URL https://opencores.org/ocsvn/forwardcom/forwardcom/trunk

Subversion Repositories forwardcom

[/] [forwardcom/] [trunk/] [mul_div.sv] - Blame information for rev 40

Go to most recent revision | Details | Compare with Previous | View Log


//////////////////////////////////////////////////////////////////////////////////
// Engineer: Agner Fog
//
// Create Date:    2021-06-06
// Last modified:  2021-06-06
// Module Name:    mul_div
// Project Name:   ForwardCom soft core
// Target Devices: Artix 7
// Tool Versions:  Vivado v. 2020.1
// License:        CERN-OHL-W v. 2 or later
// Description:    Arithmetic-logic unit for multiplication and division
// of general purpose registers.
//////////////////////////////////////////////////////////////////////////////////
 
`include "defines.vh"
 
module mul_div (
    input clock,                            // system clock (100 MHz)
    input clock_enable,                     // clock enable. Used when single-stepping
    input reset,                            // system reset
    input valid_in,                         // data from previous stage ready
    input stall_in,                         // pipeline is stalled
    input [31:0] instruction_in,            // current instruction, up to 3 words long. Only first word used here
    input [`TAG_WIDTH-1:0] tag_val_in,      // instruction tag value
    input [1:0]  category_in,               // 00: multiformat, 01: single format, 10: jump
    input        mask_alternative_in,       // mask register and fallback register used for alternative purposes
    input [1:0]  result_type_in,            // type of result: 0: register, 1: system register, 2: memory, 3: other or nothing
    input        vector_in,                 // vector registers used
    input [6:0]  opx_in,                    // operation ID in execution unit. This is mostly equal to op1 for multiformat instructions
    input [2:0]  ot_in,                     // operand type
    input [5:0]  option_bits_in,            // option bits from IM3 or mask
 
    // monitor result buses:
    input write_en1,                        // a result is written to writeport1
    input [`TAG_WIDTH-1:0] write_tag1_in,   // tag of result inwriteport1
    input [`RB1:0] writeport1_in,           // result bus 1
    input write_en2,                        // a result is written to writeport2
    input [`TAG_WIDTH-1:0] write_tag2_in,   // tag of result inwriteport2
    input [`RB1:0] writeport2_in,           // result bus 2
    input [`TAG_WIDTH-1:0] predict_tag1_in, // result tag value on writeport1 in next clock cycle
    input [`TAG_WIDTH-1:0] predict_tag2_in, // result tag value on writeport2 in next clock cycle
 
    // Register values sampled from result bus in previous stages
    input [`RB:0] operand1_in,              // first register operand or fallback
    input [`RB:0] operand2_in,              // second register operand RS
    input [`RB:0] operand3_in,              // last register operand RT
    input [`MASKSZ:0] regmask_val_in,       // mask register
    input [`RB1:0] ram_data_in,             // memory operand from data ram
    input        opr2_from_ram_in,          // value of operand 2 comes from data ram
    input        opr3_from_ram_in,          // value of last operand comes from data ram
    input        opr1_used_in,              // operand1_in is needed
    input        opr2_used_in,              // operand2_in is needed
    input        opr3_used_in,              // operand3_in is needed
    input        regmask_used_in,           // regmask_val_in is needed
 
    output reg valid_out,                   // for debug display: alu is active
    output reg register_write_out,
    output reg [4:0] register_a_out,        // register to write
    output reg [`RB1:0] result_out,         //
    output reg [`TAG_WIDTH-1:0] tag_val_out,// instruction tag value
    output reg stall_out,                   // alu is waiting for an operand or not ready to receive a new instruction
    output reg stall_next_out,              // alu will be waiting in next clock cycle
    output reg error_out,                   // unknown instruction
    output reg error_parm_out,              // wrong parameter for instruction
 
    // outputs for debugger:
    output reg [31:0] debug1_out,           // debug information
    output reg [31:0] debug2_out            // temporary debug information
);
 
logic [`RB1:0] operand1;                    // first register operand RD or RU. bit `RB is 1 if invalid
logic [`RB1:0] operand2;                    // second register operand RS. bit `RB is 1 if invalid
logic [`RB1:0] operand3;                    // last register operand RT. bit `RB is 1 if invalid
logic [`MASKSZ:0] regmask_val;              // mask register
logic [1:0]  otout;                         // operand type for output
logic [5:0]  msb;                           // index to most significant bit
logic signbit2, signbit3;                   // sign bits of three operands
logic [`RB1:0] sbit;                        // position of sign bit
logic [`RB1:0] result;                      // result for output
logic [1:0]  result_type;                   // type of result
logic [6:0]  opx;                           // operation ID in execution unit. This is mostly equal to op1 for multiformat instructions
logic mask_off;                             // result is masked off
logic stall;                                // waiting for operands
logic stall_next;                           // will be waiting for operands in next clock cycle
logic error;                                // unknown instruction
logic error_parm;                           // wrong parameter for instruction
 
// It seems to be more efficient to truncate operands locally by ANDing with sizemask than to
// make separate wires for the truncated operands, because wiring is more expensive than logic:
logic [`RB1:0] sizemask;                    // mask for operand type
 
logic [31:0] temp_debug;                    // temporary debug signals
 
always_comb begin
    // get all inputs
    stall = 0;
    stall_next = 0;
    regmask_val = 0;
    temp_debug = 0;                         // temporary debug signals
 
    if (regmask_val_in[`MASKSZ]) begin      // value missing
        if (write_en1 && regmask_val_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
            regmask_val = writeport1_in;    // obtained from result bus 1 (which may be my own output)
        end else if (write_en2 && regmask_val_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
            regmask_val = writeport2_in[(`MASKSZ-1):0]; // obtained from result bus 2
        end else begin
            if (regmask_used_in) begin
                stall = 1;                  // operand not ready
                temp_debug[0] = 1;          // debug info about cause of stall
                if (regmask_val_in[`TAG_WIDTH-1:0] != predict_tag1_in && regmask_val_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
                    stall_next = 1;         // operand not ready in next clock cycle
                end
            end
        end
    end else begin                          // value available
        regmask_val = regmask_val_in;
    end
 
    mask_off = regmask_used_in && regmask_val[`MASKSZ] == 0 && regmask_val[0] == 0 && !mask_alternative_in;
 
    operand1 = 0;
    if (operand1_in[`RB]) begin             // value missing
        if (write_en1 && operand1_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
            operand1 = writeport1_in;       // obtained from result bus 1 (which may be my own output)
        end else if (write_en2 && operand1_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
            operand1 = writeport2_in;       // obtained from result bus 2
        end else begin
            if (opr1_used_in) begin
                stall = 1;                  // operand not ready
                temp_debug[1] = 1;          // debug info about cause of stall
                if (operand1_in[`TAG_WIDTH-1:0] != predict_tag1_in && operand1_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
                    stall_next = 1;         // operand not ready in next clock cycle
                end
            end
        end
    end else begin
        operand1 = operand1_in[`RB1:0];
    end
 
    operand2 = 0;
    if (opr2_from_ram_in) begin
        operand2 = ram_data_in;
    end else if (operand2_in[`RB]) begin    // value missing
        if (write_en1 && operand2_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
            operand2 = writeport1_in;       // obtained from result bus 1 (which may be my own output)
        end else if (write_en2 && operand2_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
            operand2 =  writeport2_in;      // obtained from result bus 2
        end else begin
            if (opr2_used_in && !mask_off) begin
                stall = 1;                  // operand not ready
                temp_debug[2] = 1;          // debug info about cause of stall
                if (operand2_in[`TAG_WIDTH-1:0] != predict_tag1_in && operand2_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
                    stall_next = 1;         // operand not ready in next clock cycle
                end
            end
        end
    end else begin                          // value available
        operand2 = operand2_in[`RB1:0];
    end
 
    operand3 = 0;
    if (opr3_from_ram_in) begin
        operand3 = ram_data_in;
    end else if (operand3_in[`RB]) begin    // value missing
        if (write_en1 && operand3_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
            operand3 = writeport1_in;       // obtained from result bus 1 (which may be my own output)
        end else if (write_en2 && operand3_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
            operand3 = writeport2_in;       // obtained from result bus 2
        end else begin
            if (opr3_used_in && !mask_off) begin
                stall = 1;                  // operand not ready
                temp_debug[3] = 1;          // debug info about cause of stall
                if (operand3_in[`TAG_WIDTH-1:0] != predict_tag1_in && operand3_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
                    stall_next = 1;         // operand not ready in next clock cycle
                end
            end
        end
    end else begin                          // value available
        operand3 = operand3_in[`RB1:0];
    end
 
    opx = opx_in;                           // operation ID in execution unit. This is mostly equal to op1 for multiformat instructions
    result = 0;
    otout = ot_in[1:0];                     // operand type for output
    result_type = result_type_in;
    error = 0;
    error_parm = 0;
 
    case (ot_in[1:0])
    0: begin
        msb = 7;   // 8 bit
        sbit = 8'H80;
        sizemask = 8'HFF;
        //signbit1 = operand1[7];
        signbit2 = operand2[7];
        signbit3 = operand3[7];
        end
    1: begin
        msb = 15;   // 16 bit
        sbit     = 16'H8000;
        sizemask = 16'HFFFF;
        //signbit1 = operand1[15];
        signbit2 = operand2[15];
        signbit3 = operand3[15];
        end
    2: begin
        msb = 31;   // 32 bit
        sbit     = 32'H80000000;
        sizemask = 32'HFFFFFFFF;
        //signbit1 = operand1[31];
        signbit2 = operand2[31];
        signbit3 = operand3[31];
        end
    3: begin
        msb = `RB1;   // 64 bit
        sbit     = {1'b1,{(`RB-1){1'b0}}};
        sizemask = ~(`RB'b0);
        //signbit1 = operand1[`RB1];
        signbit2 = operand2[`RB1];
        signbit3 = operand3[`RB1];
        end
    endcase
 
 
    ////////////////////////////////////////////////
    //             Select ALU operation
    ////////////////////////////////////////////////
 
    result = 0;
 
    if (opx == `II_MUL) begin
 
       error = 1;  // instruction not supported yet
 
    end else if (opx == `II_MUL_HI || opx == `II_MUL_HI_U) begin
 
       error = 1;  // instruction not supported yet
 
    end else if (opx == `II_DIV || opx == `II_DIV_U) begin
 
       error = 1;  // instruction not supported yet
 
    end else if (opx == `II_REM || opx == `II_REM_U) begin
 
       error = 1;  // instruction not supported yet
 
    end else begin
 
        error = 1;  // unknown instruction
 
    end
 
    if (vector_in) error = 1;  // Vector instructions not supported yet
 
end
 
 
// output
always_ff @(posedge clock) if (clock_enable) begin
    if (!valid_in) begin
        register_write_out <= 0;
        // note: the FPGA has no internal tri-state buffers. We need to simulate result bus by or'ing outputs
        result_out <= 0;
        register_a_out <= 0;
        tag_val_out <= 0;
 
    // stall_in must disable the output to avoid executing the same instruction twice)
    end else if (stall || stall_in) begin
        register_write_out <= 0;
        result_out <= 0;
        register_a_out <= 0;
        tag_val_out <= 0;
 
    end else if (result_type != `RESULT_REG) begin
        // no output?
        register_write_out <= 0;
        result_out <= 0;
        register_a_out <= 0;
        tag_val_out <= 0;
 
    end else if (regmask_used_in && !regmask_val[0] & !vector_in) begin
        // mask is zero. output is fallback
        case (otout)
        0: result_out <= operand1[7:0];
        1: result_out <= operand1[15:0];
        2: result_out <= operand1[31:0];
        3: result_out <= operand1[`RB1:0];
        endcase
        register_write_out <= ~reset;
        register_a_out <= {1'b0,instruction_in[`RD]};
        tag_val_out <= tag_val_in;
 
    end else begin
        // normal register output
        case (otout)
        0: result_out <= result[7:0];
        1: result_out <= result[15:0];
        2: result_out <= result[31:0];
        3: result_out <= result[`RB1:0];
        endcase
        register_write_out <= ~reset;
        register_a_out <= {1'b0,instruction_in[`RD]};
        tag_val_out <= tag_val_in;
    end
 
    valid_out <= !stall & valid_in & !reset;
    stall_out <= stall  & valid_in & !reset;
    stall_next_out <= stall_next & valid_in & !reset;
    error_out <= error & valid_in & !reset;            // unknown instruction
    error_parm_out <= error_parm & valid_in & !reset;  // wrong parameter
 
    // outputs for debugger:
    debug1_out <= 0;
 
    debug1_out[6:0]   <= opx;
 
    debug1_out[21:20] <= category_in;
 
    debug1_out[24]    <= stall;
    debug1_out[25]    <= stall_next;
    debug1_out[27]    <= error;
 
 
    debug2_out <= temp_debug;
 
    debug2_out[16] <= opr1_used_in;
    debug2_out[17] <= opr2_used_in;
    debug2_out[18] <= opr3_used_in;
    debug2_out[19] <= regmask_used_in;
 
    debug2_out[20] <= mask_alternative_in;
    debug2_out[21] <= mask_off;
    debug2_out[22] <= regmask_val_in[0];
    debug2_out[23] <= regmask_val_in[`MASKSZ];
 
    debug2_out[27:24] <= regmask_val[3:0];
    debug2_out[28] <= regmask_val[`MASKSZ];
 
end
 
endmodule

Browse

Tools

Subversion Repositories forwardcom

[/] [forwardcom/] [trunk/] [mul_div.sv] - Blame information for rev 40

Line No.	Rev	Author	Line
1	6	Agner	`//////////////////////////////////////////////////////////////////////////////////`
2			`// Engineer: Agner Fog`
3			`//`
4			`// Create Date: 2021-06-06`
5			`// Last modified: 2021-06-06`
6			`// Module Name: mul_div`
7			`// Project Name: ForwardCom soft core`
8			`// Target Devices: Artix 7`
9			`// Tool Versions: Vivado v. 2020.1`
10			`// License: CERN-OHL-W v. 2 or later`
11			`// Description: Arithmetic-logic unit for multiplication and division`
12			`// of general purpose registers.`
13			`//////////////////////////////////////////////////////////////////////////////////`
14
15			`include "defines.vh"
16
17			`module mul_div (`
18			`input clock, // system clock (100 MHz)`
19			`input clock_enable, // clock enable. Used when single-stepping`
20			`input reset, // system reset`
21			`input valid_in, // data from previous stage ready`
22			`input stall_in, // pipeline is stalled`
23			`input [31:0] instruction_in, // current instruction, up to 3 words long. Only first word used here`
24			input [`TAG_WIDTH-1:0] tag_val_in, // instruction tag value
25			`input [1:0] category_in, // 00: multiformat, 01: single format, 10: jump`
26			`input mask_alternative_in, // mask register and fallback register used for alternative purposes`
27			`input [1:0] result_type_in, // type of result: 0: register, 1: system register, 2: memory, 3: other or nothing`
28			`input vector_in, // vector registers used`
29			`input [6:0] opx_in, // operation ID in execution unit. This is mostly equal to op1 for multiformat instructions`
30			`input [2:0] ot_in, // operand type`
31			`input [5:0] option_bits_in, // option bits from IM3 or mask`
32
33			`// monitor result buses:`
34			`input write_en1, // a result is written to writeport1`
35			input [`TAG_WIDTH-1:0] write_tag1_in, // tag of result inwriteport1
36			input [`RB1:0] writeport1_in, // result bus 1
37			`input write_en2, // a result is written to writeport2`
38			input [`TAG_WIDTH-1:0] write_tag2_in, // tag of result inwriteport2
39			input [`RB1:0] writeport2_in, // result bus 2
40			input [`TAG_WIDTH-1:0] predict_tag1_in, // result tag value on writeport1 in next clock cycle
41			input [`TAG_WIDTH-1:0] predict_tag2_in, // result tag value on writeport2 in next clock cycle
42
43			`// Register values sampled from result bus in previous stages`
44			input [`RB:0] operand1_in, // first register operand or fallback
45			input [`RB:0] operand2_in, // second register operand RS
46			input [`RB:0] operand3_in, // last register operand RT
47			input [`MASKSZ:0] regmask_val_in, // mask register
48			input [`RB1:0] ram_data_in, // memory operand from data ram
49			`input opr2_from_ram_in, // value of operand 2 comes from data ram`
50			`input opr3_from_ram_in, // value of last operand comes from data ram`
51			`input opr1_used_in, // operand1_in is needed`
52			`input opr2_used_in, // operand2_in is needed`
53			`input opr3_used_in, // operand3_in is needed`
54			`input regmask_used_in, // regmask_val_in is needed`
55
56			`output reg valid_out, // for debug display: alu is active`
57			`output reg register_write_out,`
58			`output reg [4:0] register_a_out, // register to write`
59			output reg [`RB1:0] result_out, //
60			output reg [`TAG_WIDTH-1:0] tag_val_out,// instruction tag value
61			`output reg stall_out, // alu is waiting for an operand or not ready to receive a new instruction`
62			`output reg stall_next_out, // alu will be waiting in next clock cycle`
63			`output reg error_out, // unknown instruction`
64			`output reg error_parm_out, // wrong parameter for instruction`
65
66			`// outputs for debugger:`
67			`output reg [31:0] debug1_out, // debug information`
68			`output reg [31:0] debug2_out // temporary debug information`
69			`);`
70
71			logic [`RB1:0] operand1; // first register operand RD or RU. bit `RB is 1 if invalid
72			logic [`RB1:0] operand2; // second register operand RS. bit `RB is 1 if invalid
73			logic [`RB1:0] operand3; // last register operand RT. bit `RB is 1 if invalid
74			logic [`MASKSZ:0] regmask_val; // mask register
75			`logic [1:0] otout; // operand type for output`
76			`logic [5:0] msb; // index to most significant bit`
77			`logic signbit2, signbit3; // sign bits of three operands`
78			logic [`RB1:0] sbit; // position of sign bit
79			logic [`RB1:0] result; // result for output
80			`logic [1:0] result_type; // type of result`
81			`logic [6:0] opx; // operation ID in execution unit. This is mostly equal to op1 for multiformat instructions`
82			`logic mask_off; // result is masked off`
83			`logic stall; // waiting for operands`
84			`logic stall_next; // will be waiting for operands in next clock cycle`
85			`logic error; // unknown instruction`
86			`logic error_parm; // wrong parameter for instruction`
87
88			`// It seems to be more efficient to truncate operands locally by ANDing with sizemask than to`
89			`// make separate wires for the truncated operands, because wiring is more expensive than logic:`
90			logic [`RB1:0] sizemask; // mask for operand type
91
92			`logic [31:0] temp_debug; // temporary debug signals`
93
94			`always_comb begin`
95			`// get all inputs`
96			`stall = 0;`
97			`stall_next = 0;`
98			`regmask_val = 0;`
99			`temp_debug = 0; // temporary debug signals`
100
101			if (regmask_val_in[`MASKSZ]) begin // value missing
102			if (write_en1 && regmask_val_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
103			`regmask_val = writeport1_in; // obtained from result bus 1 (which may be my own output)`
104			end else if (write_en2 && regmask_val_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
105			regmask_val = writeport2_in[(`MASKSZ-1):0]; // obtained from result bus 2
106			`end else begin`
107			`if (regmask_used_in) begin`
108			`stall = 1; // operand not ready`
109			`temp_debug[0] = 1; // debug info about cause of stall`
110			if (regmask_val_in[`TAG_WIDTH-1:0] != predict_tag1_in && regmask_val_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
111			`stall_next = 1; // operand not ready in next clock cycle`
112			`end`
113			`end`
114			`end`
115			`end else begin // value available`
116			`regmask_val = regmask_val_in;`
117			`end`
118
119			mask_off = regmask_used_in && regmask_val[`MASKSZ] == 0 && regmask_val[0] == 0 && !mask_alternative_in;
120
121			`operand1 = 0;`
122			if (operand1_in[`RB]) begin // value missing
123			if (write_en1 && operand1_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
124			`operand1 = writeport1_in; // obtained from result bus 1 (which may be my own output)`
125			end else if (write_en2 && operand1_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
126			`operand1 = writeport2_in; // obtained from result bus 2`
127			`end else begin`
128			`if (opr1_used_in) begin`
129			`stall = 1; // operand not ready`
130			`temp_debug[1] = 1; // debug info about cause of stall`
131			if (operand1_in[`TAG_WIDTH-1:0] != predict_tag1_in && operand1_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
132			`stall_next = 1; // operand not ready in next clock cycle`
133			`end`
134			`end`
135			`end`
136			`end else begin`
137			operand1 = operand1_in[`RB1:0];
138			`end`
139
140			`operand2 = 0;`
141			`if (opr2_from_ram_in) begin`
142			`operand2 = ram_data_in;`
143			end else if (operand2_in[`RB]) begin // value missing
144			if (write_en1 && operand2_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
145			`operand2 = writeport1_in; // obtained from result bus 1 (which may be my own output)`
146			end else if (write_en2 && operand2_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
147			`operand2 = writeport2_in; // obtained from result bus 2`
148			`end else begin`
149			`if (opr2_used_in && !mask_off) begin`
150			`stall = 1; // operand not ready`
151			`temp_debug[2] = 1; // debug info about cause of stall`
152			if (operand2_in[`TAG_WIDTH-1:0] != predict_tag1_in && operand2_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
153			`stall_next = 1; // operand not ready in next clock cycle`
154			`end`
155			`end`
156			`end`
157			`end else begin // value available`
158			operand2 = operand2_in[`RB1:0];
159			`end`
160
161			`operand3 = 0;`
162			`if (opr3_from_ram_in) begin`
163			`operand3 = ram_data_in;`
164			end else if (operand3_in[`RB]) begin // value missing
165			if (write_en1 && operand3_in[`TAG_WIDTH-1:0] == write_tag1_in) begin
166			`operand3 = writeport1_in; // obtained from result bus 1 (which may be my own output)`
167			end else if (write_en2 && operand3_in[`TAG_WIDTH-1:0] == write_tag2_in) begin
168			`operand3 = writeport2_in; // obtained from result bus 2`
169			`end else begin`
170			`if (opr3_used_in && !mask_off) begin`
171			`stall = 1; // operand not ready`
172			`temp_debug[3] = 1; // debug info about cause of stall`
173			if (operand3_in[`TAG_WIDTH-1:0] != predict_tag1_in && operand3_in[`TAG_WIDTH-1:0] != predict_tag2_in) begin
174			`stall_next = 1; // operand not ready in next clock cycle`
175			`end`
176			`end`
177			`end`
178			`end else begin // value available`
179			operand3 = operand3_in[`RB1:0];
180			`end`
181
182			`opx = opx_in; // operation ID in execution unit. This is mostly equal to op1 for multiformat instructions`
183			`result = 0;`
184			`otout = ot_in[1:0]; // operand type for output`
185			`result_type = result_type_in;`
186			`error = 0;`
187			`error_parm = 0;`
188
189			`case (ot_in[1:0])`
190			`0: begin`
191			`msb = 7; // 8 bit`
192			`sbit = 8'H80;`
193			`sizemask = 8'HFF;`
194			`//signbit1 = operand1[7];`
195			`signbit2 = operand2[7];`
196			`signbit3 = operand3[7];`
197			`end`
198			`1: begin`
199			`msb = 15; // 16 bit`
200			`sbit = 16'H8000;`
201			`sizemask = 16'HFFFF;`
202			`//signbit1 = operand1[15];`
203			`signbit2 = operand2[15];`
204			`signbit3 = operand3[15];`
205			`end`
206			`2: begin`
207			`msb = 31; // 32 bit`
208			`sbit = 32'H80000000;`
209			`sizemask = 32'HFFFFFFFF;`
210			`//signbit1 = operand1[31];`
211			`signbit2 = operand2[31];`
212			`signbit3 = operand3[31];`
213			`end`
214			`3: begin`
215			msb = `RB1; // 64 bit
216			sbit = {1'b1,{(`RB-1){1'b0}}};
217			sizemask = ~(`RB'b0);
218			//signbit1 = operand1[`RB1];
219			signbit2 = operand2[`RB1];
220			signbit3 = operand3[`RB1];
221			`end`
222			`endcase`
223
224
225			`////////////////////////////////////////////////`
226			`// Select ALU operation`
227			`////////////////////////////////////////////////`
228
229			`result = 0;`
230
231			if (opx == `II_MUL) begin
232
233			`error = 1; // instruction not supported yet`
234
235			end else if (opx == `II_MUL_HI \|\| opx == `II_MUL_HI_U) begin
236
237			`error = 1; // instruction not supported yet`
238
239			end else if (opx == `II_DIV \|\| opx == `II_DIV_U) begin
240
241			`error = 1; // instruction not supported yet`
242
243			end else if (opx == `II_REM \|\| opx == `II_REM_U) begin
244
245			`error = 1; // instruction not supported yet`
246
247			`end else begin`
248
249			`error = 1; // unknown instruction`
250
251			`end`
252
253			`if (vector_in) error = 1; // Vector instructions not supported yet`
254
255			`end`
256
257
258			`// output`
259			`always_ff @(posedge clock) if (clock_enable) begin`
260			`if (!valid_in) begin`
261			`register_write_out <= 0;`
262			`// note: the FPGA has no internal tri-state buffers. We need to simulate result bus by or'ing outputs`
263			`result_out <= 0;`
264			`register_a_out <= 0;`
265			`tag_val_out <= 0;`
266
267			`// stall_in must disable the output to avoid executing the same instruction twice)`
268			`end else if (stall \|\| stall_in) begin`
269			`register_write_out <= 0;`
270			`result_out <= 0;`
271			`register_a_out <= 0;`
272			`tag_val_out <= 0;`
273
274			end else if (result_type != `RESULT_REG) begin
275			`// no output?`
276			`register_write_out <= 0;`
277			`result_out <= 0;`
278			`register_a_out <= 0;`
279			`tag_val_out <= 0;`
280
281			`end else if (regmask_used_in && !regmask_val[0] & !vector_in) begin`
282			`// mask is zero. output is fallback`
283			`case (otout)`
284			`0: result_out <= operand1[7:0];`
285			`1: result_out <= operand1[15:0];`
286			`2: result_out <= operand1[31:0];`
287			3: result_out <= operand1[`RB1:0];
288			`endcase`
289			`register_write_out <= ~reset;`
290			register_a_out <= {1'b0,instruction_in[`RD]};
291			`tag_val_out <= tag_val_in;`
292
293			`end else begin`
294			`// normal register output`
295			`case (otout)`
296			`0: result_out <= result[7:0];`
297			`1: result_out <= result[15:0];`
298			`2: result_out <= result[31:0];`
299			3: result_out <= result[`RB1:0];
300			`endcase`
301			`register_write_out <= ~reset;`
302			register_a_out <= {1'b0,instruction_in[`RD]};
303			`tag_val_out <= tag_val_in;`
304			`end`
305
306			`valid_out <= !stall & valid_in & !reset;`
307			`stall_out <= stall & valid_in & !reset;`
308			`stall_next_out <= stall_next & valid_in & !reset;`
309			`error_out <= error & valid_in & !reset; // unknown instruction`
310			`error_parm_out <= error_parm & valid_in & !reset; // wrong parameter`
311
312			`// outputs for debugger:`
313			`debug1_out <= 0;`
314
315			`debug1_out[6:0] <= opx;`
316
317			`debug1_out[21:20] <= category_in;`
318
319			`debug1_out[24] <= stall;`
320			`debug1_out[25] <= stall_next;`
321			`debug1_out[27] <= error;`
322
323
324			`debug2_out <= temp_debug;`
325
326			`debug2_out[16] <= opr1_used_in;`
327			`debug2_out[17] <= opr2_used_in;`
328			`debug2_out[18] <= opr3_used_in;`
329			`debug2_out[19] <= regmask_used_in;`
330
331			`debug2_out[20] <= mask_alternative_in;`
332			`debug2_out[21] <= mask_off;`
333			`debug2_out[22] <= regmask_val_in[0];`
334			debug2_out[23] <= regmask_val_in[`MASKSZ];
335
336			`debug2_out[27:24] <= regmask_val[3:0];`
337			debug2_out[28] <= regmask_val[`MASKSZ];
338
339			`end`
340
341			`endmodule`