URL
https://opencores.org/ocsvn/sparc64soc/sparc64soc/trunk
Subversion Repositories sparc64soc
[/] [sparc64soc/] [trunk/] [T1-CPU/] [ifu/] [sparc_ifu_dcl.v] - Rev 7
Go to most recent revision | Compare with Previous | Blame | View Log
// ========== Copyright Header Begin ========================================== // // OpenSPARC T1 Processor File: sparc_ifu_dcl.v // Copyright (c) 2006 Sun Microsystems, Inc. All Rights Reserved. // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES. // // The above named program is free software; you can redistribute it and/or // modify it under the terms of the GNU General Public // License version 2 as published by the Free Software Foundation. // // The above named program 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 work; if not, write to the Free Software // Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. // // ========== Copyright Header End ============================================ //////////////////////////////////////////////////////////////////////// /* // Module Name: sparc_ifu_dcl // Description: // The decode control logic block does branch condition evaluation, // delay slot management, and appropriate condition code // selection. It also executes the tcc instruction and kills the E // stage instruction if a move did not succeed. The DCL block is // also responsible for generating the correct select signals to // choose the branch offset and immediate operand. // */ //////////////////////////////////////////////////////////////////////// `define CC_N 3 `define CC_Z 2 `define CC_V 1 `define CC_C 0 `define FP_U 3 `define FP_G 2 `define FP_L 1 `define FP_E 0 `define FSR_FCC0_HI 11 `define FSR_FCC0_LO 10 `define FSR_FCC1_HI 33 `define FSR_FCC1_LO 32 `define FSR_FCC2_HI 35 `define FSR_FCC2_LO 34 `define FSR_FCC3_HI 37 `define FSR_FCC3_LO 36 module sparc_ifu_dcl(/*AUTOARG*/ // Outputs ifu_exu_kill_e, ifu_exu_dontmv_regz0_e, ifu_exu_dontmv_regz1_e, ifu_exu_tcc_e, ifu_exu_dbrinst_d, ifu_ffu_mvcnd_m, dcl_fcl_bcregz0_e, dcl_fcl_bcregz1_e, dtu_inst_anull_e, dcl_swl_tcc_done_m, dcl_imd_immdata_sel_simm13_d_l, dcl_imd_immdata_sel_movcc_d_l, dcl_imd_immdata_sel_sethi_d_l, dcl_imd_immdata_sel_movr_d_l, dcl_imd_broff_sel_call_d_l, dcl_imd_broff_sel_br_d_l, dcl_imd_broff_sel_bcc_d_l, dcl_imd_broff_sel_bpcc_d_l, dcl_imd_immbr_sel_br_d, so, // Inputs rclk, se, si, dtu_reset, exu_ifu_cc_d, fcl_dcl_regz_e, exu_ifu_regn_e, ffu_ifu_cc_w2, ffu_ifu_cc_vld_w2, tlu_ifu_flush_pipe_w, swl_dcl_thr_d, swl_dcl_thr_w2, imd_dcl_brcond_d, imd_dcl_mvcond_d, fdp_dcl_op_s, fdp_dcl_op3_s, imd_dcl_abit_d, dec_dcl_cctype_d, dtu_dcl_opf2_d, fcl_dtu_inst_vld_e, fcl_dtu_intr_vld_e, ifu_tlu_flush_w ); input rclk, se, si, dtu_reset; input [7:0] exu_ifu_cc_d; // latest CCs from EXU input fcl_dcl_regz_e, // rs1=0 exu_ifu_regn_e; // rs1<0 input [7:0] ffu_ifu_cc_w2; input [3:0] ffu_ifu_cc_vld_w2; input tlu_ifu_flush_pipe_w; input [3:0] swl_dcl_thr_d, swl_dcl_thr_w2; input [3:0] imd_dcl_brcond_d; // branch condition type input [7:0] imd_dcl_mvcond_d; // mov condition type input [1:0] fdp_dcl_op_s; input [5:0] fdp_dcl_op3_s; input imd_dcl_abit_d; // anull bit for cond branch input [2:0] dec_dcl_cctype_d; // which cond codes to use input dtu_dcl_opf2_d; input fcl_dtu_inst_vld_e; input fcl_dtu_intr_vld_e; input ifu_tlu_flush_w; output ifu_exu_kill_e, ifu_exu_dontmv_regz0_e, ifu_exu_dontmv_regz1_e, ifu_exu_tcc_e; output ifu_exu_dbrinst_d; output ifu_ffu_mvcnd_m; output dcl_fcl_bcregz0_e, dcl_fcl_bcregz1_e; output dtu_inst_anull_e; output dcl_swl_tcc_done_m; output dcl_imd_immdata_sel_simm13_d_l, // imm data select dcl_imd_immdata_sel_movcc_d_l, dcl_imd_immdata_sel_sethi_d_l, dcl_imd_immdata_sel_movr_d_l; output dcl_imd_broff_sel_call_d_l, // dir branch offset select dcl_imd_broff_sel_br_d_l, dcl_imd_broff_sel_bcc_d_l, dcl_imd_broff_sel_bpcc_d_l; output dcl_imd_immbr_sel_br_d; output so; //---------------------------------------------------------------------- // Declarations //---------------------------------------------------------------------- wire [7:0] cc_breval_e, fp_breval_d; wire abit_e; wire cond_brtaken_e, anull_all, anull_ubr, anull_cbr; wire [3:0] anull_next_e, anull_e, thr_anull_d; wire inst_anull_d, inst_anull_e; wire [3:0] flush_abit; wire all_flush_w, all_flush_w2; wire br_always_e; wire sel_movcc, sel_movr; wire [3:0] br_cond_e, br_cond_d; wire [3:0] thr_vld_e; wire [3:0] ls_brcond_d, ls_brcond_e; wire [1:0] ccfp_sel; wire [3:0] cc_e; wire [1:0] curr_fcc_d; wire [7:0] fcc_d; wire [7:0] t0_fcc_d, t1_fcc_d, t2_fcc_d, t3_fcc_d, t0_fcc_nxt, t1_fcc_nxt, t2_fcc_nxt, t3_fcc_nxt; wire use_fcc0_d, use_fcc1_d, use_fcc2_d, use_fcc3_d; wire [3:0] thr_e, thr_dec_d; // fcc_dec_d, // fcc_dec_e; wire [1:0] op_d; wire [5:0] op3_d; wire use_xcc_d, ltz_e, cc_eval0, cc_eval1, fp_eval0_d, fp_eval1_d, fp_eval_d, fp_eval_e, r_eval1, r_eval0, ccfp_eval, ccbr_taken_e, mvbr_sel_br_d, cc_mvbr_d, cc_mvbr_e, fpcond_mvbr_d, fpcond_mvbr_e; wire call_inst_e, call_inst_d, dbr_inst_d, dbr_inst_e, ibr_inst_d, ibr_inst_e, mov_inst_d, mov_inst_e, tcc_done_e, tcc_inst_d, tcc_inst_e; wire clk; //---------------------------------------------------------------------- // Code start here //---------------------------------------------------------------------- assign clk = rclk; // S Stage Operands dff_s #(2) opreg(.din (fdp_dcl_op_s), .clk (clk), .q (op_d), .se (se), .si(), .so()); dff_s #(6) op3_reg(.din (fdp_dcl_op3_s), .clk (clk), .q (op3_d), .se (se), .si(), .so()); dff_s abite_reg(.din (imd_dcl_abit_d), .clk (clk), .q (abit_e), .se (se), .si(), .so()); // need to protect from scan contention dff_s #(4) thre_reg(.din (swl_dcl_thr_d), .q (thr_e), .clk (clk), .se(se), .si(), .so()); //------------------------------ // Choose correct immediate data //------------------------------ // movcc if op3 = 101100 assign dcl_imd_immdata_sel_movcc_d_l = ~(op_d[1] & op3_d[5] & ~op3_d[4] & op3_d[3] & ~op3_d[0]); // movr if op3 = 101111 //+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ // Reduced the number of terms in the eqn to help with timing // path, the result of which is that the immediate data sent to the // exu for a FLUSH instruction is INCORRECT! (It is decoded as a // MOVR). However, since our architecture completely ignores the // address of the flush, this should be ok. Confirmed with Sanjay // 03/31/03. (v1.29 -> 1.30) // ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ assign dcl_imd_immdata_sel_movr_d_l = ~(op_d[1] & op3_d[5] & op3_d[3] & op3_d[1] & op3_d[0]); // sethi if op3 = 100xx assign dcl_imd_immdata_sel_sethi_d_l = ~(~op_d[1]); // everything else assign dcl_imd_immdata_sel_simm13_d_l = ~(dcl_imd_immdata_sel_movcc_d_l & dcl_imd_immdata_sel_movr_d_l & dcl_imd_immdata_sel_sethi_d_l); //------------------------------ // Choose correct branch offset //------------------------------ // call or ld/store assign dcl_imd_broff_sel_call_d_l = ~(op_d[0]); // branch on register assign dcl_imd_broff_sel_br_d_l = ~(~op_d[0] & op3_d[4] & op3_d[3]); // branch w/o prediction assign dcl_imd_broff_sel_bcc_d_l = ~(~op_d[0] & op3_d[4] & ~op3_d[3]); // everything else assign dcl_imd_broff_sel_bpcc_d_l = ~(~op_d[0] & ~op3_d[4]); //------------------------------------ // mark branch/conditional instrctions //------------------------------------ // call assign call_inst_d = ~op_d[1] & op_d[0]; dff_s #(1) call_inste_reg(.din (call_inst_d), .clk (clk), .q (call_inst_e), .se (se), .si(), .so()); // call or branch but not nop/sethi assign dbr_inst_d = ~op_d[1] & (op_d[0] | op3_d[4] | op3_d[3]); // Choose between branch offset and immediate operand assign dcl_imd_immbr_sel_br_d = dbr_inst_d; // tell exu to use pc instead of rs1 assign ifu_exu_dbrinst_d = ~op_d[1]; dff_s #(1) dbr_inste_reg(.din (dbr_inst_d), .clk (clk), .q (dbr_inst_e), .se (se), .si(), .so()); // jmpl + return assign ibr_inst_d = op_d[1] & ~op_d[0] & op3_d[5] & op3_d[4] & op3_d[3] & ~op3_d[2] & ~op3_d[1]; dff_s #(1) ibr_inste_reg(.din (ibr_inst_d), .clk (clk), .q (ibr_inst_e), .se (se), .si(), .so()); // mov assign mov_inst_d = (op_d[1] & ~op_d[0] & op3_d[5] & ~op3_d[4] & op3_d[3] & op3_d[2] & (~op3_d[1] & ~op3_d[0] | op3_d[1] & op3_d[0])); dff_s #(1) mov_inste_reg(.din (mov_inst_d), .clk (clk), .q (mov_inst_e), .se (se), .si(), .so()); // tcc assign tcc_inst_d = op_d[1] & ~op_d[0] & op3_d[5] & op3_d[4] & op3_d[3] & ~op3_d[2] & op3_d[1] & ~op3_d[0]; dff_s #(1) tcc_inste_reg(.din (tcc_inst_d), .clk (clk), .q (tcc_inst_e), .se (se), .si(), .so()); assign mvbr_sel_br_d = ~op_d[1] & ~op_d[0] | // br op3_d[3] & ~op3_d[2] & op3_d[1] & ~op3_d[0]; // tcc assign cc_mvbr_d = ~(~op_d[1] & ~op_d[0] & op3_d[4] & op3_d[3] | // bpr op_d[1] & ~op_d[0] & op3_d[5] & ~op3_d[4] & op3_d[3] & op3_d[2] & op3_d[1] & op3_d[0] | // movr op_d[1] & ~op_d[0] & op3_d[5] & op3_d[4] & ~op3_d[3] & op3_d[2] & ~op3_d[1] & op3_d[0] & dtu_dcl_opf2_d); // fmovr //--------------------------- // FCC Logic //-------------------------- // choose current fcc assign use_fcc0_d = ~dec_dcl_cctype_d[1] & ~dec_dcl_cctype_d[0]; assign use_fcc1_d = ~dec_dcl_cctype_d[1] & dec_dcl_cctype_d[0]; assign use_fcc2_d = dec_dcl_cctype_d[1] & ~dec_dcl_cctype_d[0]; assign use_fcc3_d = dec_dcl_cctype_d[1] & dec_dcl_cctype_d[0]; mux4ds #(2) fcc_mux(.dout (curr_fcc_d[1:0]), .in0 (fcc_d[1:0]), .in1 (fcc_d[3:2]), .in2 (fcc_d[5:4]), .in3 (fcc_d[7:6]), .sel0 (use_fcc0_d), .sel1 (use_fcc1_d), .sel2 (use_fcc2_d), .sel3 (use_fcc3_d)); // decode to make next step easier // assign fcc_dec_d[0] = ~curr_fcc_d[1] & ~curr_fcc_d[0]; // assign fcc_dec_d[1] = ~curr_fcc_d[1] & curr_fcc_d[0]; // assign fcc_dec_d[2] = curr_fcc_d[1] & ~curr_fcc_d[0]; // assign fcc_dec_d[3] = curr_fcc_d[1] & curr_fcc_d[0]; // dff #(4) fcce_reg(.din (fcc_dec_d), // .q (fcc_dec_e), // .clk (clk), // .se (se), .si(), .so()); //------------------ // CC Logic for BCC //------------------ // Choose appropriate CCs // // dec_cctype is 3 bits // 10X icc // 11X xcc // 000 fcc0 // 001 fcc1 // 010 fcc2 // 011 fcc3 // assign use_xcc_d = (dec_dcl_cctype_d[2] | op3_d[3]) & dec_dcl_cctype_d[1]; assign use_xcc_d = dec_dcl_cctype_d[1]; assign fpcond_mvbr_d = ~dec_dcl_cctype_d[2] & ~tcc_inst_d; dff_s fpbr_reg(.din (fpcond_mvbr_d), .clk (clk), .q (fpcond_mvbr_e), .se (se), .si(), .so()); // mux between xcc and icc // assign cc_d = use_xcc_d ? exu_ifu_cc_d[7:4] : // xcc // exu_ifu_cc_d[3:0]; // icc // dff #(4) ccreg_e(.din (cc_d), // .clk (clk), // .q (cc_e), // .se (se), .si(), .so()); bw_u1_soffm2_4x UZsize_ccreg0_e(.d0 (exu_ifu_cc_d[0]), .d1 (exu_ifu_cc_d[4]), .s (use_xcc_d), .q (cc_e[0]), .ck (clk), .se(se), .sd(), .so()); bw_u1_soffm2_4x UZsize_ccreg1_e(.d0 (exu_ifu_cc_d[1]), .d1 (exu_ifu_cc_d[5]), .s (use_xcc_d), .q (cc_e[1]), .ck (clk), .se(se), .sd(), .so()); bw_u1_soffm2_4x UZsize_ccreg2_e(.d0 (exu_ifu_cc_d[2]), .d1 (exu_ifu_cc_d[6]), .s (use_xcc_d), .q (cc_e[2]), .ck (clk), .se(se), .sd(), .so()); bw_u1_soffm2_4x UZsize_ccreg3_e(.d0 (exu_ifu_cc_d[3]), .d1 (exu_ifu_cc_d[7]), .s (use_xcc_d), .q (cc_e[3]), .ck (clk), .se(se), .sd(), .so()); //------------------------------ // Evaluate Branch //------------------------------ // Select correct branch condition assign sel_movcc = ~mvbr_sel_br_d & cc_mvbr_d; assign sel_movr = ~mvbr_sel_br_d & ~cc_mvbr_d; // br_cond is the same as the "cond" field = inst[28:25] for bcc mux3ds #(4) brcond_mux(.dout (br_cond_d), .in0 (imd_dcl_brcond_d), // br or tcc .in1 (imd_dcl_mvcond_d[7:4]), // movcc .in2 (imd_dcl_mvcond_d[3:0]), // movr .sel0 (mvbr_sel_br_d), .sel1 (sel_movcc), .sel2 (sel_movr)); dff_s #(4) brcond_e_reg(.din (br_cond_d), .clk (clk), .q (br_cond_e), .se (se), .si(), .so()); // Branch Type Decode assign ls_brcond_d[0] = ~br_cond_d[1] & ~br_cond_d[0]; assign ls_brcond_d[1] = ~br_cond_d[1] & br_cond_d[0]; assign ls_brcond_d[2] = br_cond_d[1] & ~br_cond_d[0]; assign ls_brcond_d[3] = br_cond_d[1] & br_cond_d[0]; dff_s #(4) lsbrc_e_reg(.din (ls_brcond_d), .clk (clk), .q (ls_brcond_e), .se (se), .si(), .so()); // Evaluate potential integer CC branches assign ltz_e = (cc_e[`CC_N] ^ cc_e[`CC_V]); assign cc_breval_e[0] = 1'b0; // BPN assign cc_breval_e[1] = cc_e[`CC_Z]; // BPE assign cc_breval_e[2] = cc_e[`CC_Z] | ltz_e; // BPLE assign cc_breval_e[3] = ltz_e; // BPL assign cc_breval_e[4] = cc_e[`CC_Z] | cc_e[`CC_C]; // BPLEU assign cc_breval_e[5] = cc_e[`CC_C]; // BPCS assign cc_breval_e[6] = cc_e[`CC_N]; // BPNEG assign cc_breval_e[7] = cc_e[`CC_V]; // BPVS // mux to choose right condition assign cc_eval0 = cc_breval_e[0] & ls_brcond_e[0] | cc_breval_e[1] & ls_brcond_e[1] | cc_breval_e[2] & ls_brcond_e[2] | cc_breval_e[3] & ls_brcond_e[3]; assign cc_eval1 = cc_breval_e[4] & ls_brcond_e[0] | cc_breval_e[5] & ls_brcond_e[1] | cc_breval_e[6] & ls_brcond_e[2] | cc_breval_e[7] & ls_brcond_e[3]; // Evaluate FP CC branches in D stage assign fp_breval_d[0] = 1'b0; // FBN / A assign fp_breval_d[1] = (curr_fcc_d[1] | curr_fcc_d[0]); // FBNE / E assign fp_breval_d[2] = curr_fcc_d[1] ^ curr_fcc_d[0]; // FBLG / UE assign fp_breval_d[3] = curr_fcc_d[0]; // FBUL / GE assign fp_breval_d[4] = ~curr_fcc_d[1] & curr_fcc_d[0]; // FBL / UGE assign fp_breval_d[5] = curr_fcc_d[1]; // FBUG / LE assign fp_breval_d[6] = curr_fcc_d[1] & ~curr_fcc_d[0]; // FBG / ULE assign fp_breval_d[7] = curr_fcc_d[1] & curr_fcc_d[0]; // FBU / O assign fp_eval0_d = fp_breval_d[0] & ls_brcond_d[0] | fp_breval_d[1] & ls_brcond_d[1] | fp_breval_d[2] & ls_brcond_d[2] | fp_breval_d[3] & ls_brcond_d[3]; assign fp_eval1_d = fp_breval_d[4] & ls_brcond_d[0] | fp_breval_d[5] & ls_brcond_d[1] | fp_breval_d[6] & ls_brcond_d[2] | fp_breval_d[7] & ls_brcond_d[3]; assign fp_eval_d = br_cond_d[2] ? fp_eval1_d : fp_eval0_d; dff_s #(1) fpev_ff(.din (fp_eval_d), .q (fp_eval_e), .clk (clk), .se (se), .si(), .so()); // merge eval0, eval1 and fp condition codes assign ccfp_sel[0] = ~fpcond_mvbr_e & ~br_cond_e[2]; assign ccfp_sel[1] = ~fpcond_mvbr_e & br_cond_e[2]; // assign ccfp_sel[2] = fpcond_mvbr_e & ~br_cond_e[2]; // assign ccfp_sel[3] = fpcond_mvbr_e & br_cond_e[2]; assign ccfp_eval = ccfp_sel[0] & cc_eval0 | ccfp_sel[1] & cc_eval1 | fpcond_mvbr_e & fp_eval_e; // invert branch condition if this is an inverted br type // assign ccbr_taken_e = (ccfp_eval ^ br_cond_e[3]) & cc_mvbr_e; assign ccbr_taken_e = ccfp_eval ? (cc_mvbr_e & ~br_cond_e[3]) : (cc_mvbr_e & br_cond_e[3]); assign br_always_e = (~br_cond_e[0] & ~br_cond_e[1] & ~br_cond_e[2] & br_cond_e[3] & cc_mvbr_e); //-------------- // For BRZ // ------------- // Calculate Cond Assuming Z=1 And Z=0. Then Mux // assign r_eval1 = ((exu_ifu_regn_e | ~br_cond_e[1] | ~br_cond_e[0]) ^ // br_cond_e[2]) & ~cc_mvbr_e; assign r_eval1 = exu_ifu_regn_e ? (~br_cond_e[2] & ~cc_mvbr_e) : (((br_cond_e[1] & br_cond_e[0]) ^ ~br_cond_e[2]) & ~cc_mvbr_e); // assign r_eval0 = ((exu_ifu_regn_e & br_cond_e[1]) ^ // br_cond_e[2]) & ~cc_mvbr_e; assign r_eval0 = exu_ifu_regn_e ? ((br_cond_e[1] ^ br_cond_e[2]) & ~cc_mvbr_e) : (br_cond_e[2] & ~cc_mvbr_e); dff_s #(1) regcc_ff(.din (cc_mvbr_d), .clk (clk), .q (cc_mvbr_e), .se (se), .si(), .so()); // Evaluate Final Branch condition // 3:1 mux // assign cond_brtaken_e = cc_mvbr_e ? ccbr_taken_e : // exu_ifu_regz_e ? r_eval1 : // r_eval0; // 2:1 mux // assign cond_brtaken_e = exu_ifu_regz_e ? (r_eval1 | ccbr_taken_e) : // (r_eval0 | ccbr_taken_e); //////// Chandra //////// wire temp0, temp1, cond_brtaken_e_l; // limit loading on this signal // wire regz_buf_e; // bw_u1_buf_5x UZfix_regz_bf(.a (exu_ifu_regz_e), // .z (regz_buf_e)); assign temp0 = (r_eval0 | ccbr_taken_e); assign temp1 = (r_eval1 | ccbr_taken_e); bw_u1_muxi21_6x UZsize_cbtmux(.z(cond_brtaken_e_l), .d0(temp0), .d1(temp1), .s(fcl_dcl_regz_e)); bw_u1_inv_20x UZsize_cbtinv(.z(cond_brtaken_e), .a(cond_brtaken_e_l)); //////////////////////// assign dcl_fcl_bcregz0_e = (temp0 & dbr_inst_e | ibr_inst_e | call_inst_e) & ~dtu_inst_anull_e; assign dcl_fcl_bcregz1_e = (temp1 & dbr_inst_e | ibr_inst_e | call_inst_e) & ~dtu_inst_anull_e; // assign ifu_exu_dontmove_e = mov_inst_e & ~cond_brtaken_e; assign ifu_exu_dontmv_regz0_e = ~temp0 & mov_inst_e; assign ifu_exu_dontmv_regz1_e = ~temp1 & mov_inst_e; // branch condition to FPU dff_s #(1) fpcond_ff(.din (cond_brtaken_e), .q (ifu_ffu_mvcnd_m), .clk (clk), .se (se), .si(), .so()); // branch / move completion and anull signals // assign dtu_fcl_brtaken_e = ~dtu_inst_anull_e & // (ibr_inst_e | call_inst_e | // dbr_inst_e & cond_brtaken_e); // if mov didn't succeed kill write back and bypass // need to check thread as well // assign ifu_exu_kill_e = dtu_inst_anull_e | // ~fcl_dtu_inst_vld_e; // don't need this anymore assign ifu_exu_kill_e = dtu_inst_anull_e; // signal trap if tcc succeeds assign ifu_exu_tcc_e = ~dtu_inst_anull_e & tcc_inst_e & ccbr_taken_e & fcl_dtu_inst_vld_e; assign tcc_done_e = ~dtu_inst_anull_e & tcc_inst_e & ~ccbr_taken_e & fcl_dtu_inst_vld_e; dff_s #(1) tccm_ff(.din (tcc_done_e), .q (dcl_swl_tcc_done_m), .clk (clk), .se (se), .si(), .so()); // logic to anull delay slot, if this branch itsel is not anulled assign anull_cbr = abit_e & dbr_inst_e & ~br_always_e & ~call_inst_e; assign anull_ubr = abit_e & dbr_inst_e & br_always_e & ~call_inst_e; assign anull_all = anull_ubr | anull_cbr & ~cond_brtaken_e; // check which thread to anull assign thr_vld_e = thr_e & {4{fcl_dtu_inst_vld_e}}; assign all_flush_w = tlu_ifu_flush_pipe_w | ifu_tlu_flush_w; dff_s #(1) flshw2_ff(.din (all_flush_w), .q (all_flush_w2), .clk (clk), .se(se), .si(), .so()); assign flush_abit = swl_dcl_thr_w2 & {4{all_flush_w2}}; assign anull_next_e = ((~anull_e & {4{anull_all}} & thr_vld_e) | (anull_e & ~(thr_e & {4{fcl_dtu_inst_vld_e | fcl_dtu_intr_vld_e}}))) & ~flush_abit; // anull_e needs to be per thread dffr_s #(4) anull_ff(.din (anull_next_e), .clk (clk), .rst (dtu_reset), .q (anull_e), .se (se), .si(), .so()); // // assign thr_dec_e[0] = swl_dcl_thr_e[0] | rst_tri_enable; // assign thr_dec_e[3:1] = swl_dcl_thr_e[3:1] & {3{~rst_tri_enable}}; assign thr_anull_d = swl_dcl_thr_d & anull_next_e; assign inst_anull_d = (|thr_anull_d[3:0]); dff_s #(1) ina_ff(.din (inst_anull_d), .q (inst_anull_e), .clk (clk), .se (se), .si(), .so()); assign dtu_inst_anull_e = inst_anull_e; // mux4ds dcla_mux(.dout (this_inst_anull_e), // .in0 (anull_e[0]), // .in1 (anull_e[1]), // .in2 (anull_e[2]), // .in3 (anull_e[3]), // .sel0 (thr_dec_e[0]), // .sel1 (thr_dec_e[1]), // .sel2 (thr_dec_e[2]), // .sel3 (thr_dec_e[3])); // assign dtu_inst_anull_e = this_inst_anull_e & fcl_dtu_inst_vld_e; //-------------------- // Copy of FCC //-------------------- // FCC's are maintained in the ffu. A copy is kept here to run the // FP branch instructions. // load FCC from FFU mux2ds #(8) t0_fcc_mux(.dout (t0_fcc_nxt[7:0]), .in0 (t0_fcc_d[7:0]), .in1 (ffu_ifu_cc_w2[7:0]), .sel0 (~ffu_ifu_cc_vld_w2[0]), .sel1 (ffu_ifu_cc_vld_w2[0])); dffr_s #(8) t0_fcc_reg(.din (t0_fcc_nxt[7:0]), .q (t0_fcc_d[7:0]), .rst (dtu_reset), .clk (clk), .se (se), .si(), .so()); `ifdef FPGA_SYN_1THREAD assign fcc_d[7:0] = t0_fcc_d[7:0]; `else mux2ds #(8) t1_fcc_mux(.dout (t1_fcc_nxt[7:0]), .in0 (t1_fcc_d[7:0]), .in1 (ffu_ifu_cc_w2[7:0]), .sel0 (~ffu_ifu_cc_vld_w2[1]), .sel1 (ffu_ifu_cc_vld_w2[1])); mux2ds #(8) t2_fcc_mux(.dout (t2_fcc_nxt[7:0]), .in0 (t2_fcc_d[7:0]), .in1 (ffu_ifu_cc_w2[7:0]), .sel0 (~ffu_ifu_cc_vld_w2[2]), .sel1 (ffu_ifu_cc_vld_w2[2])); mux2ds #(8) t3_fcc_mux(.dout (t3_fcc_nxt[7:0]), .in0 (t3_fcc_d[7:0]), .in1 (ffu_ifu_cc_w2[7:0]), .sel0 (~ffu_ifu_cc_vld_w2[3]), .sel1 (ffu_ifu_cc_vld_w2[3])); // thread0 fcc registers dffr_s #(8) t1_fcc_reg(.din (t1_fcc_nxt[7:0]), .q (t1_fcc_d[7:0]), .rst (dtu_reset), .clk (clk), .se (se), .si(), .so()); dffr_s #(8) t2_fcc_reg(.din (t2_fcc_nxt[7:0]), .q (t2_fcc_d[7:0]), .rst (dtu_reset), .clk (clk), .se (se), .si(), .so()); dffr_s #(8) t3_fcc_reg(.din (t3_fcc_nxt[7:0]), .q (t3_fcc_d[7:0]), .rst (dtu_reset), .clk (clk), .se (se), .si(), .so()); // choose thread assign thr_dec_d[0] = swl_dcl_thr_d[0]; assign thr_dec_d[3:1] = swl_dcl_thr_d[3:1]; mux4ds #(8) fcc0d_mx(.dout (fcc_d[7:0]), .in0 (t0_fcc_d[7:0]), .in1 (t1_fcc_d[7:0]), .in2 (t2_fcc_d[7:0]), .in3 (t3_fcc_d[7:0]), .sel0 (thr_dec_d[0]), .sel1 (thr_dec_d[1]), .sel2 (thr_dec_d[2]), .sel3 (thr_dec_d[3])); `endif // !`ifdef FPGA_SYN_1THREAD endmodule // sparc_ifu_dcl
Go to most recent revision | Compare with Previous | Blame | View Log