URL
https://opencores.org/ocsvn/raytrac/raytrac/trunk
Subversion Repositories raytrac
[/] [raytrac/] [branches/] [fp_sgdma/] [ap_n_dpc.vhd] - Rev 136
Go to most recent revision | Compare with Previous | Blame | View Log
--! @file dpc.vhd --! @brief Decodificador de operacion. --! @author Julián Andrés Guarín Reyes -------------------------------------------------------------- -- RAYTRAC -- Author Julian Andres Guarin -- dpc.vhd -- This file is part of raytrac. -- -- raytrac is free software: you can redistribute it and/or 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 your option) any later version. -- -- raytrac 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 raytrac. If not, see <http://www.gnu.org/licenses/>. library ieee; use ieee.std_logic_1164.all; entity dpc is generic ( width : integer := 32 --!external_readable_widthad : integer := integer(ceil(log(real(external_readable_blocks),2.0)))) ); port ( clk : in std_logic; paraminput : in std_logic_vector ((12*width)-1 downto 0); --! Vectores A,B,C,D prd32blko : in std_logic_vector ((06*width)-1 downto 0); --! Salidas de los 6 multiplicadores. add32blko : in std_logic_vector ((04*width)-1 downto 0); --! Salidas de los 4 sumadores. sqr32blko,inv32blko : in std_logic_vector (width-1 downto 0); --! Salidas de la raiz cuadradas y el inversor. fifo32x26_q : in std_logic_vector (03*width-1 downto 0); --! Salida de la cola intermedia. fifo32x09_q : in std_logic_vector (02*width-1 downto 0); --! Salida de las colas de producto punto. unary,crossprod,addsub : in std_logic; --! Bit con el identificador del bloque AB vs CD e identificador del sub bloque (A/B) o (C/D). scalar : in std_logic; sqr32blki,inv32blki : out std_logic_vector (width-1 downto 0); --! Salidas de las 2 raices cuadradas y los 2 inversores. fifo32x26_d : out std_logic_vector (03*width-1 downto 0); --! Entrada a la cola intermedia para la normalización. fifo32x09_d : out std_logic_vector (02*width-1 downto 0); --! Entrada a las colas intermedias del producto punto. prd32blki : out std_logic_vector ((12*width)-1 downto 0); --! Entrada de los 12 factores en el bloque de multiplicación respectivamente. add32blki : out std_logic_vector ((08*width)-1 downto 0); --! Entrada de los 8 sumandos del bloque de 4 sumadores. resultoutput : out std_logic_vector ((08*width)-1 downto 0) --! 6 salidas de resultados, pues lo máximo que podrá calcularse por cada clock son 2 vectores. ); end dpc; architecture dpc_arch of dpc is constant qz : integer := 00;constant qy : integer := 01;constant qx : integer := 02; constant az : integer := 00;constant ay : integer := 01;constant ax : integer := 02;constant bz : integer := 03;constant by : integer := 04;constant bx : integer := 05; constant cz : integer := 06;constant cy : integer := 07;constant cx : integer := 08;constant dz : integer := 09;constant dy : integer := 10;constant dx : integer := 11; constant f0 : integer := 00;constant f1 : integer := 01;constant f2 : integer := 02;constant f3 : integer := 03;constant f4 : integer := 04;constant f5 : integer := 05; constant f6 : integer := 06;constant f7 : integer := 07;constant f8 : integer := 08;constant f9 : integer := 09;constant f10: integer := 10;constant f11: integer := 11; constant s0 : integer := 00;constant s1 : integer := 01;constant s2 : integer := 02;constant s3 : integer := 03;constant s4 : integer := 04;constant s5 : integer := 05; constant s6 : integer := 06;constant s7 : integer := 07; constant a0 : integer := 00;constant a1 : integer := 01;constant a2 : integer := 02;constant aa : integer := 03; constant p0 : integer := 00;constant p1 : integer := 01;constant p2 : integer := 02;constant p3 : integer := 03;constant p4 : integer := 04;constant p5 : integer := 05; constant dpfifoab : integer := 00; constant dpfifocd : integer := 01; type vectorblock12 is array (11 downto 0) of std_logic_vector(width-1 downto 0); type vectorblock08 is array (07 downto 0) of std_logic_vector(width-1 downto 0); type vectorblock06 is array (05 downto 0) of std_logic_vector(width-1 downto 0); type vectorblock04 is array (03 downto 0) of std_logic_vector(width-1 downto 0); type vectorblock03 is array (02 downto 0) of std_logic_vector(width-1 downto 0); type vectorblock02 is array (01 downto 0) of std_logic_vector(width-1 downto 0); signal sparaminput,sfactor : vectorblock12; signal ssumando,sresult : vectorblock08; signal sprd32blk : vectorblock06; signal sadd32blk : vectorblock04; signal snormfifo_q,snormfifo_d : vectorblock03; signal sdpfifo_q : vectorblock02; signal ssqr32blk,sinv32blk : std_logic_vector(width-1 downto 0); begin stuff12: for i in 11 downto 0 generate sparaminput(i) <= paraminput(i*width+width-1 downto i*width); prd32blki(i*width+width-1 downto i*width) <= sfactor(i); end generate stuff12; stuff08: for i in 07 downto 0 generate add32blki(i*width+width-1 downto i*width) <= ssumando(i); resultoutput(i*width+width-1 downto i*width) <= sresult(i); end generate stuff08; process (clk) begin if clk'event and clk='1' then for i 05 downto 0 loop sprd32blk(p0) <= prd32blko(i*width+width-1 downto i*width); end loop; end if; end process; stuff04: for i in 03 downto 1 generate sadd32blk(i) <= add32blko(i*width+width-1 downto i*width); end generate stuff04; process (clk) begin if clk'event and clk='1' then sadd32blk(a0) <= add32blko(a0*width+width-1 downto a0*width); sinv32blk <= inv32blko; end if; end process; stuff03: for i in 02 downto 0 generate snormfifo_q(i) <= fifo32x26_q(i*width+width-1 downto i*width); fifo32x26_d(i*width+width-1 downto i*width) <= snormfifo_d(i); end generate stuff03; stuff02: for i in 01 downto 0 generate sdpfifo_q(i) <= fifo32x09_q(i*width+width-1 downto i*width); end generate stuff02; fifo32x09_d <= sprd32blk(p3)&sprd32blk(p2); ssqr32blk <= sqr32blko; --! Colas de salida de los distintos resultados; sresult(0) <= ssqr32blk; sresult(1) <= sadd32blk(a0); sresult(2) <= sadd32blk(a1); sresult(3) <= sadd32blk(a2); sresult(4) <= sadd32blk(aa); sresult(5) <= sprd32blk(p3); sresult(6) <= sprd32blk(p4); sresult(7) <= sprd32blk(p5); --! Cola de normalizacion snormfifo_d(qx) <= sparaminput(ax); snormfifo_d(qy) <= sparaminput(ay); snormfifo_d(qz) <= sparaminput(az); --! La entrada al inversor SIEMPRE viene con la salida de la raiz cuadrada inv32blki <= sqr32blko; --! La entrada de la ra’z cuadrada SIEMPRE viene con la salida del sumador 1. sqr32blki <= sadd32blk(a1); --! Conectar las entradas del sumador a, a la salida ssumando(s6) <= sadd32blk(a2); ssumando(s7) <= sdpfifo_q(dpfifocd); mul:process(unary,addsub,crossprod,scalar,sparaminput,sinv32blk,sprd32blk,sadd32blk,sdpfifo_q,snormfifo_q) begin if unary='1' then --! Magnitud y normalizacion sfactor(f0) <= sparaminput(ax); sfactor(f1) <= sparaminput(ax); sfactor(f2) <= sparaminput(ay); sfactor(f3) <= sparaminput(ay); sfactor(f4) <= sparaminput(az); sfactor(f5) <= sparaminput(az); sfactor(f6) <= snormfifo_q(ax); sfactor(f7) <= sinv32blk; sfactor(f8) <= snormfifo_q(ay); sfactor(f9) <= sinv32blk; sfactor(f10) <= snormfifo_q(az); sfactor(f11) <= sinv32blk; elsif crossprod='1' then --! Solo productos punto sfactor(f0) <= sparaminput(ay); sfactor(f1) <= sparaminput(bz); sfactor(f2) <= sparaminput(az); sfactor(f3) <= sparaminput(by); sfactor(f4) <= sparaminput(az); sfactor(f5) <= sparaminput(bx); sfactor(f6) <= sparaminput(ax); sfactor(f7) <= sparaminput(bz); sfactor(f8) <= sparaminput(ax); sfactor(f9) <= sparaminput(by); sfactor(f10) <= sparaminput(ay); sfactor(f11) <= sparaminput(bx); elsif scalar='0' then --! Producto punto sfactor(f0) <= sparaminput(ax) ; sfactor(f1) <= sparaminput(bx) ; sfactor(f2) <= sparaminput(ay) ; sfactor(f3) <= sparaminput(by) ; sfactor(f4) <= sparaminput(az) ; sfactor(f5) <= sparaminput(bz) ; sfactor(f6) <= sparaminput(cx) ; sfactor(f7) <= sparaminput(dx) ; sfactor(f8) <= sparaminput(cy) ; sfactor(f9) <= sparaminput(dy) ; sfactor(f10) <= sparaminput(cz) ; sfactor(f11) <= sparaminput(dz) ; else sfactor(f0) <= sparaminput(ax) ; sfactor(f1) <= sparaminput(bx) ; sfactor(f2) <= sparaminput(ay) ; sfactor(f3) <= sparaminput(by) ; sfactor(f4) <= sparaminput(az) ; sfactor(f5) <= sparaminput(bz) ; sfactor(f6) <= sparaminput(cx) ; sfactor(f7) <= sparaminput(dx) ; sfactor(f8) <= sparaminput(cy) ; sfactor(f9) <= sparaminput(dx) ; sfactor(f10) <= sparaminput(cz) ; sfactor(f11) <= sparaminput(dx) ; end if; if addsub='1' then ssumando(s0) <= sparaminput(ax); ssumando(s1) <= sparaminput(bx); ssumando(s2) <= sparaminput(ay); ssumando(s3) <= sparaminput(by); ssumando(s4) <= sparaminput(az); ssumando(s5) <= sparaminput(bz); else ssumando(s0) <= sprd32blk(p0); ssumando(s1) <= sprd32blk(p1); if crossprod='0' then ssumando(s2) <= sadd32blk(a0); ssumando(s3) <= sdpfifo_q(dpfifoab); else ssumando(s2) <= sprd32blk(p2); ssumando(s3) <= sprd32blk(p3); end if; ssumando(s4) <= sprd32blk(p4); ssumando(s5) <= sprd32blk(p5); end if; end process; -- interconnection:process(instr3,hblockslab,abblockslab,cdblockslab,sparaminput,sprd32blk,sadd32blk,sdpfifo_q) -- begin -- --! La cola para la normalizacion de los vectores. -- snormfifo_d(qx) <= (hblockslab and ((cdblockslab and sparaminput(dx))or(not(cdblockslab) and sparaminput(cx)))) or (not(hblockslab) and ((abblockslab and sparaminput(bx))or(not(abblockslab) and sparaminput(ax)))); -- snormfifo_d(qy) <= (hblockslab and ((cdblockslab and sparaminput(dy))or(not(cdblockslab) and sparaminput(cy)))) or (not(hblockslab) and ((abblockslab and sparaminput(by))or(not(abblockslab) and sparaminput(ay)))); -- snormfifo_d(qz) <= (hblockslab and ((cdblockslab and sparaminput(dz))or(not(cdblockslab) and sparaminput(cz)))) or (not(hblockslab) and ((abblockslab and sparaminput(bz))or(not(abblockslab) and sparaminput(az)))); -- -- --! Combinatorio para decidir que operaciones realizan los sumadores / restadores. -- add32blks <= (instr3(0) xor (instr3(1) xor instr3(0)))&(instr3(0) xor (instr3(1) xor instr3(0))) ; -- -- --! Por defecto conectar los sumandos en producto punto/cruz -- ssumando(s0) <= sprd32blk(p0);ssumando(s1) <= sprd32blk(p1); -- ssumando(s6) <= sadd32blk(a0);ssumando(s7) <= sdpfifo_q(dpfifoab); -- ssumando(s10) <= sdpfifo_q(dpfifocd);ssumando(s11) <= sadd32blk(a2); -- ssumando(s4) <= sprd32blk(p4);ssumando(s5) <= sprd32blk(p5); -- ssumando(s2) <= sprd32blk(p2);ssumando(s3) <= sprd32blk(p3); -- -- --! El segundo sumador del segundo bloque siempre sera suma o resta independiente de la operacion -- ssumando(s8) <= sparaminput(cy);ssumando(s9) <= sparaminput(dy); -- -- --! Por defecto conectar los factores en producto punto -- sfactor(f0) <= sparaminput(ax);sfactor(f1) <= sparaminput(bx); -- sfactor(f2) <= sparaminput(ay);sfactor(f3) <= sparaminput(by); -- sfactor(f4) <= sparaminput(az);sfactor(f5) <= sparaminput(bz); -- sfactor(f6) <= sparaminput(bx);sfactor(f7) <= sparaminput(dx); -- sfactor(f8) <= sparaminput(by);sfactor(f9) <= sparaminput(dy); -- sfactor(f10) <= sparaminput(bz);sfactor(f11) <= sparaminput(dz); -- -- --!Los resultados por defecto se acomodan al producto punto y parcialmente a los productos simple y escalar. -- sresult(ax) <= sadd32blk(aa); -- sresult(ay) <= sprd32blk(p1); -- sresult(az) <= sprd32blk(p2); -- sresult(bx) <= sadd32blk(ac); -- sresult(by) <= sprd32blk(p4); -- sresult(bz) <= sprd32blk(p5); -- -- if (instr3(2 downto 1)="11" or instr3="100") then -- sresult(ax) <= sprd32blk(p0); -- sresult(bx) <= sprd32blk(p3); -- elsif instr3(0)='1' then -- sresult(ax) <= sprd32blk(a0); -- sresult(ay) <= sprd32blk(a1); -- sresult(az) <= sprd32blk(a2); -- sresult(bx) <= sadd32blk(aa); -- sresult(by) <= sprd32blk(ab); -- sresult(bz) <= sadd32blk(ac); -- elsif instr3(1)='1' then -- sresult(ax) <= ssqr32blk(sqrt320); -- sresult(bx) <= ssqr32blk(sqrt321); -- end if; -- -- -- if instr3(0)='1' then --! Producto Cruz, suma, resta, multiplicacion simple -- -- if (instr3(2) or instr3(1))='1' then --! Suma, Resta, Multiplicacion simple -- -- --! Conectar las entradas de los sumadores en suma o resta de vectores -- ssumando(s0) <= sparaminput(ax);ssumando(s1) <= sparaminput(bx); -- ssumando(s2) <= sparaminput(ay);ssumando(s3) <= sparaminput(by); -- ssumando(s4) <= sparaminput(az);ssumando(s5) <= sparaminput(bz); -- ssumando(s6) <= sparaminput(cx);ssumando(s7) <= sparaminput(dx); -- ssumando(s10) <= sparaminput(cz);ssumando(s11) <= sparaminput(dz); -- -- else --! Producto Cruz! -- -- if hblock='1' then --! Producto crux CxD -- --!Multiplicadores: -- sfactor(f0) <= sparaminput(cy);sfactor(f1) <= sparaminput(dz);sfactor(f2) <= sparaminput(cz);sfactor(f3) <= sparaminput(dy); -- sfactor(f4) <= sparaminput(cx);sfactor(f5) <= sparaminput(dz);sfactor(f6) <= sparaminput(cz);sfactor(f7) <= sparaminput(dx); -- sfactor(f8) <= sparaminput(cx);sfactor(f9) <= sparaminput(dy);sfactor(f10) <= sparaminput(cy);sfactor(f11) <= sparaminput(dx); -- else --! Producto crux AxD -- --!Multiplicadores: -- sfactor(f0) <= sparaminput(ay);sfactor(f1) <= sparaminput(bz);sfactor(f2) <= sparaminput(az);sfactor(f3) <= sparaminput(by); -- sfactor(f4) <= sparaminput(ax);sfactor(f5) <= sparaminput(bz);sfactor(f6) <= sparaminput(az);sfactor(f7) <= sparaminput(bx); -- sfactor(f8) <= sparaminput(ax);sfactor(f9) <= sparaminput(by);sfactor(f10) <= sparaminput(ay);sfactor(f11) <= sparaminput(bx); -- end if; -- -- end if; -- -- else --! Producto Punto, magnitud, producto escalar y normalizacion -- if instr3(2)='1' then --!Producto Escalar (INSTR3(1)=0) o Normalizacion (INSTR3(1)=1) -- -- sfactor(f0) <= (not instr31slab and sparaminput(ax)) or (instr31slab and ((not(hblockslab) and ((not(abblockslab) and sparaminput(ax)) or(abblockslab and sparaminput(bx))))or( hblockslab and snormfifo_q(qx)) ) ); -- sfactor(f1) <= (not instr31slab and sparaminput(bx)) or (instr31slab and ((not(hblockslab) and ((not(abblockslab) and sparaminput(ax)) or(abblockslab and sparaminput(bx))))or( hblockslab and sinv32blk(invr321)) ) ); -- sfactor(f2) <= (not instr31slab and sparaminput(ay)) or (instr31slab and ((not(hblockslab) and ((not(abblockslab) and sparaminput(ay)) or(abblockslab and sparaminput(by))))or( hblockslab and snormfifo_q(qy)) ) ); -- sfactor(f3) <= (not instr31slab and sparaminput(bx)) or (instr31slab and ((not(hblockslab) and ((not(abblockslab) and sparaminput(ay)) or(abblockslab and sparaminput(by))))or( hblockslab and sinv32blk(invr321)) ) ); -- sfactor(f4) <= (not instr31slab and sparaminput(az)) or (instr31slab and ((not(hblockslab) and ((not(abblockslab) and sparaminput(az)) or(abblockslab and sparaminput(bz))))or( hblockslab and snormfifo_q(qz)) ) ); -- sfactor(f5) <= (not instr31slab and sparaminput(bx)) or (instr31slab and ((not(hblockslab) and ((not(abblockslab) and sparaminput(az)) or(abblockslab and sparaminput(bz))))or( hblockslab and sinv32blk(invr321)) ) ); -- sfactor(f6) <= (not instr31slab and sparaminput(cx)) or (instr31slab and ((hblockslab and ((not(cdblockslab) and sparaminput(cx)) or(cdblockslab and sparaminput(dx))))or( not(hblockslab) and snormfifo_q(qx)) ) ); -- sfactor(f7) <= (not instr31slab and sparaminput(dx)) or (instr31slab and ((hblockslab and ((not(cdblockslab) and sparaminput(cx)) or(cdblockslab and sparaminput(dx))))or( not(hblockslab) and sinv32blk(invr320)) ) ); -- sfactor(f8) <= (not instr31slab and sparaminput(cy)) or (instr31slab and ((hblockslab and ((not(cdblockslab) and sparaminput(cy)) or(cdblockslab and sparaminput(dy))))or( not(hblockslab) and snormfifo_q(qy)) ) ); -- sfactor(f9) <= (not instr31slab and sparaminput(dx)) or (instr31slab and ((hblockslab and ((not(cdblockslab) and sparaminput(cy)) or(cdblockslab and sparaminput(dy))))or( not(hblockslab) and sinv32blk(invr320)) ) ); -- sfactor(f10) <= (not instr31slab and sparaminput(cz)) or (instr31slab and ((hblockslab and ((not(cdblockslab) and sparaminput(cz)) or(cdblockslab and sparaminput(dz))))or( not(hblockslab) and snormfifo_q(qz)) ) ); -- sfactor(f11) <= (not instr31slab and sparaminput(dx)) or (instr31slab and ((hblockslab and ((not(cdblockslab) and sparaminput(cz)) or(cdblockslab and sparaminput(dz))))or( not(hblockslab) and sinv32blk(invr320)) ) ); -- elsif instr3(1)='1' then --!Magnitud. El producto punto no se computa porque los factores estan por defecto configurados en producto punto. -- sfactor(f0) <= (not(abblockslab) and sparaminput(ax))or(abblockslab and sparaminput(bx)); -- sfactor(f1) <= (not(abblockslab) and sparaminput(ax))or(abblockslab and sparaminput(bx)); -- sfactor(f2) <= (not(abblockslab) and sparaminput(ay))or(abblockslab and sparaminput(by)); -- sfactor(f3) <= (not(abblockslab) and sparaminput(ay))or(abblockslab and sparaminput(by)); -- sfactor(f4) <= (not(abblockslab) and sparaminput(az))or(abblockslab and sparaminput(bz)); -- sfactor(f5) <= (not(abblockslab) and sparaminput(az))or(abblockslab and sparaminput(bz)); -- sfactor(f6) <= (not(cdblockslab) and sparaminput(cx))or(cdblockslab and sparaminput(dx)); -- sfactor(f7) <= (not(cdblockslab) and sparaminput(cx))or(cdblockslab and sparaminput(dx)); -- sfactor(f8) <= (not(cdblockslab) and sparaminput(cy))or(cdblockslab and sparaminput(dy)); -- sfactor(f9) <= (not(cdblockslab) and sparaminput(cy))or(cdblockslab and sparaminput(dy)); -- sfactor(f10) <= (not(cdblockslab) and sparaminput(cz))or(cdblockslab and sparaminput(dz)); -- sfactor(f11) <= (not(cdblockslab) and sparaminput(cz))or(cdblockslab and sparaminput(dz)); -- -- end if; -- end if; -- -- end process; -- end dpc_arch;
Go to most recent revision | Compare with Previous | Blame | View Log