Subversion Repositories ft816float

`timescale 1ns / 1ps

// ============================================================================

//        __

//   \\__/ o\    (C) 2006-2019  Robert Finch, Waterloo

//    \  __ /    All rights reserved.

//     \/_//     robfinch<remove>@finitron.ca

//       ||

//

//      fpAddsub_L10.v

//    - floating point adder/subtracter

//    - ten cycle latency

//    - can issue every clock cycle

//    - parameterized width

//    - IEEE 754 representation

//

//

// This source file is free software: you can redistribute it and/or modify 

// it under the terms of the GNU Lesser General Public License as published 

// by the Free Software Foundation, either version 3 of the License, or     

// (at your option) any later version.                                      

//                                                                          

// This source file is distributed in the hope that it will be useful,      

// but WITHOUT ANY WARRANTY; without even the implied warranty of           

// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the            

// GNU General Public License for more details.                             

//                                                                          

// You should have received a copy of the GNU General Public License        

// along with this program.  If not, see <http://www.gnu.org/licenses/>.    

//                                                                          

// ============================================================================

module fpAddsub_L10(clk, ce, rm, op, a, b, o);

parameter WID = 128;

localparam MSB = WID-1;

localparam EMSB = WID==128 ? 14 :

                  WID==96 ? 14 :

                  WID==80 ? 14 :

                  WID==64 ? 10 :

                                  WID==52 ? 10 :

                                  WID==48 ? 11 :

                                  WID==44 ? 10 :

                                  WID==42 ? 10 :

                                  WID==40 ?  9 :

                                  WID==32 ?  7 :

                                  WID==24 ?  6 : 4;

localparam FMSB = WID==128 ? 111 :

                  WID==96 ? 79 :

                  WID==80 ? 63 :

                  WID==64 ? 51 :

                                  WID==52 ? 39 :

                                  WID==48 ? 34 :

                                  WID==44 ? 31 :

                                  WID==42 ? 29 :

                                  WID==40 ? 28 :

                                  WID==32 ? 22 :

                                  WID==24 ? 15 : 9;

localparam FX = (FMSB+2)*2-1;   // the MSB of the expanded fraction

localparam EX = FX + 1 + EMSB + 1 + 1 - 1;

input clk;              // system clock

input ce;               // core clock enable

input [2:0] rm;  // rounding mode

input op;               // operation 0 = add, 1 = subtract

input [WID-1:0] a;       // operand a

input [WID-1:0] b;       // operand b

output [EX:0] o; // output

wire so;                        // sign output

wire [EMSB:0] xo;        // de normalized exponent output

reg [FX:0] mo;   // mantissa output

assign o = {so,xo,mo};

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #1

// - Decompose inputs into more digestible values.

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

wire [WID-1:0] a1;

wire [WID-1:0] b1;

wire sa1, sb1;

wire [EMSB:0] xa1, xb1;

wire [FMSB:0] ma1, mb1;

wire [FMSB+1:0] fracta1, fractb1;

wire adn1, bdn1;                // a,b denormalized ?

wire xaInf1, xbInf1;

wire aInf1, bInf1;

wire aNan1, bNan1;

wire az1, bz1;  // operand a,b is zero

wire op1;

fpDecompReg #(WID) u1a (.clk(clk), .ce(ce), .i(a), .o(a1), .sgn(sa1), .exp(xa1), .man(ma1), .fract(fracta1), .xz(adn1), .vz(az1), .xinf(xaInf1), .inf(aInf1), .nan(aNan1) );

fpDecompReg #(WID) u1b (.clk(clk), .ce(ce), .i(b), .o(b1), .sgn(sb1), .exp(xb1), .man(mb1), .fract(fractb1), .xz(bdn1), .vz(bz1), .xinf(xbInf1), .inf(bInf1), .nan(bNan1) );

delay1 #(1)  dop1(.clk(clk), .ce(ce), .i(op), .o(op1) );

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #2

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

reg xabeq2;

reg mabeq2;

reg anbz2;

reg xabInf2;

reg anbInf2;

wire [EMSB:0] xa2, xb2;

wire [FMSB:0] ma2, mb2;

// operands sign,exponent,mantissa

wire [FMSB+1:0] fracta2, fractb2;

wire az2, bz2;  // operand a,b is zero

reg xa_gt_xb2;

reg var2;

reg [EMSB:0] xad2;

reg [EMSB:0] xbd2;

reg realOp2;

delay1 #(EMSB+1)  dxa2(.clk(clk), .ce(ce), .i(xa1), .o(xa2) );

delay1 #(EMSB+1)  dxb2(.clk(clk), .ce(ce), .i(xb1), .o(xb2) );

delay1 #(FMSB+1)  dma2(.clk(clk), .ce(ce), .i(ma1), .o(ma2) );

delay1 #(FMSB+1)  dmb2(.clk(clk), .ce(ce), .i(mb1), .o(mb2) );

delay1 #(1)  daz2(.clk(clk), .ce(ce), .i(az1), .o(az2) );

delay1 #(1)  dbz2(.clk(clk), .ce(ce), .i(bz1), .o(bz2) );

delay1 #(FMSB+2)  dfracta2(.clk(clk), .ce(ce), .i(fracta1), .o(fracta2) );

delay1 #(FMSB+2)  dfractb2(.clk(clk), .ce(ce), .i(fractb1), .o(fractb2) );

always @(posedge clk)

        if (ce) xa_gt_xb2 <= xa1 > xb1;

always @(posedge clk)

        if (ce) var2 <= (xa1==xb1 && ma1 > mb1);

always @(posedge clk)

        if (ce) xad2 <= xa1|adn1;       // operand a exponent, compensated for denormalized numbers

always @(posedge clk)

        if (ce) xbd2 <= xb1|bdn1;       // operand b exponent, compensated for denormalized numbers

always @(posedge clk)

        if (ce) xabeq2 <= xa1==xb1;

always @(posedge clk)

        if (ce) mabeq2 <= ma1==mb1;

always @(posedge clk)

        if (ce) anbz2 <= az1 & bz1;

always @(posedge clk)

        if (ce) xabInf2 <= xaInf1 & xbInf1;

always @(posedge clk)

        if (ce) anbInf2 <= aInf1 & bInf1;

// Figure out which operation is really needed an add or

// subtract ?

// If the signs are the same, use the orignal op,

// otherwise flip the operation

//  a +  b = add,+

//  a + -b = sub, so of larger

// -a +  b = sub, so of larger

// -a + -b = add,-

//  a -  b = sub, so of larger

//  a - -b = add,+

// -a -  b = add,-

// -a - -b = sub, so of larger

always @(posedge clk)

        if (ce) realOp2 <= op1 ^ sa1 ^ sb1;

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #3

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

wire [EMSB:0] xa3, xb3;

wire xa_gt_xb3;

reg x_gt_b3;

wire xabInf3;

wire sa3,sb3;

wire op3;

wire [2:0] rm3;

reg [EMSB:0] xdiff3;

// which has greater magnitude ? Used for sign calc

reg a_gt_b3;

reg resZero3;

reg [FMSB+1:0] mfs3;

delay1 #(EMSB+1)  dxa3(.clk(clk), .ce(ce), .i(xa2), .o(xa3));

delay1 #(EMSB+1)  dxb3(.clk(clk), .ce(ce), .i(xb2), .o(xb3));

delay1 #(1) dxabInf2(.clk(clk), .ce(ce), .i(xabInf2), .o(xabInf3));

delay1 #(1) dxagtxb2(.clk(clk), .ce(ce), .i(xa_gt_xb2), .o(xa_gt_xb3));

delay2 #(1) dsa2(.clk(clk), .ce(ce), .i(sa1), .o(sa3));

delay2 #(1) dsb2(.clk(clk), .ce(ce), .i(sb1), .o(sb3));

delay2 #(1) dop2(.clk(clk), .ce(ce), .i(op1), .o(op3));

delay3 #(3) drm2(.clk(clk), .ce(ce), .i(rm), .o(rm3));

always @(posedge clk)

        if (ce) a_gt_b3 <= xa_gt_xb2 || var2;

// Find out if the result will be zero.

always @(posedge clk)

        if (ce) resZero3 <= (realOp2 & xabeq2 & mabeq2) |       anbz2;  // subtract, same magnitude,    both a,b zero

// Compute the difference in exponents, provides shift amount

always @(posedge clk)

        if (ce) xdiff3 <= xa_gt_xb2 ? xad2 - xbd2 : xbd2 - xad2;

// determine which fraction to denormalize

always @(posedge clk)

        if (ce) mfs3 <= xa_gt_xb2 ? fractb2 : fracta2;

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #4

// Compute output exponent

//

// The output exponent is the larger of the two exponents, unless a subtract

// operation is in progress and the two numbers are equal, in which case the

// exponent should be zero.

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

reg [EMSB:0] xdif4;

wire [FMSB+1:0] mfs4;

reg [EMSB:0] xo4;        // de normalized exponent output

reg so4;

always @(posedge clk)

        if (ce) xo4 <= xabInf3 ? xa3 : resZero3 ? {EMSB+1{1'b0}} : xa_gt_xb3 ? xa3 : xb3;

// Compute output sign

always @(posedge clk)

if (ce)

        case ({resZero3,sa3,op3,sb3})   // synopsys full_case parallel_case

        4'b0000: so4 <= 0;                       // + + + = +

        4'b0001: so4 <= !a_gt_b3;       // + + - = sign of larger

        4'b0010: so4 <= !a_gt_b3;       // + - + = sign of larger

        4'b0011: so4 <= 0;                       // + - - = +

        4'b0100: so4 <= a_gt_b3;                // - + + = sign of larger

        4'b0101: so4 <= 1;                      // - + - = -

        4'b0110: so4 <= 1;                      // - - + = -

        4'b0111: so4 <= a_gt_b3;                // - - - = sign of larger

        4'b1000: so4 <= 0;                       //  A +  B, sign = +

        4'b1001: so4 <= rm3==3'd3;              //  A + -B, sign = + unless rounding down

        4'b1010: so4 <= rm3==3'd3;              //  A -  B, sign = + unless rounding down

        4'b1011: so4 <= 0;                       // +A - -B, sign = +

        4'b1100: so4 <= rm3==3'd3;              // -A +  B, sign = + unless rounding down

        4'b1101: so4 <= 1;                      // -A + -B, sign = -

        4'b1110: so4 <= 1;                      // -A - +B, sign = -

        4'b1111: so4 <= rm3==3'd3;              // -A - -B, sign = + unless rounding down

        endcase

always @(posedge clk)

if (ce) xdif4 <= xdiff3 > FMSB+3 ? FMSB+3 : xdiff3;

delay1 #(FMSB+2) dmsf3(.clk(clk), .ce(ce), .i(mfs3), .o(mfs4));

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #5

// Determine the sticky bit

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

wire [EMSB:0] xdif5;

wire [FMSB+1:0] mfs5;

wire sticky, sticky5;

// register inputs to shifter and shift

delay1 #(1)      dstky4(.clk(clk), .ce(ce), .i(sticky), .o(sticky5) );

delay1 #(EMSB+1) dxdif4(.clk(clk), .ce(ce), .i(xdif4), .o(xdif5) );

delay1 #(FMSB+2) dmsf4(.clk(clk), .ce(ce), .i(mfs4), .o(mfs5));

generate

begin

if (WID==128)

    redor128 u1 (.a(xdif4), .b({mfs4,2'b0}), .o(sticky) );

else if (WID==96)

    redor96 u1 (.a(xdif4), .b({mfs4,2'b0}), .o(sticky) );

else if (WID==80)

    redor80 u1 (.a(xdif4), .b({mfs4,2'b0}), .o(sticky) );

else if (WID==64)

    redor64 u1 (.a(xdif4), .b({mfs4,2'b0}), .o(sticky) );

else if (WID==40)

    redor40 u1 (.a(xdif4), .b({mfs4,2'b0}), .o(sticky) );

else if (WID==32)

    redor32 u1 (.a(xdif4), .b({mfs4,2'b0}), .o(sticky) );

end

endgenerate

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #6

// Shift (denormalize)

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

reg [FMSB+3:0] md6;

wire xa_gt_xb6;

wire [FMSB+1:0] fracta6, fractb6;

delay3 #(1) dxagtxb5(.clk(clk), .ce(ce), .i(xa_gt_xb3), .o(xa_gt_xb6));

delay4 #(FMSB+2)  dfracta5(.clk(clk), .ce(ce), .i(fracta2), .o(fracta6) );

delay4 #(FMSB+2)  dfractb5(.clk(clk), .ce(ce), .i(fractb2), .o(fractb6) );

always @(posedge clk)

        if (ce) md6 <= ({mfs5,2'b0} >> xdif5)|sticky5;

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #7

// Sort operands

// addition can generate an extra bit, subtract can't go negative

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

reg [FMSB+3:0] oa7;

reg [FMSB+3:0] ob7;

wire a_gt_b7;

delay4 #(1) dagtb5(.clk(clk), .ce(ce), .i(a_gt_b3), .o(a_gt_b7));

always @(posedge clk)

        if (ce) oa7 <= xa_gt_xb6 ? {fracta6,2'b0} : md6;

always @(posedge clk)

        if (ce) ob7 <= xa_gt_xb6 ? md6 : {fractb6,2'b0};

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #8

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

reg [FMSB+3:0] oaa8;

reg [FMSB+3:0] obb8;

wire [EMSB:0] xo8;

wire realOp8;

vtdl #(.WID(1)) drealop7 (.clk(clk), .ce(ce), .a(4'd5), .d(realOp2), .q(realOp8));

vtdl #(.WID(EMSB+1)) dxo7(.clk(clk), .ce(ce), .a(4'd3), .d(xo4), .q(xo8));

always @(posedge clk)

        if (ce) oaa8 <= a_gt_b7 ? oa7 : ob7;

always @(posedge clk)

        if (ce) obb8 <= a_gt_b7 ? ob7 : oa7;

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #9

// perform add/subtract

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

reg [FMSB+4:0] mab9;

wire anbInf9;

wire aNan9, bNan9;

wire op9;

wire [FMSB+1:0] fracta9, fractb9;

wire xo9;

reg xinf9;

vtdl #(1) danbInf7(.clk(clk), .ce(ce), .a(4'd6), .d(anbInf2), .q(anbInf9));

vtdl #(1) danan8(.clk(clk), .ce(ce), .a(4'd7), .d(aNan1), .q(aNan9));

vtdl #(1) dbnan8(.clk(clk), .ce(ce), .a(4'd7), .d(bNan1), .q(bNan9));

vtdl #(1) dop6(.clk(clk), .ce(ce), .a(4'd5), .d(op3), .q(op9));

delay3 #(FMSB+2)  dfracta8(.clk(clk), .ce(ce), .i(fracta6), .o(fracta9) );

delay3 #(FMSB+2)  dfractb8(.clk(clk), .ce(ce), .i(fractb6), .o(fractb9) );

always @(posedge clk)

        if (ce) mab9 <= realOp8 ? oaa8 - obb8 : oaa8 + obb8;

always @(posedge clk)

        if (ce) xinf9 <= xo8 == {EMSB+1{1'b1}};

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

// Clock edge #10

// Final outputs

// - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 

vtdl #(1) dso6(.clk(clk), .ce(ce), .a(4'd5), .d(so4), .q(so));

vtdl #(.WID(EMSB+1)) dxo6(.clk(clk), .ce(ce), .a(4'd1), .d(xo8), .q(xo));

always @(posedge clk)

if (ce)

        casez({anbInf9,aNan9,bNan9,xinf9})

        4'b1???:        mo <= {1'b0,op9,{FMSB-1{1'b0}},op9,{FMSB{1'b0}}};       // inf +/- inf - generate QNaN on subtract, inf on add

        4'b01??:        mo <= {1'b0,fracta9[FMSB+1:0],{FMSB{1'b0}}};

        4'b001?:        mo <= {1'b0,fractb9[FMSB+1:0],{FMSB{1'b0}}};

        4'b0001:        mo <= 1'd0;             // exponent hit infinity -> force mantissa to zero

        default:        mo <= {mab9,{FMSB-1{1'b0}}};    // mab has an extra lead bit and two trailing bits

        endcase

endmodule

module fpAddsubnr_L10(clk, ce, rm, op, a, b, o);

parameter WID = 128;

localparam MSB = WID-1;

localparam EMSB = WID==128 ? 14 :

                  WID==96 ? 14 :

                  WID==80 ? 14 :

                  WID==64 ? 10 :

                                  WID==52 ? 10 :

                                  WID==48 ? 11 :

                                  WID==44 ? 10 :

                                  WID==42 ? 10 :

                                  WID==40 ?  9 :

                                  WID==32 ?  7 :

                                  WID==24 ?  6 : 4;

localparam FMSB = WID==128 ? 111 :

                  WID==96 ? 79 :

                  WID==80 ? 63 :

                  WID==64 ? 51 :

                                  WID==52 ? 39 :

                                  WID==48 ? 34 :

                                  WID==44 ? 31 :

                                  WID==42 ? 29 :

                                  WID==40 ? 28 :

                                  WID==32 ? 22 :

                                  WID==24 ? 15 : 9;

localparam FX = (FMSB+2)*2-1;   // the MSB of the expanded fraction

localparam EX = FX + 1 + EMSB + 1 + 1 - 1;

input clk;              // system clock

input ce;               // core clock enable

input [2:0] rm;  // rounding mode

input op;               // operation 0 = add, 1 = subtract

input [MSB:0] a; // operand a

input [MSB:0] b; // operand b

output [MSB:0] o;        // output

wire [EX:0] o1;

wire [MSB+3:0] fpn0;

fpAddsub_L10  #(WID) u1 (clk, ce, rm, op, a, b, o1);

fpNormalize #(WID) u2(.clk(clk), .ce(ce), .under(1'b0), .i(o1), .o(fpn0) );

fpRoundReg  #(WID) u3(.clk(clk), .ce(ce), .rm(rm), .i(fpn0), .o(o) );

endmodule