Github_Repos/cvw
1
0
Fork 1
mirror of https://github.com/openhwgroup/cvw synced 2025-02-11 06:05:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

erge branch 'main' of https://github.com/davidharrishmc/riscv-wally into main

Browse Source
This commit is contained in:
David Harris 2021-12-30 17:22:22 +00:00
commit 028a876a4e
8 changed files with 242 additions and 270 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
wally-pipelined/regression/wave.do
View File

@ -175,11 +175,11 @@ add wave -noupdate -group icache -expand -group {fsm out and control} /testbench
add wave -noupdate -group icache -expand -group {fsm out and control} /testbench/dut/hart/ifu/icache/controller/PreCntEn
add wave -noupdate -group icache -expand -group {fsm out and control} /testbench/dut/hart/ifu/icache/controller/CntEn
add wave -noupdate -group icache -expand -group {fsm out and control} /testbench/dut/hart/ifu/icache/FinalInstrRawF
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/controller/InstrReadF
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/InstrPAdrF
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/controller/IfuBusFetch
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/ICacheBusAdr
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/controller/FetchCountFlag
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/FetchCount
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/controller/InstrAckF
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/controller/ICacheBusAck
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/controller/ICacheMemWriteEnable
add wave -noupdate -group icache -expand -group memory /testbench/dut/hart/ifu/icache/ICacheMemWriteData
add wave -noupdate -group icache /testbench/dut/hart/ifu/icache/ICacheMemReadData
@ -187,7 +187,7 @@ add wave -noupdate -group icache /testbench/dut/hart/ifu/icache/SpillDataBlock0
add wave -noupdate -group AHB -color Gold /testbench/dut/hart/ebu/BusState
add wave -noupdate -group AHB /testbench/dut/hart/ebu/NextBusState
add wave -noupdate -group AHB -expand -group {input requests} /testbench/dut/hart/ebu/AtomicMaskedM
add wave -noupdate -group AHB -expand -group {input requests} /testbench/dut/hart/ebu/InstrReadF
add wave -noupdate -group AHB -expand -group {input requests} /testbench/dut/hart/ebu/IfuBusFetch
add wave -noupdate -group AHB -expand -group {input requests} /testbench/dut/hart/ebu/LsuBusSize
add wave -noupdate -group AHB /testbench/dut/hart/ebu/HCLK
add wave -noupdate -group AHB /testbench/dut/hart/ebu/HRESETn

96
wally-pipelined/src/cache/cacheway.sv vendored
View File

@ -26,52 +26,52 @@
`include "wally-config.vh"
module cacheway #(parameter NUMLINES=512, parameter BLOCKLEN = 256, TAGLEN = 26,
parameter OFFSETLEN = 5, parameter INDEXLEN = 9, parameter DIRTY_BITS = 1)
parameter OFFSETLEN = 5, parameter INDEXLEN = 9, parameter DIRTY_BITS = 1)
(input logic clk,
input logic reset,
input logic reset,
input logic [$clog2(NUMLINES)-1:0] RAdr,
input logic [`PA_BITS-1:0] PAdr,
input logic WriteEnable,
input logic VDWriteEnable,
input logic [`PA_BITS-1:0] PAdr,
input logic WriteEnable,
input logic VDWriteEnable,
input logic [BLOCKLEN/`XLEN-1:0] WriteWordEnable,
input logic TagWriteEnable,
input logic [BLOCKLEN-1:0] WriteData,
input logic SetValid,
input logic ClearValid,
input logic SetDirty,
input logic ClearDirty,
input logic SelEvict,
input logic VictimWay,
input logic InvalidateAll,
input logic SelFlush,
input logic FlushWay,
input logic TagWriteEnable,
input logic [BLOCKLEN-1:0] WriteData,
input logic SetValid,
input logic ClearValid,
input logic SetDirty,
input logic ClearDirty,
input logic SelEvict,
input logic VictimWay,
input logic InvalidateAll,
input logic SelFlush,
input logic FlushWay,
output logic [BLOCKLEN-1:0] ReadDataBlockWayMasked,
output logic WayHit,
output logic VictimDirtyWay,
output logic [TAGLEN-1:0] VictimTagWay
output logic [BLOCKLEN-1:0] ReadDataLineWayMasked,
output logic WayHit,
output logic VictimDirtyWay,
output logic [TAGLEN-1:0] VictimTagWay
);
logic [NUMLINES-1:0] ValidBits;
logic [NUMLINES-1:0] DirtyBits;
logic [BLOCKLEN-1:0] ReadDataBlockWay;
logic [TAGLEN-1:0] ReadTag;
logic Valid;
logic Dirty;
logic SelectedWay;
logic [TAGLEN-1:0] VicDirtyWay;
logic [TAGLEN-1:0] FlushThisWay;
logic [NUMLINES-1:0] ValidBits;
logic [NUMLINES-1:0] DirtyBits;
logic [BLOCKLEN-1:0] ReadDataBlockWay;
logic [TAGLEN-1:0] ReadTag;
logic Valid;
logic Dirty;
logic SelectedWay;
logic [TAGLEN-1:0] VicDirtyWay;
logic [TAGLEN-1:0] FlushThisWay;
logic [$clog2(NUMLINES)-1:0] RAdrD;
logic SetValidD, ClearValidD;
logic SetDirtyD, ClearDirtyD;
logic WriteEnableD, VDWriteEnableD;
logic [$clog2(NUMLINES)-1:0] RAdrD;
logic SetValidD, ClearValidD;
logic SetDirtyD, ClearDirtyD;
logic WriteEnableD, VDWriteEnableD;
genvar words;
genvar words;
generate
for(words = 0; words < BLOCKLEN/`XLEN; words++) begin : word
@ -92,25 +92,25 @@ module cacheway #(parameter NUMLINES=512, parameter BLOCKLEN = 256, TAGLEN = 26,
assign WayHit = Valid & (ReadTag == PAdr[`PA_BITS-1:OFFSETLEN+INDEXLEN]);
assign SelectedWay = SelFlush ? FlushWay :
SelEvict ? VictimWay : WayHit;
assign ReadDataBlockWayMasked = SelectedWay ? ReadDataBlockWay : '0; // first part of AO mux.
SelEvict ? VictimWay : WayHit;
assign ReadDataLineWayMasked = SelectedWay ? ReadDataBlockWay : '0; // first part of AO mux.
assign VictimDirtyWay = SelFlush ? FlushWay & Dirty & Valid :
VictimWay & Dirty & Valid;
VictimWay & Dirty & Valid;
assign VicDirtyWay = VictimWay ? ReadTag : '0;
assign FlushThisWay = FlushWay ? ReadTag : '0;
assign VictimTagWay = SelFlush ? FlushThisWay : VicDirtyWay;
always_ff @(posedge clk) begin
if (reset)
ValidBits <= {NUMLINES{1'b0}};
ValidBits <= {NUMLINES{1'b0}};
else if (InvalidateAll)
ValidBits <= {NUMLINES{1'b0}};
ValidBits <= {NUMLINES{1'b0}};
else if (SetValidD & (WriteEnableD | VDWriteEnableD)) ValidBits[RAdrD] <= 1'b1;
else if (ClearValidD & (WriteEnableD | VDWriteEnableD)) ValidBits[RAdrD] <= 1'b0;
end
end
always_ff @(posedge clk) begin
RAdrD <= RAdr;
@ -120,21 +120,21 @@ module cacheway #(parameter NUMLINES=512, parameter BLOCKLEN = 256, TAGLEN = 26,
VDWriteEnableD <= VDWriteEnable;
end
assign Valid = ValidBits[RAdrD];
generate
if(DIRTY_BITS) begin
always_ff @(posedge clk) begin
if (reset)
DirtyBits <= {NUMLINES{1'b0}};
else if (SetDirtyD & (WriteEnableD | VDWriteEnableD)) DirtyBits[RAdrD] <= 1'b1;
else if (ClearDirtyD & (WriteEnableD | VDWriteEnableD)) DirtyBits[RAdrD] <= 1'b0;
if (reset)
DirtyBits <= {NUMLINES{1'b0}};
else if (SetDirtyD & (WriteEnableD | VDWriteEnableD)) DirtyBits[RAdrD] <= 1'b1;
else if (ClearDirtyD & (WriteEnableD | VDWriteEnableD)) DirtyBits[RAdrD] <= 1'b0;
end
always_ff @(posedge clk) begin
SetDirtyD <= SetDirty;
ClearDirtyD <= ClearDirty;
SetDirtyD <= SetDirty;
ClearDirtyD <= ClearDirty;
end
assign Dirty = DirtyBits[RAdrD];

14
wally-pipelined/src/cache/dcache.sv vendored
View File

@ -79,10 +79,10 @@ module dcache
logic [BLOCKLEN-1:0] SRAMWriteData;
logic SetValid, ClearValid;
logic SetDirty, ClearDirty;
logic [BLOCKLEN-1:0] ReadDataBlockWayMaskedM [NUMWAYS-1:0];
logic [BLOCKLEN-1:0] ReadDataLineWayMasked [NUMWAYS-1:0];
logic [NUMWAYS-1:0] WayHit;
logic CacheHit;
logic [BLOCKLEN-1:0] ReadDataBlockM;
logic [BLOCKLEN-1:0] ReadDataLineM;
logic [WORDSPERLINE-1:0] SRAMWordEnable;
logic SRAMWordWriteEnableM;
@ -137,7 +137,7 @@ module dcache
.WriteData(SRAMWriteData),
.SetValid, .ClearValid, .SetDirty, .ClearDirty, .SelEvict,
.VictimWay, .FlushWay, .SelFlush,
.ReadDataBlockWayMasked(ReadDataBlockWayMaskedM),
.ReadDataLineWayMasked,
.WayHit, .VictimDirtyWay, .VictimTagWay,
.InvalidateAll(1'b0));
@ -159,10 +159,10 @@ module dcache
assign VictimDirty = | VictimDirtyWay;
// ReadDataBlockWayMaskedM is a 2d array of cache block len by number of ways.
// ReadDataLineWayMaskedM is a 2d array of cache block len by number of ways.
// Need to OR together each way in a bitwise manner.
// Final part of the AO Mux. First is the AND in the cacheway.
or_rows #(NUMWAYS, BLOCKLEN) ReadDataAOMux(.a(ReadDataBlockWayMaskedM), .y(ReadDataBlockM));
or_rows #(NUMWAYS, BLOCKLEN) ReadDataAOMux(.a(ReadDataLineWayMasked), .y(ReadDataLineM));
or_rows #(NUMWAYS, TAGLEN) VictimTagAOMux(.a(VictimTagWay), .y(VictimTag));
@ -172,13 +172,13 @@ module dcache
genvar index;
generate
for (index = 0; index < WORDSPERLINE; index++) begin
assign ReadDataBlockSetsM[index] = ReadDataBlockM[((index+1)*`XLEN)-1: (index*`XLEN)];
assign ReadDataBlockSetsM[index] = ReadDataLineM[((index+1)*`XLEN)-1: (index*`XLEN)];
end
endgenerate
// variable input mux
assign ReadDataWordM = ReadDataBlockSetsM[LsuPAdrM[$clog2(WORDSPERLINE+`XLEN/8) : $clog2(`XLEN/8)]];
assign ReadDataWordM = ReadDataBlockSetsM[LsuPAdrM[LOGWPL + LOGXLENBYTES - 1 : LOGXLENBYTES]];
// Write Path CPU (IEU) side

284
wally-pipelined/src/cache/icache.sv vendored
View File

@ -28,136 +28,135 @@
module icache
(
// Basic pipeline stuff
input logic clk, reset,
input logic StallF,
input logic [`PA_BITS-1:0] PCNextF,
input logic [`PA_BITS-1:0] PCPF,
input logic [`XLEN-1:0] PCF,
input logic clk, reset,
input logic CPUBusy,
input logic [`PA_BITS-1:0] PCNextF,
input logic [`PA_BITS-1:0] PCPF,
input logic [`XLEN-1:0] PCF,
input logic ExceptionM, PendingInterruptM,
input logic ExceptionM, PendingInterruptM,
// Data read in from the ebu unit
(* mark_debug = "true" *) input logic [`XLEN-1:0] InstrInF,
(* mark_debug = "true" *) input logic InstrAckF,
(* mark_debug = "true" *) input logic [`XLEN-1:0] IfuBusHRDATA,
(* mark_debug = "true" *) input logic ICacheBusAck,
// Read requested from the ebu unit
(* mark_debug = "true" *) output logic [`PA_BITS-1:0] InstrPAdrF,
(* mark_debug = "true" *) output logic InstrReadF,
(* mark_debug = "true" *) output logic [`PA_BITS-1:0] ICacheBusAdr,
(* mark_debug = "true" *) output logic IfuBusFetch,
// High if the instruction currently in the fetch stage is compressed
output logic CompressedF,
output logic CompressedF,
// High if the icache is requesting a stall
output logic ICacheStallF,
input logic ITLBMissF,
input logic ITLBWriteF,
input logic InvalidateICacheM,
output logic ICacheStallF,
input logic CacheableF,
input logic ITLBMissF,
input logic ITLBWriteF,
input logic InvalidateICacheM,
// The raw (not decompressed) instruction that was requested
// If this instruction is compressed, upper 16 bits may be the next 16 bits or may be zeros
(* mark_debug = "true" *) output logic [31:0] FinalInstrRawF
(* mark_debug = "true" *) output logic [31:0] FinalInstrRawF
);
// Configuration parameters
localparam integer BLOCKLEN = `ICACHE_BLOCKLENINBITS;
localparam integer NUMLINES = `ICACHE_WAYSIZEINBYTES*8/`ICACHE_BLOCKLENINBITS;
localparam integer BLOCKBYTELEN = BLOCKLEN/8;
localparam integer BLOCKLEN = `ICACHE_BLOCKLENINBITS;
localparam integer NUMLINES = `ICACHE_WAYSIZEINBYTES*8/`ICACHE_BLOCKLENINBITS;
localparam integer BLOCKBYTELEN = BLOCKLEN/8;
localparam integer OFFSETLEN = $clog2(BLOCKBYTELEN);
localparam integer INDEXLEN = $clog2(NUMLINES);
localparam integer TAGLEN = `PA_BITS - OFFSETLEN - INDEXLEN;
localparam integer OFFSETLEN = $clog2(BLOCKBYTELEN);
localparam integer INDEXLEN = $clog2(NUMLINES);
localparam integer TAGLEN = `PA_BITS - OFFSETLEN - INDEXLEN;
localparam WORDSPERLINE = BLOCKLEN/`XLEN;
localparam LOGWPL = $clog2(WORDSPERLINE);
localparam FetchCountThreshold = WORDSPERLINE - 1;
localparam integer PA_WIDTH = `PA_BITS - 2;
localparam integer NUMWAYS = `ICACHE_NUMWAYS;
localparam integer PA_WIDTH = `PA_BITS - 2;
localparam integer NUMWAYS = `ICACHE_NUMWAYS;
// Input signals to cache memory
logic ICacheMemWriteEnable;
logic [BLOCKLEN-1:0] ICacheMemWriteData;
logic [`PA_BITS-1:0] PCTagF;
logic ICacheMemWriteEnable;
logic [BLOCKLEN-1:0] ICacheMemWriteData;
logic [`PA_BITS-1:0] FinalPCPF;
// Output signals from cache memory
logic [31:0] ICacheMemReadData;
logic ICacheReadEn;
logic [BLOCKLEN-1:0] ReadLineF;
logic [31:0] ICacheMemReadData;
logic ICacheReadEn;
logic [BLOCKLEN-1:0] ReadLineF;
logic [15:0] SpillDataBlock0;
logic spill;
logic spillSave;
logic [15:0] SpillDataBlock0;
logic spill;
logic spillSave;
logic FetchCountFlag;
logic CntEn;
logic FetchCountFlag;
logic CntEn;
logic [1:1] SelAdr_q;
logic [1:1] SelAdr_q;
logic [LOGWPL-1:0] FetchCount, NextFetchCount;
logic [`PA_BITS-1:0] PCPSpillF;
logic [LOGWPL-1:0] FetchCount, NextFetchCount;
logic [`PA_BITS-1:0] PCPSpillF;
logic CntReset;
logic [1:0] SelAdr;
logic [INDEXLEN-1:0] RAdr;
logic [NUMWAYS-1:0] VictimWay;
logic LRUWriteEn;
logic [NUMWAYS-1:0] WayHit;
logic hit;
logic CntReset;
logic [1:0] SelAdr;
logic [INDEXLEN-1:0] RAdr;
logic [NUMWAYS-1:0] VictimWay;
logic LRUWriteEn;
logic [NUMWAYS-1:0] WayHit;
logic hit;
logic [BLOCKLEN-1:0] ReadDataBlockWayMasked [NUMWAYS-1:0];
logic [BLOCKLEN-1:0] ReadDataLineWayMasked [NUMWAYS-1:0];
logic CacheableF;
logic [31:0] ReadLineSetsF [`ICACHE_BLOCKLENINBITS/16-1:0];
logic [`PA_BITS-1:0] BasePAdrF, BasePAdrMaskedF;
logic [OFFSETLEN-1:0] BasePAdrOffsetF;
logic [`PA_BITS-1:0] BasePAdrMaskedF;
logic [OFFSETLEN-1:0] BasePAdrOffsetF;
logic [NUMWAYS-1:0] SRAMWayWriteEnable;
logic [NUMWAYS-1:0] SRAMWayWriteEnable;
// on spill we want to get the first 2 bytes of the next cache block.
// the spill only occurs if the PCPF mod BlockByteLength == -2. Therefore we can
// simply add 2 to land on the next cache block.
assign PCPSpillF = PCPF + {{{PA_WIDTH}{1'b0}}, 2'b10}; // *** modelsim does not allow the use of PA_BITS for literal width.
assign PCPSpillF = PCPF + {{{PA_WIDTH}{1'b0}}, 2'b10};
mux3 #(INDEXLEN)
AdrSelMux(.d0(PCNextF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.d1(PCF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.d2(PCPSpillF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.s(SelAdr),
.y(RAdr));
.d1(PCF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.d2(PCPSpillF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.s(SelAdr),
.y(RAdr));
cacheway #(.NUMLINES(NUMLINES), .BLOCKLEN(BLOCKLEN), .TAGLEN(TAGLEN),
.OFFSETLEN(OFFSETLEN), .INDEXLEN(INDEXLEN), .DIRTY_BITS(0))
MemWay[NUMWAYS-1:0](.clk, .reset, .RAdr,
.PAdr(PCTagF),
.WriteEnable(SRAMWayWriteEnable),
.VDWriteEnable(1'b0),
.WriteWordEnable({{(BLOCKLEN/`XLEN){1'b1}}}),
.TagWriteEnable(SRAMWayWriteEnable),
.WriteData(ICacheMemWriteData),
.SetValid(ICacheMemWriteEnable),
.ClearValid(1'b0), .SetDirty(1'b0), .ClearDirty(1'b0), .SelEvict(1'b0),
.VictimWay,
.FlushWay(1'b0), .SelFlush(1'b0),
.ReadDataBlockWayMasked, .WayHit,
.VictimDirtyWay(), .VictimTagWay(),
.InvalidateAll(InvalidateICacheM));
.PAdr(FinalPCPF),
.WriteEnable(SRAMWayWriteEnable),
.VDWriteEnable(1'b0),
.WriteWordEnable({{(BLOCKLEN/`XLEN){1'b1}}}),
.TagWriteEnable(SRAMWayWriteEnable),
.WriteData(ICacheMemWriteData),
.SetValid(ICacheMemWriteEnable),
.ClearValid(1'b0), .SetDirty(1'b0), .ClearDirty(1'b0), .SelEvict(1'b0),
.VictimWay,
.FlushWay(1'b0), .SelFlush(1'b0),
.ReadDataLineWayMasked, .WayHit,
.VictimDirtyWay(), .VictimTagWay(),
.InvalidateAll(InvalidateICacheM));
generate
if(NUMWAYS > 1) begin
cachereplacementpolicy #(NUMWAYS, INDEXLEN, OFFSETLEN, NUMLINES)
cachereplacementpolicy(.clk, .reset,
.WayHit,
.VictimWay,
.LsuPAdrM(PCTagF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.RAdr,
.LRUWriteEn); // *** connect
.WayHit,
.VictimWay,
.LsuPAdrM(FinalPCPF[INDEXLEN+OFFSETLEN-1:OFFSETLEN]),
.RAdr,
.LRUWriteEn);
end else begin
assign VictimWay = 1'b1; // one hot.
end
@ -165,51 +164,36 @@ module icache
assign hit = | WayHit;
// ReadDataBlockWayMasked is a 2d array of cache block len by number of ways.
// ReadDataLineWayMasked is a 2d array of cache block len by number of ways.
// Need to OR together each way in a bitwise manner.
// Final part of the AO Mux. First is the AND in the cacheway.
or_rows #(NUMWAYS, BLOCKLEN) ReadDataAOMux(.a(ReadDataBlockWayMasked), .y(ReadLineF));
or_rows #(NUMWAYS, BLOCKLEN) ReadDataAOMux(.a(ReadDataLineWayMasked), .y(ReadLineF));
genvar index;
generate
for(index = 0; index < BLOCKLEN / 16 - 1; index++) begin
assign ReadLineSetsF[index] = ReadLineF[((index+1)*16)+16-1 : (index*16)];
end
assign ReadLineSetsF[BLOCKLEN/16-1] = {16'b0, ReadLineF[BLOCKLEN-1:BLOCKLEN-16]};
endgenerate
assign ICacheMemReadData = ReadLineSetsF[FinalPCPF[$clog2(BLOCKLEN / 32) + 1 : 1]];
always_comb begin
case (PCTagF[4:1])
0: ICacheMemReadData = ReadLineF[31:0];
1: ICacheMemReadData = ReadLineF[47:16];
2: ICacheMemReadData = ReadLineF[63:32];
3: ICacheMemReadData = ReadLineF[79:48];
4: ICacheMemReadData = ReadLineF[95:64];
5: ICacheMemReadData = ReadLineF[111:80];
6: ICacheMemReadData = ReadLineF[127:96];
7: ICacheMemReadData = ReadLineF[143:112];
8: ICacheMemReadData = ReadLineF[159:128];
9: ICacheMemReadData = ReadLineF[175:144];
10: ICacheMemReadData = ReadLineF[191:160];
11: ICacheMemReadData = ReadLineF[207:176];
12: ICacheMemReadData = ReadLineF[223:192];
13: ICacheMemReadData = ReadLineF[239:208];
14: ICacheMemReadData = ReadLineF[255:224];
15: ICacheMemReadData = {16'b0, ReadLineF[255:240]};
endcase
end
// spills require storing the first cache block so it can merged
// with the second
// can optimize size, for now just make it the size of the data
// leaving the cache memory.
flopenr #(16) SpillInstrReg(.clk(clk),
.en(spillSave),
.reset(reset),
.d(ICacheMemReadData[15:0]),
.q(SpillDataBlock0));
.en(spillSave),
.reset(reset),
.d(ICacheMemReadData[15:0]),
.q(SpillDataBlock0));
assign FinalInstrRawF = spill ? {ICacheMemReadData[15:0], SpillDataBlock0} : ICacheMemReadData;
// Detect if the instruction is compressed
assign CompressedF = FinalInstrRawF[1:0] != 2'b11;
assign spill = PCF[4:1] == 4'b1111 ? 1'b1 : 1'b0;
assign spill = &PCF[$clog2(BLOCKLEN/32)+1:1];
// to compute the fetch address we need to add the bit shifted
@ -218,10 +202,10 @@ module icache
flopenr #(LOGWPL)
FetchCountReg(.clk(clk),
.reset(reset | CntReset),
.en(CntEn),
.d(NextFetchCount),
.q(FetchCount));
.reset(reset | CntReset),
.en(CntEn),
.d(NextFetchCount),
.q(FetchCount));
assign NextFetchCount = FetchCount + 1'b1;
@ -231,10 +215,10 @@ module icache
generate
for (i = 0; i < WORDSPERLINE; i++) begin:storebuffer
flopenr #(`XLEN) sb(.clk(clk),
.reset(reset),
.en(InstrAckF & (i == FetchCount)),
.d(InstrInF),
.q(ICacheMemWriteData[(i+1)*`XLEN-1:i*`XLEN]));
.reset(reset),
.en(ICacheBusAck & (i == FetchCount)),
.d(IfuBusHRDATA),
.q(ICacheMemWriteData[(i+1)*`XLEN-1:i*`XLEN]));
end
endgenerate
@ -242,54 +226,44 @@ module icache
// this mux needs to be delayed 1 cycle as it occurs 1 pipeline stage later.
// *** read enable may not be necessary.
flopenr #(1) SelAdrReg(.clk(clk),
.reset(reset),
.en(ICacheReadEn),
.d(SelAdr[1]),
.q(SelAdr_q[1]));
.reset(reset),
.en(ICacheReadEn),
.d(SelAdr[1]),
.q(SelAdr_q[1]));
assign PCTagF = SelAdr_q[1] ? PCPSpillF : PCPF;
// unlike the dcache the victim is never dirty so no eviction is necessary.
/* -----\/----- EXCLUDED -----\/-----
mux2 #(`PA_BITS) BaseAdrMux(.d0(PCTagF),
.d1({VictimTag, PCTagF[INDEXLEN+OFFSETLEN-1:OFFSETLEN], {{OFFSETLEN}{1'b0}}}),
.s(SelEvict),
.y(BasePAdrF));
-----/\----- EXCLUDED -----/\----- */
assign BasePAdrF = PCTagF;
assign FinalPCPF = SelAdr_q[1] ? PCPSpillF : PCPF;
// if not cacheable the offset bits needs to be sent to the EBU.
// if cacheable the offset bits are discarded. $ FSM will fetch the whole block.
assign CacheableF = 1'b1; // *** BUG needs to be an input from MMU.
assign BasePAdrOffsetF = CacheableF ? {{OFFSETLEN}{1'b0}} : BasePAdrF[OFFSETLEN-1:0];
assign BasePAdrMaskedF = {BasePAdrF[`PA_BITS-1:OFFSETLEN], BasePAdrOffsetF};
assign BasePAdrOffsetF = CacheableF ? {{OFFSETLEN}{1'b0}} : FinalPCPF[OFFSETLEN-1:0];
assign BasePAdrMaskedF = {FinalPCPF[`PA_BITS-1:OFFSETLEN], BasePAdrOffsetF};
assign InstrPAdrF = ({{`PA_BITS-LOGWPL{1'b0}}, FetchCount} << $clog2(`XLEN/8)) + BasePAdrMaskedF;
assign ICacheBusAdr = ({{`PA_BITS-LOGWPL{1'b0}}, FetchCount} << $clog2(`XLEN/8)) + BasePAdrMaskedF;
// truncate the offset from PCPF for memory address generation
assign SRAMWayWriteEnable = ICacheMemWriteEnable ? VictimWay : '0;
icachefsm controller(.clk,
.reset,
.StallF,
.ICacheReadEn,
.ICacheMemWriteEnable,
.ICacheStallF,
.ITLBMissF,
.ITLBWriteF,
.ExceptionM,
.PendingInterruptM,
.InstrAckF,
.InstrReadF,
.hit,
.FetchCountFlag,
.spill,
.spillSave,
.CntEn,
.CntReset,
.SelAdr,
.LRUWriteEn);
.reset,
.CPUBusy,
.ICacheReadEn,
.ICacheMemWriteEnable,
.ICacheStallF,
.ITLBMissF,
.ITLBWriteF,
.ExceptionM,
.PendingInterruptM,
.ICacheBusAck,
.IfuBusFetch,
.hit,
.FetchCountFlag,
.spill,
.spillSave,
.CntEn,
.CntReset,
.SelAdr,
.LRUWriteEn);
endmodule

48
wally-pipelined/src/cache/icachefsm.sv vendored
View File

@ -29,7 +29,7 @@ module icachefsm
(// Inputs from pipeline
input logic clk, reset,
input logic StallF,
input logic CPUBusy,
// inputs from mmu
input logic ITLBMissF,
@ -38,7 +38,7 @@ module icachefsm
input logic ExceptionM, PendingInterruptM,
// BUS interface
input logic InstrAckF,
input logic ICacheBusAck,
// icache internal inputs
input logic hit,
@ -54,7 +54,7 @@ module icachefsm
output logic ICacheStallF,
// Bus interface outputs
output logic InstrReadF,
output logic IfuBusFetch,
// icache internal outputs
output logic spillSave,
@ -105,10 +105,6 @@ module icachefsm
STATE_MISS_SPILL_FINAL, // this state replicates STATE_READY's replay of the
// spill access but does nto consider spill. It also does not do another operation.
STATE_INVALIDATE, // *** not sure if invalidate or evict? invalidate by cache block or address?
STATE_TLB_MISS,
STATE_TLB_MISS_DONE,
STATE_CPU_BUSY,
STATE_CPU_BUSY_SPILL
} statetype;
@ -125,7 +121,7 @@ module icachefsm
always_comb begin
CntReset = 1'b0;
PreCntEn = 1'b0;
//InstrReadF = 1'b0;
//IfuBusFetch = 1'b0;
ICacheMemWriteEnable = 1'b0;
spillSave = 1'b0;
SelAdr = 2'b00;
@ -149,7 +145,7 @@ module icachefsm
else if (hit & ~spill) begin
ICacheStallF = 1'b0;
LRUWriteEn = 1'b1;
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY;
SelAdr = 2'b01;
end else begin
@ -169,7 +165,7 @@ module icachefsm
SelAdr = 2'b01;
NextState = STATE_MISS_SPILL_FETCH_WDV;
end else begin
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY;
SelAdr = 2'b01;
end else begin
@ -190,9 +186,9 @@ module icachefsm
end
STATE_HIT_SPILL_MISS_FETCH_WDV: begin
SelAdr = 2'b10;
//InstrReadF = 1'b1;
//IfuBusFetch = 1'b1;
PreCntEn = 1'b1;
if (FetchCountFlag & InstrAckF) begin
if (FetchCountFlag & ICacheBusAck) begin
NextState = STATE_HIT_SPILL_MISS_FETCH_DONE;
end else begin
NextState = STATE_HIT_SPILL_MISS_FETCH_WDV;
@ -214,7 +210,7 @@ module icachefsm
ICacheStallF = 1'b0;
LRUWriteEn = 1'b1;
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY_SPILL;
SelAdr = 2'b10;
end else begin
@ -225,9 +221,9 @@ module icachefsm
// branch 3 miss no spill
STATE_MISS_FETCH_WDV: begin
SelAdr = 2'b01;
//InstrReadF = 1'b1;
//IfuBusFetch = 1'b1;
PreCntEn = 1'b1;
if (FetchCountFlag & InstrAckF) begin
if (FetchCountFlag & ICacheBusAck) begin
NextState = STATE_MISS_FETCH_DONE;
end else begin
NextState = STATE_MISS_FETCH_WDV;
@ -248,7 +244,7 @@ module icachefsm
ICacheReadEn = 1'b1;
ICacheStallF = 1'b0;
LRUWriteEn = 1'b1;
if(StallF) begin
if(CPUBusy) begin
SelAdr = 2'b01;
NextState = STATE_CPU_BUSY;
SelAdr = 2'b01;
@ -260,8 +256,8 @@ module icachefsm
STATE_MISS_SPILL_FETCH_WDV: begin
SelAdr = 2'b01;
PreCntEn = 1'b1;
//InstrReadF = 1'b1;
if (FetchCountFlag & InstrAckF) begin
//IfuBusFetch = 1'b1;
if (FetchCountFlag & ICacheBusAck) begin
NextState = STATE_MISS_SPILL_FETCH_DONE;
end else begin
NextState = STATE_MISS_SPILL_FETCH_WDV;
@ -293,7 +289,7 @@ module icachefsm
SelAdr = 2'b00;
ICacheStallF = 1'b0;
LRUWriteEn = 1'b1;
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY_SPILL;
SelAdr = 2'b10;
end else begin
@ -304,8 +300,8 @@ module icachefsm
STATE_MISS_SPILL_MISS_FETCH_WDV: begin
SelAdr = 2'b10;
PreCntEn = 1'b1;
//InstrReadF = 1'b1;
if (FetchCountFlag & InstrAckF) begin
//IfuBusFetch = 1'b1;
if (FetchCountFlag & ICacheBusAck) begin
NextState = STATE_MISS_SPILL_MISS_FETCH_DONE;
end else begin
NextState = STATE_MISS_SPILL_MISS_FETCH_WDV;
@ -326,7 +322,7 @@ module icachefsm
SelAdr = 2'b00;
ICacheStallF = 1'b0;
LRUWriteEn = 1'b1;
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY_SPILL;
SelAdr = 2'b10;
end else begin
@ -335,7 +331,7 @@ module icachefsm
end
STATE_CPU_BUSY: begin
ICacheStallF = 1'b0;
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY;
SelAdr = 2'b01;
end
@ -346,7 +342,7 @@ module icachefsm
STATE_CPU_BUSY_SPILL: begin
ICacheStallF = 1'b0;
ICacheReadEn = 1'b1;
if(StallF) begin
if(CPUBusy) begin
NextState = STATE_CPU_BUSY_SPILL;
SelAdr = 2'b10;
end
@ -362,8 +358,8 @@ module icachefsm
endcase
end
assign CntEn = PreCntEn & InstrAckF;
assign InstrReadF = (CurrState == STATE_HIT_SPILL_MISS_FETCH_WDV) ||
assign CntEn = PreCntEn & ICacheBusAck;
assign IfuBusFetch = (CurrState == STATE_HIT_SPILL_MISS_FETCH_WDV) ||
(CurrState == STATE_MISS_FETCH_WDV) ||
(CurrState == STATE_MISS_SPILL_FETCH_WDV) ||
(CurrState == STATE_MISS_SPILL_MISS_FETCH_WDV);

22
wally-pipelined/src/ebu/ahblite.sv
View File

@ -40,10 +40,10 @@ module ahblite (
input logic UnsignedLoadM,
input logic [1:0] AtomicMaskedM,
// Signals from Instruction Cache
input logic [`PA_BITS-1:0] InstrPAdrF, // *** rename these to match block diagram
input logic InstrReadF,
output logic [`XLEN-1:0] InstrRData,
output logic InstrAckF,
input logic [`PA_BITS-1:0] ICacheBusAdr, // *** rename these to match block diagram
input logic IfuBusFetch,
output logic [`XLEN-1:0] IfuBusHRDATA,
output logic ICacheBusAck,
// Signals from Data Cache
input logic [`PA_BITS-1:0] LsuBusAdr,
input logic LsuBusRead,
@ -100,23 +100,23 @@ module ahblite (
case (BusState)
IDLE: if (LsuBusRead) NextBusState = MEMREAD; // Memory has priority over instructions
else if (LsuBusWrite)NextBusState = MEMWRITE;
else if (InstrReadF) NextBusState = INSTRREAD;
else if (IfuBusFetch) NextBusState = INSTRREAD;
else NextBusState = IDLE;
MEMREAD: if (~HREADY) NextBusState = MEMREAD;
else if (InstrReadF) NextBusState = INSTRREAD;
else if (IfuBusFetch) NextBusState = INSTRREAD;
else NextBusState = IDLE;
MEMWRITE: if (~HREADY) NextBusState = MEMWRITE;
else if (InstrReadF) NextBusState = INSTRREAD;
else if (IfuBusFetch) NextBusState = INSTRREAD;
else NextBusState = IDLE;
INSTRREAD: if (~HREADY) NextBusState = INSTRREAD;
else NextBusState = IDLE; // if (InstrReadF still high)
else NextBusState = IDLE; // if (IfuBusFetch still high)
default: NextBusState = IDLE;
endcase
// bus outputs
assign #1 GrantData = (NextBusState == MEMREAD) || (NextBusState == MEMWRITE);
assign #1 AccessAddress = (GrantData) ? LsuBusAdr[31:0] : InstrPAdrF[31:0];
assign #1 AccessAddress = (GrantData) ? LsuBusAdr[31:0] : ICacheBusAdr[31:0];
assign #1 HADDR = AccessAddress;
assign ISize = 3'b010; // 32 bit instructions for now; later improve for filling cache with full width; ignored on reads anyway
assign HSIZE = (GrantData) ? {1'b0, LsuBusSize[1:0]} : ISize;
@ -136,9 +136,9 @@ module ahblite (
// *** assumes AHBW = XLEN
assign InstrRData = HRDATA;
assign IfuBusHRDATA = HRDATA;
assign LsuBusHRDATA = HRDATA;
assign InstrAckF = (BusState == INSTRREAD) && (NextBusState != INSTRREAD);
assign ICacheBusAck = (BusState == INSTRREAD) && (NextBusState != INSTRREAD);
assign LsuBusAck = (BusState == MEMREAD) && (NextBusState != MEMREAD) || (BusState == MEMWRITE) && (NextBusState != MEMWRITE);
endmodule

24
wally-pipelined/src/ifu/ifu.sv
View File

@ -31,11 +31,11 @@ module ifu (
input logic StallF, StallD, StallE, StallM, StallW,
input logic FlushF, FlushD, FlushE, FlushM, FlushW,
// Fetch
input logic [`XLEN-1:0] InstrInF,
input logic InstrAckF,
input logic [`XLEN-1:0] IfuBusHRDATA,
input logic ICacheBusAck,
(* mark_debug = "true" *) output logic [`XLEN-1:0] PCF,
output logic [`PA_BITS-1:0] InstrPAdrF,
output logic InstrReadF,
output logic [`PA_BITS-1:0] ICacheBusAdr,
output logic IfuBusFetch,
output logic ICacheStallF,
// Execute
output logic [`XLEN-1:0] PCLinkE,
@ -104,6 +104,8 @@ module ifu (
logic [`XLEN+1:0] PCFExt;
logic [`XLEN-1:0] PCBPWrongInvalidate;
logic BPPredWrongM;
logic CacheableF;
generate
@ -136,7 +138,7 @@ module ifu (
.LoadAccessFaultM(),
.StoreAccessFaultM(),
.DisableTranslation(1'b0),
.Cacheable(), .Idempotent(), .AtomicAllowed(),
.Cacheable(CacheableF), .Idempotent(), .AtomicAllowed(),
.clk, .reset,
.SATP_REGW,
@ -157,17 +159,17 @@ module ifu (
// *** put memory interface on here, InstrF becomes output
//assign InstrPAdrF = PCF; // *** no MMU
//assign InstrReadF = ~StallD; // *** & ICacheMissF; add later
// assign InstrReadF = 1; // *** & ICacheMissF; add later
//assign ICacheBusAdr = PCF; // *** no MMU
//assign IfuBusFetch = ~StallD; // *** & ICacheMissF; add later
// assign IfuBusFetch = 1; // *** & ICacheMissF; add later
// conditional
// 1. ram // controlled by `MEM_IROM
// 2. cache // `MEM_ICACHE
// 3. wire pass-through
icache icache(.clk, .reset, .StallF, .ExceptionM, .PendingInterruptM, .InstrInF, .InstrAckF,
.InstrPAdrF, .InstrReadF, .CompressedF, .ICacheStallF, .ITLBMissF, .ITLBWriteF, .FinalInstrRawF,
icache icache(.clk, .reset, .CPUBusy(StallF), .ExceptionM, .PendingInterruptM, .IfuBusHRDATA, .ICacheBusAck,
.ICacheBusAdr, .IfuBusFetch, .CompressedF, .ICacheStallF, .ITLBMissF, .ITLBWriteF, .FinalInstrRawF,
.CacheableF,
.PCNextF(PCNextFPhys),
.PCPF(PCPFmmu),
.PCF,

16
wally-pipelined/src/wally/wallypipelinedhart.sv
View File

@ -128,10 +128,10 @@ module wallypipelinedhart (
logic CommittedM;
// AHB ifu interface
logic [`PA_BITS-1:0] InstrPAdrF;
logic [`XLEN-1:0] InstrRData;
logic InstrReadF;
logic InstrAckF;
logic [`PA_BITS-1:0] ICacheBusAdr;
logic [`XLEN-1:0] IfuBusHRDATA;
logic IfuBusFetch;
logic ICacheBusAck;
// AHB LSU interface
logic [`PA_BITS-1:0] LsuBusAdr;
@ -164,8 +164,8 @@ module wallypipelinedhart (
.ExceptionM, .PendingInterruptM,
// Fetch
.InstrInF(InstrRData), .InstrAckF, .PCF, .InstrPAdrF,
.InstrReadF, .ICacheStallF,
.IfuBusHRDATA, .ICacheBusAck, .PCF, .ICacheBusAdr,
.IfuBusFetch, .ICacheStallF,
// Execute
.PCLinkE, .PCSrcE, .IEUAdrE, .PCE,
@ -277,8 +277,8 @@ module wallypipelinedhart (
ahblite ebu(// IFU connections
.clk, .reset,
.UnsignedLoadM(1'b0), .AtomicMaskedM(2'b00),
.InstrPAdrF, // *** rename these to match block diagram
.InstrReadF, .InstrRData, .InstrAckF,
.ICacheBusAdr, // *** rename these to match block diagram
.IfuBusFetch, .IfuBusHRDATA, .ICacheBusAck,
// Signals from Data Cache
.LsuBusAdr, .LsuBusRead, .LsuBusWrite, .LsuBusHWDATA,
.LsuBusHRDATA,