tmp(); emith_call(sh2_drc_log_entry); emith_restore_caller_regs(tmp); - rcache_invalidate_tmp(); #endif do_host_disasm(tcache_id); @@ -2924,9 +3064,10 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) sr = rcache_get_reg(SHR_SR, RC_GR_RMW, NULL); FLUSH_CYCLES(sr); emith_sync_t(sr); + emit_move_r_imm32(SHR_PC, pc); rcache_clean(); - tmp = rcache_used_hreg_mask(); + tmp = rcache_used_hregs_mask(); emith_save_caller_regs(tmp); emit_do_static_regs(1, 0); emith_pass_arg_r(0, CONTEXT_REG); @@ -2990,43 +3131,33 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) // dbg(1, "unhandled delay_dep_fw: %x", delay_dep_fw & ~BITMASK1(SHR_T)); if (delay_dep_bk & ~BITMASK2(SHR_PC, SHR_PR)) dbg(1, "unhandled delay_dep_bk: %x", delay_dep_bk); - rcache_set_hint_soon(0); - rcache_set_hint_late(0); - rcache_set_hint_write(0); } - else - { - // inform cache about future register usage - u32 late = 0; // regs read by future ops - u32 write = 0; // regs written to (to detect write before read) - u32 soon = 0; // regs read soon - tmp = (OP_ISBRANCH(opd[0].op) || opd[0].op == OP_RTE || // branching insns - opd[0].op == OP_TRAPA || opd[0].op == OP_UNDEFINED); - for (v = 1; v <= 9; v++) { - // no sense in looking any further than the next rcache flush - if (pc + 2*v < end_pc && !(op_flags[i+v] & OF_BTARGET) && - (!tmp || (op_flags[i+v] & OF_DELAY_OP))) { - late |= opd[v].source & ~write; - // ignore source regs after they have been written to - write |= opd[v].dest; - } else { - // upcoming rcache_flush, start writing back unused dirty stuff - tmp2 = write|opd[0].source|opd[0].dest; // insn may change reg aliases - rcache_clean_mask(rcache_dirty_mask() & ~tmp2); - break; - } - tmp |= (OP_ISBRANCH(opd[v].op) || opd[v].op == OP_RTE || - opd[v].op == OP_TRAPA || opd[v].op == OP_UNDEFINED); + + // inform cache about future register usage + u32 late = 0; // regs read by future ops + u32 write = 0; // regs written to (to detect write before read) + u32 soon = 0; // regs read soon + for (v = 1; v <= 9; v++) { + // no sense in looking any further than the next rcache flush + tmp = ((op_flags[i+v] & OF_BTARGET) || (op_flags[i+v-1] & OF_DELAY_OP) || + (OP_ISBRACND(opd[v-1].op) && !(op_flags[i+v] & OF_DELAY_OP))); + if (pc + 2*v <= end_pc && !tmp) { // (pc already incremented above) + late |= opd[v].source & ~write; + // ignore source regs after they have been written to + write |= opd[v].dest; // regs needed in the next few instructions if (v <= 4) soon = late; + } else { + // upcoming rcache_flush, start writing back unused dirty stuff + rcache_clean_masked(rcache_dirty_mask() & ~(write|opd[0].dest)); + break; } - rcache_set_hint_soon(late); // insns 1-3 - rcache_set_hint_late(late & ~soon); // insns 4-9 - rcache_set_hint_write(write & ~(late|soon) & ~opd[0].source); - // overwritten without being used } - rcache_set_locked(opd[0].source); // try not to evict src regs for this op + rcache_set_usage_now(opd[0].source); // current insn + rcache_set_usage_soon(late); // insns 1-3 + rcache_set_usage_late(late & ~soon); // insns 4-9 + rcache_set_usage_discard(write & ~(late|soon) & ~opd[0].source); switch (opd->op) { @@ -3069,7 +3200,8 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) case OP_RTE: // RTE 0000000000101011 emith_invalidate_t(); // pop PC - emit_memhandler_read_rr(sh2, SHR_PC, SHR_SP, 0, 2 | MF_POSTINCR); + tmp = emit_memhandler_read_rr(sh2, SHR_PC, SHR_SP, 0, 2 | MF_POSTINCR); + rcache_free(tmp); // pop SR tmp = emit_memhandler_read_rr(sh2, SHR_TMP, SHR_SP, 0, 2 | MF_POSTINCR); sr = rcache_get_reg(SHR_SR, RC_GR_RMW, NULL); @@ -3853,11 +3985,10 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) ///////////////////////////////////////////// case 0x07: // ADD #imm,Rn 0111nnnniiiiiiii - tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW, &tmp2); - if (op & 0x80) { // adding negative - emith_sub_r_r_imm(tmp, tmp2, -op & 0xff); - } else - emith_add_r_r_imm(tmp, tmp2, op & 0xff); + if (op & 0x80) // adding negative + emit_sub_r_imm(GET_Rn(), (u8)-op); + else + emit_add_r_imm(GET_Rn(), (u8)op); goto end_op; ///////////////////////////////////////////// @@ -3968,6 +4099,9 @@ static void REGPARM(2) *sh2_translate(SH2 *sh2, int tcache_id) end_op: rcache_unlock_all(); +#if DRC_DEBUG & 64 + RCACHE_CHECK("after insn"); +#endif cycles += opd->cycles; @@ -4007,6 +4141,7 @@ end_op: // idle or delay loop emit_sync_t_to_sr(); emith_sh2_delay_loop(cycles, drcf.delay_reg); + rcache_unlock_all(); // may lock delay_reg drcf.polling = drcf.loop_type = 0; } #endif @@ -4075,8 +4210,8 @@ end_op: emith_jump_cond_patchable(cond, target); } else if (target != NULL) { - emith_jump_patchable(target); rcache_invalidate(); + emith_jump_patchable(target); } // branch not taken, correct cycle count @@ -4099,6 +4234,7 @@ end_op: emith_sync_t(sr); rcache_clean(); tmp = rcache_get_reg_arg(0, SHR_PC, NULL); + rcache_invalidate(); #if CALL_STACK struct op_data *opd_b = (op_flags[i] & OF_DELAY_OP) ? opd-1 : opd; if (opd_b->rm == SHR_PR) { @@ -4108,6 +4244,7 @@ end_op: // JSR/BSRF tmp = rcache_get_tmp_arg(1); emith_call_link(tmp, sh2_drc_dispatcher_call); + rcache_free(tmp); } else #endif if (gconst_get(SHR_PC, &target_pc)) { @@ -4118,7 +4255,6 @@ end_op: // JMP emith_jump(sh2_drc_dispatcher); } - rcache_invalidate(); drcf.pending_branch_indirect = 0; drcf.polling = drcf.loop_type = 0; } @@ -4147,8 +4283,8 @@ end_op: target = dr_prepare_ext_branch(block->entryp, pc, sh2->is_slave, tcache_id); if (target == NULL) return NULL; - emith_jump_patchable(target); rcache_invalidate(); + emith_jump_patchable(target); } else rcache_flush(); emith_flush(); @@ -4452,14 +4588,14 @@ static void sh2_generate_utils(void) tmp = rcache_get_reg_arg(1, SHR_SR, NULL); emith_clear_msb(tmp, tmp, 22); emith_move_r_r_ptr(arg2, CONTEXT_REG); - emith_call(p32x_sh2_write32); // XXX: use sh2_drc_write32? rcache_invalidate(); + emith_call(p32x_sh2_write32); // XXX: use sh2_drc_write32? // push PC rcache_get_reg_arg(0, SHR_SP, NULL); emith_ctx_read(arg1, SHR_PC * 4); emith_move_r_r_ptr(arg2, CONTEXT_REG); - emith_call(p32x_sh2_write32); rcache_invalidate(); + emith_call(p32x_sh2_write32); // update I, cycles, do callback emith_ctx_read(arg1, offsetof(SH2, pending_level)); sr = rcache_get_reg(SHR_SR, RC_GR_RMW, NULL); @@ -4476,8 +4612,8 @@ static void sh2_generate_utils(void) if (arg0 != RET_REG) emith_move_r_r(arg0, RET_REG); emith_call_cleanup(); - emith_jump(sh2_drc_dispatcher); rcache_invalidate(); + emith_jump(sh2_drc_dispatcher); emith_flush(); // sh2_drc_entry(SH2 *sh2) -- 2.39.5