view modules/bluespec/Pygar/lab4/Processor.bsv @ 65:cf8bb3038cbd pygar svn.66

[svn r66] sim passes
author punk
date Tue, 11 May 2010 09:05:22 -0400
parents 74716e9a81cc
children
line wrap: on
line source
1 /// The MIT License
3 // Copyright (c) 2009 Massachusetts Institute of Technology
5 // Permission is hereby granted, free of charge, to any person obtaining a copy
6 // of this software and associated documentation files (the "Software"), to deal
7 // in the Software without restriction, including without limitation the rights
8 // to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9 // copies of the Software, and to permit persons to whom the Software is
10 // furnished to do so, subject to the following conditions:
12 // The above copyright notice and this permission notice shall be included in
13 // all copies or substantial portions of the Software.
15 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 // IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 // FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 // AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 // LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
21 // THE SOFTWARE.
23 import Connectable::*;
24 import GetPut::*;
25 import ClientServer::*;
26 import RegFile::*;
28 import FIFO::*;
29 import FIFOF::*;
30 import SFIFO::*;
31 import RWire::*;
33 import BFIFO::*;
34 import MemTypes::*;
35 import ProcTypes::*;
36 import BRegFile::*;
37 import BranchPred::*;
38 //import PathTypes::*; This is only there to force the debugging
40 import Trace::*;
42 //AWB includes
43 `include "asim/provides/low_level_platform_interface.bsh"
44 `include "asim/provides/soft_connections.bsh"
45 `include "asim/provides/common_services.bsh"
47 // Local includes
48 `include "asim/provides/processor_library.bsh"
49 `include "asim/rrr/remote_server_stub_PROCESSORSYSTEMRRR.bsh"
50 `include "asim/provides/common_services.bsh"
51 `include "asim/dict/STATS_PROCESSOR.bsh"
53 interface ProcStats;
54 interface Get#(Stat) num_cycles;
55 interface Get#(Stat) num_inst;
56 endinterface
58 interface CPUToHost;
59 method Bit#(32) cpuToHost(int req);
60 endinterface
62 interface Proc;
64 // Interface from processor to caches
65 interface Client#(DataReq,DataResp) dmem_client;
66 interface Client#(InstReq,InstResp) imem_client;
68 // Interface for enabling/disabling statistics on the rest of the core
69 interface Get#(Bool) statsEn_get;
71 // Interface for collecting statistics.
72 interface ProcStats stats;
74 // Interface to host
75 interface CPUToHost tohost;
77 endinterface
80 typedef enum { PCgen, Exec, Writeback } Stage deriving(Eq,Bits);
82 //-----------------------------------------------------------
83 // Register file module
84 //-----------------------------------------------------------
86 interface BRFile;
87 method Action wr( Rindx rindx, Bit#(32) data );
88 method Bit#(32) rd1( Rindx rindx );
89 method Bit#(32) rd2( Rindx rindx );
90 endinterface
92 module mkBRFile( BRFile );
94 RegFile#(Rindx,Bit#(32)) rfile <- mkBRegFile();
96 method Action wr( Rindx rindx, Bit#(32) data );
97 rfile.upd( rindx, data );
98 endmethod
100 method Bit#(32) rd1( Rindx rindx );
101 return ( rindx == 0 ) ? 0 : rfile.sub(rindx);
102 endmethod
104 method Bit#(32) rd2( Rindx rindx );
105 return ( rindx == 0 ) ? 0 : rfile.sub(rindx);
106 endmethod
108 endmodule
110 //-----------------------------------------------------------
111 // Helper functions
112 //-----------------------------------------------------------
114 function Bit#(32) slt( Bit#(32) val1, Bit#(32) val2 );
115 return zeroExtend( pack( signedLT(val1,val2) ) );
116 endfunction
118 function Bit#(32) sltu( Bit#(32) val1, Bit#(32) val2 );
119 return zeroExtend( pack( val1 < val2 ) );
120 endfunction
122 function Bit#(32) rshft( Bit#(32) val );
123 return zeroExtend(val[4:0]);
124 endfunction
127 //-----------------------------------------------------------
128 // Find funct for wbQ
129 //-----------------------------------------------------------
130 function Bool findwbf(Rindx fVal, WBResult cmpVal);
131 case (cmpVal) matches
132 tagged WB_ALU {data:.res, dest:.rd} :
133 return (fVal == rd);
134 tagged WB_Load .rd :
135 return (fVal == rd);
136 tagged WB_Store .st :
137 return False;
138 tagged WB_Host .x :
139 return False;
140 endcase
141 endfunction
144 //-----------------------------------------------------------
145 // Stall funct for wbQ
146 //-----------------------------------------------------------
147 function Bool stall(Instr inst, SFIFO#(WBResult, Rindx) f);
148 case (inst) matches
149 // -- Memory Ops ------------------------------------------------
150 tagged LW .it :
151 return f.find(it.rbase);
152 tagged SW {rsrc:.dreg, rbase:.addr, offset:.o} :
153 return (f.find(addr) || f.find2(dreg));
155 // -- Simple Ops ------------------------------------------------
156 tagged ADDIU .it : return f.find(it.rsrc);
157 tagged SLTI .it : return f.find(it.rsrc);
158 tagged SLTIU .it : return f.find(it.rsrc);
159 tagged ANDI .it : return f.find(it.rsrc);
160 tagged ORI .it : return f.find(it.rsrc);
161 tagged XORI .it : return f.find(it.rsrc);
163 tagged LUI .it : return f.find(it.rdst); //this rds/wrs itself
164 tagged SLL .it : return f.find(it.rsrc);
165 tagged SRL .it : return f.find(it.rsrc);
166 tagged SRA .it : return f.find(it.rsrc);
167 tagged SLLV .it : return (f.find(it.rsrc) || f.find(it.rshamt));
168 tagged SRLV .it : return (f.find(it.rsrc) || f.find(it.rshamt));
169 tagged SRAV .it : return (f.find(it.rsrc) || f.find(it.rshamt));
170 tagged ADDU .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
171 tagged SUBU .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
172 tagged AND .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
173 tagged OR .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
174 tagged XOR .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
175 tagged NOR .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
176 tagged SLT .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
177 tagged SLTU .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
180 // -- Branches --------------------------------------------------
182 tagged BLEZ .it : return (f.find(it.rsrc));
183 tagged BGTZ .it : return (f.find(it.rsrc));
184 tagged BLTZ .it : return (f.find(it.rsrc));
185 tagged BGEZ .it : return (f.find(it.rsrc));
186 tagged BEQ .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
187 tagged BNE .it : return (f.find(it.rsrc1) || f.find2(it.rsrc2));
189 // -- Jumps -----------------------------------------------------
191 tagged J .it : return False;
192 tagged JR .it : return f.find(it.rsrc);
193 tagged JALR .it : return f.find(it.rsrc);
194 tagged JAL .it : return False;
196 // -- Cop0 ------------------------------------------------------
198 tagged MTC0 .it : return f.find(it.rsrc);
199 tagged MFC0 .it : return False;
201 // -- Illegal ---------------------------------------------------
203 default : return False;
205 endcase
206 endfunction
207 //-----------------------------------------------------------
208 // Reference processor
209 //-----------------------------------------------------------
212 //(* doc = "synthesis attribute ram_style mkProc distributed;" *)
213 //(* synthesize *)
215 module [CONNECTED_MODULE] mkProc( Proc );
217 //-----------------------------------------------------------
218 // Debug port
220 ServerStub_PROCESSORSYSTEMRRR server_stub <- mkServerStub_PROCESSORSYSTEMRRR();
223 //-----------------------------------------------------------
224 // State
226 // Standard processor state
228 Reg#(Addr) pc <- mkReg(32'h00001000);
229 Reg#(Epoch) epoch <- mkReg(0);
230 Reg#(Stage) stage <- mkReg(PCgen);
231 BRFile rf <- mkBRFile;
233 // Branch Prediction
234 BranchPred bp <- mkBranchPred();
235 FIFO#(PCStat) execpc <- mkLFIFO();
237 // Pipelines
238 FIFO#(PCStat) pcQ <-mkSizedFIFO(3);
239 SFIFO#(WBResult, Rindx) wbQ <-mkSFIFO(findwbf);
241 Reg#(Bit#(32)) cp0_tohost <- mkReg(0);
242 Reg#(Bit#(32)) cp0_fromhost <- mkReg(0);
243 Reg#(Bool) cp0_statsEn <- mkReg(False);
245 // Memory request/response state
247 FIFO#(InstReq) instReqQ <- mkBFIFO1();
248 FIFO#(InstResp) instRespQ <- mkFIFO();
250 FIFO#(DataReq) dataReqQ <- mkBFIFO1();
251 FIFO#(DataResp) dataRespQ <- mkFIFO();
253 // Statistics state
254 Reg#(Stat) num_cycles <- mkReg(0);
255 Reg#(Stat) num_inst <- mkReg(0);
257 //Or:
258 // Statistics state
259 //STAT num_cycles <- mkStatCounter(`STATS_PROCESSOR_CYCLE_COUNT);
260 //STAT num_inst <- mkStatCounter(`STATS_PROCESSOR_INST_COUNT);
262 //-----------------------------------------------------------
263 // Rules
265 (* descending_urgency = "exec, pcgen" *)
266 rule pcgen; //( stage == PCgen );
267 let pc_plus4 = pc + 4;
269 traceTiny("mkProc", "pc",pc);
270 traceTiny("mkProc", "pcgen","P");
271 instReqQ.enq( LoadReq{ addr:pc, tag:epoch} );
273 let next_pc = bp.get(pc);
274 if (next_pc matches tagged Valid .npc)
275 begin
276 pcQ.enq(PCStat {qpc:pc, qnxtpc:npc, qepoch:epoch});
277 pc <= npc;
278 end
279 else
280 begin
281 pcQ.enq(PCStat {qpc:pc, qnxtpc:pc_plus4, qepoch:epoch});
282 pc <= pc_plus4;
283 end
285 endrule
287 rule discard (instRespQ.first() matches tagged LoadResp .ld
288 &&& ld.tag != epoch);
289 traceTiny("mkProc", "stage", "D");
290 instRespQ.deq();
291 endrule
293 (* conflict_free = "exec, writeback" *)
294 rule exec (instRespQ.first() matches tagged LoadResp.ld
295 &&& (ld.tag == epoch)
296 &&& unpack(ld.data) matches .inst
297 &&& !stall(inst, wbQ));
299 // Some abbreviations
300 let sext = signExtend;
301 let zext = zeroExtend;
302 let sra = signedShiftRight;
304 // Get the instruction
306 instRespQ.deq();
307 Instr inst
308 = case ( instRespQ.first() ) matches
309 tagged LoadResp .ld : return unpack(ld.data);
310 tagged StoreResp .st : return ?;
311 endcase;
313 // Get the PC info
314 let instrpc = pcQ.first().qpc;
315 let pc_plus4 = instrpc + 4;
317 Bool branchTaken = False;
318 Addr newPC = pc_plus4;
320 // Tracing
321 traceTiny("mkProc", "exec","X");
322 traceTiny("mkProc", "exInstTiny",inst);
323 traceFull("mkProc", "exInstFull",inst);
325 case ( inst ) matches
327 // -- Memory Ops ------------------------------------------------
329 tagged LW .it :
330 begin
331 Addr addr = rf.rd1(it.rbase) + sext(it.offset);
332 dataReqQ.enq( LoadReq{ addr:addr, tag:zeroExtend(it.rdst) } );
333 wbQ.enq(tagged WB_Load it.rdst);
334 end
336 tagged SW .it :
337 begin
338 Addr addr = rf.rd1(it.rbase) + sext(it.offset);
339 dataReqQ.enq( StoreReq{ tag:0, addr:addr, data:rf.rd2(it.rsrc) } );
340 wbQ.enq(tagged WB_Store);
341 end
343 // -- Simple Ops ------------------------------------------------
345 tagged ADDIU .it :
346 begin
347 Bit#(32) result = rf.rd1(it.rsrc) + sext(it.imm);
348 wbQ.enq(tagged WB_ALU {data:result, dest:it.rdst});
349 end
350 tagged SLTI .it : wbQ.enq(tagged WB_ALU {dest:it.rdst, data:slt( rf.rd1(it.rsrc), sext(it.imm) )});
351 tagged SLTIU .it : wbQ.enq(tagged WB_ALU {dest:it.rdst, data:sltu( rf.rd1(it.rsrc), sext(it.imm) ) });
352 tagged ANDI .it :
353 begin
354 Bit#(32) zext_it_imm = zext(it.imm);
355 wbQ.enq(tagged WB_ALU {dest:it.rdst, data:(rf.rd1(it.rsrc) & zext_it_imm)} );
356 end
357 tagged ORI .it :
358 begin
359 Bit#(32) zext_it_imm = zext(it.imm);
360 wbQ.enq(tagged WB_ALU {dest:it.rdst, data:(rf.rd1(it.rsrc) | zext_it_imm)} );
361 end
362 tagged XORI .it :
363 begin
364 Bit#(32) zext_it_imm = zext(it.imm);
365 wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc) ^ zext_it_imm )});
366 end
367 tagged LUI .it :
368 begin
369 Bit#(32) zext_it_imm = zext(it.imm);
370 wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(zext_it_imm << 32'd16) });
371 end
373 tagged SLL .it :
374 begin
375 Bit#(32) zext_it_shamt = zext(it.shamt);
376 wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc) << zext_it_shamt )} );
377 end
378 tagged SRL .it :
379 begin
380 Bit#(32) zext_it_shamt = zext(it.shamt);
381 wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc) >> zext_it_shamt )});
382 end
383 tagged SRA .it :
384 begin
385 Bit#(32) zext_it_shamt = zext(it.shamt);
386 wbQ.enq(tagged WB_ALU {dest: it.rdst, data:sra( rf.rd1(it.rsrc), zext_it_shamt )});
387 end
388 tagged SLLV .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc) << rshft(rf.rd2(it.rshamt)) )});
389 tagged SRLV .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc) >> rshft(rf.rd2(it.rshamt)) )} );
390 tagged SRAV .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:sra( rf.rd1(it.rsrc), rshft(rf.rd2(it.rshamt)) ) });
391 tagged ADDU .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc1) + rf.rd2(it.rsrc2) )} );
392 tagged SUBU .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc1) - rf.rd2(it.rsrc2) )} );
393 tagged AND .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc1) & rf.rd2(it.rsrc2) )} );
394 tagged OR .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc1) | rf.rd2(it.rsrc2) )} );
395 tagged XOR .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(rf.rd1(it.rsrc1) ^ rf.rd2(it.rsrc2) )} );
396 tagged NOR .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:(~(rf.rd1(it.rsrc1) | rf.rd2(it.rsrc2)) )} );
397 tagged SLT .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:slt( rf.rd1(it.rsrc1), rf.rd2(it.rsrc2) ) });
398 tagged SLTU .it : wbQ.enq(tagged WB_ALU {dest: it.rdst, data:sltu( rf.rd1(it.rsrc1), rf.rd2(it.rsrc2) ) });
400 // -- Branches --------------------------------------------------
402 tagged BLEZ .it :
403 if ( signedLE( rf.rd1(it.rsrc), 0 ) )
404 begin
405 newPC = pc_plus4 + (sext(it.offset) << 2);
406 branchTaken = True;
407 end
409 tagged BGTZ .it :
410 if ( signedGT( rf.rd1(it.rsrc), 0 ) )
411 begin
412 newPC = pc_plus4 + (sext(it.offset) << 2);
413 branchTaken = True;
414 end
416 tagged BLTZ .it :
417 if ( signedLT( rf.rd1(it.rsrc), 0 ) )
418 begin
419 newPC = pc_plus4 + (sext(it.offset) << 2);
420 branchTaken = True;
421 end
423 tagged BGEZ .it :
424 if ( signedGE( rf.rd1(it.rsrc), 0 ) )
425 begin
426 newPC = pc_plus4 + (sext(it.offset) << 2);
427 branchTaken = True;
428 end
430 tagged BEQ .it :
431 if ( rf.rd1(it.rsrc1) == rf.rd2(it.rsrc2) )
432 begin
433 newPC = pc_plus4 + (sext(it.offset) << 2);
434 branchTaken = True;
435 end
437 tagged BNE .it :
438 if ( rf.rd1(it.rsrc1) != rf.rd2(it.rsrc2) )
439 begin
440 newPC = pc_plus4 + (sext(it.offset) << 2);
441 branchTaken = True;
442 end
444 // -- Jumps -----------------------------------------------------
446 tagged J .it :
447 begin
448 newPC = { pc_plus4[31:28], it.target, 2'b0 };
449 branchTaken = True;
450 end
452 tagged JR .it :
453 begin
454 newPC = rf.rd1(it.rsrc);
455 branchTaken = True;
456 end
458 tagged JAL .it :
459 begin
460 wbQ.enq(tagged WB_ALU {dest:31, data:pc_plus4 });
461 newPC = { pc_plus4[31:28], it.target, 2'b0 };
462 branchTaken = True;
463 end
465 tagged JALR .it :
466 begin
467 wbQ.enq(tagged WB_ALU {dest:it.rdst, data:pc_plus4 });
468 newPC = rf.rd1(it.rsrc);
469 branchTaken = True;
470 end
472 // -- Cop0 ------------------------------------------------------
474 tagged MTC0 .it :
475 begin
476 case ( it.cop0dst )
477 5'd10 : cp0_statsEn <= unpack(truncate(rf.rd1(it.rsrc)));
478 5'd21 : cp0_tohost <= truncate(rf.rd1(it.rsrc));
479 default :
480 $display( " RTL-ERROR : %m : Illegal MTC0 cop0dst register!" );
481 endcase
482 wbQ.enq(tagged WB_Host 0); //no idea wwhat this actually should be.
483 end
485 //this is host stuff?
486 tagged MFC0 .it :
487 begin
488 case ( it.cop0src )
489 // not actually an ALU instruction but don't have the format otherwise
490 5'd10 : wbQ.enq(tagged WB_ALU {dest:it.rdst, data:zext(pack(cp0_statsEn)) });
491 5'd20 : wbQ.enq(tagged WB_ALU {dest:it.rdst, data:cp0_fromhost });
492 5'd21 : wbQ.enq(tagged WB_ALU {dest:it.rdst, data:cp0_tohost });
493 default :
494 $display( " RTL-ERROR : %m : Illegal MFC0 cop0src register!" );
495 endcase
496 end
498 // -- Illegal ---------------------------------------------------
500 default :
501 $display( " RTL-ERROR : %m : Illegal instruction !" );
503 endcase
505 //evaluate branch prediction
506 Addr ppc = pcQ.first().qnxtpc; //predicted branch
507 if (ppc != newPC) //prediction wrong
508 begin
509 epoch <= pcQ.first().qepoch + 1;
510 bp.upd(instrpc, newPC); //update branch predictor
511 pcQ.clear();
512 pc <= newPC;
513 end
514 else
515 pcQ.deq();
517 if ( cp0_statsEn )
518 num_inst <= num_inst+1;
520 endrule
522 rule writeback; // ( stage == Writeback );
523 traceTiny("mkProc", "writeback","W");
526 // get what to do off the writeback queue
527 wbQ.deq();
528 case (wbQ.first()) matches
529 tagged WB_ALU {data:.res, dest:.rdst} : rf.wr(rdst, res);
530 tagged WB_Load .regWr :
531 begin
532 dataRespQ.deq();
533 if (dataRespQ.first() matches tagged LoadResp .ld)
534 rf.wr(truncate(ld.tag), ld.data); // no need to use Rindx from queue? Duplicate?
535 end
536 tagged WB_Store : dataRespQ.deq();
537 tagged WB_Host .dat : noAction;
538 endcase
540 endrule
542 rule inc_num_cycles;
543 if ( cp0_statsEn )
544 num_cycles <= num_cycles+1;
545 endrule
546 // THis rule breaks things
547 // rule handleCPUToHost;
548 // let req <- server_stub.acceptRequest_ReadCPUToHost();
549 // case (req)
550 // 0: server_stub.sendResponse_ReadCPUToHost(cp0_tohost);
551 // 1: server_stub.sendResponse_ReadCPUToHost(pc);
552 // 2: server_stub.sendResponse_ReadCPUToHost(zeroExtend(pack(stage)));
553 // endcase
554 // endrule
555 //-----------------------------------------------------------
556 // My Adds
557 //-----------------------------------------------------------
559 //-----------------------------------------------------------
560 // Methods
562 interface Client imem_client;
563 interface Get request = toGet(instReqQ);
564 interface Put response = toPut(instRespQ);
565 endinterface
567 interface Client dmem_client;
568 interface Get request = toGet(dataReqQ);
569 interface Put response = toPut(dataRespQ);
570 endinterface
572 interface Get statsEn_get = toGet(asReg(cp0_statsEn));
574 interface ProcStats stats;
575 interface Get num_cycles = toGet(asReg(num_cycles));
576 interface Get num_inst = toGet(asReg(num_inst));
577 endinterface
579 interface CPUToHost tohost;
580 method Bit#(32) cpuToHost(int req);
581 return (case (req)
582 0: cp0_tohost;
583 1: pc;
584 2: zeroExtend(pack(stage));
585 endcase);
586 endmethod
587 endinterface
589 endmodule