riscv/rocket-chip
1
0
Fork 0
Code Releases Activity

merge I$, ITLB, BTB into Frontend

Browse Source
This commit is contained in:
Andrew Waterman 2012-10-09 21:35:03 -07:00
parent fcd69dba98
commit 661f8e635b
15 changed files with 946 additions and 1143 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

48
rocket/src/main/scala/consts.scala
View File

@ -16,25 +16,20 @@ object Constants
val HTIF_WIDTH = 16
val MEM_BACKUP_WIDTH = HTIF_WIDTH
val BR_X = Bits("b????", 4)
val BR_N = UFix(0, 4);
val BR_EQ = UFix(1, 4);
val BR_NE = UFix(2, 4);
val BR_LT = UFix(3, 4);
val BR_LTU = UFix(4, 4);
val BR_GE = UFix(5, 4);
val BR_GEU = UFix(6, 4);
val BR_J = UFix(7, 4);
val BR_JR = UFix(8, 4);
val BR_X = Bits("b???", 3)
val BR_EQ = UFix(0, 3)
val BR_NE = UFix(1, 3)
val BR_J = UFix(2, 3)
val BR_N = UFix(3, 3)
val BR_LT = UFix(4, 3)
val BR_GE = UFix(5, 3)
val BR_LTU = UFix(6, 3)
val BR_GEU = UFix(7, 3)
val PC_4 = UFix(0, 3);
val PC_BTB = UFix(1, 3);
val PC_EX4 = UFix(2, 3);
val PC_BR = UFix(3, 3);
val PC_PCR = UFix(4, 3);
val PC_WB = UFix(5, 3);
val PC_EVEC = UFix(6, 3);
val PC_JR = UFix(7, 3);
val PC_EX4 = UFix(0, 2)
val PC_EX = UFix(1, 2)
val PC_WB = UFix(2, 2)
val PC_PCR = UFix(3, 2)
val A2_X = Bits("b???", 3)
val A2_BTYPE = UFix(0, 3);
@ -70,19 +65,6 @@ object Constants
val WB_TSC = UFix(4, 3);
val WB_IRT = UFix(5, 3);
val FN_X = Bits("b????", 4)
val FN_ADD = UFix(0, 4);
val FN_SUB = UFix(1, 4);
val FN_SLT = UFix(2, 4);
val FN_SLTU = UFix(3, 4);
val FN_AND = UFix(4, 4);
val FN_OR = UFix(5, 4);
val FN_XOR = UFix(6, 4);
val FN_SL = UFix(7, 4);
val FN_SR = UFix(8, 4);
val FN_SRA = UFix(9, 4);
val FN_OP2 = UFix(10, 4);
val DW_X = X
val DW_32 = N
val DW_64 = Y
@ -175,6 +157,7 @@ object Constants
val PERM_BITS = 6;
// rocketNBDCache parameters
val INST_BITS = 32
val DCACHE_PORTS = 3
val CPU_DATA_BITS = 64;
val CPU_TAG_BITS = 9;
@ -212,8 +195,9 @@ object Constants
val MEM_DATA_BITS = 128
val REFILL_CYCLES = (1 << OFFSET_BITS)*8/MEM_DATA_BITS
val BTB_ENTRIES = 8
val ITLB_ENTRIES = 8
val DTLB_ENTRIES = 16
val ITLB_ENTRIES = 8;
val VITLB_ENTRIES = 4
val START_ADDR = 0x2000;

59
rocket/src/main/scala/cpu.scala
View File

@ -8,8 +8,8 @@ import hwacha._
class ioRocket extends Bundle()
{
val host = new ioHTIF
val imem = (new ioImem).flip
val vimem = (new ioImem).flip
val imem = new IOCPUFrontend
val vimem = new IOCPUFrontend
val dmem = new ioHellaCache
}
@ -20,8 +20,7 @@ class rocketProc extends Component
val ctrl = new rocketCtrl();
val dpath = new rocketDpath();
val dtlb = new rocketDTLB(DTLB_ENTRIES);
val itlb = new rocketITLB(ITLB_ENTRIES);
val dtlb = new rocketTLB(DTLB_ENTRIES);
val ptw = new rocketPTW(if (HAVE_VEC) 3 else 2)
val arb = new rocketHellaCacheArbiter(DCACHE_PORTS)
@ -59,7 +58,7 @@ class rocketProc extends Component
dtlbarb.io.in(DTLB_CPU).valid := ctrl.io.dtlb_val
dtlbarb.io.in(DTLB_CPU).bits.kill := ctrl.io.dtlb_kill
dtlbarb.io.in(DTLB_CPU).bits.cmd := ctrl.io.dmem.req.bits.cmd
dtlbarb.io.in(DTLB_CPU).bits.asid := Bits(0,ASID_BITS); // FIXME: connect to PCR
dtlbarb.io.in(DTLB_CPU).bits.asid := UFix(0)
dtlbarb.io.in(DTLB_CPU).bits.vpn := dpath.io.dtlb.vpn
ctrl.io.dtlb_rdy := dtlbarb.io.in(DTLB_CPU).ready
@ -75,7 +74,7 @@ class rocketProc extends Component
dtlb.io.cpu_req.valid := ctrl.io.dtlb_val
dtlb.io.cpu_req.bits.kill := ctrl.io.dtlb_kill
dtlb.io.cpu_req.bits.cmd := ctrl.io.dmem.req.bits.cmd
dtlb.io.cpu_req.bits.asid := Bits(0,ASID_BITS); // FIXME: connect to PCR
dtlb.io.cpu_req.bits.asid := UFix(0)
dtlb.io.cpu_req.bits.vpn := dpath.io.dtlb.vpn
ctrl.io.xcpt_dtlb_ld := dtlb.io.cpu_resp.xcpt_ld
ctrl.io.xcpt_dtlb_st := dtlb.io.cpu_resp.xcpt_st
@ -91,7 +90,7 @@ class rocketProc extends Component
// connect page table walker to TLBs, page table base register (from PCR)
// and D$ arbiter (selects between requests from pipeline and PTW, PTW has priority)
ptw.io.requestor(0) <> itlb.io.ptw
ptw.io.requestor(0) <> io.imem.ptw
ptw.io.requestor(1) <> dtlb.io.ptw
ptw.io.ptbr := dpath.io.ptbr;
arb.io.requestor(DMEM_PTW) <> ptw.io.mem
@ -102,20 +101,9 @@ class rocketProc extends Component
// FIXME: try to make this more compact
// connect ITLB to I$, ctrl, dpath
itlb.io.cpu.invalidate := dpath.io.ptbr_wen;
itlb.io.cpu.status := dpath.io.ctrl.status;
itlb.io.cpu.req_val := ctrl.io.imem.req_val;
itlb.io.cpu.req_asid := Bits(0,ASID_BITS); // FIXME: connect to PCR
itlb.io.cpu.req_vpn := dpath.io.imem.req_addr(VADDR_BITS,PGIDX_BITS);
io.imem.req_idx := dpath.io.imem.req_addr(PGIDX_BITS-1,0);
io.imem.req_ppn := itlb.io.cpu.resp_ppn;
io.imem.req_val := ctrl.io.imem.req_val;
io.imem.invalidate := ctrl.io.dpath.flush_inst;
ctrl.io.imem.resp_val := io.imem.resp_val;
dpath.io.imem.resp_data := io.imem.resp_data;
ctrl.io.xcpt_itlb := itlb.io.cpu.exception;
io.imem.itlb_miss := itlb.io.cpu.resp_miss;
// connect I$
ctrl.io.imem <> io.imem
dpath.io.imem <> io.imem
// connect arbiter to ctrl+dpath+DTLB
//TODO: views on nested bundles?
@ -144,22 +132,19 @@ class rocketProc extends Component
dpath.io.vec_ctrl <> ctrl.io.vec_dpath
// hooking up vector I$
val vitlb = new rocketITLB(VITLB_ENTRIES)
ptw.io.requestor(2) <> vitlb.io.ptw
vitlb.io.cpu.invalidate := dpath.io.ptbr_wen
vitlb.io.cpu.status := dpath.io.ctrl.status
vitlb.io.cpu.req_val := vu.io.imem_req.valid
vitlb.io.cpu.req_asid := Bits(0,ASID_BITS) // FIXME: connect to PCR
vitlb.io.cpu.req_vpn := vu.io.imem_req.bits(VADDR_BITS,PGIDX_BITS).toUFix
io.vimem.req_idx := vu.io.imem_req.bits(PGIDX_BITS-1,0)
io.vimem.req_ppn := vitlb.io.cpu.resp_ppn
io.vimem.req_val := vu.io.imem_req.valid
io.vimem.invalidate := ctrl.io.dpath.flush_inst
vu.io.imem_req.ready := Bool(true)
vu.io.imem_resp.valid := io.vimem.resp_val
vu.io.imem_resp.bits := io.vimem.resp_data
vu.io.vitlb_exception := vitlb.io.cpu.exception
io.vimem.itlb_miss := vitlb.io.cpu.resp_miss
ptw.io.requestor(2) <> io.vimem.ptw
io.vimem.req.bits.status := dpath.io.ctrl.status
io.vimem.req.bits.pc := vu.io.imem_req.bits.toUFix
io.vimem.req.valid := vu.io.imem_req.valid
io.vimem.req.bits.invalidate := ctrl.io.dpath.flush_inst
io.vimem.req.bits.invalidateTLB := dpath.io.ptbr_wen
vu.io.imem_req.ready := Bool(true)
vu.io.imem_resp.valid := io.vimem.resp.valid
vu.io.imem_resp.bits := io.vimem.resp.bits.data
vu.io.vitlb_exception := io.vimem.resp.bits.xcpt_if
io.vimem.resp.ready := Bool(true)
io.vimem.req.bits.mispredict := Bool(false)
io.vimem.req.bits.taken := Bool(false)
// hooking up vector command queues
vu.io.vec_cmdq.valid := ctrl.io.vec_iface.vcmdq_valid

667
rocket/src/main/scala/ctrl.scala
View File

@ -6,16 +6,13 @@ import Node._;
import Constants._
import Instructions._
import hwacha._
import ALU._
class ioCtrlDpath extends Bundle()
{
// outputs to datapath
val sel_pc = UFix(OUTPUT, 3);
val wen_btb = Bool(OUTPUT);
val clr_btb = Bool(OUTPUT);
val stallf = Bool(OUTPUT);
val stalld = Bool(OUTPUT);
val killf = Bool(OUTPUT);
val killd = Bool(OUTPUT);
val killx = Bool(OUTPUT);
val killm = Bool(OUTPUT);
@ -26,8 +23,10 @@ class ioCtrlDpath extends Bundle()
val fn_alu = UFix(OUTPUT, 4);
val mul_val = Bool(OUTPUT);
val mul_fn = UFix(OUTPUT, 2);
val mul_kill = Bool(OUTPUT)
val div_val = Bool(OUTPUT);
val div_fn = UFix(OUTPUT, 2);
val div_kill = Bool(OUTPUT)
val sel_wa = Bool(OUTPUT);
val sel_wb = UFix(OUTPUT, 3);
val pcr = UFix(OUTPUT, 3)
@ -37,6 +36,7 @@ class ioCtrlDpath extends Bundle()
val ex_fp_val= Bool(OUTPUT);
val mem_fp_val= Bool(OUTPUT);
val ex_wen = Bool(OUTPUT);
val ex_jalr = Bool(OUTPUT)
val mem_wen = Bool(OUTPUT);
val wb_wen = Bool(OUTPUT);
val wb_valid = Bool(OUTPUT)
@ -48,8 +48,6 @@ class ioCtrlDpath extends Bundle()
val badvaddr_wen = Bool(OUTPUT); // high for a load/store access fault
val vec_irq_aux_wen = Bool(OUTPUT)
// inputs from datapath
val xcpt_ma_inst = Bool(INPUT); // high on a misaligned/illegal virtual PC
val btb_hit = Bool(INPUT);
val inst = Bits(INPUT, 32);
val br_eq = Bool(INPUT);
val br_lt = Bool(INPUT);
@ -76,7 +74,7 @@ class ioCtrlDpath extends Bundle()
class ioCtrlAll extends Bundle()
{
val dpath = new ioCtrlDpath();
val imem = new ioImem().flip
val imem = new IOCPUFrontend
val dmem = new ioHellaCache
val dtlb_val = Bool(OUTPUT);
val dtlb_kill = Bool(OUTPUT);
@ -84,7 +82,6 @@ class ioCtrlAll extends Bundle()
val dtlb_miss = Bool(INPUT);
val xcpt_dtlb_ld = Bool(INPUT);
val xcpt_dtlb_st = Bool(INPUT);
val xcpt_itlb = Bool(INPUT);
val fpu = new ioCtrlFPU();
val vec_dpath = new ioCtrlDpathVec()
val vec_iface = new ioCtrlVecInterface()
@ -95,239 +92,239 @@ object rocketCtrlDecode
val xpr64 = Y;
val decode_default =
// eret
// fp_val renx2 | syscall
// | vec_val | renx1 mem_val mul_val div_val wen pcr | | privileged
// val | | brtype | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | div_fn | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
List(N, X,X,BR_X, X,X,A2_X, DW_X, FN_X, N,M_X, MT_X, X,MUL_X, X,DIV_X, X,WA_X, WB_X, PCR_X,SYNC_X,X,X,X,X)
val xdecode = Array(
// eret
// fp_val renx2 | syscall
// | vec_val | renx1 mem_val mul_val div_val wen pcr | | privileged
// val | | brtype | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | div_fn | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
BNE-> List(Y, N,N,BR_NE, Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BEQ-> List(Y, N,N,BR_EQ, Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BLT-> List(Y, N,N,BR_LT, Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BLTU-> List(Y, N,N,BR_LTU,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BGE-> List(Y, N,N,BR_GE, Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BGEU-> List(Y, N,N,BR_GEU,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
J-> List(Y, N,N,BR_J, N,N,A2_JTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
JAL-> List(Y, N,N,BR_J, N,N,A2_JTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RA,WB_PC, PCR_N,SYNC_N,N,N,N,N),
JALR_C-> List(Y, N,N,BR_JR, N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
JALR_J-> List(Y, N,N,BR_JR, N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
JALR_R-> List(Y, N,N,BR_JR, N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
RDNPC-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
LB-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_B, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LH-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_H, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LW-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LD-> List(xpr64,N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LBU-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_BU,N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LHU-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_HU,N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LWU-> List(xpr64,N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_WU,N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SB-> List(Y, N,N,BR_N, Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_B, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SH-> List(Y, N,N,BR_N, Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_H, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SW-> List(Y, N,N,BR_N, Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_W, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SD-> List(xpr64,N,N,BR_N, Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_D, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOADD_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_ADD, MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOSWAP_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_SWAP,MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOAND_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_AND, MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOOR_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_OR, MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMIN_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MIN, MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMINU_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MINU,MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAX_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAX, MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAXU_W-> List(Y, N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAXU,MT_W, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOADD_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_ADD, MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOSWAP_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_SWAP,MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOAND_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_AND, MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOOR_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_OR, MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMIN_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MIN, MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMINU_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MINU,MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAX_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAX, MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAXU_D-> List(xpr64,N,N,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAXU,MT_D, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LUI-> List(Y, N,N,BR_N, N,N,A2_LTYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADDI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLTI -> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_SLT, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLTIU-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_SLTU,N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ANDI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_AND, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ORI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_OR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
XORI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_XOR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLLI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_SL, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRLI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_SR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRAI-> List(Y, N,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_SRA, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADD-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SUB-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_SUB, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLT-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_SLT, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLTU-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_SLTU,N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
riscvAND-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_AND, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
riscvOR-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_OR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
riscvXOR-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_XOR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLL-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_SL, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRL-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_SR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRA-> List(Y, N,N,BR_N, Y,Y,A2_RTYPE,DW_XPR,FN_SRA, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADDIW-> List(xpr64,N,N,BR_N, N,Y,A2_ITYPE,DW_32,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLLIW-> List(xpr64,N,N,BR_N, N,Y,A2_ITYPE,DW_32,FN_SL, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRLIW-> List(xpr64,N,N,BR_N, N,Y,A2_ITYPE,DW_32,FN_SR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRAIW-> List(xpr64,N,N,BR_N, N,Y,A2_ITYPE,DW_32,FN_SRA, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADDW-> List(xpr64,N,N,BR_N, Y,Y,A2_RTYPE,DW_32,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SUBW-> List(xpr64,N,N,BR_N, Y,Y,A2_RTYPE,DW_32,FN_SUB, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLLW-> List(xpr64,N,N,BR_N, Y,Y,A2_RTYPE,DW_32,FN_SL, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRLW-> List(xpr64,N,N,BR_N, Y,Y,A2_RTYPE,DW_32,FN_SR, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRAW-> List(xpr64,N,N,BR_N, Y,Y,A2_RTYPE,DW_32,FN_SRA, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
MUL-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_LO, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULH-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_H, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULHU-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_HU, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULHSU-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_HSU,N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULW-> List(xpr64,N,N,BR_N, Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, Y,MUL_LO, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIV-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_D, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIVU-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_DU,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REM-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_R, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REMU-> List(Y, N,N,BR_N, Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_RU,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIVW-> List(xpr64,N,N,BR_N, Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_D, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIVUW-> List(xpr64,N,N,BR_N, Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_DU,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REMW-> List(xpr64,N,N,BR_N, Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_R, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REMUW-> List(xpr64,N,N,BR_N, Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,MUL_X, Y,DIV_RU,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
SYSCALL-> List(Y, N,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,Y,N,N),
SETPCR-> List(Y, N,N,BR_N, N,N,A2_ITYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_S,SYNC_N,N,N,Y,Y),
CLEARPCR-> List(Y, N,N,BR_N, N,N,A2_ITYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_C,SYNC_N,N,N,Y,Y),
ERET-> List(Y, N,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,Y,N,Y,N),
FENCE-> List(Y, N,N,BR_N, N,N,A2_X, DW_X, FN_X, Y,M_FENCE, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_D,N,N,N,N),
FENCE_I-> List(Y, N,N,BR_N, N,N,A2_X, DW_X, FN_X, Y,M_FLA, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_I,N,N,N,Y),
CFLUSH-> List(Y, N,N,BR_N, N,N,A2_X, DW_X, FN_X, Y,M_FLA, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,Y,Y),
MFPCR-> List(Y, N,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_F,SYNC_N,N,N,Y,Y),
MTPCR-> List(Y, N,N,BR_N, Y,N,A2_RTYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_T,SYNC_N,N,N,Y,Y),
RDTIME-> List(Y, N,N,BR_N, N,N,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_TSC,PCR_N,SYNC_N,N,N,N,N),
RDCYCLE-> List(Y, N,N,BR_N, N,N,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_TSC,PCR_N,SYNC_N,N,N,N,N),
RDINSTRET-> List(Y, N,N,BR_N, N,N,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_IRT,PCR_N,SYNC_N,N,N,N,N))
val fdecode = Array(
// eret
// fp_val renx2 | syscall
// | vec_val | renx1 mem_val mul_val div_val wen pcr | | privileged
// val | | brtype | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | div_fn | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
FCVT_S_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJ_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJ_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJX_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJX_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJN_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJN_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMIN_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMIN_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMAX_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMAX_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FADD_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FADD_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSUB_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSUB_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMUL_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMUL_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMADD_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMADD_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMSUB_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMSUB_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMADD_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMADD_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMSUB_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMSUB_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MFTX_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MFTX_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_W_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_W_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_WU_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_WU_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_L_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_L_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_LU_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_LU_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FEQ_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FEQ_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLT_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLT_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLE_S-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLE_D-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MXTF_S-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MXTF_D-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_W-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_W-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_WU-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_WU-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_L-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_L-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_LU-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_LU-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MFFSR-> List(FPU_Y,Y,N,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MTFSR-> List(FPU_Y,Y,N,BR_N, N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLW-> List(FPU_Y,Y,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_W, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
FLD-> List(FPU_Y,Y,N,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
FSW-> List(FPU_Y,Y,N,BR_N, N,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_W, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
FSD-> List(FPU_Y,Y,N,BR_N, N,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_D, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N))
val vdecode = Array(
// eret
// fp_val renx2 | syscall
// | vec_val | renx1 mem_val mul_val div_val wen pcr | | privileged
// val | | brtype | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | div_fn | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
VVCFGIVL-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,Y),
VVCFG-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,Y),
VSETVL-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,Y),
VF-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VMVV-> List(VEC_Y,N,Y,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
VMSV-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFMVV-> List(VEC_Y,N,Y,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FENCE_V_L-> List(VEC_Y,N,Y,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
FENCE_V_G-> List(VEC_Y,N,Y,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_D,N,N,N,N),
VLD-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLW-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLWU-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLH-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLHU-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLB-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLBU-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSD-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSW-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSH-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSB-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLD-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLW-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSD-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSW-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTD-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTW-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTWU-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTH-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTHU-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTB-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTBU-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTD-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTW-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTH-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTB-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLSTD-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLSTW-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSSTD-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSSTW-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VENQCMD-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VENQIMM1-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VENQIMM2-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VENQCNT-> List(VEC_Y,N,Y,BR_N, Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VXCPTEVAC-> List(VEC_Y,N,Y,BR_N, N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VXCPTKILL-> List(VEC_Y,N,Y,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,Y,N),
VXCPTHOLD-> List(VEC_Y,N,Y,BR_N, N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,DIV_X, N,WA_X, WB_X, PCR_N,SYNC_N,N,N,Y,N))
// jalr eret
// fp_val | renx2 div_val | syscall
// | vec_val | | renx1 mem_val mul_val | wen pcr | | privileged
// val | | brtype | | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
List(N, X,X,BR_X, X,X,X,A2_X, DW_X, FN_X, N,M_X, MT_X, X,MUL_X, X,X,WA_X, WB_X, PCR_X,SYNC_X,X,X,X,X)
val xdecode = Array(
// jalr eret
// fp_val | renx2 div_val | syscall
// | vec_val | | renx1 mem_val mul_val | wen pcr | | privileged
// val | | brtype | | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
BNE-> List(Y, N,N,BR_NE, N,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BEQ-> List(Y, N,N,BR_EQ, N,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BLT-> List(Y, N,N,BR_LT, N,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BLTU-> List(Y, N,N,BR_LTU,N,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BGE-> List(Y, N,N,BR_GE, N,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
BGEU-> List(Y, N,N,BR_GEU,N,Y,Y,A2_BTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
J-> List(Y, N,N,BR_J, N,N,N,A2_JTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
JAL-> List(Y, N,N,BR_J, N,N,N,A2_JTYPE,DW_X, FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RA,WB_PC, PCR_N,SYNC_N,N,N,N,N),
JALR_C-> List(Y, N,N,BR_N, Y,N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
JALR_J-> List(Y, N,N,BR_N, Y,N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
JALR_R-> List(Y, N,N,BR_N, Y,N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
RDNPC-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_PC, PCR_N,SYNC_N,N,N,N,N),
LB-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_B, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LH-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_H, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LW-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LD-> List(xpr64,N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LBU-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_BU,N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LHU-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_HU,N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LWU-> List(xpr64,N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_WU,N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SB-> List(Y, N,N,BR_N, N,Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_B, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SH-> List(Y, N,N,BR_N, N,Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_H, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SW-> List(Y, N,N,BR_N, N,Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_W, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SD-> List(xpr64,N,N,BR_N, N,Y,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_D, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOADD_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_ADD, MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOSWAP_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_SWAP,MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOAND_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_AND, MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOOR_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_OR, MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMIN_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MIN, MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMINU_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MINU,MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAX_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAX, MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAXU_W-> List(Y, N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAXU,MT_W, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOADD_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_ADD, MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOSWAP_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_SWAP,MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOAND_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_AND, MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOOR_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_OR, MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMIN_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MIN, MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMINU_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MINU,MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAX_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAX, MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
AMOMAXU_D-> List(xpr64,N,N,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, Y,M_XA_MAXU,MT_D, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
LUI-> List(Y, N,N,BR_N, N,N,N,A2_LTYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADDI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLTI -> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_SLT, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLTIU-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_SLTU,N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ANDI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_AND, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ORI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_OR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
XORI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_XOR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLLI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_SL, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRLI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_SR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRAI-> List(Y, N,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_SRA, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADD-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SUB-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_SUB, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLT-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_SLT, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLTU-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_SLTU,N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
riscvAND-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_AND, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
riscvOR-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_OR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
riscvXOR-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_XOR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLL-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_SL, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRL-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_SR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRA-> List(Y, N,N,BR_N, N,Y,Y,A2_RTYPE,DW_XPR,FN_SRA, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADDIW-> List(xpr64,N,N,BR_N, N,N,Y,A2_ITYPE,DW_32,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLLIW-> List(xpr64,N,N,BR_N, N,N,Y,A2_ITYPE,DW_32,FN_SL, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRLIW-> List(xpr64,N,N,BR_N, N,N,Y,A2_ITYPE,DW_32,FN_SR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRAIW-> List(xpr64,N,N,BR_N, N,N,Y,A2_ITYPE,DW_32,FN_SRA, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
ADDW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_RTYPE,DW_32,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SUBW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_RTYPE,DW_32,FN_SUB, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SLLW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_RTYPE,DW_32,FN_SL, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRLW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_RTYPE,DW_32,FN_SR, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
SRAW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_RTYPE,DW_32,FN_SRA, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
MUL-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_LO, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULH-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_H, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULHU-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_HU, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULHSU-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, Y,MUL_HSU,N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MULW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, Y,MUL_LO, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIV-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,DIV_D, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIVU-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,DIV_DU, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REM-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,DIV_R, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REMU-> List(Y, N,N,BR_N, N,Y,Y,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,DIV_RU, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIVW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,DIV_D, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
DIVUW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,DIV_DU, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REMW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,DIV_R, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
REMUW-> List(xpr64,N,N,BR_N, N,Y,Y,A2_X, DW_32, FN_X, N,M_X, MT_X, N,DIV_RU, Y,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
SYSCALL-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,Y,N,N),
SETPCR-> List(Y, N,N,BR_N, N,N,N,A2_ITYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_S,SYNC_N,N,N,Y,Y),
CLEARPCR-> List(Y, N,N,BR_N, N,N,N,A2_ITYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_C,SYNC_N,N,N,Y,Y),
ERET-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,Y,N,Y,N),
FENCE-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_X, FN_X, Y,M_FENCE, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_D,N,N,N,N),
FENCE_I-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_X, FN_X, Y,M_FLA, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_I,N,N,N,Y),
CFLUSH-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_X, FN_X, Y,M_FLA, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,Y,Y),
MFPCR-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_F,SYNC_N,N,N,Y,Y),
MTPCR-> List(Y, N,N,BR_N, N,Y,N,A2_RTYPE,DW_XPR,FN_OP2, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_T,SYNC_N,N,N,Y,Y),
RDTIME-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_TSC,PCR_N,SYNC_N,N,N,N,N),
RDCYCLE-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_TSC,PCR_N,SYNC_N,N,N,N,N),
RDINSTRET-> List(Y, N,N,BR_N, N,N,N,A2_X, DW_XPR,FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_IRT,PCR_N,SYNC_N,N,N,N,N))
val fdecode = Array(
// jalr eret
// fp_val | renx2 div_val | syscall
// | vec_val | | renx1 mem_val mul_val | wen pcr | | privileged
// val | | brtype | | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
FCVT_S_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJ_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJ_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJX_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJX_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJN_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSGNJN_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMIN_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMIN_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMAX_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMAX_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FADD_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FADD_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSUB_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FSUB_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMUL_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMUL_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMADD_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMADD_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMSUB_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FMSUB_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMADD_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMADD_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMSUB_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FNMSUB_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MFTX_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MFTX_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_W_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_W_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_WU_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_WU_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_L_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_L_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_LU_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_LU_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FEQ_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FEQ_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLT_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLT_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLE_S-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLE_D-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MXTF_S-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MXTF_D-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_W-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_W-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_WU-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_WU-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_L-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_L-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_S_LU-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FCVT_D_LU-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MFFSR-> List(FPU_Y,Y,N,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
MTFSR-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FLW-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_W, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
FLD-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, Y,M_XRD, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
FSW-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_W, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
FSD-> List(FPU_Y,Y,N,BR_N, N,N,Y,A2_BTYPE,DW_XPR,FN_ADD, Y,M_XWR, MT_D, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N))
val vdecode = Array(
// jalr eret
// fp_val | renx2 div_val | syscall
// | vec_val | | renx1 mem_val mul_val | wen pcr | | privileged
// val | | brtype | | | s_alu2 dw alu | mem_cmd mem_type| mul_fn | | s_wa s_wb | sync | | | replay_next
// | | | | | | | | | | | | | | | | | | | | | | | | |
VVCFGIVL-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,Y),
VVCFG-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,Y),
VSETVL-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,Y,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,Y),
VF-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ITYPE,DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VMVV-> List(VEC_Y,N,Y,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
VMSV-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFMVV-> List(VEC_Y,N,Y,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_X, PCR_N,SYNC_N,N,N,N,N),
FENCE_V_L-> List(VEC_Y,N,Y,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,N,N),
FENCE_V_G-> List(VEC_Y,N,Y,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_D,N,N,N,N),
VLD-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLW-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLWU-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLH-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLHU-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLB-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLBU-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSD-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSW-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSH-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSB-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLD-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLW-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSD-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSW-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTD-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTW-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTWU-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTH-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTHU-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTB-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VLSTBU-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTD-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTW-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTH-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VSSTB-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLSTD-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFLSTW-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSSTD-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VFSSTW-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_D, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,N,N),
VENQCMD-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VENQIMM1-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VENQIMM2-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VENQCNT-> List(VEC_Y,N,Y,BR_N, N,Y,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VXCPTEVAC-> List(VEC_Y,N,Y,BR_N, N,N,Y,A2_ZERO, DW_XPR,FN_ADD, N,M_X, MT_X, N,MUL_X, N,N,WA_RD,WB_ALU,PCR_N,SYNC_N,N,N,Y,N),
VXCPTKILL-> List(VEC_Y,N,Y,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,Y,N),
VXCPTHOLD-> List(VEC_Y,N,Y,BR_N, N,N,N,A2_X, DW_X, FN_X, N,M_X, MT_X, N,MUL_X, N,N,WA_X, WB_X, PCR_N,SYNC_N,N,N,Y,N))
}
class rocketCtrl extends Component
@ -340,32 +337,22 @@ class rocketCtrl extends Component
val cs = DecodeLogic(io.dpath.inst, rocketCtrlDecode.decode_default, decode_table)
val id_int_val :: id_fp_val :: id_vec_val :: id_br_type :: id_renx2 :: id_renx1 :: id_sel_alu2 :: id_fn_dw :: id_fn_alu :: cs0 = cs
val id_mem_val :: id_mem_cmd :: id_mem_type :: id_mul_val :: id_mul_fn :: id_div_val :: id_div_fn :: id_wen :: id_sel_wa :: id_sel_wb :: cs1 = cs0
val id_int_val :: id_fp_val :: id_vec_val :: id_br_type :: id_jalr :: id_renx2 :: id_renx1 :: id_sel_alu2 :: id_fn_dw :: id_fn_alu :: cs0 = cs
val id_mem_val :: id_mem_cmd :: id_mem_type :: id_mul_val :: id_mul_fn :: id_div_val :: id_wen :: id_sel_wa :: id_sel_wb :: cs1 = cs0
val id_pcr :: id_sync :: id_eret :: id_syscall :: id_privileged :: id_replay_next :: Nil = cs1
val if_reg_xcpt_ma_inst = Reg(io.dpath.xcpt_ma_inst, resetVal = Bool(false));
val id_raddr3 = io.dpath.inst(16,12);
val id_raddr2 = io.dpath.inst(21,17);
val id_raddr1 = io.dpath.inst(26,22);
val id_waddr = Mux(id_sel_wa === WA_RA, RA, io.dpath.inst(31,27));
val wb_reg_div_mul_val = Reg(resetVal = Bool(false))
val wb_reg_dcache_miss = Reg(io.dmem.resp.bits.miss || io.dmem.resp.bits.nack, resetVal = Bool(false));
val id_reg_valid = Reg(resetVal = Bool(false));
val id_reg_btb_hit = Reg(resetVal = Bool(false));
val id_reg_xcpt_itlb = Reg(resetVal = Bool(false));
val id_reg_xcpt_ma_inst = Reg(resetVal = Bool(false));
val id_reg_icmiss = Reg(resetVal = Bool(false));
val id_reg_replay = Reg(resetVal = Bool(false));
val id_load_use = Bool();
val id_load_use = Bool();
val ex_reg_br_type = Reg(){Bits()}
val ex_reg_jalr = Reg(){Bool()}
val ex_reg_btb_hit = Reg(){Bool()};
val ex_reg_div_val = Reg(){Bool()};
val ex_reg_mul_val = Reg(){Bool()};
val ex_reg_mul_fn = Reg(){UFix()};
val ex_reg_mem_val = Reg(){Bool()};
val ex_reg_mem_cmd = Reg(){Bits()};
val ex_reg_mem_type = Reg(){UFix(width = 3)};
@ -385,7 +372,7 @@ class rocketCtrl extends Component
val ex_reg_fp_val = Reg(resetVal = Bool(false));
val ex_reg_fp_sboard_set = Reg(resetVal = Bool(false));
val ex_reg_vec_val = Reg(resetVal = Bool(false));
val ex_reg_replay = Reg(resetVal = Bool(false));
val ex_reg_replay_next = Reg(resetVal = Bool(false));
val ex_reg_load_use = Reg(resetVal = Bool(false));
val mem_reg_valid = Reg(resetVal = Bool(false));
@ -404,6 +391,7 @@ class rocketCtrl extends Component
val mem_reg_xcpt_syscall = Reg(resetVal = Bool(false));
val mem_reg_fp_val = Reg(resetVal = Bool(false));
val mem_reg_replay = Reg(resetVal = Bool(false));
val mem_reg_replay_next = Reg(resetVal = Bool(false));
val mem_reg_kill = Reg(resetVal = Bool(false));
val mem_reg_fp_sboard_set = Reg(resetVal = Bool(false));
@ -415,30 +403,16 @@ class rocketCtrl extends Component
val wb_reg_eret = Reg(resetVal = Bool(false));
val wb_reg_exception = Reg(resetVal = Bool(false));
val wb_reg_replay = Reg(resetVal = Bool(false));
val wb_reg_replay_next = Reg(resetVal = Bool(false));
val wb_reg_cause = Reg(){UFix()};
val wb_reg_fp_val = Reg(resetVal = Bool(false));
val wb_reg_fp_sboard_set = Reg(resetVal = Bool(false));
val wb_reg_dcache_miss = Reg(io.dmem.resp.bits.miss || io.dmem.resp.bits.nack, resetVal = Bool(false));
val wb_reg_div_mul_val = Reg(resetVal = Bool(false))
val take_pc = Bool()
val take_pc_wb = Bool()
when (!io.dpath.stalld) {
when (io.dpath.killf) {
id_reg_valid := Bool(false)
id_reg_btb_hit := Bool(false);
id_reg_xcpt_ma_inst := Bool(false);
id_reg_xcpt_itlb := Bool(false);
id_reg_replay := !take_pc; // replay on I$ miss
}
.otherwise{
id_reg_valid := Bool(true)
id_reg_btb_hit := io.dpath.btb_hit;
id_reg_xcpt_ma_inst := if_reg_xcpt_ma_inst;
id_reg_xcpt_itlb := io.xcpt_itlb;
id_reg_replay := id_replay_next
}
id_reg_icmiss := !io.imem.resp_val;
}
val ctrl_killm = Bool()
var vec_replay = Bool(false)
var vec_stalld = Bool(false)
@ -499,6 +473,7 @@ class rocketCtrl extends Component
when (reset.toBool || io.dpath.killd) {
ex_reg_br_type := BR_N;
ex_reg_jalr := Bool(false)
ex_reg_btb_hit := Bool(false);
ex_reg_div_val := Bool(false);
ex_reg_mul_val := Bool(false);
@ -517,55 +492,50 @@ class rocketCtrl extends Component
ex_reg_fp_val := Bool(false);
ex_reg_fp_sboard_set := Bool(false);
ex_reg_vec_val := Bool(false);
ex_reg_replay := Bool(false);
ex_reg_replay_next := Bool(false);
ex_reg_load_use := Bool(false);
}
.otherwise {
ex_reg_br_type := id_br_type;
ex_reg_btb_hit := id_reg_btb_hit;
ex_reg_jalr := id_jalr
ex_reg_btb_hit := io.imem.resp.bits.taken
ex_reg_div_val := id_div_val.toBool && id_waddr != UFix(0);
ex_reg_mul_val := id_mul_val.toBool && id_waddr != UFix(0);
ex_reg_mul_fn := id_mul_fn.toUFix
ex_reg_mem_val := id_mem_val.toBool;
ex_reg_valid := id_reg_valid
ex_reg_valid := Bool(true)
ex_reg_pcr := id_pcr
ex_reg_wen := id_wen.toBool && id_waddr != UFix(0);
ex_reg_fp_wen := id_fp_val && io.fpu.dec.wen
ex_reg_eret := id_eret.toBool;
ex_reg_flush_inst := (id_sync === SYNC_I);
ex_reg_xcpt_ma_inst := id_reg_xcpt_ma_inst;
ex_reg_xcpt_itlb := id_reg_xcpt_itlb;
ex_reg_xcpt_ma_inst := io.imem.resp.bits.xcpt_ma
ex_reg_xcpt_itlb := io.imem.resp.bits.xcpt_if
ex_reg_xcpt_illegal := illegal_inst;
ex_reg_xcpt_privileged := (id_privileged & ~io.dpath.status(SR_S)).toBool;
ex_reg_xcpt_syscall := id_syscall.toBool;
ex_reg_fp_val := id_fp_val
ex_reg_fp_sboard_set := io.fpu.dec.sboard
ex_reg_vec_val := id_vec_val.toBool
ex_reg_replay := id_reg_replay
ex_reg_replay_next := id_replay_next
ex_reg_load_use := id_load_use;
}
ex_reg_xcpt_interrupt := !take_pc && id_interrupt
ex_reg_mem_cmd := id_mem_cmd
ex_reg_mem_type := id_mem_type.toUFix
ex_reg_xcpt_interrupt := id_reg_valid && id_interrupt && !take_pc
ex_reg_cause := id_cause
val beq = io.dpath.br_eq;
val bne = ~io.dpath.br_eq;
val blt = io.dpath.br_lt;
val bltu = io.dpath.br_ltu;
val bge = ~io.dpath.br_lt;
val bgeu = ~io.dpath.br_ltu;
val br_taken = !(wb_reg_dcache_miss && ex_reg_load_use) &&
((ex_reg_br_type === BR_EQ) && beq ||
(ex_reg_br_type === BR_NE) && bne ||
(ex_reg_br_type === BR_LT) && blt ||
(ex_reg_br_type === BR_LTU) && bltu ||
(ex_reg_br_type === BR_GE) && bge ||
(ex_reg_br_type === BR_GEU) && bgeu ||
(ex_reg_br_type === BR_J)) // treat J/JAL like taken branches
val jr_taken = !(wb_reg_dcache_miss && ex_reg_load_use) && ex_reg_br_type === BR_JR
val br_taken =
Mux(ex_reg_br_type === BR_EQ, io.dpath.br_eq,
Mux(ex_reg_br_type === BR_NE, ~io.dpath.br_eq,
Mux(ex_reg_br_type === BR_LT, io.dpath.br_lt,
Mux(ex_reg_br_type === BR_GE, ~io.dpath.br_lt,
Mux(ex_reg_br_type === BR_LTU, io.dpath.br_ltu,
Mux(ex_reg_br_type === BR_GEU, ~io.dpath.br_ltu,
ex_reg_br_type === BR_J))))))
val mem_reg_div_mul_val = Reg(){Bool()};
val mem_reg_div_val = Reg(){Bool()}
val mem_reg_mul_val = Reg(){Bool()}
val mem_reg_eret = Reg(){Bool()};
val mem_reg_mem_val = Reg(){Bool()};
val mem_reg_mem_cmd = Reg(){Bits()}
@ -574,7 +544,8 @@ class rocketCtrl extends Component
when (reset.toBool || io.dpath.killx) {
mem_reg_valid := Bool(false);
mem_reg_pcr := PCR_N
mem_reg_div_mul_val := Bool(false);
mem_reg_div_val := Bool(false)
mem_reg_mul_val := Bool(false)
mem_reg_wen := Bool(false);
mem_reg_fp_wen := Bool(false);
mem_reg_eret := Bool(false);
@ -589,11 +560,13 @@ class rocketCtrl extends Component
mem_reg_xcpt_syscall := Bool(false);
mem_reg_fp_val := Bool(false);
mem_reg_fp_sboard_set := Bool(false)
mem_reg_replay_next := Bool(false)
}
.otherwise {
mem_reg_valid := ex_reg_valid
mem_reg_pcr := ex_reg_pcr
mem_reg_div_mul_val := ex_reg_div_val || ex_reg_mul_val;
mem_reg_div_val := ex_reg_div_val && io.dpath.div_rdy
mem_reg_mul_val := ex_reg_mul_val && io.dpath.mul_rdy
mem_reg_wen := ex_reg_wen;
mem_reg_fp_wen := ex_reg_fp_wen;
mem_reg_eret := ex_reg_eret;
@ -608,13 +581,14 @@ class rocketCtrl extends Component
mem_reg_xcpt_syscall := ex_reg_xcpt_syscall;
mem_reg_fp_val := ex_reg_fp_val
mem_reg_fp_sboard_set := ex_reg_fp_sboard_set
mem_reg_replay_next := ex_reg_replay_next
}
mem_reg_mem_cmd := ex_reg_mem_cmd;
mem_reg_mem_type := ex_reg_mem_type;
mem_reg_xcpt_interrupt := ex_reg_xcpt_interrupt && !take_pc_wb
mem_reg_cause := ex_reg_cause
when (io.dpath.killm) {
when (ctrl_killm) {
wb_reg_valid := Bool(false)
wb_reg_pcr := PCR_N
wb_reg_wen := Bool(false);
@ -624,17 +598,19 @@ class rocketCtrl extends Component
wb_reg_div_mul_val := Bool(false);
wb_reg_fp_val := Bool(false)
wb_reg_fp_sboard_set := Bool(false)
wb_reg_replay_next := Bool(false)
}
.otherwise {
wb_reg_valid := mem_reg_valid
wb_reg_pcr := mem_reg_pcr
wb_reg_wen := mem_reg_wen;
wb_reg_fp_wen := mem_reg_fp_wen;
wb_reg_eret := mem_reg_eret;
wb_reg_eret := mem_reg_eret && !mem_reg_replay
wb_reg_flush_inst := mem_reg_flush_inst;
wb_reg_div_mul_val := mem_reg_div_mul_val;
wb_reg_div_mul_val := mem_reg_div_val || mem_reg_mul_val
wb_reg_fp_val := mem_reg_fp_val
wb_reg_fp_sboard_set := mem_reg_fp_sboard_set
wb_reg_replay_next := mem_reg_replay_next
}
val sboard = new rocketCtrlSboard(32, 3, 2);
@ -717,7 +693,7 @@ class rocketCtrl extends Component
UFix(0,5)))))))))))); // instruction address misaligned
// control transfer from ex/mem
val take_pc_ex = ex_reg_btb_hit != br_taken || jr_taken
val take_pc_ex = ex_reg_btb_hit != br_taken || ex_reg_jalr
take_pc_wb := wb_reg_replay || vec_replay || wb_reg_exception || wb_reg_eret
take_pc := take_pc_ex || take_pc_wb;
@ -726,22 +702,23 @@ class rocketCtrl extends Component
val dmem_kill_mem = mem_reg_valid && (io.dtlb_miss || io.dmem.resp.bits.nack)
val fpu_kill_mem = mem_reg_fp_val && io.fpu.nack_mem
val replay_mem = dmem_kill_mem || mem_reg_wen && mem_ll_wb || mem_reg_replay || fpu_kill_mem
val kill_mem = dmem_kill_mem || mem_reg_wen && mem_ll_wb || take_pc_wb || mem_exception || mem_reg_kill || fpu_kill_mem
val kill_dcache = io.dtlb_miss || mem_reg_wen && mem_ll_wb || take_pc_wb || mem_exception || mem_reg_kill
val killm_common = mem_reg_wen && mem_ll_wb || take_pc_wb || mem_exception || mem_reg_kill
ctrl_killm := killm_common || dmem_kill_mem || fpu_kill_mem
// replay execute stage PC when the D$ is blocked, when the D$ misses,
// for privileged instructions, and for fence.i instructions
val replay_ex = wb_reg_dcache_miss && ex_reg_load_use || mem_reg_flush_inst ||
ex_reg_replay || ex_reg_mem_val && !(io.dmem.req.ready && io.dtlb_rdy) ||
ex_reg_mem_val && !(io.dmem.req.ready && io.dtlb_rdy) ||
ex_reg_div_val && !io.dpath.div_rdy ||
ex_reg_mul_val && !io.dpath.mul_rdy
ex_reg_mul_val && !io.dpath.mul_rdy ||
mem_reg_replay_next
val kill_ex = take_pc_wb || replay_ex
mem_reg_replay := replay_ex && !take_pc_wb;
mem_reg_kill := kill_ex;
wb_reg_replay := replay_mem && !take_pc_wb
wb_reg_exception := mem_exception && !take_pc_wb;
wb_reg_exception := mem_exception && !take_pc_wb && !wb_reg_replay_next
wb_reg_cause := mem_cause;
val replay_wb = wb_reg_replay || vec_replay || io.dpath.pcr_replay
@ -755,19 +732,16 @@ class rocketCtrl extends Component
io.dpath.vec_irq_aux_wen := wb_reg_exception && wb_reg_cause >= UFix(24) && wb_reg_cause < UFix(32)
io.dpath.sel_pc :=
Mux(wb_reg_exception, PC_PCR, // exception
Mux(replay_wb, PC_WB, // replay
Mux(wb_reg_eret, PC_PCR, // eret instruction
Mux(ex_reg_btb_hit && !br_taken, PC_EX4, // mispredicted not taken branch
Mux(!ex_reg_btb_hit && br_taken, PC_BR, // mispredicted taken branch
Mux(jr_taken, PC_JR, // taken JALR
Mux(io.dpath.btb_hit, PC_BTB, // predicted PC from BTB
PC_4))))))); // PC+4
Mux(wb_reg_exception, PC_PCR, // exception
Mux(wb_reg_eret, PC_PCR, // eret instruction
Mux(replay_wb, PC_WB, // replay
Mux(ex_reg_jalr, PC_EX, // JALR
Mux(!ex_reg_btb_hit, PC_EX, // mispredicted taken branch
PC_EX4))))) // mispredicted not taken branch
io.dpath.wen_btb := !ex_reg_btb_hit && br_taken
io.dpath.clr_btb := ex_reg_btb_hit && !br_taken || id_reg_icmiss;
io.imem.req_val := !reset.toBool && (take_pc_wb || !mem_reg_replay && !ex_reg_replay && (take_pc_ex || !id_reg_replay))
io.imem.req.bits.mispredict := !take_pc_wb && !ex_reg_jalr && ex_reg_btb_hit != br_taken
io.imem.req.bits.taken := !ex_reg_btb_hit
io.imem.req.valid := take_pc
// stall for RAW/WAW hazards on loads, AMOs, and mul/div in execute stage.
val data_hazard_ex = ex_reg_wen &&
@ -795,7 +769,7 @@ class rocketCtrl extends Component
io.fpu.dec.ren2 && id_raddr2 === io.dpath.mem_waddr ||
io.fpu.dec.ren3 && id_raddr3 === io.dpath.mem_waddr ||
io.fpu.dec.wen && id_waddr === io.dpath.mem_waddr)
val id_mem_hazard = data_hazard_mem && (mem_reg_mem_val && mem_mem_cmd_bh || mem_reg_div_mul_val || mem_reg_fp_val) ||
val id_mem_hazard = data_hazard_mem && (mem_reg_mem_val && mem_mem_cmd_bh || mem_reg_div_val || mem_reg_mul_val || mem_reg_fp_val) ||
fp_data_hazard_mem && mem_reg_fp_val
id_load_use := mem_reg_mem_val && (data_hazard_mem || fp_data_hazard_mem)
@ -812,28 +786,24 @@ class rocketCtrl extends Component
val id_wb_hazard = data_hazard_wb && (wb_reg_dcache_miss || wb_reg_div_mul_val) ||
fp_data_hazard_wb && (wb_reg_dcache_miss || wb_reg_fp_val)
val killd_common = take_pc || id_interrupt || ex_reg_replay_next
val ctrl_killd = killd_common || !io.imem.resp.valid
val ctrl_stalld =
!take_pc &&
(
id_ex_hazard || id_mem_hazard || id_wb_hazard ||
id_stall_raddr1 || id_stall_raddr2 || id_stall_waddr ||
id_fp_val && id_stall_fpu ||
id_mem_val.toBool && !(io.dmem.req.ready && io.dtlb_rdy) ||
((id_sync === SYNC_D) || (id_sync === SYNC_I)) && !io.dmem.req.ready ||
vec_stalld
);
val ctrl_stallf = ctrl_stalld;
val ctrl_killd = take_pc || ctrl_stalld;
val ctrl_killf = take_pc || !io.imem.resp_val;
id_ex_hazard || id_mem_hazard || id_wb_hazard ||
id_stall_raddr1 || id_stall_raddr2 || id_stall_waddr ||
id_fp_val && id_stall_fpu ||
id_mem_val && !(io.dmem.req.ready && io.dtlb_rdy) ||
(id_sync === SYNC_D || id_sync === SYNC_I) && !io.dmem.req.ready ||
vec_stalld
io.dpath.flush_inst := wb_reg_flush_inst;
io.dpath.stallf := ctrl_stallf;
io.dpath.stalld := ctrl_stalld;
io.dpath.killf := ctrl_killf;
io.dpath.killd := ctrl_killd;
io.dpath.stalld := !ctrl_killd && ctrl_stalld;
io.dpath.killd := ctrl_killd || ctrl_stalld
io.dpath.killx := kill_ex;
io.dpath.killm := kill_mem;
io.dpath.killm := killm_common
io.imem.resp.ready := killd_common || !ctrl_stalld
io.imem.req.bits.invalidate := wb_reg_flush_inst
io.dpath.mem_load := mem_reg_mem_val && mem_reg_wen
io.dpath.ren2 := id_renx2.toBool;
@ -841,12 +811,15 @@ class rocketCtrl extends Component
io.dpath.sel_alu2 := id_sel_alu2.toUFix
io.dpath.fn_dw := id_fn_dw.toBool;
io.dpath.fn_alu := id_fn_alu.toUFix
io.dpath.div_fn := id_div_fn.toUFix
io.dpath.div_val := id_div_val.toBool && id_waddr != UFix(0);
io.dpath.mul_fn := id_mul_fn.toUFix
io.dpath.mul_val := id_mul_val.toBool && id_waddr != UFix(0);
io.dpath.div_fn := ex_reg_mul_fn
io.dpath.div_val := ex_reg_div_val
io.dpath.div_kill := mem_reg_div_val && killm_common
io.dpath.mul_fn := ex_reg_mul_fn
io.dpath.mul_val := ex_reg_mul_val
io.dpath.mul_kill := mem_reg_mul_val && killm_common
io.dpath.ex_fp_val:= ex_reg_fp_val;
io.dpath.mem_fp_val:= mem_reg_fp_val;
io.dpath.ex_jalr := ex_reg_jalr
io.dpath.ex_wen := ex_reg_wen;
io.dpath.mem_wen := mem_reg_wen;
io.dpath.wb_wen := wb_reg_wen;
@ -860,12 +833,12 @@ class rocketCtrl extends Component
io.fpu.valid := !io.dpath.killd && id_fp_val
io.fpu.killx := kill_ex
io.fpu.killm := kill_mem
io.fpu.killm := killm_common
io.dtlb_val := ex_reg_mem_val
io.dtlb_kill := mem_reg_kill
io.dmem.req.valid := ex_reg_mem_val
io.dmem.req.bits.kill := kill_dcache
io.dmem.req.bits.kill := killm_common || io.dtlb_miss
io.dmem.req.bits.cmd := ex_reg_mem_cmd
io.dmem.req.bits.typ := ex_reg_mem_type
}

84
rocket/src/main/scala/divider.scala
View File

@ -4,8 +4,10 @@ import Chisel._
import Node._
import Constants._
class rocketDivider(w: Int, earlyOut: Boolean = false) extends Component {
class rocketDivider(earlyOut: Boolean = false) extends Component {
val io = new ioMultiplier
val w0 = io.req.bits.in0.getWidth
val w = w0+1 // sign bit
val s_ready :: s_neg_inputs :: s_busy :: s_neg_outputs :: s_done :: Nil = Enum(5) { UFix() };
val state = Reg(resetVal = s_ready);
@ -26,57 +28,16 @@ class rocketDivider(w: Int, earlyOut: Boolean = false) extends Component {
val fn = io.req.bits.fn(io.req.bits.fn.width-2,0)
val tc = (fn === DIV_D) || (fn === DIV_R);
switch (state) {
is (s_ready) {
when (io.req.valid) {
state := Mux(tc, s_neg_inputs, s_busy)
}
}
is (s_neg_inputs) {
state := Mux(io.req_kill, s_ready, s_busy)
}
is (s_busy) {
when (io.req_kill && Reg(io.req.ready)) {
state := s_ready
}
.elsewhen (count === UFix(w)) {
state := Mux(neg_quo || neg_rem, s_neg_outputs, s_done)
}
}
is (s_neg_outputs) {
state := s_done
}
is (s_done) {
when (io.resp_rdy) {
state := s_ready
}
}
}
// state machine
val lhs_sign = tc && Mux(dw === DW_64, io.req.bits.in0(w0-1), io.req.bits.in0(w0/2-1))
val lhs_hi = Mux(dw === DW_64, io.req.bits.in0(w0-1,w0/2), Fill(w0/2, lhs_sign))
val lhs_in = Cat(lhs_sign, lhs_hi, io.req.bits.in0(w0/2-1,0))
val lhs_sign = tc && Mux(dw === DW_64, io.req.bits.in0(w-1), io.req.bits.in0(w/2-1))
val lhs_hi = Mux(dw === DW_64, io.req.bits.in0(w-1,w/2), Fill(w/2, lhs_sign))
val lhs_in = Cat(lhs_hi, io.req.bits.in0(w/2-1,0))
val rhs_sign = tc && Mux(dw === DW_64, io.req.bits.in1(w-1), io.req.bits.in1(w/2-1))
val rhs_hi = Mux(dw === DW_64, io.req.bits.in1(w-1,w/2), Fill(w/2, rhs_sign))
val rhs_in = Cat(rhs_hi, io.req.bits.in1(w/2-1,0))
val rhs_sign = tc && Mux(dw === DW_64, io.req.bits.in1(w0-1), io.req.bits.in1(w0/2-1))
val rhs_hi = Mux(dw === DW_64, io.req.bits.in1(w0-1,w0/2), Fill(w0/2, rhs_sign))
val rhs_in = Cat(rhs_sign, rhs_hi, io.req.bits.in1(w0/2-1,0))
when (io.req.fire()) {
count := UFix(0)
half := (dw === DW_32);
neg_quo := Bool(false);
neg_rem := Bool(false);
rem := (fn === DIV_R) || (fn === DIV_RU);
reg_tag := io.req_tag;
divby0 := Bool(true);
divisor := rhs_in
remainder := lhs_in
}
when (state === s_neg_inputs) {
neg_rem := remainder(w-1)
neg_quo := (remainder(w-1) != divisor(w-1))
state := s_busy
when (remainder(w-1)) {
remainder := Cat(remainder(2*w, w), -remainder(w-1,0))
}
@ -85,6 +46,7 @@ class rocketDivider(w: Int, earlyOut: Boolean = false) extends Component {
}
}
when (state === s_neg_outputs) {
state := s_done
when (neg_rem && neg_quo && !divby0) {
remainder := Cat(-remainder(2*w, w+1), remainder(w), -remainder(w-1,0))
}
@ -96,6 +58,9 @@ class rocketDivider(w: Int, earlyOut: Boolean = false) extends Component {
}
}
when (state === s_busy) {
when (count === UFix(w)) {
state := Mux(neg_quo || neg_rem, s_neg_outputs, s_done)
}
count := count + UFix(1)
val msb = subtractor(w)
@ -112,11 +77,26 @@ class rocketDivider(w: Int, earlyOut: Boolean = false) extends Component {
remainder := remainder << shift
count := shift
}
}
}
when (state === s_done && io.resp_rdy || io.req_kill) {
state := s_ready
}
when (io.req.fire()) {
state := Mux(lhs_sign || rhs_sign, s_neg_inputs, s_busy)
count := UFix(0)
half := (dw === DW_32);
neg_quo := lhs_sign != rhs_sign
neg_rem := lhs_sign
rem := (fn === DIV_R) || (fn === DIV_RU);
reg_tag := io.req_tag;
divby0 := Bool(true);
divisor := rhs_in
remainder := lhs_in
}
val result = Mux(rem, remainder(2*w, w+1), remainder(w-1,0))
val result = Mux(rem, remainder(w+w0, w+1), remainder(w0-1,0))
io.resp_bits := Mux(half, Cat(Fill(w/2, result(w/2-1)), result(w/2-1,0)), result)
io.resp_bits := Mux(half, Cat(Fill(w0/2, result(w0/2-1)), result(w0/2-1,0)), result)
io.resp_tag := reg_tag
io.resp_val := state === s_done
io.req.ready := state === s_ready

127
rocket/src/main/scala/dpath.scala
View File

@ -7,19 +7,13 @@ import Constants._
import Instructions._
import hwacha._
class ioDpathImem extends Bundle()
{
val req_addr = UFix(OUTPUT, VADDR_BITS+1);
val resp_data = Bits(INPUT, 32);
}
class ioDpathAll extends Bundle()
{
val host = new ioHTIF();
val ctrl = new ioCtrlDpath().flip
val dmem = new ioHellaCache
val dtlb = new ioDTLB_CPU_req_bundle().asOutput()
val imem = new ioDpathImem();
val imem = new IOCPUFrontend
val ptbr_wen = Bool(OUTPUT);
val ptbr = UFix(OUTPUT, PADDR_BITS);
val fpu = new ioDpathFPU();
@ -32,27 +26,16 @@ class ioDpathAll extends Bundle()
class rocketDpath extends Component
{
val io = new ioDpathAll();
val btb = new rocketDpathBTB(4); // # of entries in BTB
val if_btb_target = btb.io.target;
val pcr = new rocketDpathPCR();
val ex_pcr = pcr.io.r.data;
val alu = new rocketDpathALU();
val alu = new ALU
val ex_alu_out = alu.io.out;
val ex_alu_adder_out = alu.io.adder_out;
val rfile = new rocketDpathRegfile();
// instruction fetch definitions
val if_reg_pc = Reg(resetVal = UFix(START_ADDR,VADDR_BITS+1));
// instruction decode definitions
val id_reg_inst = Reg(resetVal = NOP);
val id_reg_pc = Reg() { UFix(width = VADDR_BITS+1) };
// execute definitions
val ex_reg_pc = Reg() { UFix() };
val ex_reg_inst = Reg() { Bits() };
@ -62,13 +45,8 @@ class rocketDpath extends Component
val ex_reg_rs2 = Reg() { Bits() };
val ex_reg_rs1 = Reg() { Bits() };
val ex_reg_waddr = Reg() { UFix() };
val ex_reg_ctrl_eret = Reg(resetVal = Bool(false));
val ex_reg_ctrl_fn_dw = Reg() { UFix() };
val ex_reg_ctrl_fn_alu = Reg() { UFix() };
val ex_reg_ctrl_mul_val = Reg(resetVal = Bool(false));
val ex_reg_ctrl_mul_fn = Reg() { UFix() };
val ex_reg_ctrl_div_val = Reg(resetVal = Bool(false));
val ex_reg_ctrl_div_fn = Reg() { UFix() };
val ex_reg_ctrl_sel_wb = Reg() { UFix() };
val ex_wdata = Bits();
@ -99,9 +77,6 @@ class rocketDpath extends Component
val r_dmem_resp_replay = Reg(resetVal = Bool(false));
val r_dmem_fp_replay = Reg(resetVal = Bool(false));
val r_dmem_resp_waddr = Reg() { UFix() };
// instruction fetch stage
val if_pc_plus4 = if_reg_pc + UFix(4);
val ex_pc_plus4 = ex_reg_pc + UFix(4);
val ex_branch_target = ex_reg_pc + Cat(ex_reg_op2(VADDR_BITS-1,0), Bits(0,1)).toUFix
@ -109,41 +84,24 @@ class rocketDpath extends Component
val ex_ea_sign = Mux(ex_alu_adder_out(VADDR_BITS-1), ~ex_alu_adder_out(63,VADDR_BITS) === UFix(0), ex_alu_adder_out(63,VADDR_BITS) != UFix(0))
val ex_effective_address = Cat(ex_ea_sign, ex_alu_adder_out(VADDR_BITS-1,0)).toUFix
val if_next_pc =
Mux(io.ctrl.sel_pc === PC_BTB, Cat(if_btb_target(VADDR_BITS-1), if_btb_target),
Mux(io.ctrl.sel_pc === PC_EX4, ex_pc_plus4,
Mux(io.ctrl.sel_pc === PC_BR, ex_branch_target,
Mux(io.ctrl.sel_pc === PC_JR, ex_effective_address,
Mux(io.ctrl.sel_pc === PC_PCR, Cat(pcr.io.evec(VADDR_BITS-1), pcr.io.evec),
Mux(io.ctrl.sel_pc === PC_WB, wb_reg_pc,
if_pc_plus4)))))) // PC_4
when (!io.ctrl.stallf) {
if_reg_pc := if_next_pc.toUFix;
}
io.ctrl.xcpt_ma_inst := if_next_pc(1,0) != Bits(0)
io.imem.req_addr :=
Mux(io.ctrl.stallf, if_reg_pc,
if_next_pc.toUFix);
btb.io.current_pc := if_reg_pc;
btb.io.hit <> io.ctrl.btb_hit;
btb.io.wen <> io.ctrl.wen_btb;
btb.io.clr <> io.ctrl.clr_btb;
btb.io.correct_pc := ex_reg_pc;
btb.io.correct_target := ex_branch_target
btb.io.invalidate := io.ctrl.flush_inst
// hook up I$
io.imem.req.bits.invalidateTLB := pcr.io.ptbr_wen
io.imem.req.bits.currentpc := ex_reg_pc
io.imem.req.bits.status := pcr.io.status
io.imem.req.bits.pc :=
Mux(io.ctrl.sel_pc === PC_EX4, ex_pc_plus4,
Mux(io.ctrl.sel_pc === PC_EX, Mux(io.ctrl.ex_jalr, ex_effective_address, ex_branch_target),
Mux(io.ctrl.sel_pc === PC_PCR, Cat(pcr.io.evec(VADDR_BITS-1), pcr.io.evec).toUFix,
wb_reg_pc))) // PC_WB
// instruction decode stage
when (!io.ctrl.stalld) {
id_reg_pc := if_reg_pc;
id_reg_inst := Mux(io.ctrl.killf, NOP, io.imem.resp_data)
}
val id_inst = io.imem.resp.bits.data
val id_pc = io.imem.resp.bits.pc
debug(id_inst)
debug(id_pc)
val id_raddr1 = id_reg_inst(26,22).toUFix;
val id_raddr2 = id_reg_inst(21,17).toUFix;
val id_raddr1 = id_inst(26,22).toUFix;
val id_raddr2 = id_inst(21,17).toUFix;
// regfile read
rfile.io.r0.en <> io.ctrl.ren2;
@ -156,7 +114,7 @@ class rocketDpath extends Component
// destination register selection
val id_waddr =
Mux(io.ctrl.sel_wa === WA_RD, id_reg_inst(31,27).toUFix,
Mux(io.ctrl.sel_wa === WA_RD, id_inst(31,27).toUFix,
RA); // WA_RA
// bypass muxes
@ -185,26 +143,26 @@ class rocketDpath extends Component
val id_imm_l = io.ctrl.sel_alu2 === A2_LTYPE
val id_imm_zero = io.ctrl.sel_alu2 === A2_ZERO || io.ctrl.sel_alu2 === A2_RTYPE
val id_imm_ibz = io.ctrl.sel_alu2 === A2_ITYPE || io.ctrl.sel_alu2 === A2_BTYPE || id_imm_zero
val id_imm_sign = Mux(id_imm_bj, id_reg_inst(31),
Mux(id_imm_l, id_reg_inst(26),
val id_imm_sign = Mux(id_imm_bj, id_inst(31),
Mux(id_imm_l, id_inst(26),
Mux(id_imm_zero, Bits(0,1),
id_reg_inst(21)))) // IMM_ITYPE
id_inst(21)))) // IMM_ITYPE
val id_imm_small = Mux(id_imm_zero, Bits(0,12),
Cat(Mux(id_imm_bj, id_reg_inst(31,27), id_reg_inst(21,17)), id_reg_inst(16,10)))
Cat(Mux(id_imm_bj, id_inst(31,27), id_inst(21,17)), id_inst(16,10)))
val id_imm = Cat(Fill(32, id_imm_sign),
Mux(id_imm_l, Cat(id_reg_inst(26,7), Bits(0,12)),
Mux(id_imm_l, Cat(id_inst(26,7), Bits(0,12)),
Mux(id_imm_ibz, Cat(Fill(20, id_imm_sign), id_imm_small),
Cat(Fill(7, id_imm_sign), id_reg_inst(31,7))))) // A2_JTYPE
Cat(Fill(7, id_imm_sign), id_inst(31,7))))) // A2_JTYPE
val id_op2_dmem_bypass = id_rs2_dmem_bypass && io.ctrl.sel_alu2 === A2_RTYPE
val id_op2 = Mux(io.ctrl.sel_alu2 === A2_RTYPE, id_rs2, id_imm)
io.ctrl.inst := id_reg_inst
io.fpu.inst := id_reg_inst
io.ctrl.inst := id_inst
io.fpu.inst := id_inst
// execute stage
ex_reg_pc := id_reg_pc;
ex_reg_inst := id_reg_inst
ex_reg_pc := id_pc
ex_reg_inst := id_inst
ex_reg_raddr1 := id_raddr1
ex_reg_raddr2 := id_raddr2;
ex_reg_op2 := id_op2;
@ -213,21 +171,8 @@ class rocketDpath extends Component
ex_reg_waddr := id_waddr;
ex_reg_ctrl_fn_dw := io.ctrl.fn_dw.toUFix;
ex_reg_ctrl_fn_alu := io.ctrl.fn_alu;
ex_reg_ctrl_mul_fn := io.ctrl.mul_fn;
ex_reg_ctrl_div_fn := io.ctrl.div_fn;
ex_reg_ctrl_sel_wb := io.ctrl.sel_wb;
when(io.ctrl.killd) {
ex_reg_ctrl_div_val := Bool(false);
ex_reg_ctrl_mul_val := Bool(false);
ex_reg_ctrl_eret := Bool(false);
}
.otherwise {
ex_reg_ctrl_div_val := io.ctrl.div_val;
ex_reg_ctrl_mul_val := io.ctrl.mul_val;
ex_reg_ctrl_eret := io.ctrl.id_eret;
}
val ex_rs1 = Mux(Reg(id_rs1_dmem_bypass), wb_reg_dmem_wdata, ex_reg_rs1)
val ex_rs2 = Mux(Reg(id_rs2_dmem_bypass), wb_reg_dmem_wdata, ex_reg_rs2)
val ex_op2 = Mux(Reg(id_op2_dmem_bypass), wb_reg_dmem_wdata, ex_reg_op2)
@ -240,19 +185,19 @@ class rocketDpath extends Component
io.fpu.fromint_data := ex_rs1
// divider
val div = new rocketDivider(64)
div.io.req.valid := ex_reg_ctrl_div_val
div.io.req.bits.fn := Cat(ex_reg_ctrl_fn_dw, ex_reg_ctrl_div_fn)
val div = new rocketDivider(earlyOut = true)
div.io.req.valid := io.ctrl.div_val
div.io.req.bits.fn := Cat(ex_reg_ctrl_fn_dw, io.ctrl.div_fn)
div.io.req.bits.in0 := ex_rs1
div.io.req.bits.in1 := ex_rs2
div.io.req_tag := ex_reg_waddr
div.io.req_kill := io.ctrl.killm
div.io.req_kill := io.ctrl.div_kill
div.io.resp_rdy := !dmem_resp_replay
io.ctrl.div_rdy := div.io.req.ready
io.ctrl.div_result_val := div.io.resp_val
// multiplier
var mul_io = new rocketMultiplier(unroll = 6).io
var mul_io = new rocketMultiplier(unroll = 4, earlyOut = true).io
if (HAVE_VEC)
{
val vu_mul = new rocketVUMultiplier(nwbq = 1)
@ -260,12 +205,12 @@ class rocketDpath extends Component
vu_mul.io.vu.resp <> io.vec_imul_resp
mul_io = vu_mul.io.cpu
}
mul_io.req.valid := ex_reg_ctrl_mul_val;
mul_io.req.bits.fn := Cat(ex_reg_ctrl_fn_dw, ex_reg_ctrl_mul_fn)
mul_io.req.valid := io.ctrl.mul_val
mul_io.req.bits.fn := Cat(ex_reg_ctrl_fn_dw, io.ctrl.mul_fn)
mul_io.req.bits.in0 := ex_rs1
mul_io.req.bits.in1 := ex_rs2
mul_io.req_tag := ex_reg_waddr
mul_io.req_kill := io.ctrl.killm
mul_io.req_kill := io.ctrl.mul_kill
mul_io.resp_rdy := !dmem_resp_replay && !div.io.resp_val
io.ctrl.mul_rdy := mul_io.req.ready
io.ctrl.mul_result_val := mul_io.resp_val

32
rocket/src/main/scala/dpath_alu.scala
View File

@ -15,27 +15,45 @@ class ioALU extends Bundle(){
val adder_out = UFix(OUTPUT, 64);
}
class rocketDpathALU extends Component
object ALU
{
val FN_X = Bits("b????")
val FN_ADD = UFix(0)
val FN_SL = UFix(1)
val FN_XOR = UFix(4)
val FN_OR = UFix(6)
val FN_AND = UFix(7)
val FN_SR = UFix(5)
val FN_SUB = UFix(8)
val FN_SLT = UFix(10)
val FN_SLTU = UFix(11)
val FN_SRA = UFix(13)
val FN_OP2 = UFix(15)
def isSub(cmd: Bits) = cmd(3)
def isSLTU(cmd: Bits) = cmd(0)
}
class ALU extends Component
{
import ALU._
val io = new ioALU();
// ADD, SUB
val sub = (io.fn === FN_SUB) || (io.fn === FN_SLT) || (io.fn === FN_SLTU)
val sub = isSub(io.fn)
val adder_rhs = Mux(sub, ~io.in2, io.in2)
val sum = (io.in1 + adder_rhs + sub.toUFix)(63,0)
// SLT, SLTU
val less = Mux(io.in1(63) === io.in2(63), sum(63),
Mux(io.fn === FN_SLT, io.in1(63), io.in2(63)))
Mux(isSLTU(io.fn), io.in2(63), io.in1(63)))
// SLL, SRL, SRA
val sra = (io.fn === FN_SRA)
val shamt = Cat(io.in2(5) & (io.dw === DW_64), io.in2(4,0)).toUFix
val shright = sra || (io.fn === FN_SR)
val shin_hi_32 = Mux(sra, Fill(32, io.in1(31)), UFix(0,32))
val shin_hi_32 = Mux(isSub(io.fn), Fill(32, io.in1(31)), UFix(0,32))
val shin_hi = Mux(io.dw === DW_64, io.in1(63,32), shin_hi_32)
val shin = Cat(shin_hi, io.in1(31,0))
val shout_r = (Cat(sra & shin(63), shin).toFix >> shamt)(63,0)
val shout_r = (Cat(isSub(io.fn) & shin(63), shin).toFix >> shamt)(63,0)
val shout_l = (shin << shamt)(63,0)
val bitwise_logic =

13
rocket/src/main/scala/dpath_util.scala
View File

@ -36,13 +36,14 @@ class rocketDpathBTB(entries: Int) extends Component
val valid = Reg(resetVal = Bool(false))
val my_hit = valid && tag === io.current_pc
val my_update = valid && tag === io.correct_pc
val my_clr = io.clr && my_update || io.invalidate
val my_wen = io.wen && (my_update || !update && UFix(i) === repl_way)
valid := !my_clr && (valid || my_wen)
when (my_wen) {
tag := io.correct_pc
target := io.correct_target
when (io.wen && (my_update || !update && UFix(i) === repl_way)) {
valid := Bool(false)
when (!io.clr) {
valid := Bool(true)
tag := io.correct_pc
target := io.correct_target
}
}
hit_reduction = hit_reduction || my_hit

183
rocket/src/main/scala/dtlb.scala
View File

@ -1,183 +0,0 @@
package rocket
import Chisel._;
import Node._;
import Constants._;
import scala.math._;
import hwacha._
// ioDTLB_CPU also located in hwacha/src/vuVXU-Interface.scala
// should keep them in sync
class ioDTLB_CPU_req_bundle extends Bundle
{
// lookup requests
val kill = Bool()
val cmd = Bits(width=4) // load/store/amo
val asid = Bits(width=ASID_BITS)
val vpn = Bits(width=VPN_BITS+1)
}
class ioDTLB_CPU_req extends FIFOIO()( { new ioDTLB_CPU_req_bundle() } )
class ioDTLB_CPU_resp extends Bundle
{
// lookup responses
val miss = Bool(OUTPUT)
val ppn = Bits(OUTPUT, PPN_BITS)
val xcpt_ld = Bool(OUTPUT)
val xcpt_st = Bool(OUTPUT)
val xcpt_pf = Bool(OUTPUT)
}
class ioDTLB extends Bundle
{
// status bits (from PCR), to check current permission and whether VM is enabled
val status = Bits(INPUT, 32)
// invalidate all TLB entries
val invalidate = Bool(INPUT)
val cpu_req = new ioDTLB_CPU_req().flip
val cpu_resp = new ioDTLB_CPU_resp()
val ptw = new ioTLB_PTW()
}
class rocketDTLB(entries: Int) extends Component
{
val io = new ioDTLB();
val addr_bits = ceil(log10(entries)/log10(2)).toInt;
val s_ready :: s_request :: s_wait :: Nil = Enum(3) { UFix() };
val state = Reg(resetVal = s_ready);
val r_cpu_req_val = Reg(resetVal = Bool(false));
val r_cpu_req_vpn = Reg() { Bits() }
val r_cpu_req_cmd = Reg() { Bits() }
val r_cpu_req_asid = Reg() { Bits() }
val r_refill_tag = Reg() { Bits() }
val r_refill_waddr = Reg() { UFix() }
when (io.cpu_req.valid && io.cpu_req.ready) {
r_cpu_req_vpn := io.cpu_req.bits.vpn;
r_cpu_req_cmd := io.cpu_req.bits.cmd;
r_cpu_req_asid := io.cpu_req.bits.asid;
r_cpu_req_val := Bool(true);
}
.otherwise {
r_cpu_req_val := Bool(false);
}
val req_load = (r_cpu_req_cmd === M_XRD);
val req_store = (r_cpu_req_cmd === M_XWR);
val req_amo = r_cpu_req_cmd(3).toBool;
val req_pf = (r_cpu_req_cmd === M_PFR) || (r_cpu_req_cmd === M_PFW)
val bad_va = r_cpu_req_vpn(VPN_BITS) != r_cpu_req_vpn(VPN_BITS-1);
val tag_cam = new rocketCAM(entries, ASID_BITS+VPN_BITS);
val tag_ram = Mem(entries) { io.ptw.resp_ppn.clone }
when (io.ptw.resp_val) { tag_ram(r_refill_waddr) := io.ptw.resp_ppn }
val lookup_tag = Cat(r_cpu_req_asid, r_cpu_req_vpn);
tag_cam.io.clear := io.invalidate;
tag_cam.io.clear_hit := io.cpu_resp.xcpt_ld || io.cpu_resp.xcpt_st || io.cpu_resp.xcpt_pf
tag_cam.io.tag := lookup_tag;
tag_cam.io.write := io.ptw.resp_val || io.ptw.resp_err;
tag_cam.io.write_tag := r_refill_tag;
tag_cam.io.write_addr := r_refill_waddr;
val tag_hit = tag_cam.io.hit || bad_va;
val tag_hit_addr = tag_cam.io.hit_addr;
// extract fields from status register
val status_s = io.status(SR_S).toBool; // user/supervisor mode
val status_u = !status_s;
val status_vm = io.status(SR_VM).toBool // virtual memory enable
// extract fields from PT permission bits
val ptw_perm_ur = io.ptw.resp_perm(2);
val ptw_perm_uw = io.ptw.resp_perm(1);
val ptw_perm_sr = io.ptw.resp_perm(5);
val ptw_perm_sw = io.ptw.resp_perm(4);
// permission bit arrays
val ur_array = Reg(resetVal = Bits(0, entries)); // user read permission
val uw_array = Reg(resetVal = Bits(0, entries)); // user write permission
val sr_array = Reg(resetVal = Bits(0, entries)); // supervisor read permission
val sw_array = Reg(resetVal = Bits(0, entries)); // supervisor write permission
when (io.ptw.resp_val) {
ur_array := ur_array.bitSet(r_refill_waddr, ptw_perm_ur);
uw_array := uw_array.bitSet(r_refill_waddr, ptw_perm_uw);
sr_array := sr_array.bitSet(r_refill_waddr, ptw_perm_sr);
sw_array := sw_array.bitSet(r_refill_waddr, ptw_perm_sw);
}
// when the page table lookup reports an error, set all permission
// bits to 0 so the next access will cause an exception
when (io.ptw.resp_err) {
ur_array := ur_array.bitSet(r_refill_waddr, Bool(false));
uw_array := uw_array.bitSet(r_refill_waddr, Bool(false));
sr_array := sr_array.bitSet(r_refill_waddr, Bool(false));
sw_array := sw_array.bitSet(r_refill_waddr, Bool(false));
}
// high if there are any unused (invalid) entries in the TLB
val has_invalid_entry = !tag_cam.io.valid_bits.andR
val invalid_entry = PriorityEncoder(~tag_cam.io.valid_bits)
val plru = new PseudoLRU(entries)
val repl_waddr = Mux(has_invalid_entry, invalid_entry, plru.replace).toUFix;
val lookup = (state === s_ready) && status_vm && r_cpu_req_val && (req_load || req_store || req_amo || req_pf);
val lookup_hit = lookup && tag_hit;
val lookup_miss = lookup && !tag_hit;
val tlb_hit = !io.cpu_req.bits.kill && lookup_hit;
val tlb_miss = !io.cpu_req.bits.kill && lookup_miss;
// currently replace TLB entries in LIFO order
// TODO: implement LRU replacement policy
when (tlb_miss) {
r_refill_tag := lookup_tag;
r_refill_waddr := repl_waddr;
}
when (tlb_hit) {
plru.access(tag_hit_addr)
}
val load_fault_common = tlb_hit &&
((status_s && !sr_array(tag_hit_addr)) ||
(status_u && !ur_array(tag_hit_addr)) ||
bad_va)
val store_fault_common = tlb_hit &&
((status_s && !sw_array(tag_hit_addr)) ||
(status_u && !uw_array(tag_hit_addr)) ||
bad_va)
io.cpu_resp.xcpt_ld := load_fault_common && (req_load || req_amo)
io.cpu_resp.xcpt_st := store_fault_common && (req_store || req_amo)
io.cpu_resp.xcpt_pf := load_fault_common && req_pf
io.cpu_req.ready := (state === s_ready) && !lookup_miss;
io.cpu_resp.miss := tlb_miss;
io.cpu_resp.ppn :=
Mux(status_vm, tag_ram(tag_hit_addr), r_cpu_req_vpn(PPN_BITS-1,0));
io.ptw.req_val := (state === s_request);
io.ptw.req_vpn := r_refill_tag(VPN_BITS-1,0);
// control state machine
switch (state) {
is (s_ready) {
when (tlb_miss) {
state := s_request;
}
}
is (s_request) {
when (io.ptw.req_rdy) {
state := s_wait;
}
}
is (s_wait) {
when (io.ptw.resp_val || io.ptw.resp_err) {
state := s_ready;
}
}
}
}

333
rocket/src/main/scala/icache.scala
View File

@ -6,172 +6,249 @@ import Constants._;
import scala.math._;
import uncore._
// interface between I$ and pipeline/ITLB (32 bits wide)
class ioImem extends Bundle
case class ICacheConfig(co: CoherencePolicyWithUncached, sets: Int, assoc: Int, parity: Boolean = false)
{
val invalidate = Bool(INPUT);
val itlb_miss = Bool(INPUT);
val req_val = Bool(INPUT);
val req_idx = Bits(INPUT, PGIDX_BITS);
val req_ppn = Bits(INPUT, PPN_BITS);
val resp_data = Bits(OUTPUT, 32);
val resp_val = Bool(OUTPUT);
val w = 1
val ibytes = INST_BITS/8
val dm = assoc == 1
val lines = sets * assoc
val databits = MEM_DATA_BITS
val datawidth = databits + (if (parity) 1 else 0)
val idxbits = log2Up(sets)
val offbits = OFFSET_BITS
val untagbits = idxbits + offbits
val tagbits = PADDR_BITS - untagbits
val tagwidth = tagbits + (if (parity) 1 else 0)
require(isPow2(sets) && isPow2(assoc))
require(isPow2(w) && isPow2(ibytes))
require(PGIDX_BITS >= untagbits)
}
class ioRocketICache extends Bundle()
{
val cpu = new ioImem();
val mem = new ioUncachedRequestor
class FrontendReq extends Bundle {
val pc = UFix(width = VADDR_BITS+1)
val status = Bits(width = 32)
val invalidate = Bool()
val invalidateTLB = Bool()
val mispredict = Bool()
val taken = Bool()
val currentpc = UFix(width = VADDR_BITS+1)
}
// basic direct mapped instruction cache
// 32 bit wide cpu port, 128 bit wide memory port, 64 byte cachelines
// parameters :
// lines = # cache lines
class rocketICache(sets: Int, assoc: Int, co: CoherencePolicyWithUncached) extends Component
class FrontendResp extends Bundle {
val pc = UFix(width = VADDR_BITS+1) // ID stage PC
val data = Bits(width = INST_BITS)
val taken = Bool()
val xcpt_ma = Bool()
val xcpt_if = Bool()
}
class IOCPUFrontend extends Bundle {
val req = new PipeIO()(new FrontendReq)
val resp = new FIFOIO()(new FrontendResp).flip
val ptw = new IOTLBPTW().flip
}
class Frontend(c: ICacheConfig) extends Component
{
val io = new ioRocketICache();
val lines = sets * assoc;
val addrbits = PADDR_BITS;
val indexbits = log2Up(sets);
val offsetbits = OFFSET_BITS;
val tagmsb = addrbits - 1;
val taglsb = indexbits+offsetbits;
val tagbits = addrbits-taglsb;
val indexmsb = taglsb-1;
val indexlsb = offsetbits;
val offsetmsb = indexlsb-1;
val databits = 32;
val offsetlsb = log2Up(databits/8);
val rf_cnt_bits = log2Up(REFILL_CYCLES);
require(PGIDX_BITS >= taglsb); // virtually-indexed, physically-tagged constraint
require(isPow2(sets) && isPow2(assoc));
val io = new Bundle {
val cpu = new IOCPUFrontend().flip
val mem = new ioUncachedRequestor
}
val s_reset :: s_ready :: s_request :: s_refill_wait :: s_refill :: Nil = Enum(5) { UFix() };
val state = Reg(resetVal = s_reset);
val btb = new rocketDpathBTB(BTB_ENTRIES)
val icache = new ICache(c)
val tlb = new TLB(ITLB_ENTRIES)
val s1_pc = Reg() { UFix() }
val s2_valid = Reg(resetVal = Bool(true))
val s2_pc = Reg(resetVal = UFix(START_ADDR))
val s2_btb_hit = Reg(resetVal = Bool(false))
val s2_xcpt_if = Reg(resetVal = Bool(false))
val btbTarget = Cat(btb.io.target(VADDR_BITS-1), btb.io.target)
val pcp4_0 = s1_pc + UFix(c.ibytes)
val pcp4 = Cat(s1_pc(VADDR_BITS-1) & pcp4_0(VADDR_BITS-1), pcp4_0(VADDR_BITS-1,0))
val icmiss = s2_valid && !icache.io.resp.valid
val npc = Mux(icmiss, s2_pc, Mux(btb.io.hit, btbTarget, pcp4)).toUFix
val stall = !io.cpu.resp.ready
when (!stall) {
s1_pc := npc
s2_valid := !icmiss
s2_pc := s1_pc
s2_btb_hit := btb.io.hit
s2_xcpt_if := tlb.io.resp.xcpt_if
}
when (io.cpu.req.valid) {
s1_pc := io.cpu.req.bits.pc
s2_valid := Bool(false)
}
btb.io.current_pc := s1_pc
btb.io.wen := io.cpu.req.bits.mispredict
btb.io.clr := !io.cpu.req.bits.taken
btb.io.correct_pc := io.cpu.req.bits.currentpc
btb.io.correct_target := io.cpu.req.bits.pc
btb.io.invalidate := io.cpu.req.bits.invalidate || io.cpu.req.bits.invalidateTLB
tlb.io.ptw <> io.cpu.ptw
tlb.io.req.valid := !stall && !icmiss
tlb.io.req.bits.vpn := s1_pc >> UFix(PGIDX_BITS)
tlb.io.req.bits.status := io.cpu.req.bits.status
tlb.io.req.bits.asid := UFix(0)
tlb.io.req.bits.invalidate := io.cpu.req.bits.invalidateTLB
tlb.io.req.bits.instruction := Bool(true)
icache.io.mem <> io.mem
icache.io.req.valid := !stall
icache.io.req.bits.idx := Mux(io.cpu.req.valid, io.cpu.req.bits.pc, npc)
icache.io.req.bits.invalidate := io.cpu.req.bits.invalidate
icache.io.req.bits.ppn := tlb.io.resp.ppn
icache.io.req.bits.kill := io.cpu.req.valid || tlb.io.resp.miss
icache.io.resp.ready := io.cpu.resp.ready
io.cpu.resp.valid := s2_valid && (s2_xcpt_if || icache.io.resp.valid)
io.cpu.resp.bits.pc := s2_pc
io.cpu.resp.bits.data := icache.io.resp.bits.data
io.cpu.resp.bits.taken := s2_btb_hit
io.cpu.resp.bits.xcpt_ma := s2_pc(log2Up(c.ibytes)-1,0) != UFix(0)
io.cpu.resp.bits.xcpt_if := s2_xcpt_if
}
class ICache(c: ICacheConfig) extends Component
{
val io = new Bundle {
val req = new PipeIO()(new Bundle {
val idx = UFix(width = PGIDX_BITS)
val invalidate = Bool()
val ppn = UFix(width = PPN_BITS) // delayed one cycle
val kill = Bool() // delayed one cycle
}).flip
val resp = new FIFOIO()(new Bundle {
val data = Bits(width = INST_BITS)
val datablock = Bits(width = c.databits)
})
val mem = new ioUncachedRequestor
}
val s_ready :: s_request :: s_refill_wait :: s_refill :: Nil = Enum(4) { UFix() }
val state = Reg(resetVal = s_ready)
val invalidated = Reg() { Bool() }
val r_cpu_req_idx = Reg { Bits() }
val r_cpu_req_ppn = Reg { Bits() }
val r_cpu_req_val = Reg(resetVal = Bool(false));
val stall = !io.resp.ready
val rdy = Bool()
val tag_hit = Bool()
when (io.cpu.req_val && rdy) {
r_cpu_req_val := Bool(true)
r_cpu_req_idx := io.cpu.req_idx
}
.otherwise {
r_cpu_req_val := Bool(false)
}
when (state === s_ready && r_cpu_req_val && !io.cpu.itlb_miss) {
r_cpu_req_ppn := io.cpu.req_ppn
val s2_valid = Reg(resetVal = Bool(false))
val s2_addr = Reg { UFix(width = PADDR_BITS) }
val s1_valid = Reg(resetVal = Bool(false))
val s1_pgoff = Reg() { UFix(width = PGIDX_BITS) }
val s0_valid = io.req.valid && rdy || s1_valid && stall && !io.req.bits.kill
val s0_pgoff = Mux(io.req.valid, io.req.bits.idx, s1_pgoff)
s1_valid := s0_valid
when (io.req.valid && rdy) {
s1_pgoff := s0_pgoff
}
val r_cpu_hit_addr = Cat(io.cpu.req_ppn, r_cpu_req_idx)
val r_cpu_hit_tag = r_cpu_hit_addr(tagmsb,taglsb)
val r_cpu_miss_addr = Cat(r_cpu_req_ppn, r_cpu_req_idx)
val r_cpu_miss_tag = r_cpu_miss_addr(tagmsb,taglsb)
// refill counter
val refill_count = Reg(resetVal = UFix(0, rf_cnt_bits));
when (io.mem.xact_rep.valid) {
refill_count := refill_count + UFix(1);
s2_valid := s1_valid && rdy && !io.req.bits.kill || stall
when (s1_valid && rdy && !stall) {
s2_addr := Cat(io.req.bits.ppn, s1_pgoff).toUFix
}
val refill_done = io.mem.xact_rep.valid && refill_count.andR
val repl_way = if (assoc == 1) UFix(0) else LFSR16(state === s_ready && r_cpu_req_val && !io.cpu.itlb_miss && !tag_hit)(log2Up(assoc)-1,0)
val word_shift = Cat(r_cpu_req_idx(offsetmsb-rf_cnt_bits,offsetlsb), UFix(0, log2Up(databits))).toUFix
val tag_we = refill_done
val tag_addr =
Mux((state === s_refill), r_cpu_req_idx(indexmsb,indexlsb),
io.cpu.req_idx(indexmsb,indexlsb)).toUFix;
val data_addr =
Mux((state === s_refill_wait) || (state === s_refill), Cat(r_cpu_req_idx(indexmsb,offsetbits), refill_count),
io.cpu.req_idx(indexmsb, offsetbits-rf_cnt_bits)).toUFix;
val s2_tag = s2_addr(c.tagbits+c.untagbits-1,c.untagbits)
val s2_idx = s2_addr(c.untagbits-1,c.offbits)
val s2_offset = s2_addr(c.offbits-1,0)
val s2_any_tag_hit = Bool()
val s2_hit = s2_valid && s2_any_tag_hit
val s2_miss = s2_valid && !s2_any_tag_hit
rdy := state === s_ready && !s2_miss
val tag_array = Mem(sets, seqRead = true) { Bits(width = tagbits*assoc) }
val (rf_cnt, refill_done) = Counter(io.mem.xact_rep.valid, REFILL_CYCLES)
val repl_way = if (c.dm) UFix(0) else LFSR16(s2_miss)(log2Up(c.assoc)-1,0)
val tag_array = Mem(c.sets, seqRead = true) { Bits(width = c.tagwidth*c.assoc) }
val tag_rdata = Reg() { Bits() }
when (tag_we) {
tag_array.write(tag_addr, Fill(assoc, r_cpu_miss_tag), FillInterleaved(tagbits, if (assoc > 1) UFixToOH(repl_way) else Bits(1)))
}.otherwise {
tag_rdata := tag_array(tag_addr)
when (refill_done) {
val wmask = FillInterleaved(c.tagwidth, if (c.dm) Bits(1) else UFixToOH(repl_way))
val tag = Cat(if (c.parity) s2_tag.xorR else null, s2_tag)
tag_array.write(s2_idx, Fill(c.assoc, tag), wmask)
}.elsewhen (s0_valid) {
tag_rdata := tag_array(s0_pgoff(c.untagbits-1,c.offbits))
}
val vb_array = Reg(resetVal = Bits(0, lines))
when (io.cpu.invalidate) {
val vb_array = Reg(resetVal = Bits(0, c.lines))
when (refill_done && !invalidated) {
vb_array := vb_array.bitSet(Cat(repl_way, s2_idx), Bool(true))
}
when (io.req.bits.invalidate) {
vb_array := Bits(0)
}.elsewhen (tag_we) {
vb_array := vb_array.bitSet(Cat(r_cpu_req_idx(indexmsb,indexlsb), if (assoc > 1) repl_way else null), !invalidated)
invalidated := Bool(true)
}
val s2_disparity = Vec(c.assoc) { Bool() }
for (i <- 0 until c.assoc)
when (s2_valid && s2_disparity(i)) { vb_array := vb_array.bitSet(Cat(UFix(i), s2_idx), Bool(false)) }
val data_mux = (new Mux1H(assoc)){Bits(width = databits)}
var any_hit = Bool(false)
for (i <- 0 until assoc)
{
val valid = vb_array(Cat(r_cpu_req_idx(indexmsb,indexlsb), if (assoc > 1) UFix(i, log2Up(assoc)) else null))
val hit = valid && tag_rdata(tagbits*(i+1)-1, tagbits*i) === r_cpu_hit_addr(tagmsb,taglsb)
// data array
val data_array = Mem(sets*REFILL_CYCLES, seqRead = true){ io.mem.xact_rep.bits.data.clone }
val data_out = Reg(){ io.mem.xact_rep.bits.data.clone }
when (io.mem.xact_rep.valid && repl_way === UFix(i)) { data_array(data_addr) := io.mem.xact_rep.bits.data }
.otherwise { data_out := data_array(data_addr) }
data_mux.io.sel(i) := hit
data_mux.io.in(i) := (data_out >> word_shift)(databits-1,0);
any_hit = any_hit || hit
val s2_tag_hit = Vec(c.assoc) { Bool() }
val s2_data_disparity = Vec(c.assoc) { Bool() }
for (i <- 0 until c.assoc) {
val s1_vb = vb_array(Cat(UFix(i), s1_pgoff(c.untagbits-1,c.offbits))).toBool
val s2_vb = Reg() { Bool() }
val s2_tag_out = Reg() { Bits() }
when (s1_valid && rdy && !stall) {
s2_vb := s1_vb
s2_tag_out := tag_rdata(c.tagwidth*(i+1)-1, c.tagwidth*i)
}
s2_tag_hit(i) := s2_vb && s2_tag_out(c.tagbits-1,0) === s2_tag
s2_disparity(i) := Bool(c.parity) && s2_vb && (s2_tag_out.xorR || s2_data_disparity(i))
}
tag_hit := any_hit
s2_any_tag_hit := s2_tag_hit.reduceLeft(_||_) && !s2_disparity.reduceLeft(_||_)
val s2_dout = Vec(c.assoc) { Reg() { Bits(width = c.databits) } }
for (i <- 0 until c.assoc) {
val data_array = Mem(c.sets*REFILL_CYCLES, seqRead = true){ Bits(width = c.datawidth) }
val s1_dout = Reg(){ Bits() }
when (io.mem.xact_rep.valid && repl_way === UFix(i)) {
val d = io.mem.xact_rep.bits.data
val wdata = if (c.parity) Cat(d.xorR, d) else d
data_array(Cat(s2_idx,rf_cnt)) := wdata
}.elsewhen (s0_valid) {
s1_dout := data_array(s0_pgoff(c.untagbits-1,c.offbits-rf_cnt.getWidth))
}
when (s1_valid && rdy && !stall) { s2_dout(i) := s1_dout }
s2_data_disparity(i) := s2_dout(i).xorR
}
val s2_dout_word = s2_dout.map(x => (x >> Cat(s2_offset(log2Up(c.databits/8)-1,log2Up(c.ibytes)), Bits(0,log2Up(c.ibytes*8))))(c.ibytes*8-1,0))
io.resp.bits.data := Mux1H(s2_tag_hit, s2_dout_word)
io.resp.bits.datablock := Mux1H(s2_tag_hit, s2_dout)
val finish_q = (new Queue(1)) { new TransactionFinish }
finish_q.io.enq.valid := refill_done && io.mem.xact_rep.bits.require_ack
finish_q.io.enq.bits.global_xact_id := io.mem.xact_rep.bits.global_xact_id
// output signals
io.cpu.resp_val := !io.cpu.itlb_miss && (state === s_ready) && r_cpu_req_val && tag_hit;
rdy := !io.cpu.itlb_miss && (state === s_ready) && (!r_cpu_req_val || tag_hit);
io.cpu.resp_data := data_mux.io.out
io.resp.valid := s2_hit
io.mem.xact_init.valid := (state === s_request) && finish_q.io.enq.ready
io.mem.xact_init.bits := co.getUncachedReadTransactionInit(r_cpu_miss_addr(tagmsb,indexlsb).toUFix, UFix(0))
io.mem.xact_init.bits := c.co.getUncachedReadTransactionInit(s2_addr >> UFix(c.offbits), UFix(0))
io.mem.xact_finish <> finish_q.io.deq
// control state machine
when (io.cpu.invalidate) {
invalidated := Bool(true)
}
switch (state) {
is (s_reset) {
state := s_ready;
}
is (s_ready) {
when (r_cpu_req_val && !tag_hit && !io.cpu.itlb_miss) {
state := s_request;
}
when (s2_miss) { state := s_request }
invalidated := Bool(false)
}
is (s_request)
{
when (io.mem.xact_init.ready && finish_q.io.enq.ready) {
state := s_refill_wait;
}
is (s_request) {
when (io.mem.xact_init.ready && finish_q.io.enq.ready) { state := s_refill_wait }
}
is (s_refill_wait) {
when (io.mem.xact_abort.valid) {
state := s_request
}
when (io.mem.xact_rep.valid) {
state := s_refill;
}
when (io.mem.xact_abort.valid) { state := s_request }
when (io.mem.xact_rep.valid) { state := s_refill }
}
is (s_refill) {
when (refill_done) {
state := s_ready;
}
when (refill_done) { state := s_ready }
}
}
}
}

229
rocket/src/main/scala/itlb.scala
View File

@ -1,229 +0,0 @@
package rocket
import Chisel._;
import Node._;
import Constants._;
import scala.math._;
class ioCAM(entries: Int, addr_bits: Int, tag_bits: Int) extends Bundle {
val clear = Bool(INPUT);
val clear_hit = Bool(INPUT)
val tag = Bits(INPUT, tag_bits);
val hit = Bool(OUTPUT);
val hit_addr = UFix(OUTPUT, addr_bits);
val valid_bits = Bits(OUTPUT, entries);
val write = Bool(INPUT);
val write_tag = Bits(INPUT, tag_bits);
val write_addr = UFix(INPUT, addr_bits);
}
class rocketCAM(entries: Int, tag_bits: Int) extends Component {
val addr_bits = ceil(log(entries)/log(2)).toInt;
val io = new ioCAM(entries, addr_bits, tag_bits);
val cam_tags = Vec(entries) { Reg() { Bits(width = tag_bits) } }
val mux = (new Mux1H(entries)) { Bits(width = addr_bits) }
val vb_array = Reg(resetVal = Bits(0, entries));
when (io.write) {
vb_array := vb_array.bitSet(io.write_addr, Bool(true));
cam_tags(io.write_addr) := io.write_tag
}
when (io.clear) {
vb_array := Bits(0, entries);
}
.elsewhen (io.clear_hit) {
vb_array := vb_array & ~mux.io.sel.toBits
}
var l_hit = Bool(false)
for (i <- 0 to entries-1) {
val my_hit = vb_array(UFix(i)).toBool && (cam_tags(i) === io.tag)
l_hit = l_hit || my_hit
mux.io.in(i) := Bits(i)
mux.io.sel(i) := my_hit
}
io.valid_bits := vb_array;
io.hit := l_hit;
io.hit_addr := mux.io.out.toUFix;
}
class PseudoLRU(n: Int)
{
val state = Reg() { Bits(width = n) }
def access(way: UFix) = {
var next_state = state
var idx = UFix(1,1)
for (i <- log2Up(n)-1 to 0 by -1) {
val bit = way(i)
val mask = (UFix(1,n) << idx)(n-1,0)
next_state = next_state & ~mask | Mux(bit, UFix(0), mask)
//next_state.bitSet(idx, !bit)
idx = Cat(idx, bit)
}
state := next_state
}
def replace = {
var idx = UFix(1,1)
for (i <- 0 until log2Up(n))
idx = Cat(idx, state(idx))
idx(log2Up(n)-1,0)
}
}
// interface between TLB and PTW
class ioTLB_PTW extends Bundle
{
// requests
val req_val = Bool(OUTPUT);
val req_rdy = Bool(INPUT);
val req_vpn = Bits(OUTPUT, VPN_BITS);
// responses
val resp_val = Bool(INPUT);
val resp_err = Bool(INPUT);
val resp_ppn = Bits(INPUT, PPN_BITS);
val resp_perm = Bits(INPUT, PERM_BITS);
}
// interface between ITLB and fetch stage of pipeline
class ioITLB_CPU extends Bundle
{
// status bits (from PCR), to check current permission and whether VM is enabled
val status = Bits(INPUT, 32);
// invalidate all TLB entries
val invalidate = Bool(INPUT);
// lookup requests
val req_val = Bool(INPUT);
val req_rdy = Bool(OUTPUT);
val req_asid = Bits(INPUT, ASID_BITS);
val req_vpn = UFix(INPUT, VPN_BITS+1);
// lookup responses
val resp_miss = Bool(OUTPUT);
// val resp_val = Bool(OUTPUT);
val resp_ppn = UFix(OUTPUT, PPN_BITS);
val exception = Bool(OUTPUT);
}
class ioITLB extends Bundle
{
val cpu = new ioITLB_CPU();
val ptw = new ioTLB_PTW();
}
class rocketITLB(entries: Int) extends Component
{
val addr_bits = ceil(log10(entries)/log10(2)).toInt;
val io = new ioITLB();
val s_ready :: s_request :: s_wait :: Nil = Enum(3) { UFix() };
val state = Reg(resetVal = s_ready);
val r_cpu_req_val = Reg(resetVal = Bool(false));
val r_cpu_req_vpn = Reg() { Bits() };
val r_cpu_req_asid = Reg() { Bits() };
val r_refill_tag = Reg() { Bits() };
val r_refill_waddr = Reg() { UFix() };
when (io.cpu.req_val && io.cpu.req_rdy) {
r_cpu_req_vpn := io.cpu.req_vpn;
r_cpu_req_asid := io.cpu.req_asid;
r_cpu_req_val := Bool(true);
}
.otherwise {
r_cpu_req_val := Bool(false);
}
val bad_va = r_cpu_req_vpn(VPN_BITS) != r_cpu_req_vpn(VPN_BITS-1);
val tag_cam = new rocketCAM(entries, ASID_BITS+VPN_BITS);
val tag_ram = Mem(entries) { io.ptw.resp_ppn.clone }
when (io.ptw.resp_val) { tag_ram(r_refill_waddr) := io.ptw.resp_ppn }
val lookup_tag = Cat(r_cpu_req_asid, r_cpu_req_vpn);
tag_cam.io.clear := io.cpu.invalidate;
tag_cam.io.clear_hit := io.cpu.exception
tag_cam.io.tag := lookup_tag;
tag_cam.io.write := io.ptw.resp_val || io.ptw.resp_err;
tag_cam.io.write_tag := r_refill_tag;
tag_cam.io.write_addr := r_refill_waddr;
val tag_hit = tag_cam.io.hit || bad_va;
val tag_hit_addr = tag_cam.io.hit_addr;
// extract fields from status register
val status_s = io.cpu.status(SR_S).toBool; // user/supervisor mode
val status_u = !status_s;
val status_vm = io.cpu.status(SR_VM).toBool // virtual memory enable
// extract fields from PT permission bits
val ptw_perm_ux = io.ptw.resp_perm(0);
val ptw_perm_sx = io.ptw.resp_perm(3);
// permission bit arrays
val ux_array = Reg(resetVal = Bits(0, entries)); // user execute permission
val sx_array = Reg(resetVal = Bits(0, entries)); // supervisor execute permission
when (io.ptw.resp_val) {
ux_array := ux_array.bitSet(r_refill_waddr, ptw_perm_ux);
sx_array := sx_array.bitSet(r_refill_waddr, ptw_perm_sx);
}
// when the page table lookup reports an error, set both execute permission
// bits to 0 so the next access will cause an exceptions
when (io.ptw.resp_err) {
ux_array := ux_array.bitSet(r_refill_waddr, Bool(false));
sx_array := sx_array.bitSet(r_refill_waddr, Bool(false));
}
// high if there are any unused entries in the ITLB
val has_invalid_entry = !tag_cam.io.valid_bits.andR
val invalid_entry = PriorityEncoder(~tag_cam.io.valid_bits)
val plru = new PseudoLRU(entries)
val repl_waddr = Mux(has_invalid_entry, invalid_entry, plru.replace).toUFix;
val lookup = (state === s_ready) && r_cpu_req_val;
val lookup_hit = lookup && tag_hit;
val lookup_miss = lookup && !tag_hit;
val tlb_hit = status_vm && lookup_hit;
val tlb_miss = status_vm && lookup_miss;
when (tlb_miss) {
r_refill_tag := lookup_tag;
r_refill_waddr := repl_waddr;
}
when (tlb_hit) {
plru.access(tag_hit_addr)
}
val access_fault =
tlb_hit &&
((status_s && !sx_array(tag_hit_addr).toBool) ||
(status_u && !ux_array(tag_hit_addr).toBool) ||
bad_va);
io.cpu.exception := access_fault;
io.cpu.req_rdy := Mux(status_vm, (state === s_ready) && (!r_cpu_req_val || tag_hit), Bool(true));
io.cpu.resp_miss := tlb_miss || (state != s_ready);
io.cpu.resp_ppn := Mux(status_vm, tag_ram(tag_hit_addr), r_cpu_req_vpn(PPN_BITS-1,0)).toUFix;
io.ptw.req_val := (state === s_request);
io.ptw.req_vpn := r_refill_tag(VPN_BITS-1,0);
// control state machine
switch (state) {
is (s_ready) {
when (tlb_miss) {
state := s_request;
}
}
is (s_request) {
when (io.ptw.req_rdy) {
state := s_wait;
}
}
is (s_wait) {
when (io.ptw.resp_val || io.ptw.resp_err) {
state := s_ready;
}
}
}
}

2
rocket/src/main/scala/multiplier.scala
View File

@ -100,7 +100,7 @@ class rocketMultiplier(unroll: Int = 1, earlyOut: Boolean = false) extends Compo
r_prod:= rhs_in
r_lsb := Bool(false)
}
.elsewhen (io.resp_val && io.resp_rdy || io.req_kill && r_cnt === UFix(0)) { // can only kill on first cycle
.elsewhen (io.resp_val && io.resp_rdy || io.req_kill) {
r_val := Bool(false)
}

37
rocket/src/main/scala/ptw.scala
View File

@ -73,7 +73,7 @@ class rocketHellaCacheArbiter(n: Int) extends Component
class ioPTW(n: Int) extends Bundle
{
val requestor = Vec(n) { new ioTLB_PTW }.flip
val requestor = Vec(n) { new IOTLBPTW }.flip
val mem = new ioHellaCache
val ptbr = UFix(INPUT, PADDR_BITS)
}
@ -99,20 +99,15 @@ class rocketPTW(n: Int) extends Component
val vpn_idxs = (1 until levels).map(i => r_req_vpn((levels-i)*bitsPerLevel-1, (levels-i-1)*bitsPerLevel))
val vpn_idx = (2 until levels).foldRight(vpn_idxs(0))((i,j) => Mux(count === UFix(i-1), vpn_idxs(i-1), j))
val req_rdy = state === s_ready
var req_val = Bool(false)
for (r <- io.requestor) {
r.req_rdy := req_rdy && !req_val
req_val = req_val || r.req_val
}
val req_dest = PriorityEncoder(io.requestor.map(_.req_val))
val req_vpn = io.requestor.slice(0, n-1).foldRight(io.requestor(n-1).req_vpn)((r, v) => Mux(r.req_val, r.req_vpn, v))
when (state === s_ready && req_val) {
r_req_vpn := req_vpn
r_req_dest := req_dest
req_addr := Cat(io.ptbr(PADDR_BITS-1,PGIDX_BITS), req_vpn(VPN_BITS-1,VPN_BITS-bitsPerLevel), Bits(0,3))
val arb = new Arbiter(n)(UFix(width = VPN_BITS))
arb.io.in <> io.requestor.map(_.req)
arb.io.out.ready := state === s_ready
when (arb.io.out.fire()) {
r_req_vpn := arb.io.out.bits
r_req_dest := arb.io.chosen
req_addr := Cat(io.ptbr(PADDR_BITS-1,PGIDX_BITS), arb.io.out.bits(VPN_BITS-1,VPN_BITS-bitsPerLevel), Bits(0,3))
}
val dmem_resp_val = Reg(io.mem.resp.valid, resetVal = Bool(false))
@ -129,8 +124,8 @@ class rocketPTW(n: Int) extends Component
io.mem.req.bits.ppn := Reg(req_addr(PADDR_BITS-1,PGIDX_BITS))
io.mem.req.bits.kill := Bool(false)
val resp_val = state === s_done
val resp_err = state === s_error
val resp_val = state === s_done || state === s_error
val resp_err = state === s_error || state === s_wait
val resp_ptd = io.mem.resp.bits.data_subword(1,0) === Bits(1)
val resp_pte = io.mem.resp.bits.data_subword(1,0) === Bits(2)
@ -140,16 +135,16 @@ class rocketPTW(n: Int) extends Component
for (i <- 0 until io.requestor.size) {
val me = r_req_dest === UFix(i)
io.requestor(i).resp_val := resp_val && me
io.requestor(i).resp_err := resp_err && me
io.requestor(i).resp_perm := r_resp_perm
io.requestor(i).resp_ppn := resp_ppn
io.requestor(i).resp.valid := resp_val && me
io.requestor(i).resp.bits.error := resp_err
io.requestor(i).resp.bits.perm := r_resp_perm
io.requestor(i).resp.bits.ppn := resp_ppn.toUFix
}
// control state machine
switch (state) {
is (s_ready) {
when (req_val) {
when (arb.io.out.valid) {
state := s_req;
}
count := UFix(0)

16
rocket/src/main/scala/queues.scala
View File

@ -3,25 +3,27 @@ package rocket
import Chisel._
import Node._;
class SkidBuffer[T <: Data](resetSignal: Bool = null)(data: => T) extends Component(resetSignal)
class SkidBuffer[T <: Data](entries: Int, lateEnq: Boolean = false)(data: => T) extends Component
{
val io = new Bundle {
val enq = new FIFOIO()(data).flip
val deq = new FIFOIO()(data)
}
require(entries >= 2)
val fq = new Queue(1, flow = true)(data)
val pq = new Queue(1, pipe = true)(data)
val pq = new Queue(entries-1, pipe = true)(data)
val (iq, oq) = if (lateEnq) (pq, fq) else (fq, pq)
fq.io.enq <> io.enq
pq.io.enq <> fq.io.deq
io.deq <> pq.io.deq
iq.io.enq <> io.enq
oq.io.enq <> iq.io.deq
io.deq <> oq.io.deq
}
object SkidBuffer
{
def apply[T <: Data](enq: FIFOIO[T]): FIFOIO[T] = {
val s = new SkidBuffer()(enq.bits.clone)
def apply[T <: Data](enq: FIFOIO[T], entries: Int = 2): FIFOIO[T] = {
val s = new SkidBuffer(entries)(enq.bits.clone)
s.io.enq <> enq
s.io.deq
}

4
rocket/src/main/scala/tile.scala
View File

@ -13,7 +13,7 @@ class Tile(co: CoherencePolicyWithUncached, resetSignal: Bool = null) extends Co
}
val cpu = new rocketProc
val icache = new rocketICache(128, 4, co) // 128 sets x 4 ways (32KB)
val icache = new Frontend(ICacheConfig(co, 128, 4)) // 128 sets x 4 ways (32KB)
val dcache = new HellaCache(co)
val arbiter = new rocketMemArbiter(2 + (if (HAVE_VEC) 1 else 0))
@ -31,7 +31,7 @@ class Tile(co: CoherencePolicyWithUncached, resetSignal: Bool = null) extends Co
if (HAVE_VEC)
{
val vicache = new rocketICache(128, 1, co) // 128 sets x 1 ways (8KB)
val vicache = new Frontend(ICacheConfig(co, 128, 1)) // 128 sets x 1 ways (8KB)
arbiter.io.requestor(2) <> vicache.io.mem
cpu.io.vimem <> vicache.io.cpu
}

255
rocket/src/main/scala/tlb.scala Normal file
View File

@ -0,0 +1,255 @@
package rocket
import Chisel._;
import Node._;
import Constants._;
import scala.math._;
import hwacha._
class ioCAM(entries: Int, addr_bits: Int, tag_bits: Int) extends Bundle {
val clear = Bool(INPUT);
val clear_hit = Bool(INPUT)
val tag = Bits(INPUT, tag_bits);
val hit = Bool(OUTPUT);
val hits = UFix(OUTPUT, entries);
val valid_bits = Bits(OUTPUT, entries);
val write = Bool(INPUT);
val write_tag = Bits(INPUT, tag_bits);
val write_addr = UFix(INPUT, addr_bits);
}
class rocketCAM(entries: Int, tag_bits: Int) extends Component {
val addr_bits = ceil(log(entries)/log(2)).toInt;
val io = new ioCAM(entries, addr_bits, tag_bits);
val cam_tags = Vec(entries) { Reg() { Bits(width = tag_bits) } }
val vb_array = Reg(resetVal = Bits(0, entries));
when (io.write) {
vb_array := vb_array.bitSet(io.write_addr, Bool(true));
cam_tags(io.write_addr) := io.write_tag
}
when (io.clear) {
vb_array := Bits(0, entries);
}
.elsewhen (io.clear_hit) {
vb_array := vb_array & ~io.hits
}
val hits = (0 until entries).map(i => vb_array(i) && cam_tags(i) === io.tag)
io.valid_bits := vb_array;
io.hits := Vec(hits){Bool()}.toBits.toUFix
io.hit := io.hits.orR
}
class PseudoLRU(n: Int)
{
val state = Reg() { Bits(width = n) }
def access(way: UFix) = {
var next_state = state
var idx = UFix(1,1)
for (i <- log2Up(n)-1 to 0 by -1) {
val bit = way(i)
val mask = (UFix(1,n) << idx)(n-1,0)
next_state = next_state & ~mask | Mux(bit, UFix(0), mask)
//next_state.bitSet(idx, !bit)
idx = Cat(idx, bit)
}
state := next_state
}
def replace = {
var idx = UFix(1,1)
for (i <- 0 until log2Up(n))
idx = Cat(idx, state(idx))
idx(log2Up(n)-1,0)
}
}
class IOTLBPTW extends Bundle {
val req = new FIFOIO()(UFix(width = VPN_BITS))
val resp = new PipeIO()(new Bundle {
val error = Bool()
val ppn = UFix(width = PPN_BITS)
val perm = Bits(width = PERM_BITS)
}).flip
}
class TLBReq extends Bundle
{
val asid = UFix(width = ASID_BITS)
val vpn = UFix(width = VPN_BITS+1)
val status = Bits(width = 32)
val invalidate = Bool()
val instruction = Bool()
}
class TLBResp(entries: Int) extends Bundle
{
// lookup responses
val miss = Bool(OUTPUT)
val hit_idx = UFix(OUTPUT, entries)
val ppn = UFix(OUTPUT, PPN_BITS)
val xcpt_ld = Bool(OUTPUT)
val xcpt_st = Bool(OUTPUT)
val xcpt_pf = Bool(OUTPUT)
val xcpt_if = Bool(OUTPUT)
override def clone = new TLBResp(entries).asInstanceOf[this.type]
}
class TLB(entries: Int) extends Component
{
val io = new Bundle {
val req = new FIFOIO()(new TLBReq).flip
val resp = new TLBResp(entries)
val ptw = new IOTLBPTW
}
val s_ready :: s_request :: s_wait :: s_wait_invalidate :: Nil = Enum(4) { UFix() }
val state = Reg(resetVal = s_ready)
val r_refill_tag = Reg() { UFix() }
val r_refill_waddr = Reg() { UFix() }
val tag_cam = new rocketCAM(entries, ASID_BITS+VPN_BITS);
val tag_ram = Vec(entries) { Reg() { io.ptw.resp.bits.ppn.clone } }
when (io.ptw.resp.valid) { tag_ram(r_refill_waddr) := io.ptw.resp.bits.ppn }
val lookup_tag = Cat(io.req.bits.asid, io.req.bits.vpn).toUFix
tag_cam.io.clear := io.req.bits.invalidate
tag_cam.io.clear_hit := io.req.fire() && Mux(io.req.bits.instruction, io.resp.xcpt_if, io.resp.xcpt_ld && io.resp.xcpt_st)
tag_cam.io.tag := lookup_tag
tag_cam.io.write := state === s_wait && io.ptw.resp.valid
tag_cam.io.write_tag := r_refill_tag
tag_cam.io.write_addr := r_refill_waddr
val tag_hit = tag_cam.io.hit
val tag_hit_addr = OHToUFix(tag_cam.io.hits)
// permission bit arrays
val ur_array = Reg(resetVal = Bits(0, entries)) // user read permission
val uw_array = Reg(resetVal = Bits(0, entries)) // user write permission
val ux_array = Reg(resetVal = Bits(0, entries)) // user execute permission
val sr_array = Reg(resetVal = Bits(0, entries)) // supervisor read permission
val sw_array = Reg(resetVal = Bits(0, entries)) // supervisor write permission
val sx_array = Reg(resetVal = Bits(0, entries)) // supervisor execute permission
when (tag_cam.io.write) {
val perm = (!io.ptw.resp.bits.error).toFix & io.ptw.resp.bits.perm(5,0)
ur_array := ur_array.bitSet(r_refill_waddr, perm(2))
uw_array := uw_array.bitSet(r_refill_waddr, perm(1))
ux_array := ux_array.bitSet(r_refill_waddr, perm(0))
sr_array := sr_array.bitSet(r_refill_waddr, perm(5))
sw_array := sw_array.bitSet(r_refill_waddr, perm(4))
sx_array := sx_array.bitSet(r_refill_waddr, perm(3))
}
// high if there are any unused (invalid) entries in the TLB
val has_invalid_entry = !tag_cam.io.valid_bits.andR
val invalid_entry = PriorityEncoder(~tag_cam.io.valid_bits)
val plru = new PseudoLRU(entries)
val repl_waddr = Mux(has_invalid_entry, invalid_entry, plru.replace)
val status_s = io.req.bits.status(SR_S) // user/supervisor mode
val status_vm = io.req.bits.status(SR_VM) // virtual memory enable
val bad_va = io.req.bits.vpn(VPN_BITS) != io.req.bits.vpn(VPN_BITS-1)
val tlb_hit = status_vm && tag_hit
val tlb_miss = status_vm && !tag_hit && !bad_va
when (io.req.valid && tlb_hit) {
plru.access(tag_hit_addr)
}
io.req.ready := state === s_ready
io.resp.xcpt_ld := bad_va || tlb_hit && !Mux(status_s, sr_array(tag_hit_addr), ur_array(tag_hit_addr))
io.resp.xcpt_st := bad_va || tlb_hit && !Mux(status_s, sw_array(tag_hit_addr), uw_array(tag_hit_addr))
io.resp.xcpt_if := bad_va || tlb_hit && !Mux(status_s, sx_array(tag_hit_addr), ux_array(tag_hit_addr))
io.resp.miss := tlb_miss
io.resp.ppn := Mux(status_vm, Mux1H(tag_cam.io.hits, tag_ram), io.req.bits.vpn(PPN_BITS-1,0))
io.resp.hit_idx := tag_cam.io.hits
io.ptw.req.valid := state === s_request
io.ptw.req.bits := r_refill_tag
when (io.req.fire() && tlb_miss) {
state := s_request
r_refill_tag := lookup_tag
r_refill_waddr := repl_waddr
}
when (state === s_request) {
when (io.req.bits.invalidate) {
state := s_ready
}
when (io.ptw.req.ready) {
state := s_wait
when (io.req.bits.invalidate) { state := s_wait_invalidate }
}
}
when (state === s_wait && io.req.bits.invalidate) {
state := s_wait_invalidate
}
when ((state === s_wait || state === s_wait_invalidate) && io.ptw.resp.valid) {
state := s_ready
}
}
// ioDTLB_CPU also located in hwacha/src/vuVXU-Interface.scala
// should keep them in sync
class ioDTLB_CPU_req_bundle extends TLBReq
{
val kill = Bool()
val cmd = Bits(width=4) // load/store/amo
}
class ioDTLB_CPU_req extends FIFOIO()( { new ioDTLB_CPU_req_bundle() } )
class ioDTLB_CPU_resp extends TLBResp(1)
class ioDTLB extends Bundle
{
// status bits (from PCR), to check current permission and whether VM is enabled
val status = Bits(INPUT, 32)
// invalidate all TLB entries
val invalidate = Bool(INPUT)
val cpu_req = new ioDTLB_CPU_req().flip
val cpu_resp = new ioDTLB_CPU_resp()
val ptw = new IOTLBPTW
}
class rocketTLB(entries: Int) extends Component
{
val io = new ioDTLB();
val r_cpu_req_val = Reg(resetVal = Bool(false));
val r_cpu_req_vpn = Reg() { UFix() }
val r_cpu_req_cmd = Reg() { Bits() }
val r_cpu_req_asid = Reg() { UFix() }
val tlb = new TLB(entries)
tlb.io.req.valid := r_cpu_req_val && !io.cpu_req.bits.kill
tlb.io.req.bits.instruction := Bool(false)
tlb.io.req.bits.invalidate := io.invalidate
tlb.io.req.bits.status := io.status
tlb.io.req.bits.vpn := r_cpu_req_vpn
tlb.io.req.bits.asid := r_cpu_req_asid
def cmdIsRead(cmd: Bits) = cmd === M_XRD || cmd(3)
def cmdIsWrite(cmd: Bits) = cmd === M_XWR || cmd(3)
def cmdIsPrefetch(cmd: Bits) = cmd === M_PFR || cmd === M_PFW
def cmdNeedsTLB(cmd: Bits) = cmdIsRead(cmd) || cmdIsWrite(cmd) || cmdIsPrefetch(cmd)
when (io.cpu_req.fire() && cmdNeedsTLB(io.cpu_req.bits.cmd)) {
r_cpu_req_vpn := io.cpu_req.bits.vpn;
r_cpu_req_cmd := io.cpu_req.bits.cmd;
r_cpu_req_asid := io.cpu_req.bits.asid;
r_cpu_req_val := Bool(true);
}
.otherwise {
r_cpu_req_val := Bool(false);
}
io.cpu_req.ready := tlb.io.req.ready && !io.cpu_resp.miss
io.cpu_resp.ppn := tlb.io.resp.ppn
io.cpu_resp.miss := r_cpu_req_val && tlb.io.resp.miss
io.cpu_resp.xcpt_ld := r_cpu_req_val && tlb.io.resp.xcpt_ld && cmdIsRead(r_cpu_req_cmd)
io.cpu_resp.xcpt_st := r_cpu_req_val && tlb.io.resp.xcpt_st && cmdIsWrite(r_cpu_req_cmd)
io.cpu_resp.xcpt_pf := r_cpu_req_val && tlb.io.resp.xcpt_ld && cmdIsPrefetch(r_cpu_req_cmd)
io.ptw <> tlb.io.ptw
}