cvw/src/uncore/clint_apb.sv
2023-03-24 17:32:25 -05:00

249 lines
9.4 KiB
Systemverilog

///////////////////////////////////////////
// clint_apb.sv
//
// Written: David_Harris@hmc.edu 14 January 2021
// Modified:
//
// Purpose: Core-Local Interruptor
// See FE310-G002-Manual-v19p05 for specifications
//
// Documentation: RISC-V System on Chip Design Chapter 15
//
// A component of the CORE-V-WALLY configurable RISC-V project.
//
// Copyright (C) 2021-23 Harvey Mudd College & Oklahoma State University
//
// SPDX-License-Identifier: Apache-2.0 WITH SHL-2.1
//
// Licensed under the Solderpad Hardware License v 2.1 (the “License”); you may not use this file
// except in compliance with the License, or, at your option, the Apache License version 2.0. You
// may obtain a copy of the License at
//
// https://solderpad.org/licenses/SHL-2.1/
//
// Unless required by applicable law or agreed to in writing, any work distributed under the
// License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND,
// either express or implied. See the License for the specific language governing permissions
// and limitations under the License.
////////////////////////////////////////////////////////////////////////////////////////////////
`include "wally-config.vh"
module clint_apb (
input logic PCLK, PRESETn,
input logic PSEL,
input logic [15:0] PADDR,
input logic [`XLEN-1:0] PWDATA,
input logic [`XLEN/8-1:0] PSTRB,
input logic PWRITE,
input logic PENABLE,
output logic [`XLEN-1:0] PRDATA,
output logic PREADY,
output logic [63:0] MTIME,
output logic MTimerInt, MSwInt
);
logic MSIP;
logic [15:0] entry;
logic memwrite;
logic [63:0] MTIMECMP;
integer i, j;
assign memwrite = PWRITE & PENABLE & PSEL; // only write in access phase
assign PREADY = 1'b1; // CLINT never takes >1 cycle to respond
// word aligned reads
if (`XLEN==64) assign #2 entry = {PADDR[15:3], 3'b000};
else assign #2 entry = {PADDR[15:2], 2'b00};
// DH 2/20/21: Eventually allow MTIME to run off a separate clock
// This will require synchronizing MTIME to the system clock
// before it is read or compared to MTIMECMP.
// It will also require synchronizing the write to MTIMECMP.
// Use req and ack signals synchronized across the clock domains.
// register access
if (`XLEN==64) begin:clint // 64-bit
always @(posedge PCLK) begin
case(entry)
16'h0000: PRDATA <= {63'b0, MSIP};
16'h4000: PRDATA <= MTIMECMP;
16'hBFF8: PRDATA <= MTIME;
default: PRDATA <= 0;
endcase
end
always_ff @(posedge PCLK or negedge PRESETn)
if (~PRESETn) begin
MSIP <= 0;
MTIMECMP <= 64'hFFFFFFFFFFFFFFFF; // Spec says MTIMECMP is not reset, but we reset to maximum value to prevent spurious timer interrupts
end else if (memwrite) begin
if (entry == 16'h0000) MSIP <= PWDATA[0];
if (entry == 16'h4000) begin
for(i=0;i<`XLEN/8;i++)
if(PSTRB[i])
MTIMECMP[i*8 +: 8] <= PWDATA[i*8 +: 8]; // ***dh: this notation isn't in book yet - maybe from Ross
end
end
// eventually replace MTIME logic below with timereg
// timereg tr(PCLK, PRESETn, TIMECLK, memwrite & (entry==16'hBFF8), 1'b0, PWDATA, MTIME, done);
always_ff @(posedge PCLK or negedge PRESETn)
if (~PRESETn) begin
MTIME <= 0;
end else if (memwrite & entry == 16'hBFF8) begin
// MTIME Counter. Eventually change this to run off separate clock. Synchronization then needed
for(j=0;j<`XLEN/8;j++)
if(PSTRB[j])
MTIME[j*8 +: 8] <= PWDATA[j*8 +: 8];
end else MTIME <= MTIME + 1;
end else begin:clint // 32-bit
always @(posedge PCLK) begin
case(entry)
16'h0000: PRDATA <= {31'b0, MSIP};
16'h4000: PRDATA <= MTIMECMP[31:0];
16'h4004: PRDATA <= MTIMECMP[63:32];
16'hBFF8: PRDATA <= MTIME[31:0];
16'hBFFC: PRDATA <= MTIME[63:32];
default: PRDATA <= 0;
endcase
end
always_ff @(posedge PCLK or negedge PRESETn)
if (~PRESETn) begin
MSIP <= 0;
MTIMECMP <= 0;
// MTIMECMP is not reset ***?
end else if (memwrite) begin
if (entry == 16'h0000) MSIP <= PWDATA[0];
if (entry == 16'h4000)
for(j=0;j<`XLEN/8;j++)
if(PSTRB[j])
MTIMECMP[j*8 +: 8] <= PWDATA[j*8 +: 8];
if (entry == 16'h4004)
for(j=0;j<`XLEN/8;j++)
if(PSTRB[j])
MTIMECMP[32 + j*8 +: 8] <= PWDATA[j*8 +: 8];
// MTIME Counter. Eventually change this to run off separate clock. Synchronization then needed
end
// eventually replace MTIME logic below with timereg
// timereg tr(PCLK, PRESETn, TIMECLK, memwrite & (entry==16'hBFF8), memwrite & (entry == 16'hBFFC), PWDATA, MTIME, done);
always_ff @(posedge PCLK or negedge PRESETn)
if (~PRESETn) begin
MTIME <= 0;
// MTIMECMP is not reset
end else if (memwrite & (entry == 16'hBFF8)) begin
for(i=0;i<`XLEN/8;i++)
if(PSTRB[i])
MTIME[i*8 +: 8] <= PWDATA[i*8 +: 8];
end else if (memwrite & (entry == 16'hBFFC)) begin
// MTIME Counter. Eventually change this to run off separate clock. Synchronization then needed
for(i=0;i<`XLEN/8;i++)
if(PSTRB[i])
MTIME[32 + i*8 +: 8]<= PWDATA[i*8 +: 8];
end else MTIME <= MTIME + 1;
end
// Software interrupt when MSIP is set
assign MSwInt = MSIP;
// Timer interrupt when MTIME >= MTIMECMP
assign MTimerInt = ({1'b0, MTIME} >= {1'b0, MTIMECMP}); // unsigned comparison
endmodule
module timeregsync(
input logic clk, resetn,
input logic we0, we1,
input logic [`XLEN-1:0] wd,
output logic [63:0] q);
if (`XLEN==64)
always_ff @(posedge clk or negedge resetn)
if (~resetn) q <= 0;
else if (we0) q <= wd;
else q <= q + 1;
else
always_ff @(posedge clk or negedge resetn)
if (~resetn) q <= 0;
else if (we0) q[31:0] <= wd;
else if (we1) q[63:32] <= wd;
else q <= q + 1;
endmodule
module timereg(
input logic PCLK, PRESETn, TIMECLK,
input logic we0, we1,
input logic [`XLEN-1:0] PWDATA,
output logic [63:0] MTIME,
output logic done);
// if (`TIMEBASE_SYNC) begin:timereg // use PCLK for MTIME
if (1) begin:timereg // use PCLK for MTIME
timregsync timeregsync(.clk(PCLK), .resetn(PRESETn), .we0, .we1, .wd(PWDATA), .q(MTIME));
assign done = 1; // immediately completes
end else begin // use asynchronous TIMECLK
// TIME counter runs on TIMECLK but bus interface runs on PCLK
// Need to synchronize reads and writes
// This is subtle because synchronizing a binary counter on a per-bit basis could give a mix of old and new bits
// Instead, we use a Gray coded counter that only changes one bit per cycle
// Synchronizing this for a read is safe because we are guaranteed to get either the old or the new value.
// Writing to the counter requires a request/acknowledge handshake to ensure the write value is held long enough.
// The handshake signals are synchronized in each direction across the interface
// There is no back pressure on instructions, so if multiple counter writes occur ***
logic req, req_sync, ack, we0_stored, we1_stored, ack_stored, resetn_sync;
logic [`XLEN-1:0] wd_stored;
logic [63:0] time_int, time_int_gc, time_gc, MTIME_GC;
// When a write enable is asserted for a cycle, sample the enables and data and raise a request until it is acknowledged
// When the acknowledge falls, the transaction is done and the system is ready for another write.
// ***look at redoing this assuming write enable and data are held rather than pulsed.
always_ff @(posedge PCLK or negedge PRESETn)
if (~PRESETn)
req <= 0; // don't bother resetting wd
else begin
req <= we0 | we1 | req & ~ack;
we0_stored <= we0;
we1_stored <= we1;
wd_stored <= PWDATA;
ack_stored <= ack;
done <= ack_stored & ~ack;
end
// synchronize the reset and reqest into the TIMECLK domain
sync resetsync(TIMECLK, PRESETn, resetn_sync);
sync rsync(TIMECLK, req, req_sync);
// synchronize the acknowledge back to the PCLK domain to indicate the request was handled and can be lowered
sync async(PCLK, req_sync, ack);
timeregsync timeregsync(.clk(TIMECLK), .resetn(resetn_sync), .we0(we0_stored), .we1(we1_stored), .wd(wd_stored), .q(time_int));
binarytogray b2g(time_int, time_int_gc);
flop gcreg(TIMECLK, time_int_gc, time_gc);
sync timesync[63:0](PCLK, time_gc, MTIME_GC);
graytobinary g2b(MTIME_GC, MTIME);
end
endmodule
module binarytogray #(parameter N = `XLEN) (
input logic [N-1:0] b,
output logic [N-1:0] g);
// G[N-1] = B[N-1]; G[i] = B[i] ^ B[i+1] for 0 <= i < N-1
// requires single layer of N-1 XOR gates
assign g = b ^ {1'b0, b[N-1:1]};
endmodule
module graytobinary #(parameter N = `XLEN) (
input logic [N-1:0] g,
output logic [N-1:0] b);
// B[N-1] = G[N-1]; B[i] = G[i] ^ B[i+1] for 0 <= i < N-1
// requires rippling through N-1 XOR gates
genvar i;
assign b[N-1] = g[N-1];
for (i=N-2; i >= 0; i--) begin:g2b
assign b[i] = g[i] ^ b[i+1];
end
endmodule