From 4a0351ab45fc3772a1c16f8f0efc0f1037df6414 Mon Sep 17 00:00:00 2001 From: Michael Meissner Date: Tue, 21 Nov 1995 21:41:25 +0000 Subject: [PATCH] Add floating point model specific support; Redo method model specific support is done; Add remaining floating add/subtract-multiply --- sim/ppc/ChangeLog | 12 + sim/ppc/igen.c | 58 +- sim/ppc/ppc-instructions | 1122 +++++++++++++++++++++++++++++++------- 3 files changed, 949 insertions(+), 243 deletions(-) diff --git a/sim/ppc/ChangeLog b/sim/ppc/ChangeLog index 72d86eb4e1..89d46361e3 100644 --- a/sim/ppc/ChangeLog +++ b/sim/ppc/ChangeLog @@ -1,3 +1,15 @@ +Tue Nov 21 16:31:25 1995 Michael Meissner + + * igen.c (insn_table_load_insns): Add support for model-static for + internal functions that should not be inlined. + (lf_print_c_semantic): Remove model_cleanup. + (gen_model_{c,h}): Ditto. + + * ppc-instructions: Redo model specific support once again. Add + floating point support to the model specific information. Flesh + out all of the floating mutiply add/subtract instructions. Add + better tracing support to the model specific information. + Sun Nov 19 23:00:52 1995 Michael Meissner * ppc-instructions (model data, model_busy): Rather than using a diff --git a/sim/ppc/igen.c b/sim/ppc/igen.c index 79010edcd2..57211db86d 100644 --- a/sim/ppc/igen.c +++ b/sim/ppc/igen.c @@ -634,6 +634,9 @@ static insn *last_model_function; static insn *model_internal; static insn *last_model_internal; +static insn *model_static; +static insn *last_model_static; + static insn *model_data; static insn *last_model_data; @@ -1033,6 +1036,9 @@ insn_table_load_insns(char *file_name) else if (it_is("model-internal", file_entry->fields[insn_flags])) { model_table_insert_specific(table, file_entry, &model_internal, &last_model_internal); } + else if (it_is("model-static", file_entry->fields[insn_flags])) { + model_table_insert_specific(table, file_entry, &model_static, &last_model_static); + } else if (it_is("model-data", file_entry->fields[insn_flags])) { model_table_insert_specific(table, file_entry, &model_data, &last_model_data); } @@ -1316,6 +1322,12 @@ lf_print_my_prefix(lf *file, (idecode ? "idecode" : "semantics"), file_entry->fields[insn_name], file_entry->line_nr); + lf_printf(file, "const itable_index my_index = "); + lf_print_function_name(file, + file_entry->fields[insn_name], + NULL, + function_name_prefix_itable); + lf_printf(file, ";\n"); } @@ -2133,16 +2145,6 @@ lf_print_c_semantic(lf *file, NULL, function_name_prefix_itable); lf_printf(file, ", processor, cia);\n"); - lf_printf(file, "\n"); - lf_printf(file, " if (WITH_MODEL_ISSUE)\n"); - lf_printf(file, " model_issue("); - lf_print_function_name(file, - instruction->file_entry->fields[insn_name], - NULL, - function_name_prefix_itable); - lf_printf(file, ", cpu_model(processor), cia, %s);\n", - ((idecode_cache) ? "cache_entry, 0" : "(idecode_cache *)0, instruction")); - lf_printf(file, "}\n"); lf_printf(file, "\n"); @@ -2177,7 +2179,6 @@ lf_print_c_semantic(lf *file, lf_printf(file, "\n"); } - /* the function footer */ lf_printf(file, "return nia;\n"); lf_indent(file, -2); lf_printf(file, "}\n"); @@ -2979,7 +2980,6 @@ gen_model_h(insn_table *table, lf *file) int model_create_p = 0; int model_init_p = 0; int model_halt_p = 0; - int model_issue_p = 0; int model_mon_info_p = 0; int model_mon_info_free_p = 0; @@ -3040,8 +3040,6 @@ gen_model_h(insn_table *table, lf *file) model_init_p = 1; else if (strcmp (name, "model_halt") == 0) model_halt_p = 1; - else if (strcmp (name, "model_issue") == 0) - model_issue_p = 1; else if (strcmp (name, "model_mon_info") == 0) model_mon_info_p = 1; else if (strcmp (name, "model_mon_info_free") == 0) @@ -3066,16 +3064,6 @@ gen_model_h(insn_table *table, lf *file) lf_printf(file, "\n"); } - if (!model_issue_p) { - lf_printf(file, "INLINE_MODEL void model_issue\n"); - lf_printf(file, "(itable_index index,\n"); - lf_printf(file, " model_data *model_ptr,\n"); - lf_printf(file, " unsigned_word cia,\n"); - lf_printf(file, " idecode_cache *cache_entry,\n"); - lf_printf(file, " instruction_word instruction);\n"); - lf_printf(file, "\n"); - } - if (!model_mon_info_p) { lf_printf(file, "INLINE_MODEL model_print *model_mon_info\n"); lf_printf(file, "(model_data *model_ptr);\n"); @@ -3170,7 +3158,6 @@ gen_model_c(insn_table *table, lf *file) int model_create_p = 0; int model_init_p = 0; int model_halt_p = 0; - int model_issue_p = 0; int model_mon_info_p = 0; int model_mon_info_free_p = 0; @@ -3191,10 +3178,18 @@ gen_model_c(insn_table *table, lf *file) model_c_or_h_data(table, file, insn_ptr->file_entry); } + for(insn_ptr = model_static; insn_ptr; insn_ptr = insn_ptr->next) { + model_c_or_h_function(table, file, insn_ptr->file_entry, "STATIC_MODEL"); + } + for(insn_ptr = model_internal; insn_ptr; insn_ptr = insn_ptr->next) { model_c_or_h_function(table, file, insn_ptr->file_entry, "STATIC_INLINE_MODEL"); } + for(insn_ptr = model_static; insn_ptr; insn_ptr = insn_ptr->next) { + model_c_function(table, file, insn_ptr->file_entry, "STATIC_MODEL"); + } + for(insn_ptr = model_internal; insn_ptr; insn_ptr = insn_ptr->next) { model_c_function(table, file, insn_ptr->file_entry, "STATIC_INLINE_MODEL"); } @@ -3208,8 +3203,6 @@ gen_model_c(insn_table *table, lf *file) model_init_p = 1; else if (strcmp (name, "model_halt") == 0) model_halt_p = 1; - else if (strcmp (name, "model_issue") == 0) - model_issue_p = 1; else if (strcmp (name, "model_mon_info") == 0) model_mon_info_p = 1; else if (strcmp (name, "model_mon_info_free") == 0) @@ -3238,17 +3231,6 @@ gen_model_c(insn_table *table, lf *file) lf_printf(file, "\n"); } - if (!model_issue_p) { - lf_printf(file, "INLINE_MODEL void model_issue(itable_index index,\n"); - lf_printf(file, " model_data *model_ptr,\n"); - lf_printf(file, " unsigned_word cia,\n"); - lf_printf(file, " idecode_cache *cache_entry,\n"); - lf_printf(file, " instruction_word instruction)\n"); - lf_printf(file, "{\n"); - lf_printf(file, "}\n"); - lf_printf(file, "\n"); - } - if (!model_mon_info_p) { lf_printf(file, "INLINE_MODEL model_print *model_mon_info(model_data *model_ptr)\n"); lf_printf(file, "{\n"); diff --git a/sim/ppc/ppc-instructions b/sim/ppc/ppc-instructions index e0f0a1d749..ffdcbeffd3 100644 --- a/sim/ppc/ppc-instructions +++ b/sim/ppc/ppc-instructions @@ -65,6 +65,7 @@ # # 6 Comma separated list of functional units + # PowerPC models ::model:604:ppc604: PPC_UNIT_BAD, PPC_UNIT_BAD, 1, 1, 0 ::model:603e:ppc603e:PPC_UNIT_BAD, PPC_UNIT_BAD, 1, 1, 0 @@ -95,29 +96,23 @@ unsigned32 flags; /* any flags that are needed */ }; - /* Register mappings */ - #define PPC_INT_REG 0 /* start of integer registers */ - #define PPC_FLOAT_REG (PPC_INT_REG + 32) /* start of floating point registers */ - #define PPC_CR_REG (PPC_FLOAT_REG + 32) /* start of CR0 .. CR7 */ - #define PPC_SPR_REG (PPC_CR_REG + 8) /* start of special purpose registers */ - #define PPC_FPSCR_REG (PPC_SPR_REG + 1024) /* start of fpscr register */ - #define NR_PPC_REGS (PPC_FPSCR_REG + 1) /* # of registers to allocate */ + /* Register mappings in status masks */ + #define PPC_CR_REG 0 /* start of CR0 .. CR7 */ + #define PPC_FPSCR_REG (PPC_CR_REG + 8) /* start of fpscr register */ - /* Structure for each register to indicate whether it is free or not */ - typedef struct _model_reg model_reg; - struct _model_reg { - model_reg *next; /* next register to be released */ - int in_use; /* non zero if register is used */ - }; + #define PPC_NO_SPR (-1) /* flag for no SPR register */ /* Structure for each functional unit that is busy */ typedef struct _model_busy model_busy; struct _model_busy { model_busy *next; /* next function unit */ - model_reg *reg; /* list of registers to release */ ppc_function_unit unit; /* function unit name */ - signed16 issue; /* # of cycles until unit can accept another insn */ - signed16 done; /* # of cycles until insn is done */ + unsigned32 int_busy; /* int registers that are busy */ + unsigned32 fp_busy; /* floating point registers that are busy */ + unsigned32 cr_fpscr_busy; /* CR/FPSCR registers that are busy */ + signed16 spr_busy; /* SPR register that is busy or PPC_NO_SPR */ + signed8 issue; /* # of cycles until unit can accept another insn */ + signed8 done; /* # of cycles until insn is done */ }; /* Structure to hold the current state information for the simulated CPU model */ @@ -127,7 +122,6 @@ const model_time *timing; /* timing information */ model_busy *busy_list; /* list of busy function units */ model_busy *free_list; /* list of model_busy structs not in use */ - model_reg registers[NR_PPC_REGS]; /* register status */ count_type nr_cycles; /* # cycles */ count_type nr_branches; /* # branches */ count_type nr_branches_fallthrough; /* # conditional branches that fell through */ @@ -136,10 +130,12 @@ count_type nr_stalls_data; /* # of stalls for data */ count_type nr_stalls_unit; /* # of stalls waiting for a function unit */ count_type nr_stalls_serialize; /* # of stalls waiting for things to quiet down */ - count_type nr_insns_not_handled; /* # of instructions not handled */ count_type nr_units[nr_ppc_function_units]; /* function unit counts */ - int insn_handled; /* whether last insn was handled */ - unsigned_1 busy[nr_ppc_function_units]; /* whether a function is busy or not */ + unsigned32 int_busy; /* int registers that are busy */ + unsigned32 fp_busy; /* floating point registers that are busy */ + unsigned32 cr_fpscr_busy; /* CR/FPSCR registers that are busy */ + unsigned8 spr_busy[nr_of_sprs]; /* SPR registers that are busy */ + unsigned8 busy[nr_ppc_function_units]; /* whether a function is busy or not */ }; STATIC_MODEL const char *const ppc_function_unit_name[ (int)nr_ppc_function_units ] = { @@ -154,15 +150,69 @@ "branch functional unit instruction", }; -# Code called after executing the semantics of the function -void::model-function::model_cleanup:itable_index index, model_data *model_ptr - if (model_ptr->insn_handled) - model_ptr->insn_handled = 0; - else { - model_ptr->nr_insns_not_handled++; - TRACE(trace_tbd,("model specific code for %s not done\n", itable[index].name)); + +# Trace releasing resources +void::model-static::model_trace_release:model_data *model_ptr, model_busy *busy + int i; + TRACE(trace_model,("done, %s\n", ppc_function_unit_name[busy->unit])); + if (busy->int_busy) { + for(i = 0; i < 32; i++) { + if (((1 << i) & busy->int_busy) != 0) { + TRACE(trace_model, ("Register r%d is now available.\n", i)); + } + } } + if (busy->fp_busy) { + for(i = 0; i < 32; i++) { + if (((1 << i) & busy->fp_busy) != 0) { + TRACE(trace_model, ("Register f%d is now available.\n", i)); + } + } + } + if (busy->cr_fpscr_busy) { + for(i = 0; i < 8; i++) { + if (((1 << i) & busy->cr_fpscr_busy) != 0) { + TRACE(trace_model, ("Register cr%d is now available.\n", i)); + } + } + if (busy->cr_fpscr_busy & 0x100) + TRACE(trace_model, ("Register fpscr is now available.\n")); + } + if (busy->spr_busy != PPC_NO_SPR) + TRACE(trace_model, ("Register %s is now available.\n", spr_name(busy->spr_busy))); +# Trace waiting for registers to become available +void::model-static::model_trace_busy_p:model_data *model_ptr, unsigned32 int_busy, unsigned32 fp_busy, unsigned32 cr_or_fpscr_busy, int spr_busy + int i; + if (int_busy) { + int_busy &= model_ptr->int_busy; + for(i = 0; i < 32; i++) { + if (((1 << i) & int_busy) != 0) { + TRACE(trace_model, ("Waiting for register r%d.\n", i)); + } + } + } + if (fp_busy) { + fp_busy &= model_ptr->fp_busy; + for(i = 0; i < 32; i++) { + if (((1 << i) & fp_busy) != 0) { + TRACE(trace_model, ("Waiting for register f%d.\n", i)); + } + } + } + if (cr_or_fpscr_busy) { + cr_or_fpscr_busy &= model_ptr->cr_fpscr_busy; + for(i = 0; i < 8; i++) { + if (((1 << i) & cr_or_fpscr_busy) != 0) { + TRACE(trace_model, ("Waiting for register cr%d.\n", i)); + } + } + if (cr_or_fpscr_busy & 0x100) + TRACE(trace_model, ("Waiting for register fpscr.\n")); + } + if (spr_busy != PPC_NO_SPR && model_ptr->spr_busy[spr_busy]) + TRACE(trace_model, ("Waiting for register %s.\n", spr_name(spr_busy))); + # Advance state to next cycle, releasing any registers allocated void::model-internal::model_new_cycle:model_data *model_ptr model_busy *cur_busy = model_ptr->busy_list; @@ -174,13 +224,15 @@ void::model-internal::model_new_cycle:model_data *model_ptr for ( ; cur_busy; cur_busy = next) { next = cur_busy->next; if (--cur_busy->done <= 0) { /* function unit done, release registers */ - model_reg *reg = cur_busy->reg; - TRACE(trace_model,("done, retiring %s\n", ppc_function_unit_name[cur_busy->unit])); - while (reg) { - TRACE(trace_model,("freeing up reg, address 0x%lx (%d)\n", (long)reg, reg - &model_ptr->registers[0])); - reg->in_use = 0; - reg = reg->next; - } + model_ptr->int_busy &= ~cur_busy->int_busy; + model_ptr->fp_busy &= ~cur_busy->fp_busy; + model_ptr->cr_fpscr_busy &= ~cur_busy->cr_fpscr_busy; + if (cur_busy->spr_busy != PPC_NO_SPR) + model_ptr->spr_busy[cur_busy->spr_busy] = 0; + + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_release(model_ptr, cur_busy); + model_ptr->busy[cur_busy->unit] = 0; cur_busy->next = free_list; free_list = cur_busy; @@ -204,12 +256,11 @@ void::model-internal::model_new_cycle:model_data *model_ptr model_ptr->busy_list = next_busy; model_ptr->free_list = free_list; -# Mark a function unit as busy, return the busy structure so regs can be added to be released +# Mark a function unit as busy, return the busy structure model_busy *::model-internal::model_make_busy:model_data *model_ptr, ppc_function_unit unit, int issue, int done model_busy *busy; TRACE(trace_model,("unit = %s, issue = %d, done = %d\n", ppc_function_unit_name[unit], issue, done)); - model_ptr->insn_handled = 1; if (!model_ptr->free_list) { busy = ZALLOC(model_busy); @@ -217,17 +268,45 @@ model_busy *::model-internal::model_make_busy:model_data *model_ptr, ppc_functio else { busy = model_ptr->free_list; model_ptr->free_list = busy->next; - busy->reg = (model_reg *)0; } busy->next = model_ptr->busy_list; busy->unit = unit; busy->issue = issue; busy->done = done; + busy->int_busy = 0; + busy->fp_busy = 0; + busy->cr_fpscr_busy = 0; + busy->spr_busy = PPC_NO_SPR; model_ptr->busy_list = busy; model_ptr->busy[unit] = 1; model_ptr->nr_units[unit]++; return busy; + +# Make a given integer register busy +void::model-internal::model_make_int_reg_busy:model_data *model_ptr, model_busy *busy_ptr, int regno + TRACE(trace_model,("Marking register r%d as busy\n", regno)); + busy_ptr->int_busy |= (1 << regno); + model_ptr->int_busy |= (1 << regno); +# Make a given floating point register busy +void::model-internal::model_make_fp_reg_busy:model_data *model_ptr, model_busy *busy_ptr, int regno + TRACE(trace_model,("Marking register f%d as busy\n", regno)); + busy_ptr->fp_busy |= (1 << regno); + model_ptr->fp_busy |= (1 << regno); + +# Make a given CR register busy +void::model-internal::model_make_cr_reg_busy:model_data *model_ptr, model_busy *busy_ptr, int regno + TRACE(trace_model,("Marking register cr%d as busy\n", regno)); + busy_ptr->cr_fpscr_busy |= (1 << regno); + model_ptr->cr_fpscr_busy |= (1 << regno); + +# Make a given SPR register busy +void::model-internal::model_make_spr_reg_busy:model_data *model_ptr, model_busy *busy_ptr, int regno + TRACE(trace_model,("Marking register %s as busy\n", spr_name(regno))); + busy_ptr->spr_busy = regno; + model_ptr->spr_busy[regno] = 1; + + # Wait until a function unit is non-busy, and then allocate a busy pointer & return the pointer model_busy *::model-internal::model_wait_for_unit:itable_index index, model_data *const model_ptr, const model_time *const time_ptr ppc_function_unit first_unit = time_ptr->first_unit; @@ -266,11 +345,13 @@ void::model-function::model_serialize:itable_index index, model_data *model_ptr # Wait for a CR to become unbusy void::model-function::model_wait_for_cr:model_data *model_ptr, unsigned CRBIT unsigned u; + unsigned32 cr_mask; int cr_var = 0; for (u = 0xc0000000; (u != 0) && (CRBIT & u) == 0; u >>= 4 ) cr_var++; - while (model_ptr->registers[cr_var + PPC_CR_REG].in_use) { + cr_mask = (1 << cr_var); + while ((model_ptr->cr_fpscr_busy & cr_mask) != 0) { TRACE(trace_model,("waiting for CR %d\n", cr_var)); model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); @@ -278,92 +359,144 @@ void::model-function::model_wait_for_cr:model_data *model_ptr, unsigned CRBIT # Schedule an instruction that takes 2 integer input registers and produces an output register & possibly sets CR0 void::model-function::ppc_insn_int2:itable_index index, cpu *processor, model_data *model_ptr, signed_word *rD, signed_word *rA, signed_word *rB, unsigned Rc + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA) | (1 << ppc_RB) | (1 << ppc_RD); + model_busy *busy_ptr; + if (!WITH_MODEL_ISSUE) return; - else { - registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; - model_busy *busy_ptr; - - if (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + else if (!Rc) { + if ((model_ptr->int_busy & int_mask) != 0) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RD].next = (model_reg *)0; - ppc_regs[ppc_RD].in_use = 1; - if (!Rc) - busy_ptr->reg = &ppc_regs[ppc_RD]; - else { - model_reg *reg_CR0 = &ppc_regs[0 + PPC_CR_REG]; - reg_CR0->next = &ppc_regs[ppc_RD]; - busy_ptr->reg = reg_CR0; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); + return; + } + + else { + const unsigned32 cr_mask = (1 << 0); + + if ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); + model_make_cr_reg_busy(model_ptr, busy_ptr, 0); + return; } # Schedule an instruction that takes 1 integer input registers and produces an output register & possibly sets CR0 void::model-function::ppc_insn_int1:itable_index index, cpu *processor, model_data *model_ptr, signed_word *rD, signed_word *rA, unsigned Rc + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA) | (1 << ppc_RD); + model_busy *busy_ptr; + if (!WITH_MODEL_ISSUE) return; - else { - registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; - model_busy *busy_ptr; - - if (ppc_regs[ppc_RA].in_use) { + else if (!Rc) { + if ((model_ptr->int_busy & int_mask) != 0) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RD].next = (model_reg *)0; - ppc_regs[ppc_RD].in_use = 1; - if (!Rc) - busy_ptr->reg = &ppc_regs[ppc_RD]; - else { - model_reg *reg_CR0 = &ppc_regs[0 + PPC_CR_REG]; - reg_CR0->next = &ppc_regs[ppc_RD]; - busy_ptr->reg = reg_CR0; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); + return; + } + + else { + const unsigned32 cr_mask = (1 << 0); + + if ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); + model_make_cr_reg_busy(model_ptr, busy_ptr, 0); + return; } # Schedule an instruction that takes no integer input registers and produces an output register & possibly sets CR0 void::model-function::ppc_insn_int0:itable_index index, cpu *processor, model_data *model_ptr, signed_word *rD, unsigned Rc + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RD); + model_busy *busy_ptr; + if (!WITH_MODEL_ISSUE) return; - else { - registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; - model_busy *busy_ptr; + else if (!Rc) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RD].next = (model_reg *)0; - ppc_regs[ppc_RD].in_use = 1; - if (!Rc) - busy_ptr->reg = &ppc_regs[ppc_RD]; - else { - model_reg *reg_CR0 = &ppc_regs[0 + PPC_CR_REG]; - reg_CR0->next = &ppc_regs[ppc_RD]; - busy_ptr->reg = reg_CR0; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); + return; + } + + else { + const unsigned32 cr_mask = (1 << 0); + + while ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); + model_make_cr_reg_busy(model_ptr, busy_ptr, 0); + return; } # Schedule an instruction that takes 2 integer input registers and produces an output register & updates a second register @@ -373,28 +506,27 @@ void::model-function::ppc_insn_int2_update:itable_index index, cpu *processor, m else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA) | (1 << ppc_RB) | (1 << ppc_RD); model_busy *busy_ptr; - if (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + if ((model_ptr->int_busy & int_mask) != 0) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RA].next = (model_reg *)0; - ppc_regs[ppc_RA].in_use = 1; - - ppc_regs[ppc_RD].next = &ppc_regs[ppc_RA]; - ppc_regs[ppc_RD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_RD]; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RA); + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); } # Schedule an instruction that takes 1 integer input registers and produces an output register & updates the other register @@ -404,27 +536,26 @@ void::model-function::ppc_insn_int1_update:itable_index index, cpu *processor, m else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA) | (1 << ppc_RD); model_busy *busy_ptr; - if (ppc_regs[ppc_RA].in_use) { + if ((model_ptr->int_busy & int_mask) != 0) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RA].next = (model_reg *)0; - ppc_regs[ppc_RA].in_use = 1; - - ppc_regs[ppc_RD].next = &ppc_regs[ppc_RA]; - ppc_regs[ppc_RD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_RD]; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RA); + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); } # Schedule an instruction that takes 2 integer input registers and produces no output register @@ -434,14 +565,17 @@ void::model-function::ppc_insn_int2_noout:itable_index index, cpu *processor, mo else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA) | (1 << ppc_RB); - if (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + if ((model_ptr->int_busy & int_mask) != 0) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } @@ -457,13 +591,16 @@ void::model-function::ppc_insn_int1_noout:itable_index index, cpu *processor, mo else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA); - if (ppc_regs[ppc_RA].in_use) { + if ((model_ptr->int_busy & int_mask) != 0) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use) { + while ((model_ptr->int_busy & int_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } @@ -488,25 +625,26 @@ void::model-function::ppc_insn_int2_cr:itable_index index, cpu *processor, model else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_CRD = CRD + PPC_CR_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA) | (1 << ppc_RB); + const unsigned32 cr_mask = (1 << CRD); model_busy *busy_ptr; - if (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + if ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use | ppc_regs[ppc_RB].in_use) { + while ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_CRD].next = (model_reg *)0; - ppc_regs[ppc_CRD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_CRD]; + model_make_cr_reg_busy(model_ptr, busy_ptr, CRD); } # Schedule an instruction that takes 1 integer input register and produces a CR output register @@ -516,24 +654,271 @@ void::model-function::ppc_insn_int1_cr:itable_index index, cpu *processor, model else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_CRD = CRD + PPC_CR_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RA); + const unsigned32 cr_mask = (1 << CRD); model_busy *busy_ptr; - if (ppc_regs[ppc_RA].in_use) { + if ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { model_new_cycle(model_ptr); /* don't count first dependency as a stall */ - while (ppc_regs[ppc_RA].in_use) { + while ((model_ptr->int_busy & int_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_CRD].next = (model_reg *)0; - ppc_regs[ppc_CRD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_CRD]; + model_make_cr_reg_busy(model_ptr, busy_ptr, CRD); + } + +# Schedule an instruction that takes 3 floating point input registers and produces an output register & possibly sets CR1 +void::model-function::ppc_insn_fp3:itable_index index, cpu *processor, model_data *model_ptr, unsigned64 *rD, unsigned64 *rA, unsigned64 *rB, unsigned64 *rC, unsigned Rc + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->fpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->fpr[0]); + const unsigned ppc_RC = (rC - &cpu_regs->fpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->fpr[0]); + const unsigned32 fp_mask = (1 << ppc_RA) | (1 << ppc_RB) | (1 << ppc_RC) | (1 << ppc_RD); + model_busy *busy_ptr; + + if (!WITH_MODEL_ISSUE) + return; + + else if (!Rc) { + if ((model_ptr->fp_busy & fp_mask) != 0) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->fp_busy & fp_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, 0, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + return; + } + + else { + const unsigned32 cr_mask = (1 << 1); + if ((model_ptr->fp_busy & fp_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->fp_busy & fp_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + model_make_cr_reg_busy(model_ptr, busy_ptr, 1); + return; + } + +# Schedule an instruction that takes 2 floating point input registers and produces an output register & possibly sets CR1 +void::model-function::ppc_insn_fp2:itable_index index, cpu *processor, model_data *model_ptr, unsigned64 *rD, unsigned64 *rA, unsigned64 *rB, unsigned Rc + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->fpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->fpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->fpr[0]); + const unsigned32 fp_mask = (1 << ppc_RA) | (1 << ppc_RB) | (1 << ppc_RD); + model_busy *busy_ptr; + + if (!WITH_MODEL_ISSUE) + return; + + else if (!Rc) { + if ((model_ptr->fp_busy & fp_mask) != 0) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->fp_busy & fp_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, 0, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + } + + else { + const unsigned32 cr_mask = (1 << 1); + + if ((model_ptr->fp_busy & fp_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->fp_busy & fp_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + model_make_cr_reg_busy(model_ptr, busy_ptr, 1); + } + +# Schedule an instruction that takes 1 floating point input registers and produces an output register & possibly sets CR1 +void::model-function::ppc_insn_fp1:itable_index index, cpu *processor, model_data *model_ptr, unsigned64 *rD, unsigned64 *rA, unsigned Rc + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->fpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->fpr[0]); + const unsigned32 fp_mask = (1 << ppc_RA) | (1 << ppc_RD); + model_busy *busy_ptr; + + if (!WITH_MODEL_ISSUE) + return; + + else if (!Rc) { + if ((model_ptr->fp_busy & fp_mask) != 0) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->fp_busy & fp_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, 0, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + } + + else { + const unsigned32 cr_mask = (1 << 1); + + if ((model_ptr->fp_busy & fp_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->fp_busy & fp_mask) || (model_ptr->cr_fpscr_busy & cr_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + model_make_cr_reg_busy(model_ptr, busy_ptr, 1); + } + +# Schedule an instruction that takes 1 floating point input register & 2 integer registers and does not produce an output +# (or takes 2 integer registers and produces an output in the floating point register) +void::model-function::ppc_insn_int2_fp1:itable_index index, cpu *processor, model_data *model_ptr, unsigned64 *rD, signed_word *rA, signed_word *rB, int RD_is_output, int RA_is_update + if (!WITH_MODEL_ISSUE) + return; + + else { + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->gpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->fpr[0]); + const unsigned32 int_mask = (1 << ppc_RB) | ((ppc_RA == 0) ? 0 : (1 << ppc_RA)); + const unsigned32 fp_mask = (1 << ppc_RD); + model_busy *busy_ptr; + + if ((model_ptr->int_busy & int_mask) || (model_ptr->fp_busy & fp_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->int_busy & int_mask) || (model_ptr->fp_busy & fp_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, fp_mask, 0, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + if (RD_is_output) + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + + if (RA_is_update) + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RA); + } + +# Schedule an instruction that takes 1 floating point input register & 1 integer register and does not produce an output +# (or takes 1 integer register and produces an output in the floating point register) +void::model-function::ppc_insn_int1_fp1:itable_index index, cpu *processor, model_data *model_ptr, unsigned64 *rD, signed_word *rA, int RD_is_output, int RA_is_update + if (!WITH_MODEL_ISSUE) + return; + + else { + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->gpr[0]); + const unsigned ppc_RD = (rD - &cpu_regs->fpr[0]); + const unsigned32 int_mask = ((ppc_RA == 0) ? 0 : (1 << ppc_RA)); + const unsigned32 fp_mask = (1 << ppc_RD); + model_busy *busy_ptr; + + if ((model_ptr->int_busy & int_mask) || (model_ptr->fp_busy & fp_mask)) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while ((model_ptr->int_busy & int_mask) || (model_ptr->fp_busy & fp_mask)) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, fp_mask, 0, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + if (RD_is_output) + model_make_fp_reg_busy(model_ptr, busy_ptr, ppc_RD); + + if (RA_is_update) + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RA); + } + +# Schedule an instruction that takes 2 floating input registers and produces a CR output register +void::model-function::ppc_insn_fp2_cr:itable_index index, cpu *processor, model_data *model_ptr, unsigned CRD, unsigned64 *rA, unsigned64 *rB + if (!WITH_MODEL_ISSUE) + return; + + else { + registers *cpu_regs = cpu_registers(processor); + const unsigned ppc_RA = (rA - &cpu_regs->fpr[0]); + const unsigned ppc_RB = (rB - &cpu_regs->fpr[0]); + const unsigned32 fp_mask = (1 << ppc_RA) | (1 << ppc_RB); + const unsigned32 cr_mask = (1 << CRD); + model_busy *busy_ptr; + + if (((model_ptr->fp_busy & fp_mask) | (model_ptr->cr_fpscr_busy & cr_mask)) != 0) { + model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + + while (((model_ptr->fp_busy & fp_mask) | (model_ptr->cr_fpscr_busy & cr_mask)) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, fp_mask, cr_mask, PPC_NO_SPR); + + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); + } + } + + busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); + model_make_cr_reg_busy(model_ptr, busy_ptr, CRD); } # Schedule an MFSPR instruction that takes 1 special purpose register and produces an integer output register @@ -543,20 +928,20 @@ void::model-function::ppc_insn_from_spr:itable_index index, cpu *processor, mode else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_SPR = nSPR + PPC_SPR_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RD); model_busy *busy_ptr; - while (ppc_regs[ppc_SPR].in_use) { + while ((model_ptr->int_busy & int_mask) != 0 || model_ptr->spr_busy[nSPR] != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, nSPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RD].next = (model_reg *)0; - ppc_regs[ppc_RD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_RD]; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); } # Schedule an MTSPR instruction that takes 1 integer register and produces a special purpose output register @@ -566,20 +951,20 @@ void::model-function::ppc_insn_to_spr:itable_index index, cpu *processor, model_ else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RS = (rS - &cpu_regs->gpr[0]) + PPC_INT_REG; - const unsigned ppc_SPR = nSPR + PPC_SPR_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RS = (rS - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RS); model_busy *busy_ptr; - while (ppc_regs[ppc_RS].in_use) { + while ((model_ptr->int_busy & int_mask) != 0 || model_ptr->spr_busy[nSPR] != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, 0, nSPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_SPR].next = (model_reg *)0; - ppc_regs[ppc_SPR].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_SPR]; + model_make_spr_reg_busy(model_ptr, busy_ptr, nSPR); } # Schedule a MFCR instruction that moves the CR into an integer regsiter @@ -589,59 +974,55 @@ void::model-function::ppc_insn_mfcr:itable_index index, cpu *processor, model_da else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RD = (rD - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RD); + const unsigned32 cr_mask = 0xff; model_busy *busy_ptr; - while (ppc_regs[0 + PPC_CR_REG].in_use - | ppc_regs[1 + PPC_CR_REG].in_use - | ppc_regs[2 + PPC_CR_REG].in_use - | ppc_regs[3 + PPC_CR_REG].in_use - | ppc_regs[4 + PPC_CR_REG].in_use - | ppc_regs[5 + PPC_CR_REG].in_use - | ppc_regs[6 + PPC_CR_REG].in_use - | ppc_regs[7 + PPC_CR_REG].in_use) { + while (((model_ptr->int_busy & int_mask) | (model_ptr->cr_fpscr_busy & cr_mask)) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_RD].next = (model_reg *)0; - ppc_regs[ppc_RD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_RD]; + model_make_int_reg_busy(model_ptr, busy_ptr, ppc_RD); } # Schedule a MTCR instruction that moves an integer register into the CR -void::model-function::ppc_insn_mtcr:itable_index index, cpu *processor, model_data *model_ptr, signed_word *rT +void::model-function::ppc_insn_mtcr:itable_index index, cpu *processor, model_data *model_ptr, signed_word *rT, unsigned FXM if (!WITH_MODEL_ISSUE) return; else { registers *cpu_regs = cpu_registers(processor); - const unsigned ppc_RT = (rT - &cpu_regs->gpr[0]) + PPC_INT_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_RT = (rT - &cpu_regs->gpr[0]); + const unsigned32 int_mask = (1 << ppc_RT); + const unsigned32 cr_mask = 0xff; + const model_time *normal_time = &model_ptr->timing[index]; + static const model_time ppc604_1bit_time = { PPC_UNIT_SCIU1, PPC_UNIT_SCIU2, 1, 1, 0 }; model_busy *busy_ptr; - model_reg *prev_reg; int i; - if (ppc_regs[ppc_RT].in_use) { - model_new_cycle(model_ptr); /* don't count first dependency as a stall */ + while (((model_ptr->int_busy & int_mask) | (model_ptr->cr_fpscr_busy & cr_mask)) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, int_mask, 0, cr_mask, PPC_NO_SPR); - while (ppc_regs[ppc_RT].in_use) { - model_ptr->nr_stalls_data++; - model_new_cycle(model_ptr); - } + model_ptr->nr_stalls_data++; + model_new_cycle(model_ptr); } - busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - prev_reg = (model_reg *)0; + /* If only one bit is being moved, use the SCIU, not the MCIU on the 604 */ + if (CURRENT_MODEL == MODEL_ppc604 && (FXM & (FXM-1)) == 0) { + normal_time = &ppc604_1bit_time; + } + + busy_ptr = model_wait_for_unit(index, model_ptr, normal_time); for (i = 0; i < 8; i++) { - ppc_regs[i + PPC_CR_REG].next = prev_reg; - ppc_regs[i + PPC_CR_REG].in_use = 1; - prev_reg = &ppc_regs[i + PPC_CR_REG]; + model_make_cr_reg_busy(model_ptr, busy_ptr, i); } - busy_ptr->reg = prev_reg; } # Convert a BIT32(x) number back into the original number @@ -661,21 +1042,21 @@ void::model-function::ppc_insn_cr2:itable_index index, cpu *processor, model_dat return; else { - const unsigned ppc_CRA = ppc_undo_bit32(crA_bit) + PPC_CR_REG; - const unsigned ppc_CRB = ppc_undo_bit32(crB_bit) + PPC_CR_REG; - const unsigned ppc_CRD = crD + PPC_CR_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned ppc_CRA = ppc_undo_bit32(crA_bit); + const unsigned ppc_CRB = ppc_undo_bit32(crB_bit); + const unsigned32 cr_mask = (1 << ppc_CRA) | (1 << ppc_CRB) | (1 << crD); model_busy *busy_ptr; - while (ppc_regs[ppc_CRA].in_use | ppc_regs[ppc_CRB].in_use) { + while ((model_ptr->cr_fpscr_busy & cr_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, 0, cr_mask, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_CRD].next = (model_reg *)0; - ppc_regs[ppc_CRD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_CRD]; + model_make_cr_reg_busy(model_ptr, busy_ptr, crD); } # Schedule an instruction that takes 1 CR input registers and produces an output CR register @@ -684,20 +1065,19 @@ void::model-function::ppc_insn_cr1:itable_index index, cpu *processor, model_dat return; else { - const unsigned ppc_CRA = CRA + PPC_CR_REG; - const unsigned ppc_CRD = crD + PPC_CR_REG; - model_reg *ppc_regs = model_ptr->registers; + const unsigned32 cr_mask = (1 << CRA) | (1 << crD); model_busy *busy_ptr; - while (ppc_regs[ppc_CRA].in_use) { + while ((model_ptr->cr_fpscr_busy & cr_mask) != 0) { + if (WITH_TRACE && ppc_trace[trace_model]) + model_trace_busy_p(model_ptr, 0, 0, cr_mask, PPC_NO_SPR); + model_ptr->nr_stalls_data++; model_new_cycle(model_ptr); } busy_ptr = model_wait_for_unit(index, model_ptr, &model_ptr->timing[index]); - ppc_regs[ppc_CRD].next = (model_reg *)0; - ppc_regs[ppc_CRD].in_use = 1; - busy_ptr->reg = &ppc_regs[ppc_CRD]; + model_make_cr_reg_busy(model_ptr, busy_ptr, crD); } model_data *::model-function::model_create:cpu *processor @@ -720,6 +1100,7 @@ model_print *::model-function::model_mon_info:model_data *model_ptr model_print *head; model_print *tail; ppc_function_unit i; + count_type nr_insns; head = tail = ZALLOC(model_print); tail->count = model_ptr->nr_cycles; @@ -754,15 +1135,6 @@ model_print *::model-function::model_mon_info:model_data *model_ptr tail->suffix_singular = " waiting for serialization"; } - if (model_ptr->nr_insns_not_handled) { - tail->next = ZALLOC(model_print); - tail = tail->next; - tail->count = model_ptr->nr_insns_not_handled; - tail->name = "instruction"; - tail->suffix_plural = "s that were not accounted for in timing info"; - tail->suffix_singular = " that was not accounted for in timing info"; - } - if (model_ptr->nr_branches) { tail->next = ZALLOC(model_print); tail = tail->next; @@ -799,8 +1171,10 @@ model_print *::model-function::model_mon_info:model_data *model_ptr tail->suffix_singular = ""; } + nr_insns = 0; for (i = PPC_UNIT_BAD; i < nr_ppc_function_units; i++) { if (model_ptr->nr_units[i]) { + nr_insns += model_ptr->nr_units[i]; tail->next = ZALLOC(model_print); tail = tail->next; tail->count = model_ptr->nr_units[i]; @@ -810,6 +1184,13 @@ model_print *::model-function::model_mon_info:model_data *model_ptr } } + tail->next = ZALLOC(model_print); + tail = tail->next; + tail->count = nr_insns; + tail->name = "instruction"; + tail->suffix_plural = "s that were accounted for in timing info"; + tail->suffix_singular = " that was accounted for in timing info"; + tail->next = (model_print *)0; return head; @@ -822,7 +1203,6 @@ void::model-function::model_mon_info_free:model_data *model_ptr, model_print *pt void::model-function::model_branches:model_data *model_ptr, int failed model_ptr->nr_units[PPC_UNIT_BPU]++; - model_ptr->insn_handled = 1; if (failed) model_ptr->nr_branches_fallthrough++; else @@ -835,6 +1215,7 @@ void::model-function::model_branch_predict:model_data *model_ptr, int success else model_ptr->nr_branch_predict_falses++; + # The following (illegal) instruction is `known' by gen and is # called when ever an illegal instruction is encountered ::internal::illegal @@ -3060,7 +3441,6 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, } ppc_insn_from_spr(my_index, processor, cpu_model(processor), rT, n); -# FIXME: 604 uses SCIU{1,2} if only one bit is being set 0.31,6.RS,11./,12.FXM,20./,21.144,31./:XFX::mtfcr:Move to Condition Register Fields *601: PPC_UNIT_IU, PPC_UNIT_IU, 2, 2, 0 *603: PPC_UNIT_SRU, PPC_UNIT_SRU, 1, 1, 0 @@ -3078,7 +3458,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, } CR = (MASKED(*rS, 32, 63) & mask) | (CR & ~mask); } - ppc_insn_mtcr(my_index, processor, cpu_model(processor), rS); + ppc_insn_mtcr(my_index, processor, cpu_model(processor), rS, FXM); 0.31,6.BF,9./,11./,16./,21.512,31./:X:::Move to Condition Register from XER @@ -3105,6 +3485,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + EXTS(D); *frT = DOUBLE(MEM(unsigned, EA, 4)); + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frT, rA, 0/*RD_is_output*/, 0/*RA_is_update*/); 0.31,6.FRT,11.RA,16.RB,21.535,31./:X:f::Load Floating-Point Single Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3117,6 +3498,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + *rB; *frT = DOUBLE(MEM(unsigned, EA, 4)); + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frT, rA, rB, 0/*RD_is_output*/, 0/*RA_is_update*/); 0.49,6.FRT,11.RA,16.D:D:f::Load Floating-Point Single with Update *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3130,6 +3512,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + EXTS(D); *frT = DOUBLE(MEM(unsigned, EA, 4)); *rA = EA; + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frT, rA, 0/*RD_is_output*/, 1/*RA_is_update*/); 0.31,6.FRT,11.RA,16.RB,21.576,31./:X:f::Load Floating-Point Single with Update Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3143,6 +3526,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + *rB; *frT = DOUBLE(MEM(unsigned, EA, 4)); *rA = EA; + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frT, rA, rB, 0/*RD_is_output*/, 1/*RA_is_update*/); 0.50,6.FRT,11.RA,16.D:D:f::Load Floating-Point Double *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3155,6 +3539,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + EXTS(D); *frT = MEM(unsigned, EA, 8); + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frT, rA, 0/*RD_is_output*/, 0/*RA_is_update*/); 0.31,6.FRT,11.RA,16.RB,21.599,31./:X:f::Load Floating-Point Double Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3167,6 +3552,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + *rB; *frT = MEM(unsigned, EA, 8); + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frT, rA, rB, 0/*RD_is_output*/, 0/*RA_is_update*/); 0.51,6.FRT,11.RA,16.D:D:f::Load Floating-Point Double with Update *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3180,6 +3566,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + EXTS(D); *frT = MEM(unsigned, EA, 8); *rA = EA; + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frT, rA, 0/*RD_is_output*/, 1/*RA_is_update*/); 0.31,6.FRT,11.RA,16.RB,21.631,31./:X:f::Load Floating-Point Double with Update Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 3, 3, 0 @@ -3193,6 +3580,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + *rB; *frT = MEM(unsigned, EA, 8); *rA = EA; + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frT, rA, rB, 0/*RD_is_output*/, 1/*RA_is_update*/); # @@ -3210,6 +3598,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + EXTS(D); STORE(EA, 4, SINGLE(*frS)); + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frS, rA, 1/*RD_is_output*/, 0/*RA_is_update*/); 0.31,6.FRS,11.RA,16.RB,21.663,31./:X:f::Store Floating-Point Single Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3222,6 +3611,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + *rB; STORE(EA, 4, SINGLE(*frS)); + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frS, rA, rB, 1/*RD_is_output*/, 0/*RA_is_update*/); 0.53,6.FRS,11.RA,16.D:D:f::Store Floating-Point Single with Update *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3235,6 +3625,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + EXTS(D); STORE(EA, 4, SINGLE(*frS)); *rA = EA; + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frS, rA, 1/*RD_is_output*/, 1/*RA_is_update*/); 0.31,6.FRS,11.RA,16.RB,21.695,31./:X:f::Store Floating-Point Single with Update Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3248,6 +3639,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + *rB; STORE(EA, 4, SINGLE(*frS)); *rA = EA; + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frS, rA, rB, 1/*RD_is_output*/, 1/*RA_is_update*/); 0.54,6.FRS,11.RA,16.D:D:f::Store Floating-Point Double *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3260,6 +3652,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + EXTS(D); STORE(EA, 8, *frS); + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frS, rA, 1/*RD_is_output*/, 0/*RA_is_update*/); 0.31,6.FRS,11.RA,16.RB,21.727,31./:X:f::Store Floating-Point Double Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3272,6 +3665,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, else b = *rA; EA = b + *rB; STORE(EA, 8, *frS); + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frS, rA, rB, 1/*RD_is_output*/, 0/*RA_is_update*/); 0.55,6.FRS,11.RA,16.D:D:f::Store Floating-Point Double with Update *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3285,6 +3679,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + EXTS(D); STORE(EA, 8, *frS); *rA = EA; + ppc_insn_int1_fp1(my_index, processor, cpu_model(processor), frS, rA, 1/*RD_is_output*/, 1/*RA_is_update*/); 0.31,6.FRS,11.RA,16.RB,21.759,31./:X:f::Store Floating-Point Double with Update Indexed *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3298,6 +3693,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, EA = *rA + *rB; STORE(EA, 8, *frS); *rA = EA; + ppc_insn_int2_fp1(my_index, processor, cpu_model(processor), frS, rA, rB, 1/*RD_is_output*/, 1/*RA_is_update*/); # @@ -3311,6 +3707,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *frT = *frB; CR1_UPDATE(Rc); + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); 0.63,6.FRT,11./,16.FRB,21.40,31.Rc:X:f::Floating Negate *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3319,6 +3716,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *frT = *frB ^ BIT64(0); CR1_UPDATE(Rc); + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); 0.63,6.FRT,11./,16.FRB,21.264,31.Rc:X:f::Floating Absolute Value *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3327,6 +3725,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *frT = *frB & ~BIT64(0); CR1_UPDATE(Rc); + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); 0.63,6.FRT,11./,16.FRB,21.136,31.Rc:X:f::Floating Negative Absolute Value *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3335,6 +3734,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *frT = *frB | BIT64(0); CR1_UPDATE(Rc); + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); @@ -3366,6 +3766,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frB, Rc); 0.59,6.FRT,11.FRA,16.FRB,21./,26.21,31.Rc:A:f:fadds:Floating Add Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3391,6 +3792,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frB, Rc); 0.63,6.FRT,11.FRA,16.FRB,21./,26.20,31.Rc:A:f:fsub:Floating Subtract *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3416,6 +3818,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frB, Rc); 0.59,6.FRT,11.FRA,16.FRB,21./,26.20,31.Rc:A:f:fsubs:Floating Subtract Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3440,6 +3843,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, float s = *(double*)frA - *(double*)frB; *(double*)frT = s; } + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frB, Rc); FPSCR_END(Rc); 0.63,6.FRT,11.FRA,16./,21.FRC,26.25,31.Rc:A:f:fmul:Floating Multiply @@ -3466,6 +3870,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frC, Rc); 0.59,6.FRT,11.FRA,16./,21.FRC,26.25,31.Rc:A:f:fmuls:Floating Multiply Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3491,6 +3896,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frC, Rc); 0.63,6.FRT,11.FRA,16.FRB,21./,26.18,31.Rc:A:f:fdiv:Floating Divide *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 31, 31, 0 @@ -3516,6 +3922,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frB, Rc); 0.59,6.FRT,11.FRA,16.FRB,21./,26.18,31.Rc:A:f:fdivs:Floating Divide Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 17, 17, 0 @@ -3541,6 +3948,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp2(my_index, processor, cpu_model(processor), frT, frA, frB, Rc); 0.63,6.FRT,11.FRA,16.FRB,21.FRC,26.29,31.Rc:A:f:fmadd:Floating Multiply-Add *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 5, 5, 0 @@ -3585,48 +3993,322 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *(double*)frT = s; } FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.59,6.FRT,11.FRA,16.FRB,21.FRC,26.29,31.Rc:A:f::Floating Multiply-Add Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + float product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the add */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + float s = product + *(double*)frB; + *(double*)frT = (double)s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.63,6.FRT,11.FRA,16.FRB,21.FRC,26.28,31.Rc:A:f::Floating Multiply-Subtract *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 5, 5, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 2, 4, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 2, 4, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + double product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 0, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the subtract */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 0, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + double s = product - *(double*)frB; + *(double*)frT = s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.59,6.FRT,11.FRA,16.FRB,21.FRC,26.28,31.Rc:A:f::Floating Multiply-Subtract Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + float product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the subtract */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + float s = product - *(double*)frB; + *(double*)frT = (double)s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.63,6.FRT,11.FRA,16.FRB,21.FRC,26.31,31.Rc:A:f::Floating Negative Multiply-Add *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 5, 5, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 2, 4, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 2, 4, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + double product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 0, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the add */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 0, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + double s = -(product + *(double*)frB); + *(double*)frT = s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.59,6.FRT,11.FRA,16.FRB,21.FRC,26.31,31.Rc:A:f::Floating Negative Multiply-Add Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + float product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the add */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + float s = -(product + *(double*)frB); + *(double*)frT = (double)s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.63,6.FRT,11.FRA,16.FRB,21.FRC,26.30,31.Rc:A:f::Floating Negative Multiply-Subtract *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 5, 5, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 2, 4, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 2, 4, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + double product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 0, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the subtract */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 0, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + double s = -(product - *(double*)frB); + *(double*)frT = s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); 0.59,6.FRT,11.FRA,16.FRB,21.FRC,26.30,31.Rc:A:f::Floating Negative Multiply-Subtract Single *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 *603: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *603e:PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 *604: PPC_UNIT_FPU, PPC_UNIT_FPU, 1, 3, 0 + FPSCR_BEGIN; + float product; /*HACK! - incorrectly loosing precision ... */ + /* compute the multiply */ + if (is_invalid_operation(processor, cia, + *frA, *frC, + fpscr_vxsnan | fpscr_vximz, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + (unsigned64*)&product, *frA, 0, *frC, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + product = *(double*)frA * *(double*)frC; + } + /* compute the subtract */ + if (is_invalid_operation(processor, cia, + product, *frB, + fpscr_vxsnan | fpscr_vxisi, + 1, /*single?*/ + 0) /*negate?*/) { + invalid_arithemetic_operation(processor, cia, + frT, product, *frB, 0, + 0, /*instruction_is_frsp*/ + 0, /*instruction_is_convert_to_64bit*/ + 0, /*instruction_is_convert_to_32bit*/ + 0); /*single-precision*/ + } + else { + /*HACK!*/ + float s = -(product - *(double*)frB); + *(double*)frT = (double)s; + } + FPSCR_END(Rc); + ppc_insn_fp3(my_index, processor, cpu_model(processor), frT, frA, frB, frC, Rc); # @@ -3839,6 +4521,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, if (sign == 1) FPSCR_SET_FPRF(fpscr_rf_neg_normal_number); goto Done; Done: + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); 0.63,6.FRT,11./,16.FRB,21.814,31.Rc:X:64,f::Floating Convert To Integer Doubleword @@ -3856,6 +4539,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, frT, *frB, fpscr_rn_round_towards_zero, 32); FPSCR_END(Rc); + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); 0.63,6.FRT,11./,16.FRB,21.846,31.Rc:X:64,f::Floating Convert from Integer Doubleword int sign = EXTRACTED64(*frB, 0, 0); @@ -3884,6 +4568,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, goto Done; /**/ Done: + ppc_insn_fp1(my_index, processor, cpu_model(processor), frT, frB, Rc); # # I.4.6.7 Floating-Point Compare Instructions @@ -3909,6 +4594,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, if (is_SNaN(*frA, 0) || is_SNaN(*frB, 0)) FPSCR_OR_VX(fpscr_vxsnan); FPSCR_END(0); + ppc_insn_fp2_cr(my_index, processor, cpu_model(processor), BF, frA, frB); 0.63,6.BF,9./,11.FRA,16.FRB,21.32,31./:X:f:fcmpo:Floating Compare Ordered *601: PPC_UNIT_FPU, PPC_UNIT_FPU, 4, 4, 0 @@ -3936,6 +4622,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, FPSCR_OR_VX(fpscr_vxvc); } FPSCR_END(0); + ppc_insn_fp2_cr(my_index, processor, cpu_model(processor), BF, frA, frB); # @@ -3992,6 +4679,10 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, #if WITH_IDECODE_CACHE_SIZE cpu_flush_icache(processor); #endif + if (RA == 0) + ppc_insn_int1_noout(my_index, processor, cpu_model(processor), rB); + else + ppc_insn_int2_noout(my_index, processor, cpu_model(processor), rA, rB); 0.19,6./,11./,16./,21.150,31./:XL::isync:Instruction Synchronize *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -3999,6 +4690,7 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *603e:PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 3, 0 *604: PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 1, 0 cpu_synchronize_context(processor); + ppc_insn_int0_noout(my_index, processor, cpu_model(processor)); # @@ -4011,6 +4703,10 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *603e:PPC_UNIT_LSU, PPC_UNIT_LSU, 2, 2, 0 *604: PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 1, 0 TRACE(trace_tbd,("Data Cache Block Touch\n")); + if (RA == 0) + ppc_insn_int1_noout(my_index, processor, cpu_model(processor), rB); + else + ppc_insn_int2_noout(my_index, processor, cpu_model(processor), rA, rB); 0.31,6./,11.RA,16.RB,21.246,31./:X:::Data Cache Block Touch for Store *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -4018,6 +4714,10 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *603e:PPC_UNIT_LSU, PPC_UNIT_LSU, 2, 2, 0 *604: PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 3, 0 TRACE(trace_tbd,("Data Cache Block Touch for Store\n")); + if (RA == 0) + ppc_insn_int1_noout(my_index, processor, cpu_model(processor), rB); + else + ppc_insn_int2_noout(my_index, processor, cpu_model(processor), rA, rB); 0.31,6./,11.RA,16.RB,21.1014,31./:X:::Data Cache Block set to Zero *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -4025,6 +4725,10 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *603e:PPC_UNIT_LSU, PPC_UNIT_LSU, 10, 10, 0 *604: PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 3, 0 TRACE(trace_tbd,("Data Cache Block set to Zero\n")); + if (RA == 0) + ppc_insn_int1_noout(my_index, processor, cpu_model(processor), rB); + else + ppc_insn_int2_noout(my_index, processor, cpu_model(processor), rA, rB); 0.31,6./,11.RA,16.RB,21.54,31./:X:::Data Cache Block Store *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -4032,6 +4736,10 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *603e:PPC_UNIT_LSU, PPC_UNIT_LSU, 5, 5, 0 *604: PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 1, 0 TRACE(trace_tbd,("Data Cache Block Store\n")); + if (RA == 0) + ppc_insn_int1_noout(my_index, processor, cpu_model(processor), rB); + else + ppc_insn_int2_noout(my_index, processor, cpu_model(processor), rA, rB); 0.31,6./,11.RA,16.RB,21.86,31./:X:::Data Cache Block Flush *601: PPC_UNIT_IU, PPC_UNIT_IU, 1, 1, 0 @@ -4039,6 +4747,10 @@ void::function::invalid_arithemetic_operation:cpu *processor, unsigned_word cia, *603e:PPC_UNIT_LSU, PPC_UNIT_LSU, 5, 5, 0 *604: PPC_UNIT_LSU, PPC_UNIT_LSU, 1, 1, 0 TRACE(trace_tbd,("Data Cache Block Flush\n")); + if (RA == 0) + ppc_insn_int1_noout(my_index, processor, cpu_model(processor), rB); + else + ppc_insn_int2_noout(my_index, processor, cpu_model(processor), rA, rB); # # II.3.3 Enforce In-order Execution of I/O Instruction