| Line 12... |
Line 12... |
// Creator: Dan Gisselquist, Ph.D.
|
// Creator: Dan Gisselquist, Ph.D.
|
// Gisselquist Technology, LLC
|
// Gisselquist Technology, LLC
|
//
|
//
|
///////////////////////////////////////////////////////////////////////////
|
///////////////////////////////////////////////////////////////////////////
|
//
|
//
|
// Copyright (C) 2015, Gisselquist Technology, LLC
|
// Copyright (C) 2015-2016, 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.
|
| Line 30... |
Line 30... |
// http://www.gnu.org/licenses/gpl.html
|
// http://www.gnu.org/licenses/gpl.html
|
//
|
//
|
//
|
//
|
///////////////////////////////////////////////////////////////////////////
|
///////////////////////////////////////////////////////////////////////////
|
//
|
//
|
`define LONG_MPY
|
`include "cpudefs.v"
|
module cpuops(i_clk,i_rst, i_ce, i_valid, i_op, i_a, i_b, o_c, o_f, o_valid,
|
//
|
o_illegal, o_busy);
|
module cpuops(i_clk,i_rst, i_ce, i_op, i_a, i_b, o_c, o_f, o_valid,
|
parameter IMPLEMENT_MPY = 1;
|
o_busy);
|
|
parameter IMPLEMENT_MPY = `OPT_MULTIPLY;
|
input i_clk, i_rst, i_ce;
|
input i_clk, i_rst, i_ce;
|
input [3:0] i_op;
|
input [3:0] i_op;
|
input [31:0] i_a, i_b;
|
input [31:0] i_a, i_b;
|
input i_valid;
|
|
output reg [31:0] o_c;
|
output reg [31:0] o_c;
|
output wire [3:0] o_f;
|
output wire [3:0] o_f;
|
output reg o_valid;
|
output reg o_valid;
|
output wire o_illegal;
|
|
output wire o_busy;
|
output wire o_busy;
|
|
|
// Rotate-left pre-logic
|
// Rotate-left pre-logic
|
wire [63:0] w_rol_tmp;
|
wire [63:0] w_rol_tmp;
|
assign w_rol_tmp = { i_a, i_a } << i_b[4:0];
|
assign w_rol_tmp = { i_a, i_a } << i_b[4:0];
|
| Line 97... |
Line 96... |
set_ovfl =(((i_op==4'h0)&&(i_a[31] != i_b[31]))//SUB&CMP
|
set_ovfl =(((i_op==4'h0)&&(i_a[31] != i_b[31]))//SUB&CMP
|
||((i_op==4'h2)&&(i_a[31] == i_b[31])) // ADD
|
||((i_op==4'h2)&&(i_a[31] == i_b[31])) // ADD
|
||(i_op == 4'h6) // LSL
|
||(i_op == 4'h6) // LSL
|
||(i_op == 4'h5)); // LSR
|
||(i_op == 4'h5)); // LSR
|
|
|
`ifdef LONG_MPY
|
wire [63:0] mpy_result; // Where we dump the multiply result
|
reg mpyhi;
|
reg mpyhi; // Return the high half of the multiply
|
wire mpybusy;
|
wire mpybusy; // The multiply is busy if true
|
`endif
|
wire mpydone; // True if we'll be valid on the next clock;
|
|
|
// A 4-way multiplexer can be done in one 6-LUT.
|
// A 4-way multiplexer can be done in one 6-LUT.
|
// A 16-way multiplexer can therefore be done in 4x 6-LUT's with
|
// A 16-way multiplexer can therefore be done in 4x 6-LUT's with
|
// the Xilinx multiplexer fabric that follows.
|
// the Xilinx multiplexer fabric that follows.
|
// Given that we wish to apply this multiplexer approach to 33-bits,
|
// Given that we wish to apply this multiplexer approach to 33-bits,
|
// this will cost a minimum of 132 6-LUTs.
|
// this will cost a minimum of 132 6-LUTs.
|
|
|
|
wire this_is_a_multiply_op;
|
|
assign this_is_a_multiply_op = (i_ce)&&((i_op[3:1]==3'h5)||(i_op[3:0]==4'h8));
|
|
|
generate
|
generate
|
if (IMPLEMENT_MPY == 0)
|
if (IMPLEMENT_MPY == 0)
|
begin
|
begin // No multiply support.
|
|
assign mpy_result = 63'h00;
|
|
end else if (IMPLEMENT_MPY == 1)
|
|
begin // Our single clock option (no extra clocks)
|
|
wire signed [63:0] w_mpy_a_input, w_mpy_b_input;
|
|
assign w_mpy_a_input = {{(32){(i_a[31])&(i_op[0])}},i_a[31:0]};
|
|
assign w_mpy_b_input = {{(32){(i_b[31])&(i_op[0])}},i_b[31:0]};
|
|
assign mpy_result = w_mpy_a_input * w_mpy_b_input;
|
|
assign mpybusy = 1'b0;
|
|
assign mpydone = 1'b0;
|
|
always @(*) mpyhi = 1'b0; // Not needed
|
|
end else if (IMPLEMENT_MPY == 2)
|
|
begin // Our two clock option (ALU must pause for 1 clock)
|
|
reg signed [63:0] r_mpy_a_input, r_mpy_b_input;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_ce)
|
|
begin
|
begin
|
pre_sign <= (i_a[31]);
|
r_mpy_a_input <={{(32){(i_a[31])&(i_op[0])}},i_a[31:0]};
|
c <= 1'b0;
|
r_mpy_b_input <={{(32){(i_b[31])&(i_op[0])}},i_b[31:0]};
|
casez(i_op)
|
|
4'b0000:{c,o_c } <= {1'b0,i_a}-{1'b0,i_b};// CMP/SUB
|
|
4'b0001: o_c <= i_a & i_b; // BTST/And
|
|
4'b0010:{c,o_c } <= i_a + i_b; // Add
|
|
4'b0011: o_c <= i_a | i_b; // Or
|
|
4'b0100: o_c <= i_a ^ i_b; // Xor
|
|
4'b0101:{o_c,c } <= w_lsr_result[32:0]; // LSR
|
|
4'b0110:{c,o_c } <= w_lsl_result[32:0]; // LSL
|
|
4'b0111:{o_c,c } <= w_asr_result[32:0]; // ASR
|
|
`ifndef LONG_MPY
|
|
4'b1000: o_c <= { i_b[15: 0], i_a[15:0] }; // LODIHI
|
|
`endif
|
|
4'b1001: o_c <= { i_a[31:16], i_b[15:0] }; // LODILO
|
|
// 4'h1010: The unimplemented MPYU,
|
|
// 4'h1011: and here for the unimplemented MPYS
|
|
4'b1100: o_c <= w_brev_result; // BREV
|
|
4'b1101: o_c <= w_popc_result; // POPC
|
|
4'b1110: o_c <= w_rol_result; // ROL
|
|
default: o_c <= i_b; // MOV, LDI
|
|
endcase
|
|
end
|
end
|
|
|
assign o_busy = 1'b0;
|
assign mpy_result = r_mpy_a_input * r_mpy_b_input;
|
|
assign mpybusy = 1'b0;
|
|
|
reg r_illegal;
|
|
always @(posedge i_clk)
|
|
r_illegal <= (i_ce)&&((i_op == 4'ha)||(i_op == 4'hb)
|
|
`ifdef LONG_MPY
|
|
||(i_op == 4'h8)
|
|
`endif
|
|
);
|
|
assign o_illegal = r_illegal;
|
|
end else begin
|
|
//
|
|
// Multiply pre-logic
|
|
//
|
|
`ifdef LONG_MPY
|
|
reg [63:0] r_mpy_result;
|
|
if (IMPLEMENT_MPY == 1)
|
|
begin // Our two clock option (one clock extra)
|
|
reg signed [64:0] r_mpy_a_input, r_mpy_b_input;
|
|
reg mpypipe, x;
|
|
initial mpypipe = 1'b0;
|
initial mpypipe = 1'b0;
|
|
reg mpypipe;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
mpypipe <= (i_ce)&&((i_op[3:1]==3'h5)||(i_op[3:0]==4'h8));
|
if (i_rst)
|
|
mpypipe <= 1'b0;
|
|
else
|
|
mpypipe <= (this_is_a_multiply_op);
|
|
|
|
assign mpydone = mpypipe; // this_is_a_multiply_op;
|
|
always @(posedge i_clk)
|
|
if (this_is_a_multiply_op)
|
|
mpyhi = i_op[1];
|
|
end else if (IMPLEMENT_MPY == 3)
|
|
begin // Our three clock option (ALU pauses for 2 clocks)
|
|
reg signed [63:0] r_smpy_result;
|
|
reg [63:0] r_umpy_result;
|
|
reg signed [31:0] r_mpy_a_input, r_mpy_b_input;
|
|
reg [1:0] mpypipe;
|
|
reg [1:0] r_sgn;
|
|
|
|
initial mpypipe = 2'b0;
|
|
always @(posedge i_clk)
|
|
if (i_rst)
|
|
mpypipe <= 2'b0;
|
|
else
|
|
mpypipe <= { mpypipe[0], this_is_a_multiply_op };
|
|
|
|
// First clock
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_ce)
|
|
begin
|
begin
|
r_mpy_a_input <= {{(33){(i_a[31])&(i_op[0])}},
|
r_mpy_a_input <= i_a[31:0];
|
i_a[31:0]};
|
r_mpy_b_input <= i_b[31:0];
|
r_mpy_b_input <= {{(33){(i_b[31])&(i_op[0])}},
|
r_sgn <= { r_sgn[0], i_op[0] };
|
i_b[31:0]};
|
|
end
|
end
|
|
|
|
// Second clock
|
|
`ifdef VERILATOR
|
|
wire signed [63:0] s_mpy_a_input, s_mpy_b_input;
|
|
wire [63:0] u_mpy_a_input, u_mpy_b_input;
|
|
|
|
assign s_mpy_a_input = {{(32){r_mpy_a_input[31]}},r_mpy_a_input};
|
|
assign s_mpy_b_input = {{(32){r_mpy_b_input[31]}},r_mpy_b_input};
|
|
assign u_mpy_a_input = {32'h00,r_mpy_a_input};
|
|
assign u_mpy_b_input = {32'h00,r_mpy_b_input};
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (mpypipe)
|
r_smpy_result = s_mpy_a_input * s_mpy_b_input;
|
{x, r_mpy_result} = r_mpy_a_input
|
|
* r_mpy_b_input;
|
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_ce)
|
r_umpy_result = u_mpy_a_input * u_mpy_b_input;
|
|
`else
|
|
|
|
wire [31:0] u_mpy_a_input, u_mpy_b_input;
|
|
|
|
assign u_mpy_a_input = r_mpy_a_input;
|
|
assign u_mpy_b_input = r_mpy_b_input;
|
|
|
|
always @(posedge i_clk)
|
|
r_smpy_result = r_mpy_a_input * r_mpy_b_input;
|
|
always @(posedge i_clk)
|
|
r_umpy_result = u_mpy_a_input * u_mpy_b_input;
|
|
`endif
|
|
|
|
always @(posedge i_clk)
|
|
if (this_is_a_multiply_op)
|
mpyhi = i_op[1];
|
mpyhi = i_op[1];
|
assign mpybusy = mpypipe;
|
assign mpybusy = mpypipe[0];
|
end else if (IMPLEMENT_MPY == 2)
|
assign mpy_result = (r_sgn[1])?r_smpy_result:r_umpy_result;
|
|
assign mpydone = mpypipe[1];
|
|
|
|
// Results are then set on the third clock
|
|
end else // if (IMPLEMENT_MPY <= 4)
|
begin // The three clock option
|
begin // The three clock option
|
|
reg [63:0] r_mpy_result;
|
reg [31:0] r_mpy_a_input, r_mpy_b_input;
|
reg [31:0] r_mpy_a_input, r_mpy_b_input;
|
reg r_mpy_signed;
|
reg r_mpy_signed;
|
reg [1:0] mpypipe;
|
reg [2:0] mpypipe;
|
|
|
// First clock, latch in the inputs
|
// First clock, latch in the inputs
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
begin
|
begin
|
// mpypipe indicates we have a multiply in the
|
// mpypipe indicates we have a multiply in the
|
// pipeline. In this case, the multiply
|
// pipeline. In this case, the multiply
|
// pipeline is a two stage pipeline, so we need
|
// pipeline is a two stage pipeline, so we need
|
// two bits in the pipe.
|
// two bits in the pipe.
|
mpypipe[0] <= (i_ce)&&((i_op[3:1]==3'h5)
|
if (i_rst)
|
||(i_op[3:0]==4'h8));
|
mpypipe <= 3'h0;
|
|
else begin
|
|
mpypipe[0] <= this_is_a_multiply_op;
|
mpypipe[1] <= mpypipe[0];
|
mpypipe[1] <= mpypipe[0];
|
|
mpypipe[2] <= mpypipe[1];
|
|
end
|
|
|
if (i_op[0]) // i.e. if signed multiply
|
if (i_op[0]) // i.e. if signed multiply
|
begin
|
begin
|
r_mpy_a_input <= {(~i_a[31]),i_a[30:0]};
|
r_mpy_a_input <= {(~i_a[31]),i_a[30:0]};
|
r_mpy_b_input <= {(~i_b[31]),i_b[30:0]};
|
r_mpy_b_input <= {(~i_b[31]),i_b[30:0]};
|
| Line 207... |
Line 240... |
// case of 64 bit multiply. We'll keep track
|
// case of 64 bit multiply. We'll keep track
|
// of it, though, and pretend in all other
|
// of it, though, and pretend in all other
|
// cases.
|
// cases.
|
r_mpy_signed <= i_op[0];
|
r_mpy_signed <= i_op[0];
|
|
|
if (i_ce)
|
if (this_is_a_multiply_op)
|
mpyhi = i_op[1];
|
mpyhi = i_op[1];
|
end
|
end
|
|
|
assign mpybusy = |mpypipe;
|
assign mpybusy = |mpypipe[1:0];
|
|
assign mpydone = mpypipe[2];
|
|
|
// Second clock, do the multiplies, get the "partial
|
// Second clock, do the multiplies, get the "partial
|
// products". Here, we break our input up into two
|
// products". Here, we break our input up into two
|
// halves,
|
// halves,
|
//
|
//
|
| Line 274... |
Line 308... |
{ 32'h00, pp_l[31:16] }
|
{ 32'h00, pp_l[31:16] }
|
+ { 15'h00, pp_oi }
|
+ { 15'h00, pp_oi }
|
+ { pp_s, 15'h00 }
|
+ { pp_s, 15'h00 }
|
+ { pp_f, 16'h00 };
|
+ { pp_f, 16'h00 };
|
end
|
end
|
end // Fourth clock -- results are available for writeback.
|
|
`else
|
assign mpy_result = r_mpy_result;
|
wire signed [16:0] w_mpy_a_input, w_mpy_b_input;
|
// Fourth clock -- results are clocked into writeback
|
wire [33:0] w_mpy_result;
|
end
|
reg [31:0] r_mpy_result;
|
endgenerate // All possible multiply results have been determined
|
assign w_mpy_a_input ={ ((i_a[15])&(i_op[0])), i_a[15:0] };
|
|
assign w_mpy_b_input ={ ((i_b[15])&(i_op[0])), i_b[15:0] };
|
|
assign w_mpy_result = w_mpy_a_input * w_mpy_b_input;
|
|
always @(posedge i_clk)
|
|
if (i_ce)
|
|
r_mpy_result = w_mpy_result[31:0];
|
|
`endif
|
|
|
|
//
|
//
|
// The master ALU case statement
|
// The master ALU case statement
|
//
|
//
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
| Line 304... |
Line 331... |
4'b0011: o_c <= i_a | i_b; // Or
|
4'b0011: o_c <= i_a | i_b; // Or
|
4'b0100: o_c <= i_a ^ i_b; // Xor
|
4'b0100: o_c <= i_a ^ i_b; // Xor
|
4'b0101:{o_c,c } <= w_lsr_result[32:0]; // LSR
|
4'b0101:{o_c,c } <= w_lsr_result[32:0]; // LSR
|
4'b0110:{c,o_c } <= w_lsl_result[32:0]; // LSL
|
4'b0110:{c,o_c } <= w_lsl_result[32:0]; // LSL
|
4'b0111:{o_c,c } <= w_asr_result[32:0]; // ASR
|
4'b0111:{o_c,c } <= w_asr_result[32:0]; // ASR
|
`ifdef LONG_MPY
|
4'b1000: o_c <= mpy_result[31:0]; // MPY
|
4'b1000: o_c <= r_mpy_result[31:0]; // MPY
|
|
`else
|
|
4'b1000: o_c <= { i_b[15: 0], i_a[15:0] }; // LODIHI
|
|
`endif
|
|
4'b1001: o_c <= { i_a[31:16], i_b[15:0] }; // LODILO
|
4'b1001: o_c <= { i_a[31:16], i_b[15:0] }; // LODILO
|
`ifdef LONG_MPY
|
4'b1010: o_c <= mpy_result[63:32]; // MPYHU
|
4'b1010: o_c <= r_mpy_result[63:32]; // MPYHU
|
4'b1011: o_c <= mpy_result[63:32]; // MPYHS
|
4'b1011: o_c <= r_mpy_result[63:32]; // MPYHS
|
|
`else
|
|
4'b1010: o_c <= r_mpy_result; // MPYU
|
|
4'b1011: o_c <= r_mpy_result; // MPYS
|
|
`endif
|
|
4'b1100: o_c <= w_brev_result; // BREV
|
4'b1100: o_c <= w_brev_result; // BREV
|
4'b1101: o_c <= w_popc_result; // POPC
|
4'b1101: o_c <= w_popc_result; // POPC
|
4'b1110: o_c <= w_rol_result; // ROL
|
4'b1110: o_c <= w_rol_result; // ROL
|
default: o_c <= i_b; // MOV, LDI
|
default: o_c <= i_b; // MOV, LDI
|
endcase
|
endcase
|
end else if (r_busy)
|
end else // if (mpydone)
|
`ifdef LONG_MPY
|
o_c <= (mpyhi)?mpy_result[63:32]:mpy_result[31:0];
|
o_c <= (mpyhi)?r_mpy_result[63:32]:r_mpy_result[31:0];
|
|
`else
|
|
o_c <= r_mpy_result;
|
|
`endif
|
|
|
|
reg r_busy;
|
reg r_busy;
|
initial r_busy = 1'b0;
|
initial r_busy = 1'b0;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
r_busy <= (~i_rst)&&(i_ce)&&(i_valid)
|
if (i_rst)
|
`ifdef LONG_MPY
|
r_busy <= 1'b0;
|
&&((i_op[3:1] == 3'h5)
|
else
|
||(i_op[3:0] == 4'h8))||mpybusy;
|
r_busy <= ((IMPLEMENT_MPY > 1)
|
`else
|
&&(this_is_a_multiply_op))||mpybusy;
|
&&(i_op[3:1] == 3'h5);
|
assign o_busy = (r_busy); // ||((IMPLEMENT_MPY>1)&&(this_is_a_multiply_op));
|
`endif
|
|
|
|
assign o_busy = r_busy;
|
|
|
|
assign o_illegal = 1'b0;
|
|
end endgenerate
|
|
|
|
assign z = (o_c == 32'h0000);
|
assign z = (o_c == 32'h0000);
|
assign n = (o_c[31]);
|
assign n = (o_c[31]);
|
assign v = (set_ovfl)&&(pre_sign != o_c[31]);
|
assign v = (set_ovfl)&&(pre_sign != o_c[31]);
|
|
|
| Line 355... |
Line 364... |
|
|
initial o_valid = 1'b0;
|
initial o_valid = 1'b0;
|
always @(posedge i_clk)
|
always @(posedge i_clk)
|
if (i_rst)
|
if (i_rst)
|
o_valid <= 1'b0;
|
o_valid <= 1'b0;
|
|
else if (IMPLEMENT_MPY <= 1)
|
|
o_valid <= (i_ce);
|
else
|
else
|
o_valid <= (i_ce)&&(i_valid)
|
o_valid <=((i_ce)&&(!this_is_a_multiply_op))||(mpydone);
|
`ifdef LONG_MPY
|
|
&&(i_op[3:1] != 3'h5)&&(i_op[3:0] != 4'h8)
|
|
||(o_busy)&&(~mpybusy);
|
|
`else
|
|
&&(i_op[3:1] != 3'h5)||(o_busy);
|
|
`endif
|
|
endmodule
|
endmodule
|
|
|
No newline at end of file
|
No newline at end of file
|
