////////////////////////////////////////////////////////////////////////////////
|
////////////////////////////////////////////////////////////////////////////////
|
//
|
//
|
// Filename: pipefetch.v
|
// Filename: pipefetch.v
|
//
|
//
|
// Project: Zip CPU -- a small, lightweight, RISC CPU soft core
|
// Project: Zip CPU -- a small, lightweight, RISC CPU soft core
|
//
|
//
|
// Purpose: Keeping our CPU fed with instructions, at one per clock and
|
// Purpose: Keeping our CPU fed with instructions, at one per clock and
|
// with no stalls, can be quite a chore. Worse, the Wishbone
|
// with no stalls, can be quite a chore. Worse, the Wishbone
|
// takes a couple of cycles just to read one instruction from
|
// takes a couple of cycles just to read one instruction from
|
// the bus. However, if we use pipeline accesses to the Wishbone
|
// the bus. However, if we use pipeline accesses to the Wishbone
|
// bus, then we can read more and faster. Further, if we cache
|
// bus, then we can read more and faster. Further, if we cache
|
// these results so that we have them before we need them, then
|
// these results so that we have them before we need them, then
|
// we have a chance of keeping our CPU from stalling. Those are
|
// we have a chance of keeping our CPU from stalling. Those are
|
// the purposes of this instruction fetch module: 1) Pipeline
|
// the purposes of this instruction fetch module: 1) Pipeline
|
// wishbone accesses, and 2) an instruction cache.
|
// wishbone accesses, and 2) an instruction cache.
|
//
|
//
|
// Creator: Dan Gisselquist, Ph.D.
|
// Creator: Dan Gisselquist, Ph.D.
|
// Gisselquist Tecnology, LLC
|
// Gisselquist Tecnology, LLC
|
//
|
//
|
////////////////////////////////////////////////////////////////////////////////
|
////////////////////////////////////////////////////////////////////////////////
|
//
|
//
|
// Copyright (C) 2015, Gisselquist Technology, LLC
|
// Copyright (C) 2015, Gisselquist Technology, LLC
|
//
|
//
|
// This program is free software (firmware): you can redistribute it and/or
|
// This program is free software (firmware): you can redistribute it and/or
|
// modify it under the terms of the GNU General Public License as published
|
// modify it under the terms of the GNU General Public License as published
|
// by the Free Software Foundation, either version 3 of the License, or (at
|
// by the Free Software Foundation, either version 3 of the License, or (at
|
// your option) any later version.
|
// your option) any later version.
|
//
|
//
|
// This program is distributed in the hope that it will be useful, but WITHOUT
|
// This program is distributed in the hope that it will be useful, but WITHOUT
|
// ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY or
|
// ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY or
|
// FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
// FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
// for more details.
|
// for more details.
|
//
|
//
|
// License: GPL, v3, as defined and found on www.gnu.org,
|
// License: GPL, v3, as defined and found on www.gnu.org,
|
// http://www.gnu.org/licenses/gpl.html
|
// http://www.gnu.org/licenses/gpl.html
|
//
|
//
|
//
|
//
|
////////////////////////////////////////////////////////////////////////////////
|
////////////////////////////////////////////////////////////////////////////////
|
//
|
//
|
module pipefetch(i_clk, i_rst, i_new_pc, i_stall_n, i_pc,
|
module pipefetch(i_clk, i_rst, i_new_pc, i_stall_n, i_pc,
|
o_i, o_pc, o_v,
|
o_i, o_pc, o_v,
|
o_wb_cyc, o_wb_stb, o_wb_we, o_wb_addr, o_wb_data,
|
o_wb_cyc, o_wb_stb, o_wb_we, o_wb_addr, o_wb_data,
|
i_wb_ack, i_wb_stall, i_wb_data, i_wb_request);
|
i_wb_ack, i_wb_stall, i_wb_data, i_wb_request);
|
parameter RESET_ADDRESS=32'h0010_0000,
|
parameter RESET_ADDRESS=32'h0010_0000,
|
LGCACHELEN = 6, CACHELEN=(1<<LGCACHELEN),
|
LGCACHELEN = 6, CACHELEN=(1<<LGCACHELEN),
|
BUSW=32;
|
BUSW=32;
|
input i_clk, i_rst, i_new_pc, i_stall_n;
|
input i_clk, i_rst, i_new_pc, i_stall_n;
|
input [(BUSW-1):0] i_pc;
|
input [(BUSW-1):0] i_pc;
|
output reg [(BUSW-1):0] o_i;
|
output reg [(BUSW-1):0] o_i;
|
output reg [(BUSW-1):0] o_pc;
|
output reg [(BUSW-1):0] o_pc;
|
output wire o_v;
|
output wire o_v;
|
//
|
//
|
output reg o_wb_cyc, o_wb_stb;
|
output reg o_wb_cyc, o_wb_stb;
|
output wire o_wb_we;
|
output wire o_wb_we;
|
output reg [(BUSW-1):0] o_wb_addr;
|
output reg [(BUSW-1):0] o_wb_addr;
|
output wire [(BUSW-1):0] o_wb_data;
|
output wire [(BUSW-1):0] o_wb_data;
|
//
|
//
|
input i_wb_ack, i_wb_stall;
|
input i_wb_ack, i_wb_stall;
|
input [(BUSW-1):0] i_wb_data;
|
input [(BUSW-1):0] i_wb_data;
|
//
|
//
|
// Is the (data) memory unit also requesting access to the bus?
|
// Is the (data) memory unit also requesting access to the bus?
|
input i_wb_request;
|
input i_wb_request;
|
|
|
// Fixed bus outputs: we read from the bus only, never write.
|
// Fixed bus outputs: we read from the bus only, never write.
|
// Thus the output data is ... irrelevant and don't care. We set it
|
// Thus the output data is ... irrelevant and don't care. We set it
|
// to zero just to set it to something.
|
// to zero just to set it to something.
|
assign o_wb_we = 1'b0;
|
assign o_wb_we = 1'b0;
|
assign o_wb_data = 0;
|
assign o_wb_data = 0;
|
|
|
reg [(BUSW-1):0] r_cache_base;
|
reg [(BUSW-1):0] r_cache_base;
|
reg [(LGCACHELEN):0] r_nvalid, r_acks_waiting;
|
reg [(LGCACHELEN):0] r_nvalid, r_acks_waiting;
|
reg [(BUSW-1):0] cache[0:(CACHELEN-1)];
|
reg [(BUSW-1):0] cache[0:(CACHELEN-1)];
|
|
|
reg [(LGCACHELEN-1):0] r_cache_offset;
|
reg [(LGCACHELEN-1):0] r_cache_offset;
|
|
|
reg r_addr_set;
|
reg r_addr_set;
|
reg [(BUSW-1):0] r_addr;
|
reg [(BUSW-1):0] r_addr;
|
|
|
wire [(BUSW-1):0] bus_nvalid;
|
wire [(BUSW-1):0] bus_nvalid;
|
assign bus_nvalid = { {(BUSW-LGCACHELEN-1){1'b0}}, r_nvalid };
|
assign bus_nvalid = { {(BUSW-LGCACHELEN-1){1'b0}}, r_nvalid };
|
|
|
// What are some of the conditions for which we need to restart the
|
// What are some of the conditions for which we need to restart the
|
// cache?
|
// cache?
|
wire w_pc_out_of_bounds;
|
wire w_pc_out_of_bounds;
|
assign w_pc_out_of_bounds = ((i_new_pc)&&((r_nvalid == 0)
|
assign w_pc_out_of_bounds = ((i_new_pc)&&((r_nvalid == 0)
|
||(i_pc < r_cache_base)
|
||(i_pc < r_cache_base)
|
||(i_pc >= r_cache_base + CACHELEN)));
|
||(i_pc >= r_cache_base + CACHELEN)));
|
wire w_ran_off_end_of_cache;
|
wire w_ran_off_end_of_cache;
|
assign w_ran_off_end_of_cache =((r_addr_set)&&((r_addr < r_cache_base)
|
assign w_ran_off_end_of_cache =((r_addr_set)&&((r_addr < r_cache_base)
|
||(r_addr >= r_cache_base + CACHELEN)));
|
||(r_addr >= r_cache_base + CACHELEN)));
|
wire w_running_out_of_cache;
|
wire w_running_out_of_cache;
|
assign w_running_out_of_cache = (r_addr_set)
|
assign w_running_out_of_cache = (r_addr_set)
|
&&(r_addr >= r_cache_base + (1<<(LGCACHELEN-2))
|
&&(r_addr >= r_cache_base + (1<<(LGCACHELEN-2))
|
+ (1<<(LGCACHELEN-1)));
|
+ (1<<(LGCACHELEN-1)));
|
initial r_nvalid = 0;
|
initial r_nvalid = 0;
|
initial r_cache_base = RESET_ADDRESS;
|
initial r_cache_base = RESET_ADDRESS;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
begin
|
begin
|
if (i_rst)
|
if (i_rst)
|
begin
|
begin
|
o_wb_cyc <= 1'b0;
|
o_wb_cyc <= 1'b0;
|
// r_cache_base <= RESET_ADDRESS;
|
// r_cache_base <= RESET_ADDRESS;
|
// end else if ((~o_wb_cyc)&&(i_new_pc)&&(r_nvalid != 0)
|
// end else if ((~o_wb_cyc)&&(i_new_pc)&&(r_nvalid != 0)
|
// &&(i_pc >= r_cache_base)
|
// &&(i_pc >= r_cache_base)
|
// &&(i_pc < r_cache_base + bus_nvalid))
|
// &&(i_pc < r_cache_base + bus_nvalid))
|
// begin
|
// begin
|
// The new instruction is in our cache, do nothing
|
// The new instruction is in our cache, do nothing
|
// with the bus here.
|
// with the bus here.
|
end else if ((o_wb_cyc)&&(w_pc_out_of_bounds))
|
end else if ((o_wb_cyc)&&(w_pc_out_of_bounds))
|
begin
|
begin
|
// We need to abandon our bus action to start over in
|
// We need to abandon our bus action to start over in
|
// a new region, setting up a new cache. This may
|
// a new region, setting up a new cache. This may
|
// happen mid cycle while waiting for a result. By
|
// happen mid cycle while waiting for a result. By
|
// dropping o_wb_cyc, we state that we are no longer
|
// dropping o_wb_cyc, we state that we are no longer
|
// interested in that result--whatever it might be.
|
// interested in that result--whatever it might be.
|
o_wb_cyc <= 1'b0;
|
o_wb_cyc <= 1'b0;
|
o_wb_stb <= 1'b0;
|
o_wb_stb <= 1'b0;
|
end else if ((~o_wb_cyc)&&(~r_nvalid[LGCACHELEN])&&(~i_wb_request)&&(r_addr_set))
|
end else if ((~o_wb_cyc)&&(~r_nvalid[LGCACHELEN])&&(~i_wb_request)&&(r_addr_set))
|
begin
|
begin
|
// Restart a bus cycle that was interrupted when the
|
// Restart a bus cycle that was interrupted when the
|
// data section wanted access to our bus.
|
// data section wanted access to our bus.
|
o_wb_cyc <= 1'b1;
|
o_wb_cyc <= 1'b1;
|
o_wb_stb <= 1'b1;
|
o_wb_stb <= 1'b1;
|
// o_wb_addr <= r_cache_base + bus_nvalid;
|
// o_wb_addr <= r_cache_base + bus_nvalid;
|
end else if ((~o_wb_cyc)&&(
|
end else if ((~o_wb_cyc)&&(
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
begin
|
begin
|
// Start a bus transaction
|
// Start a bus transaction
|
o_wb_cyc <= 1'b1;
|
o_wb_cyc <= 1'b1;
|
o_wb_stb <= 1'b1;
|
o_wb_stb <= 1'b1;
|
// o_wb_addr <= (i_new_pc) ? i_pc : r_addr;
|
// o_wb_addr <= (i_new_pc) ? i_pc : r_addr;
|
// r_nvalid <= 0;
|
// r_nvalid <= 0;
|
// r_cache_base <= (i_new_pc) ? i_pc : r_addr;
|
// r_cache_base <= (i_new_pc) ? i_pc : r_addr;
|
// r_cache_offset <= 0;
|
// r_cache_offset <= 0;
|
end else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
end else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
begin
|
begin
|
// If we're using the last quarter of the cache, then
|
// If we're using the last quarter of the cache, then
|
// let's start a bus transaction to extend the cache.
|
// let's start a bus transaction to extend the cache.
|
o_wb_cyc <= 1'b1;
|
o_wb_cyc <= 1'b1;
|
o_wb_stb <= 1'b1;
|
o_wb_stb <= 1'b1;
|
// o_wb_addr <= r_cache_base + (1<<(LGCACHELEN));
|
// o_wb_addr <= r_cache_base + (1<<(LGCACHELEN));
|
// r_nvalid <= r_nvalid - (1<<(LGCACHELEN-2));
|
// r_nvalid <= r_nvalid - (1<<(LGCACHELEN-2));
|
// r_cache_base <= r_cache_base + (1<<(LGCACHELEN-2));
|
// r_cache_base <= r_cache_base + (1<<(LGCACHELEN-2));
|
// r_cache_offset <= r_cache_offset + (1<<(LGCACHELEN-2));
|
// r_cache_offset <= r_cache_offset + (1<<(LGCACHELEN-2));
|
end else if (o_wb_cyc)
|
end else if (o_wb_cyc)
|
begin
|
begin
|
// This handles everything ... but the case where
|
// This handles everything ... but the case where
|
// while reading we need to extend our cache.
|
// while reading we need to extend our cache.
|
if ((o_wb_stb)&&(~i_wb_stall))
|
if ((o_wb_stb)&&(~i_wb_stall))
|
begin
|
begin
|
// o_wb_addr <= o_wb_addr + 1;
|
// o_wb_addr <= o_wb_addr + 1;
|
if ((o_wb_addr - r_cache_base >= CACHELEN-1)
|
if ((o_wb_addr - r_cache_base >= CACHELEN-1)
|
||(i_wb_request))
|
||(i_wb_request))
|
o_wb_stb <= 1'b0;
|
o_wb_stb <= 1'b0;
|
end
|
end
|
|
|
if (i_wb_ack)
|
if (i_wb_ack)
|
begin
|
begin
|
// r_nvalid <= r_nvalid + 1;
|
// r_nvalid <= r_nvalid + 1;
|
if ((r_acks_waiting == 1)&&(~o_wb_stb))
|
if ((r_acks_waiting == 1)&&(~o_wb_stb))
|
o_wb_cyc <= 1'b0;
|
o_wb_cyc <= 1'b0;
|
end
|
end
|
end
|
end
|
end
|
end
|
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if ((~o_wb_cyc)&&(
|
if (i_rst) // Required, so we can reload memoy and then reset
|
|
r_nvalid <= 0;
|
|
else if ((~o_wb_cyc)&&(
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
r_nvalid <= 0;
|
r_nvalid <= 0;
|
else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
r_nvalid <= r_nvalid - (1<<(LGCACHELEN-2));
|
r_nvalid <= r_nvalid - (1<<(LGCACHELEN-2));
|
else if ((o_wb_cyc)&&(i_wb_ack))
|
else if ((o_wb_cyc)&&(i_wb_ack))
|
r_nvalid <= r_nvalid+1;
|
r_nvalid <= r_nvalid+1;
|
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if ((~o_wb_cyc)&&(
|
if ((~o_wb_cyc)&&(
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
r_cache_base <= (i_new_pc) ? i_pc : r_addr;
|
r_cache_base <= (i_new_pc) ? i_pc : r_addr;
|
else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
r_cache_base <= r_cache_base + (1<<(LGCACHELEN-2));
|
r_cache_base <= r_cache_base + (1<<(LGCACHELEN-2));
|
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if ((~o_wb_cyc)&&(
|
if ((~o_wb_cyc)&&(
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
(w_pc_out_of_bounds)||(w_ran_off_end_of_cache)))
|
r_cache_offset <= 0;
|
r_cache_offset <= 0;
|
else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
else if ((~o_wb_cyc)&&(w_running_out_of_cache))
|
r_cache_offset <= r_cache_offset + (1<<(LGCACHELEN-2));
|
r_cache_offset <= r_cache_offset + (1<<(LGCACHELEN-2));
|
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if ((~o_wb_cyc)&&((w_pc_out_of_bounds)
|
if ((~o_wb_cyc)&&((w_pc_out_of_bounds)
|
||(w_ran_off_end_of_cache)))
|
||(w_ran_off_end_of_cache)))
|
o_wb_addr <= (i_new_pc) ? i_pc : r_addr;
|
o_wb_addr <= (i_new_pc) ? i_pc : r_addr;
|
else if ((o_wb_cyc)&&(o_wb_stb)&&(~i_wb_stall))
|
else if ((o_wb_cyc)&&(o_wb_stb)&&(~i_wb_stall))
|
o_wb_addr <= o_wb_addr + 1;
|
o_wb_addr <= o_wb_addr + 1;
|
|
|
initial r_acks_waiting = 0;
|
initial r_acks_waiting = 0;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (~o_wb_cyc)
|
if (~o_wb_cyc)
|
r_acks_waiting <= 0;
|
r_acks_waiting <= 0;
|
else if ((o_wb_stb)&&(~i_wb_stall)&&(~i_wb_ack))
|
else if ((o_wb_stb)&&(~i_wb_stall)&&(~i_wb_ack))
|
r_acks_waiting <= r_acks_waiting + ((i_wb_ack)? 0:1);
|
r_acks_waiting <= r_acks_waiting + ((i_wb_ack)? 0:1);
|
else if ((i_wb_ack)&&((~o_wb_stb)||(i_wb_stall)))
|
else if ((i_wb_ack)&&((~o_wb_stb)||(i_wb_stall)))
|
r_acks_waiting <= r_acks_waiting - 1;
|
r_acks_waiting <= r_acks_waiting - 1;
|
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if ((o_wb_cyc)&&(i_wb_ack))
|
if ((o_wb_cyc)&&(i_wb_ack))
|
cache[r_nvalid[(LGCACHELEN-1):0]+r_cache_offset]
|
cache[r_nvalid[(LGCACHELEN-1):0]+r_cache_offset]
|
<= i_wb_data;
|
<= i_wb_data;
|
|
|
initial r_addr_set = 1'b0;
|
initial r_addr_set = 1'b0;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_rst)
|
if (i_rst)
|
r_addr_set <= 1'b0;
|
r_addr_set <= 1'b0;
|
else if (i_new_pc)
|
else if (i_new_pc)
|
r_addr_set <= 1'b1;
|
r_addr_set <= 1'b1;
|
|
|
// Now, read from the cache
|
// Now, read from the cache
|
wire w_cv; // Cache valid, address is in the cache
|
wire w_cv; // Cache valid, address is in the cache
|
reg r_cv;
|
reg r_cv;
|
assign w_cv = ((r_nvalid != 0)&&(r_addr>=r_cache_base)
|
assign w_cv = ((r_nvalid != 0)&&(r_addr>=r_cache_base)
|
&&(r_addr-r_cache_base < bus_nvalid));
|
&&(r_addr-r_cache_base < bus_nvalid));
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
r_cv <= (~i_new_pc)&&(w_cv);
|
r_cv <= (~i_new_pc)&&(w_cv);
|
assign o_v = (r_cv)&&(~i_new_pc);
|
assign o_v = (r_cv)&&(~i_new_pc);
|
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_new_pc)
|
if (i_new_pc)
|
r_addr <= i_pc;
|
r_addr <= i_pc;
|
else if ((i_stall_n)&&(w_cv))
|
else if ((i_stall_n)&&(w_cv))
|
r_addr <= r_addr + 1;
|
r_addr <= r_addr + 1;
|
|
|
wire [(LGCACHELEN-1):0] c_rdaddr, c_cache_base;
|
wire [(LGCACHELEN-1):0] c_rdaddr, c_cache_base;
|
assign c_cache_base = r_cache_base[(LGCACHELEN-1):0];
|
assign c_cache_base = r_cache_base[(LGCACHELEN-1):0];
|
assign c_rdaddr = r_addr[(LGCACHELEN-1):0]-c_cache_base+r_cache_offset;
|
assign c_rdaddr = r_addr[(LGCACHELEN-1):0]-c_cache_base+r_cache_offset;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_stall_n)
|
if (i_stall_n)
|
o_i <= cache[c_rdaddr];
|
o_i <= cache[c_rdaddr];
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_stall_n)
|
if (i_stall_n)
|
o_pc <= r_addr;
|
o_pc <= r_addr;
|
|
|
|
|
endmodule
|
endmodule
|
|
|
