URL
https://opencores.org/ocsvn/aemb/aemb/trunk
Subversion Repositories aemb
[/] [aemb/] [tags/] [AEMB_7_05/] [rtl/] [verilog/] [aeMB_aslu.v] - Rev 36
Go to most recent revision | Compare with Previous | Blame | View Log
/* * $Id: aeMB_aslu.v,v 1.9 2007-05-17 09:08:21 sybreon Exp $ * * AEMB Arithmetic Shift Logic Unit * Copyright (C) 2004-2007 Shawn Tan Ser Ngiap <shawn.tan@aeste.net> * * This library 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 2.1 of * the License, or (at your option) any later version. * * This library 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 * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 * USA * * DESCRIPTION * Arithmetic, shift and logic execution unit. It also performs the * necessary calculations for branch and load/store targets. * * HISTORY * $Log: not supported by cvs2svn $ * Revision 1.8 2007/04/30 15:56:50 sybreon * Removed byte acrobatics. * * Revision 1.7 2007/04/27 00:23:55 sybreon * Added code documentation. * Improved size & speed of rtl/verilog/aeMB_aslu.v * * Revision 1.6 2007/04/26 14:29:53 sybreon * Made minor performance optimisations. * * Revision 1.5 2007/04/25 22:15:04 sybreon * Added support for 8-bit and 16-bit data types. * * Revision 1.4 2007/04/11 04:30:43 sybreon * Added pipeline stalling from incomplete bus cycles. * Separated sync and async portions of code. * * Revision 1.3 2007/04/04 06:11:05 sybreon * Added CMP instruction * * Revision 1.2 2007/04/03 14:46:26 sybreon * Fixed endian correction issues on data bus. * * Revision 1.1 2007/03/09 17:52:17 sybreon * initial import * */ module aeMB_aslu (/*AUTOARG*/ // Outputs dwb_adr_o, dwb_sel_o, rRESULT, rDWBSEL, // Inputs sDWBDAT, rBRA, rDLY, rREGA, rREGB, rSIMM, rMXSRC, rMXTGT, rMXALU, rOPC, rPC, rIMM, rRD, rRA, rMXLDST, nclk, prst, drun, prun ); parameter DSIZ = 32; output [DSIZ-1:0] dwb_adr_o; output [3:0] dwb_sel_o; output [31:0] rRESULT; output [3:0] rDWBSEL; input [31:0] sDWBDAT; input rBRA, rDLY; input [31:0] rREGA, rREGB; input [31:0] rSIMM; input [1:0] rMXSRC,rMXTGT; input [1:0] rMXALU; input [5:0] rOPC; input [31:0] rPC; input [15:0] rIMM; input [4:0] rRD, rRA; input [1:0] rMXLDST; input nclk, prst, drun, prun; reg [31:0] rRESULT, xRESULT; reg rMSR_C, xMSR_C; /** Operand Selection ----------------- Selects the A and B operands depending on the source. All forwarding information is controlled by the decode unit. */ wire [31:0] wOPA, wOPB; assign wOPA = (rMXSRC == 2'b11) ? sDWBDAT : (rMXSRC == 2'b10) ? rRESULT : (rMXSRC == 2'b01) ? rPC : rREGA; assign wOPB = (rMXTGT == 2'b11) ? sDWBDAT : (rMXTGT == 2'b10) ? rRESULT : (rMXTGT == 2'b01) ? rSIMM : rREGB; /** Simple Arithmetic ----------------- Performs addition and subtraction using a single 32-bit adder with carry in and out. This is done in parallel with other operations. The adder is also used to calculate branch target addresses as well as load/store memory addresses. TODO: Verify signed compare */ wire wADDC, wSUBC, wRES_AC, wCMPC, wOPC; wire [31:0] wADD, wSUB, wRES_A, wCMP, wOPX; wire wCMP0 = (wOPA[7:0] > wOPB[7:0]); wire wCMP1 = (wOPA[15:8] > wOPB[15:8]); wire wCMP2 = (wOPA[23:16] > wOPB[23:16]); wire wCMP3 = (wOPA[31:24] > wOPB[31:24]); wire wCMPU = (wOPA > wOPB); wire wCMPF = (rIMM[1]) ? wCMPU : ((wCMPU & ~(wOPB[31] ^ wOPA[31])) | (wOPB[31] & ~wOPA[31])); assign {wCMPC,wCMP} = {wSUBC,wCMPF,wSUB[30:0]}; assign wOPX = (rOPC[0] & !rOPC[5]) ? ~wOPA : wOPA ; assign wOPC = ((rMSR_C & rOPC[1]) | (rOPC[0] & !rOPC[1])) & (!rOPC[5] & !rMXLDST[1]); assign {wSUBC,wSUB} = {wADDC,wADD}; assign {wADDC,wADD} = (wOPB + wOPX) + wOPC; reg rRES_AC; reg [31:0] rRES_A; always @(/*AUTOSENSE*/rIMM or rOPC or wADD or wADDC or wCMP or wCMPC or wSUB or wSUBC) case ({rOPC[3],rOPC[0],rIMM[0]}) 4'h2, 4'h6, 4'h7: {rRES_AC,rRES_A} <= #1 {~wSUBC,wSUB}; // SUB 4'h3: {rRES_AC,rRES_A} <= #1 {~wCMPC,wCMP}; // CMP default: {rRES_AC,rRES_A} <= #1 {wADDC,wADD}; endcase // case ({rOPC[3],rOPC[0],rIMM[0]}) /** Logic ----- Performs all the simple logic functions in parallel with other operations. */ wire [31:0] wOR = wOPA | wOPB; wire [31:0] wAND = wOPA & wOPB; wire [31:0] wXOR = wOPA ^ wOPB; wire [31:0] wANDN = wOPA & ~wOPB; reg [31:0] rRES_L; always @(/*AUTOSENSE*/rOPC or wAND or wANDN or wOR or wXOR) case (rOPC[1:0]) 2'o0: rRES_L <= #1 wOR; 2'o1: rRES_L <= #1 wAND; 2'o2: rRES_L <= #1 wXOR; 2'o3: rRES_L <= #1 wANDN; endcase // case (rOPC[1:0]) /** Shifter ------- Performs shift instructions as well as sign extension. This is done in parallel with the other operations. */ wire wSRAC, wSRCC, wSRLC, wRES_SC; wire [31:0] wSRA,wSRC, wSRL, wSEXT8, wSEXT16, wRES_S; assign {wSRAC,wSRA} = {wOPA[0],wOPA[0],wOPA[31:1]}; assign {wSRCC,wSRC} = {wOPA[0],rMSR_C,wOPA[31:1]}; assign {wSRLC,wSRL} = {wOPA[0],1'b0,wOPA[31:1]}; assign wSEXT8 = {{(24){wOPA[7]}},wOPA[7:0]}; assign wSEXT16 = {{(16){wOPA[15]}},wOPA[15:0]}; reg rRES_SC; reg [31:0] rRES_S; always @(/*AUTOSENSE*/rIMM or rMSR_C or wSEXT16 or wSEXT8 or wSRA or wSRAC or wSRC or wSRCC or wSRL or wSRLC) case (rIMM[6:5]) 2'o0: {rRES_SC,rRES_S} <= #1 {wSRAC,wSRA}; 2'o1: {rRES_SC,rRES_S} <= #1 {wSRCC,wSRC}; 2'o2: {rRES_SC,rRES_S} <= #1 {wSRLC,wSRL}; 2'o3: {rRES_SC,rRES_S} <= #1 (rIMM[0]) ? {rMSR_C,wSEXT16} : {rMSR_C,wSEXT8}; endcase // case (rIMM[6:5]) /** Mover ----- Moves an operand from source to destination for move instructions done in parallel with other operations. */ reg [31:0] rRES_M; always @(/*AUTOSENSE*/rRA or wOPA or wOPB) rRES_M <= #1 (rRA[3]) ? wOPB : wOPA; /** Data WISHBONE Bus ----------------- Asserts the appropriate byte select signals depending on the size of the operation and the address location. For faster operation, the result of the adder is used as the end selector. FIXME: It does not check for invalid memory locations. FIXME: Endian correction! */ reg [3:0] rDWBSEL, xDWBSEL; assign dwb_adr_o = {rRESULT[DSIZ-1:2],2'b00}; assign dwb_sel_o = rDWBSEL; always @(/*AUTOSENSE*/rOPC or wADD) case (wADD[1:0]) 2'o0: case (rOPC[1:0]) 2'o0: xDWBSEL <= 4'h8; 2'o1: xDWBSEL <= 4'hC; default: xDWBSEL <= 4'hF; endcase // case (rOPC[1:0]) 2'o1: xDWBSEL <= 4'h4; 2'o2: xDWBSEL <= (rOPC[0]) ? 4'h3 : 4'h2; 2'o3: xDWBSEL <= 4'h1; endcase // case (wADD[1:0]) /** RESULT + C ---------- The RESULT and MSR[C] are collected at the end of the pipeline, depending on the operation selected. This was done in order to allow the operations to proceed in parallel for faster speed. FIXME: rMSR[C] might need to be blocked (drun) during a branch. TODO: MTS/MFS instruction */ always @(/*AUTOSENSE*/rMXALU or rRES_A or rRES_L or rRES_M or rRES_S) begin case (rMXALU) 2'o0: xRESULT <= #1 rRES_A; 2'o1: xRESULT <= #1 rRES_L; 2'o2: xRESULT <= #1 rRES_S; 2'o3: xRESULT <= #1 rRES_M; endcase // case (rMXALU) end always @(/*AUTOSENSE*/rMSR_C or rMXALU or rOPC or rRES_AC or rRES_SC) begin case (rMXALU) 2'o0: xMSR_C <= #1 (rOPC[2]) ? rMSR_C : rRES_AC; 2'o2: xMSR_C <= #1 rRES_SC; default: xMSR_C <= #1 rMSR_C; endcase // case (rMXALU) end // PIPELINE REGISTER ////////////////////////////////////////////////// always @(negedge nclk) if (prst) begin /*AUTORESET*/ // Beginning of autoreset for uninitialized flops rDWBSEL <= 4'h0; rMSR_C <= 1'h0; rRESULT <= 32'h0; // End of automatics end else if (prun) begin rRESULT <= #1 xRESULT; rMSR_C <= #1 xMSR_C; rDWBSEL <= #1 xDWBSEL; end endmodule // aeMB_aslu
Go to most recent revision | Compare with Previous | Blame | View Log