X-Git-Url: http://git.joshuawise.com/mandelfpga.git/blobdiff_plain/9032b2b5f7475dc07709aa2e51f9a014f76760c0..08b195f395e9ed559004efdc953c3afaf8ba0444:/Main.v?ds=inline diff --git a/Main.v b/Main.v index 42e614a..aa3d861 100644 --- a/Main.v +++ b/Main.v @@ -5,10 +5,14 @@ * An implementation of a pipelined algorithm to calculate the Mandelbrot set * in real time on an FPGA. */ + +/* verilator lint_off WIDTH */ `define XRES 640 `define YRES 480 -`define WHIRRRRR 27 +`define WHIRRRRR 47 + +`define TOPBIT 13 module SyncGen( input pixclk, @@ -60,28 +64,31 @@ endmodule module NaiveMultiplier( input clk, - input [12:0] x, y, + input [`TOPBIT:0] x, y, input xsign, ysign, - output reg [12:0] out, + output reg [`TOPBIT:0] out, output reg sign, - output reg [1:0] ovf); + output reg ovf); always @(posedge clk) begin {ovf,out} <= - (((y[12] ? (x ) : 0) + - (y[11] ? (x >> 1) : 0) + - (y[10] ? (x >> 2) : 0) + - (y[9] ? (x >> 3) : 0)) + - ((y[8] ? (x >> 4) : 0) + - (y[7] ? (x >> 5) : 0) + - (y[6] ? (x >> 6) : 0)))+ - (((y[5] ? (x >> 7) : 0) + - (y[4] ? (x >> 8) : 0) + - (y[3] ? (x >> 9) : 0)) + - ((y[2] ? (x >> 10): 0) + - (y[1] ? (x >> 11): 0) + - (y[0] ? (x >> 12): 0))); + ((((0) + // 15 + (0)) + // 14 + ((y[13] ? (x ) : 0) + + (y[12] ? (x[`TOPBIT:1]) : 0))) + + (((y[11] ? (x[`TOPBIT:2]) : 0) + + (y[10] ? (x[`TOPBIT:3]) : 0)) + + ((y[9] ? (x[`TOPBIT:4]) : 0) + + (y[8] ? (x[`TOPBIT:5]) : 0))))+ + ((((y[7] ? (x[`TOPBIT:6]) : 0) + + (y[6] ? (x[`TOPBIT:7]) : 0)) + + ((y[5] ? (x[`TOPBIT:8]) : 0) + + (y[4] ? (x[`TOPBIT:9]) : 0))) + + (((y[3] ? (x[`TOPBIT:10]): 0) + + (y[2] ? (x[`TOPBIT:11]): 0)) + + ((y[1] ? (x[`TOPBIT:12]): 0) + + (y[0] ? (x[`TOPBIT]) : 0)))); sign <= xsign ^ ysign; end @@ -89,53 +96,57 @@ endmodule module Multiplier( input clk, - input [12:0] x, y, + input [`TOPBIT:0] x, y, input xsign, ysign, - output wire [12:0] out, + output wire [`TOPBIT:0] out, output wire sign, - output wire [1:0] overflow); + output wire overflow); NaiveMultiplier nm(clk, x, y, xsign, ysign, out, sign, overflow); endmodule +// Yuq. module MandelUnit( input clk, - input [12:0] x, y, + input [`TOPBIT:0] x, y, input xsign, ysign, - input [14:0] r, i, + input [`TOPBIT+2:0] r, i, input rsign, isign, input [7:0] ibail, icuriter, - output reg [12:0] xout, yout, + output reg [`TOPBIT:0] xout, yout, output reg xsout, ysout, - output reg [14:0] rout, iout, + output reg [`TOPBIT+2:0] rout, iout, output reg rsout, isout, output reg [7:0] obail, ocuriter); - wire [14:0] r2, i2, ri, diff; - wire [15:0] twocdiff; + wire [`TOPBIT+1:0] r2, i2; + wire [`TOPBIT+2:0] ri, diff; + wire [`TOPBIT+3:0] twocdiff; wire r2sign, i2sign, risign, dsign; - wire [16:0] bigsum; - wire bigsum_ovf, rin_ovf, iin_ovf, throwaway; + wire [`TOPBIT+2:0] bigsum; + wire bigsum_ovf; - reg [12:0] xd, yd; - reg rd, id; + reg [`TOPBIT:0] xd, yd; + reg ineedbaild; reg xsd, ysd; reg [7:0] ibaild, curiterd; assign ri[0] = 0; - Multiplier r2m(clk, r[12:0], r[12:0], rsign, rsign, r2[12:0], r2sign, r2[14:13]); - Multiplier i2m(clk, i[12:0], i[12:0], isign, isign, i2[12:0], i2sign, i2[14:13]); - Multiplier rim(clk, r[12:0], i[12:0], rsign, isign, ri[13:1], risign, {throwaway,ri[14]}); + Multiplier r2m(clk, r[`TOPBIT:0], r[`TOPBIT:0], rsign, rsign, r2[`TOPBIT:0], r2sign, r2[`TOPBIT+1]); + Multiplier i2m(clk, i[`TOPBIT:0], i[`TOPBIT:0], isign, isign, i2[`TOPBIT:0], i2sign, i2[`TOPBIT+1]); + Multiplier rim(clk, r[`TOPBIT:0], i[`TOPBIT:0], rsign, isign, ri[`TOPBIT+1:1], risign, ri[`TOPBIT+2]); - assign bigsum = r2 + i2; - assign bigsum_ovf = bigsum[16] | bigsum[15] | bigsum[14]; - assign rin_ovf = rd; - assign iin_ovf = id; + assign bigsum = r2[`TOPBIT+1:0] + i2[`TOPBIT+1:0]; + assign bigsum_ovf = bigsum[`TOPBIT+2]; + assign twocdiff = r2 - i2; - assign diff = twocdiff[15] ? -twocdiff : twocdiff; - assign dsign = twocdiff[15]; + assign diff = twocdiff[`TOPBIT+3] ? -twocdiff : twocdiff; + assign dsign = twocdiff[`TOPBIT+3]; + + wire [`TOPBIT+3:0] twocrout = xd - diff; + wire [`TOPBIT+3:0] twociout = yd - ri; always @ (posedge clk) begin @@ -149,28 +160,29 @@ module MandelUnit( ysout <= ysd; ibaild <= ibail; curiterd <= icuriter; - rd <= r[13] | r[14]; - id <= i[13] | i[14]; + ineedbaild <= r[`TOPBIT+1] | r[`TOPBIT+2] | i[`TOPBIT+1] | i[`TOPBIT+2]; + // r^2 - i^2 + x if (xsd ^ dsign) begin - if (diff > xd) begin - rout <= diff - xd; + if (twocrout[`TOPBIT+3]) begin // diff > xd + rout <= -twocrout; rsout <= dsign; end else begin - rout <= xd - diff; + rout <= twocrout; rsout <= xsd; end end else begin rout <= diff + xd; - rsout <= xsd; + rsout <= xsd; // xsd == dsign end + // 2 * r * i + y if (ysd ^ risign) begin - if (ri > yd) begin - iout <= ri - yd; + if (twociout[`TOPBIT+3]) begin // ri > yd + iout <= -twociout; isout <= risign; end else begin - iout <= yd - ri; + iout <= twociout; isout <= ysd; end end else begin @@ -180,7 +192,7 @@ module MandelUnit( // If we haven't bailed out, and we meet any of the bailout conditions, // bail out now. Otherwise, leave the bailout at whatever it was before. - if ((ibaild == 255) && (bigsum_ovf | rin_ovf | iin_ovf)) + if ((ibaild == 255) && (bigsum_ovf | ineedbaild)) obail <= curiterd; else obail <= ibaild; @@ -193,81 +205,136 @@ module Mandelbrot( input mclk, input pixclk, input [11:0] x, y, - input [13:0] xofs, yofs, + input [`TOPBIT+1:0] xofs, yofs, input [7:0] colorofs, input [2:0] scale, output reg [2:0] red, green, output reg [1:0] blue); -`define MAXOUTN 12 +`define MAXOUTN 21 - wire [12:0] rx, ry; - wire [13:0] nx, ny; + wire [`TOPBIT:0] rx, ry; + wire [`TOPBIT+1:0] nx, ny; wire rxsign, rysign; - assign nx = x + xofs; - assign ny = y + yofs; - assign rx = (nx[13] ? -nx[12:0] : nx[12:0]) << scale; - assign rxsign = nx[13]; - assign ry = (ny[13] ? -ny[12:0] : ny[12:0]) << scale; - assign rysign = ny[13]; - + assign nx = {2'b0,x} + {2'b0,xofs}; + assign ny = {2'b0,y} + {2'b0,yofs}; + assign rx = (nx[`TOPBIT+1] ? -nx[`TOPBIT:0] : nx[`TOPBIT:0]) << scale; + assign rxsign = nx[`TOPBIT+1]; + assign ry = (ny[`TOPBIT+1] ? -ny[`TOPBIT:0] : ny[`TOPBIT:0]) << scale; + assign rysign = ny[`TOPBIT+1]; - wire [14:0] mr[`MAXOUTN:0], mi[`MAXOUTN:0]; + wire [`TOPBIT+2:0] mr[`MAXOUTN:0], mi[`MAXOUTN:0]; wire mrs[`MAXOUTN:0], mis[`MAXOUTN:0]; wire [7:0] mb[`MAXOUTN:0]; - wire [12:0] xprop[`MAXOUTN:0], yprop[`MAXOUTN:0]; + wire [`TOPBIT:0] xprop[`MAXOUTN:0], yprop[`MAXOUTN:0]; wire xsprop[`MAXOUTN:0], ysprop[`MAXOUTN:0]; wire [7:0] curiter[`MAXOUTN:0]; - wire [14:0] initx, inity, initr, initi; - wire [7:0] initci, initb; - wire initxs, initys, initrs, initis; + reg [`TOPBIT:0] initx, inity; + reg [`TOPBIT+2:0] initr, initi; + reg [7:0] initci, initb; + reg initxs, initys, initrs, initis; - reg [14:0] loopx, loopy, loopr, loopi; - reg [7:0] loopci, loopb; - reg loopxs, loopys, looprs, loopis; + // Values after the number of iterations denoted by the subscript. + reg [`TOPBIT:0] stagex [2:1], stagey [2:1]; + reg [`TOPBIT+2:0] stager [2:1], stagei [2:1]; + reg [7:0] stageci [2:1], stageb [2:1]; + reg stagexs [2:1], stageys [2:1], stagers [2:1], stageis [2:1]; - reg state = 0; + reg [2:0] state = 3'b001; // One-hot encoded state. - // On pixclk = 1, - // A new value to be loaded comes in, and a value in need of loopback comes out. - // On pixclk = 0, - // A new value in need of loopback comes in, and a completed value comes out. - - assign initx = state ? rx : loopx; - assign inity = state ? ry : loopy; - assign initr = state ? rx : loopr; - assign initi = state ? ry : loopi; - assign initxs = state ? rxsign : loopxs; - assign initys = state ? rysign : loopys; - assign initrs = state ? rxsign : looprs; - assign initis = state ? rysign : loopis; - assign initb = state ? 8'b11111111 : loopb; - assign initci = state ? 8'b00000000 : loopci; + // States are advanced one from what they should be, so that they'll + // get there on the _next_ mclk. + always @(posedge mclk) + begin + initx <= (state[2]) ? rx : + (state[0]) ? stagex[1] : + (state[1]) ? stagex[2] : 0; + inity <= (state[2]) ? ry : + (state[0]) ? stagey[1] : + (state[1]) ? stagey[2] : 0; + initr <= (state[2]) ? {2'b0,rx} : + (state[0]) ? stager[1] : + (state[1]) ? stager[2] : 0; + initi <= (state[2]) ? {2'b0,ry} : + (state[0]) ? stagei[1] : + (state[1]) ? stagei[2] : 0; + initxs <= (state[2]) ? rxsign : + (state[0]) ? stagexs[1] : + (state[1]) ? stagexs[2] : 0; + initys <= (state[2]) ? rysign : + (state[0]) ? stageys[1] : + (state[1]) ? stageys[2] : 0; + initrs <= (state[2]) ? rxsign : + (state[0]) ? stagers[1] : + (state[1]) ? stagers[2] : 0; + initis <= (state[2]) ? rysign : + (state[0]) ? stageis[1] : + (state[1]) ? stageis[2] : 0; + initb <= (state[2]) ? 8'b11111111 : + (state[0]) ? stageb[1] : + (state[1]) ? stageb[2] : 0; + initci <= (state[2]) ? 8'b00000000 : + (state[0]) ? stageci[1] : + (state[1]) ? stageci[2] : 0; + end reg [7:0] out; - reg pixclksync; + + // We detect when the state should be poked by a high negedge followed + // by a high posedge -- if that happens, then we're guaranteed that the + // state following the current state will be 3'b100. + reg lastneg; always @(negedge mclk) - pixclksync <= ~pixclk; + lastneg <= pixclk; always @(posedge mclk) begin - if (!state) begin - out <= ~mb[`MAXOUTN] + colorofs; - end else begin + if (lastneg && pixclk) // If a pixclk has happened, the state should be reset. + state <= 3'b100; + else // Otherwise, just poke it forward. + case(state) + 3'b001: state <= 3'b010; + 3'b010: state <= 3'b100; + 3'b100: state <= 3'b001; + `ifdef isim + default: begin $display("invalid state"); $finish; end + `endif + endcase + + // Data output handling + if (state[0]) begin {red, green, blue} <= {out[0],out[3],out[6],out[1],out[4],out[7],out[2],out[5]}; - loopx <= xprop[`MAXOUTN]; - loopy <= yprop[`MAXOUTN]; - loopr <= mr[`MAXOUTN]; - loopi <= mi[`MAXOUTN]; - loopxs <= xsprop[`MAXOUTN]; - loopys <= ysprop[`MAXOUTN]; - looprs <= mrs[`MAXOUTN]; - loopis <= mis[`MAXOUTN]; - loopb <= mb[`MAXOUTN]; - loopci <= curiter[`MAXOUTN]; end - state <= ~pixclksync; + if (state[1]) begin + out <= ~mb[`MAXOUTN] + colorofs; + end + + if (state[0]) begin // PnR0 in, PnR2 out + stagex[2] <= xprop[`MAXOUTN]; + stagey[2] <= yprop[`MAXOUTN]; + stager[2] <= mr[`MAXOUTN]; + stagei[2] <= mi[`MAXOUTN]; + stagexs[2] <= xsprop[`MAXOUTN]; + stageys[2] <= ysprop[`MAXOUTN]; + stagers[2] <= mrs[`MAXOUTN]; + stageis[2] <= mis[`MAXOUTN]; + stageb[2] <= mb[`MAXOUTN]; + stageci[2] <= curiter[`MAXOUTN]; + end + + if (state[2]) begin // PnR2 in, PnR1 out + stagex[1] <= xprop[`MAXOUTN]; + stagey[1] <= yprop[`MAXOUTN]; + stager[1] <= mr[`MAXOUTN]; + stagei[1] <= mi[`MAXOUTN]; + stagexs[1] <= xsprop[`MAXOUTN]; + stageys[1] <= ysprop[`MAXOUTN]; + stagers[1] <= mrs[`MAXOUTN]; + stageis[1] <= mis[`MAXOUTN]; + stageb[1] <= mb[`MAXOUTN]; + stageci[1] <= curiter[`MAXOUTN]; + end end MandelUnit mu0( @@ -278,44 +345,33 @@ module Mandelbrot( xprop[0], yprop[0], xsprop[0], ysprop[0], mr[0], mi[0], mrs[0], mis[0], mb[0], curiter[0]); + +`define MAKE_UNIT(name, num) \ + MandelUnit name(mclk, \ + xprop[(num)], yprop[(num)], xsprop[(num)], ysprop[(num)], mr[(num)], mi[(num)], mrs[(num)], mis[(num)], mb[(num)], curiter[(num)], \ + xprop[(num)+1], yprop[(num)+1], xsprop[(num)+1], ysprop[(num)+1], mr[(num)+1], mi[(num)+1], mrs[(num)+1], mis[(num)+1], mb[(num)+1], curiter[(num)+1]) - MandelUnit mu1(mclk, - xprop[0], yprop[0], xsprop[0], ysprop[0], mr[0], mi[0], mrs[0], mis[0], mb[0], curiter[0], - xprop[1], yprop[1], xsprop[1], ysprop[1], mr[1], mi[1], mrs[1], mis[1], mb[1], curiter[1]); - MandelUnit mu2(mclk, - xprop[1], yprop[1], xsprop[1], ysprop[1], mr[1], mi[1], mrs[1], mis[1], mb[1], curiter[1], - xprop[2], yprop[2], xsprop[2], ysprop[2], mr[2], mi[2], mrs[2], mis[2], mb[2], curiter[2]); - MandelUnit mu3(mclk, - xprop[2], yprop[2], xsprop[2], ysprop[2], mr[2], mi[2], mrs[2], mis[2], mb[2], curiter[2], - xprop[3], yprop[3], xsprop[3], ysprop[3], mr[3], mi[3], mrs[3], mis[3], mb[3], curiter[3]); - MandelUnit mu4(mclk, - xprop[3], yprop[3], xsprop[3], ysprop[3], mr[3], mi[3], mrs[3], mis[3], mb[3], curiter[3], - xprop[4], yprop[4], xsprop[4], ysprop[4], mr[4], mi[4], mrs[4], mis[4], mb[4], curiter[4]); - MandelUnit mu5(mclk, - xprop[4], yprop[4], xsprop[4], ysprop[4], mr[4], mi[4], mrs[4], mis[4], mb[4], curiter[4], - xprop[5], yprop[5], xsprop[5], ysprop[5], mr[5], mi[5], mrs[5], mis[5], mb[5], curiter[5]); - MandelUnit mu6(mclk, - xprop[5], yprop[5], xsprop[5], ysprop[5], mr[5], mi[5], mrs[5], mis[5], mb[5], curiter[5], - xprop[6], yprop[6], xsprop[6], ysprop[6], mr[6], mi[6], mrs[6], mis[6], mb[6], curiter[6]); - MandelUnit mu7(mclk, - xprop[6], yprop[6], xsprop[6], ysprop[6], mr[6], mi[6], mrs[6], mis[6], mb[6], curiter[6], - xprop[7], yprop[7], xsprop[7], ysprop[7], mr[7], mi[7], mrs[7], mis[7], mb[7], curiter[7]); - MandelUnit mu8(mclk, - xprop[7], yprop[7], xsprop[7], ysprop[7], mr[7], mi[7], mrs[7], mis[7], mb[7], curiter[7], - xprop[8], yprop[8], xsprop[8], ysprop[8], mr[8], mi[8], mrs[8], mis[8], mb[8], curiter[8]); - MandelUnit mu9(mclk, - xprop[8], yprop[8], xsprop[8], ysprop[8], mr[8], mi[8], mrs[8], mis[8], mb[8], curiter[8], - xprop[9], yprop[9], xsprop[9], ysprop[9], mr[9], mi[9], mrs[9], mis[9], mb[9], curiter[9]); - MandelUnit mua(mclk, - xprop[9], yprop[9], xsprop[9], ysprop[9], mr[9], mi[9], mrs[9], mis[9], mb[9], curiter[9], - xprop[10], yprop[10], xsprop[10], ysprop[10], mr[10], mi[10], mrs[10], mis[10], mb[10], curiter[10]); - MandelUnit mub(mclk, - xprop[10], yprop[10], xsprop[10], ysprop[10], mr[10], mi[10], mrs[10], mis[10], mb[10], curiter[10], - xprop[11], yprop[11], xsprop[11], ysprop[11], mr[11], mi[11], mrs[11], mis[11], mb[11], curiter[11]); - MandelUnit muc(mclk, - xprop[11], yprop[11], xsprop[11], ysprop[11], mr[11], mi[11], mrs[11], mis[11], mb[11], curiter[11], - xprop[12], yprop[12], xsprop[12], ysprop[12], mr[12], mi[12], mrs[12], mis[12], mb[12], curiter[12]); - + `MAKE_UNIT(mu1, 0); + `MAKE_UNIT(mu2, 1); + `MAKE_UNIT(mu3, 2); + `MAKE_UNIT(mu4, 3); + `MAKE_UNIT(mu5, 4); + `MAKE_UNIT(mu6, 5); + `MAKE_UNIT(mu7, 6); + `MAKE_UNIT(mu8, 7); + `MAKE_UNIT(mu9, 8); + `MAKE_UNIT(mua, 9); + `MAKE_UNIT(mub, 10); + `MAKE_UNIT(muc, 11); + `MAKE_UNIT(mud, 12); + `MAKE_UNIT(mue, 13); + `MAKE_UNIT(muf, 14); + `MAKE_UNIT(mug, 15); + `MAKE_UNIT(muh, 16); + `MAKE_UNIT(mui, 17); + `MAKE_UNIT(muj, 18); + `MAKE_UNIT(muk, 19); + `MAKE_UNIT(mul, 20); endmodule module Logo( @@ -338,18 +394,40 @@ module Logo( endmodule module MandelTop( +`ifdef verilator + input pixclk, mclk, +`else input gclk, output wire dcmok, +`endif output wire vs, hs, output wire [2:0] red, green, output [1:0] blue, input left, right, up, down, rst, cycle, logooff, input [2:0] scale); + +`ifdef verilator +`else + wire pixclk, mclk, clk; + wire dcm1ok, dcm2ok; + assign dcmok = dcm1ok && dcm2ok; + + IBUFG iclkbuf(.O(clk), .I(gclk)); + + pixDCM dcm( // CLKIN is 50MHz xtal, CLKFX_OUT is 25MHz + .CLKIN_IN(clk), + .CLKFX_OUT(pixclk), + .LOCKED_OUT(dcm1ok) + ); + + mandelDCM dcm2( + .CLKIN_IN(clk), + .CLKFX_OUT(mclk), + .LOCKED_OUT(dcm2ok) + ); +`endif wire border; - wire pixclk; - wire [7:0] zero = 8'b0; - wire clk; wire [11:0] x, y; - reg [13:0] xofs = -`XRES/2, yofs = -`YRES/2; + reg [`TOPBIT+1:0] xofs = -`XRES/2, yofs = -`YRES/2; reg [5:0] slowctr = 0; reg [7:0] colorcycle = 0; wire [11:0] realx, realy; @@ -358,15 +436,8 @@ module MandelTop( wire [2:0] mandelr, mandelg, logor, logog; wire [1:0] mandelb, logob; - pixDCM dcm( // CLKIN is 50MHz xtal, CLKFX_OUT is 25MHz - .CLKIN_IN(gclk), - .CLKFX_OUT(pixclk), - .CLKIN_IBUFG_OUT(clk), - .LOCKED_OUT(dcmok) - ); - SyncGen sync(pixclk, vs, hs, x, y, realx, realy, border); - Mandelbrot mandel(clk, pixclk, x, y, xofs, yofs, cycle ? colorcycle : 0, scale, mandelr, mandelg, mandelb); + Mandelbrot mandel(mclk, pixclk, x, y, xofs, yofs, cycle ? colorcycle : 8'b0, scale, mandelr, mandelg, mandelb); Logo logo(pixclk, realx, realy, logoenb, logor, logog, logob); assign {red,green,blue} =