// ========== Copyright Header Begin ========================================== // // OpenSPARC T1 Processor File: sparc_exu_rml.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_exu_rml // Description: Register management logic. Contains CWP, CANSAVE, CANRESTORE // and other window management registers. Generates RF related traps // and switches the global registers to alternate globals. All the registers // are written in the W stage (there is no bypassing so they must // swap out) and will either get a new value generated by a window management // Instruction or by a WRPS instruction. The following traps can be generated: // Fill: restore with canrestore == 0 // clean_window: save with cleanwin-canrestore == 0 // spill: flushw with cansave != nwindows -2 or // save with cansave == 0 // It is assumed that the contents of the new window will get squashed // on a clean_window or fill trap so the save or restore gets executed // normally. Spill traps or WRCWPs mean that all 16 windowed registers // must be saved and restored (a 4 cycle operation). */ module sparc_exu_rml (/*AUTOARG*/ // Outputs exu_tlu_spill_wtype, exu_tlu_spill_other, exu_tlu_cwp_retry, exu_tlu_cwp3_w, exu_tlu_cwp2_w, exu_tlu_cwp1_w, exu_tlu_cwp0_w, so, exu_tlu_cwp_cmplt, exu_tlu_cwp_cmplt_tid, rml_ecl_cwp_d, rml_ecl_cansave_d, rml_ecl_canrestore_d, rml_ecl_otherwin_d, rml_ecl_wstate_d, rml_ecl_cleanwin_d, rml_ecl_fill_e, rml_ecl_clean_window_e, rml_ecl_other_e, rml_ecl_wtype_e, exu_ifu_spill_e, rml_ecl_gl_e, rml_irf_old_lo_cwp_e, rml_irf_new_lo_cwp_e, rml_irf_old_e_cwp_e, rml_irf_new_e_cwp_e, rml_irf_swap_even_e, rml_irf_swap_odd_e, rml_irf_swap_local_e, rml_irf_kill_restore_w, rml_irf_cwpswap_tid_e, rml_ecl_swap_done, rml_ecl_rmlop_done_e, exu_ifu_oddwin_s, exu_tlu_spill, exu_tlu_spill_tid, rml_ecl_kill_m, rml_irf_old_agp, rml_irf_new_agp, rml_irf_swap_global, rml_irf_global_tid, // Inputs tlu_exu_cwp_retry_m, rst_tri_en, rclk, se, si, grst_l, arst_l, ifu_exu_tid_s2, ifu_exu_save_d, ifu_exu_restore_d, ifu_exu_saved_e, ifu_exu_restored_e, ifu_exu_flushw_e, ecl_rml_thr_m, ecl_rml_thr_w, ecl_rml_cwp_wen_e, ecl_rml_cansave_wen_w, ecl_rml_canrestore_wen_w, ecl_rml_otherwin_wen_w, ecl_rml_wstate_wen_w, ecl_rml_cleanwin_wen_w, ecl_rml_xor_data_e, ecl_rml_kill_e, ecl_rml_kill_w, ecl_rml_early_flush_w, exu_tlu_wsr_data_w, tlu_exu_agp, tlu_exu_agp_swap, tlu_exu_agp_tid, tlu_exu_cwp_m, tlu_exu_cwpccr_update_m, ecl_rml_inst_vld_w,current_cwp ) ; input rclk; input se; input si; input grst_l; input arst_l; input [1:0] ifu_exu_tid_s2; input ifu_exu_save_d; input ifu_exu_restore_d; input ifu_exu_saved_e; input ifu_exu_restored_e; input ifu_exu_flushw_e; input [3:0] ecl_rml_thr_m; input [3:0] ecl_rml_thr_w; input ecl_rml_cwp_wen_e; input ecl_rml_cansave_wen_w; input ecl_rml_canrestore_wen_w; input ecl_rml_otherwin_wen_w; input ecl_rml_wstate_wen_w; input ecl_rml_cleanwin_wen_w; input [2:0] ecl_rml_xor_data_e; input ecl_rml_kill_e;// needed for oddwin updates input ecl_rml_kill_w; input ecl_rml_early_flush_w; input [5:0] exu_tlu_wsr_data_w; // for wstate input [1:0] tlu_exu_agp; // alternate global pointer input tlu_exu_agp_swap;// switch globals input [1:0] tlu_exu_agp_tid;// thread that agp refers to input [2:0] tlu_exu_cwp_m; // for switching cwp on return from trap input tlu_exu_cwpccr_update_m; input ecl_rml_inst_vld_w; /*AUTOINPUT*/ // Beginning of automatic inputs (from unused autoinst inputs) input rst_tri_en; // To cwp of sparc_exu_rml_cwp.v input tlu_exu_cwp_retry_m; // To cwp of sparc_exu_rml_cwp.v // End of automatics /*AUTOOUTPUT*/ // Beginning of automatic outputs (from unused autoinst outputs) output [2:0] exu_tlu_cwp0_w; // From cwp of sparc_exu_rml_cwp.v output [2:0] exu_tlu_cwp1_w; // From cwp of sparc_exu_rml_cwp.v output [2:0] exu_tlu_cwp2_w; // From cwp of sparc_exu_rml_cwp.v output [2:0] exu_tlu_cwp3_w; // From cwp of sparc_exu_rml_cwp.v output exu_tlu_cwp_retry; // From cwp of sparc_exu_rml_cwp.v output exu_tlu_spill_other; // From cwp of sparc_exu_rml_cwp.v output [2:0] exu_tlu_spill_wtype; // From cwp of sparc_exu_rml_cwp.v // End of automatics output so; output exu_tlu_cwp_cmplt; output [1:0] exu_tlu_cwp_cmplt_tid; output [2:0] rml_ecl_cwp_d; output [2:0] rml_ecl_cansave_d; output [2:0] rml_ecl_canrestore_d; output [2:0] rml_ecl_otherwin_d; output [5:0] rml_ecl_wstate_d; output [2:0] rml_ecl_cleanwin_d; output rml_ecl_fill_e; output rml_ecl_clean_window_e; output rml_ecl_other_e; output [2:0] rml_ecl_wtype_e; output exu_ifu_spill_e; output [1:0] rml_ecl_gl_e; output [2:0] rml_irf_old_lo_cwp_e; // current window pointer for locals and odds output [2:0] rml_irf_new_lo_cwp_e; // current window pointer for locals and odd output [1:0] rml_irf_old_e_cwp_e; // current window pointer for evens output [1:0] rml_irf_new_e_cwp_e; // current window pointer for evens output rml_irf_swap_even_e; output rml_irf_swap_odd_e; output rml_irf_swap_local_e; output rml_irf_kill_restore_w; output [1:0] rml_irf_cwpswap_tid_e; output [3:0] rml_ecl_swap_done; output rml_ecl_rmlop_done_e; output [3:0] exu_ifu_oddwin_s; output exu_tlu_spill; output [1:0] exu_tlu_spill_tid; output rml_ecl_kill_m; output [1:0] rml_irf_old_agp; // alternate global pointer output [1:0] rml_irf_new_agp; // alternate global pointer output rml_irf_swap_global; output [1:0] rml_irf_global_tid; output reg [11:0] current_cwp; wire clk; wire [1:0] tid_d; wire [3:0] thr_d; wire [1:0] tid_e; wire rml_reset_l; wire reset; wire save_e; wire save_m; wire restore_e; wire swap_e; wire agp_wen; wire [1:0] agp_thr0; wire [1:0] agp_thr1; wire [1:0] agp_thr2; wire [1:0] agp_thr3; wire [1:0] agp_thr0_next; wire [1:0] agp_thr1_next; wire [1:0] agp_thr2_next; wire [1:0] agp_thr3_next; wire agp_wen_thr0_w; wire agp_wen_thr1_w; wire agp_wen_thr2_w; wire agp_wen_thr3_w; wire [1:0] new_agp; wire [1:0] agp_tid; wire [3:0] agp_thr; wire full_swap_e; wire did_restore_m; wire did_restore_w; wire kill_restore_m; wire kill_restore_w; wire [2:0] rml_ecl_cwp_e; wire [2:0] rml_ecl_cansave_e; wire [2:0] rml_ecl_canrestore_e; wire [2:0] rml_ecl_otherwin_e; wire [2:0] rml_ecl_cleanwin_e; wire [2:0] rml_next_cwp_e; wire [2:0] rml_next_cansave_e;// e-stage of rml generated new data wire [2:0] rml_next_canrestore_e; wire [2:0] rml_next_otherwin_e; wire [2:0] rml_next_cleanwin_e; wire [2:0] next_cwp_e; wire [2:0] next_cansave_e; // e-stage of new data wire [2:0] next_canrestore_e; wire [2:0] next_otherwin_e; wire [2:0] next_cleanwin_e; wire [2:0] next_cwp_m; // m-stage of new data wire [2:0] next_cansave_m; wire [2:0] next_canrestore_m; wire [2:0] next_otherwin_m; wire [2:0] next_cleanwin_m; wire [2:0] next_cansave_w;// w-stage of new data wire [2:0] next_canrestore_w; wire [2:0] next_otherwin_w; wire [2:0] next_cleanwin_w; wire [2:0] next_cwp_noreset_w; wire [2:0] next_cwp_w; wire rml_cwp_wen_e; // wen for cwp from rml wire rml_cwp_wen_m; // wen for cwp from rml wire [2:0] spill_cwp_e; // next cwp if there is a spill trap wire spill_cwp_carry0; // carry bit from spill cwp computations wire spill_cwp_carry1; wire next_cwp_sel_inc; // select line to next_cwp mux wire rml_cansave_wen_w;// rml generated wen wire rml_canrestore_wen_w; wire rml_otherwin_wen_w; wire rml_cleanwin_wen_w; wire cansave_wen_w;// wen to registers wire canrestore_wen_w; wire otherwin_wen_w; wire cleanwin_wen_w; wire cwp_wen_nokill_w; wire cwp_wen_w; wire wstate_wen_w; wire cwp_wen_m; // rml generated wen w/o kills wire cansave_wen_m; wire canrestore_wen_m; wire otherwin_wen_m; wire cleanwin_wen_m; wire cansave_wen_valid_m; // rml generated wen w/ kills wire canrestore_wen_valid_m; wire otherwin_wen_valid_m; wire cleanwin_wen_valid_m; wire cwp_wen_e; // rml generated wen_e wire cansave_wen_e; wire canrestore_wen_e; wire otherwin_wen_e; wire cleanwin_wen_e; wire cansave_inc_e; wire canrestore_inc_e; wire spill_trap_save; wire spill_trap_flush; wire spill_m; wire [2:0] cleanwin_xor_canrestore; wire otherwin_is0_e; wire cansave_is0_e; wire canrestore_is0_e; wire swap_locals_ins; wire swap_outs; wire [2:0] old_cwp_e; wire [2:0] new_cwp_e; wire [2:0] rml_ecl_wtype_d; wire [2:0] rml_ecl_wtype_e; wire rml_ecl_other_d; wire rml_ecl_other_e; wire exu_tlu_spill_e; wire rml_ecl_kill_e; wire rml_kill_w; wire vld_w; wire win_trap_e; wire win_trap_m; wire win_trap_w; assign clk = rclk; // Reset flop dffrl_async rstff(.din (grst_l), .q (rml_reset_l), .clk (clk), .rst_l (arst_l), .se(se), .si(), .so()); assign reset = ~rml_reset_l; dff_s #(2) tid_s2d(.din(ifu_exu_tid_s2[1:0]), .clk(clk), .q(tid_d[1:0]), .se(se), .si(), .so()); dff_s #(2) tid_d2e(.din(tid_d[1:0]), .clk(clk), .q(tid_e[1:0]), .se(se), .si(), .so()); assign thr_d[3] = tid_d[1] & tid_d[0]; assign thr_d[2] = tid_d[1] & ~tid_d[0]; assign thr_d[1] = ~tid_d[1] & tid_d[0]; assign thr_d[0] = ~tid_d[1] & ~tid_d[0]; dff_s save_d2e(.din(ifu_exu_save_d), .clk(clk), .q(save_e), .se(se), .si(), .so()); dff_s save_e2m(.din(save_e), .clk(clk), .q(save_m), .se(se), .si(), .so()); dff_s restore_d2e(.din(ifu_exu_restore_d), .clk(clk), .q(restore_e), .se(se), .si(), .so()); // don't check flush_pipe in w if caused by rml trap. Things with a higher priority // than a window trap have been accumulated into ecl_rml_kill_w assign vld_w = ecl_rml_inst_vld_w & (~ecl_rml_early_flush_w | win_trap_w); assign rml_kill_w = ecl_rml_kill_w | ~vld_w; assign win_trap_e = rml_ecl_fill_e | exu_tlu_spill_e | rml_ecl_clean_window_e; dff_s win_trap_e2m(.din(win_trap_e), .clk(clk), .q(win_trap_m), .se(se), .si(), .so()); dff_s win_trap_m2w(.din(win_trap_m), .clk(clk), .q(win_trap_w), .se(se), .si(), .so()); assign canrestore_is0_e = (~rml_ecl_canrestore_e[0] & ~rml_ecl_canrestore_e[1] & ~rml_ecl_canrestore_e[2]); assign cansave_is0_e = (~rml_ecl_cansave_e[0] & ~rml_ecl_cansave_e[1] & ~rml_ecl_cansave_e[2]); assign otherwin_is0_e = ~rml_ecl_other_e; /////////////////////////////////////// // Signals that operations are done // restore/return is not signalled here // because it depends on the write to the // irf (computed in ecl_wb) //////////////////////////////////////// assign rml_ecl_rmlop_done_e = (ifu_exu_saved_e | ifu_exu_restored_e | (ifu_exu_flushw_e & ~spill_trap_flush)); ////////////////////////// // Trap generation ////////////////////////// // Fill trap generated on restore and canrestore == 0 assign rml_ecl_fill_e = restore_e & canrestore_is0_e; // Spill trap on save with cansave == 0 assign spill_trap_save = save_e & cansave_is0_e; assign exu_ifu_spill_e = spill_trap_save; // Spill trap on wflush with cansave != (NWINDOWS - 2 = 6) assign spill_trap_flush = (ifu_exu_flushw_e & ~(rml_ecl_cansave_e[2] & rml_ecl_cansave_e[1] & ~rml_ecl_cansave_e[0])); assign exu_tlu_spill_e = (spill_trap_save | spill_trap_flush); dff_s spill_e2m(.din(exu_tlu_spill_e), .clk(clk), .q(spill_m), .se(se), .si(), .so()); // Clean window trap on save w/ cleanwin - canrestore == 0 // or cleanwin == canrestore // (not signalled on spill traps because spill is higher priority) assign cleanwin_xor_canrestore = rml_ecl_cleanwin_e ^ rml_ecl_canrestore_e; assign rml_ecl_clean_window_e = ~(cleanwin_xor_canrestore[2] | cleanwin_xor_canrestore[1] | cleanwin_xor_canrestore[0]) & save_e & ~exu_tlu_spill_e; // Kill signal for w1 wen bit (all others don't care) assign rml_ecl_kill_e = rml_ecl_fill_e | exu_tlu_spill_e; dff_s rml_kill_e2m(.din(rml_ecl_kill_e), .clk(clk), .q(rml_ecl_kill_m), .se(se), .si(), .so()); // WTYPE generation assign rml_ecl_other_d = (rml_ecl_otherwin_d[0] | rml_ecl_otherwin_d[1] | rml_ecl_otherwin_d[2]); dff_s other_d2e(.din(rml_ecl_other_d), .clk(clk), .q(rml_ecl_other_e), .se(se), .si(), .so()); mux2ds #(3) wtype_mux(.dout(rml_ecl_wtype_d[2:0]), .in0(rml_ecl_wstate_d[2:0]), .in1(rml_ecl_wstate_d[5:3]), .sel0(~rml_ecl_other_d), .sel1(rml_ecl_other_d)); dff_s #(3) wtype_d2e(.din(rml_ecl_wtype_d[2:0]), .clk(clk), .q(rml_ecl_wtype_e[2:0]), .se(se), .si(), .so()); //////////////////////////// // Interface with IRF //////////////////////////// assign rml_irf_old_lo_cwp_e[2:0] = old_cwp_e[2:0]; assign rml_irf_new_lo_cwp_e[2:0] = new_cwp_e[2:0]; assign rml_irf_old_e_cwp_e[1:0] = (old_cwp_e[0])? old_cwp_e[2:1] + 2'b01: old_cwp_e[2:1]; assign rml_irf_new_e_cwp_e[1:0] = (new_cwp_e[0])? new_cwp_e[2:1] + 2'b01: new_cwp_e[2:1]; assign rml_irf_swap_local_e = (swap_e | swap_locals_ins); assign rml_irf_swap_odd_e = ((save_e | ecl_rml_cwp_wen_e | spill_trap_flush | swap_locals_ins) & old_cwp_e[0]) | ((restore_e | swap_outs) & ~old_cwp_e[0]); assign rml_irf_swap_even_e = ((save_e | ecl_rml_cwp_wen_e | spill_trap_flush | swap_locals_ins) & ~old_cwp_e[0]) | ((restore_e | swap_outs) & old_cwp_e[0]); assign swap_e = save_e | restore_e | ecl_rml_cwp_wen_e | spill_trap_flush; dff_s dff_did_restore_e2m(.din(swap_e), .clk(clk), .q(did_restore_m), .se(se), .si(), .so()); dff_s dff_did_restore_m2w(.din(did_restore_m), .clk(clk), .q(did_restore_w), .se(se), .si(), .so()); // kill restore on all saves (except those that spill) and any swaps that // get kill signals assign kill_restore_m = (~spill_m & save_m); dff_s dff_kill_restore_m2w(.din(kill_restore_m), .clk(clk), .q(kill_restore_w), .se(se), .si(), .so()); assign rml_irf_kill_restore_w = kill_restore_w | (did_restore_w & rml_kill_w); /////////////////////////////// // CWP logic /////////////////////////////// // Logic to compute next_cwp on spill trap. // CWP = CWP + CANSAVE + 2 assign spill_cwp_e[0] = rml_ecl_cwp_e[0] ^ rml_ecl_cansave_e[0]; assign spill_cwp_carry0 = rml_ecl_cwp_e[0] & rml_ecl_cansave_e[0]; assign spill_cwp_e[1] = rml_ecl_cwp_e[1] ^ rml_ecl_cansave_e[1] ^ ~spill_cwp_carry0; assign spill_cwp_carry1 = (rml_ecl_cwp_e[1] | rml_ecl_cansave_e[1] | spill_cwp_carry0) & ~(rml_ecl_cwp_e[1] & rml_ecl_cansave_e[1] & spill_cwp_carry0); assign spill_cwp_e[2] = rml_ecl_cwp_e[2] ^ rml_ecl_cansave_e[2] ^ spill_cwp_carry1; assign rml_cwp_wen_e = (save_e | restore_e) & ~exu_tlu_spill_e; assign cwp_wen_e = (rml_cwp_wen_e | ecl_rml_cwp_wen_e) & ~ecl_rml_kill_e; sparc_exu_rml_inc3 cwp_inc(.dout(rml_next_cwp_e[2:0]), .din(rml_ecl_cwp_e[2:0]), .inc(save_e)); assign next_cwp_sel_inc = ~(ecl_rml_cwp_wen_e | exu_tlu_spill_e); mux3ds #(3) next_cwp_mux(.dout(next_cwp_e[2:0]), .in0(rml_next_cwp_e[2:0]), .in1(ecl_rml_xor_data_e[2:0]), .in2(spill_cwp_e[2:0]), .sel0(next_cwp_sel_inc), .sel1(ecl_rml_cwp_wen_e), .sel2(exu_tlu_spill_e)); dff_s cwp_wen_e2m(.din(cwp_wen_e), .clk(clk), .q(rml_cwp_wen_m), .se(se), .si(), .so()); dff_s #(3) next_cwp_e2m(.din(next_cwp_e[2:0]), .clk(clk), .q(next_cwp_m[2:0]), .se(se), .si(), .so()); assign cwp_wen_m = rml_cwp_wen_m; dff_s #(3) next_cwp_m2w(.din(next_cwp_m[2:0]), .clk(clk), .q(next_cwp_noreset_w[2:0]), .se(se), .si(), .so()); dff_s cwp_wen_m2w(.din(cwp_wen_m), .clk(clk), .q(cwp_wen_nokill_w), .se(se), .si(), .so()); assign cwp_wen_w = cwp_wen_nokill_w & ~rml_kill_w; assign next_cwp_w[2:0] = next_cwp_noreset_w[2:0]; assign full_swap_e = (exu_tlu_spill_e | ecl_rml_cwp_wen_e); // oddwin signal for ifu needs bypass from w. It is done in M and staged for timing. // This is possible because the thread is switched out so there is only one bypass condition. // Only save/return will switch in fast enough for a bypass so this is the only write condition // we need to check wire [3:0] oddwin_m; wire [3:0] oddwin_w; assign oddwin_m[3] = (cwp_wen_m & ecl_rml_thr_m[3])? next_cwp_m[0]: oddwin_w[3]; assign oddwin_m[2] = (cwp_wen_m & ecl_rml_thr_m[2])? next_cwp_m[0]: oddwin_w[2]; assign oddwin_m[1] = (cwp_wen_m & ecl_rml_thr_m[1])? next_cwp_m[0]: oddwin_w[1]; assign oddwin_m[0] = (cwp_wen_m & ecl_rml_thr_m[0])? next_cwp_m[0]: oddwin_w[0]; dff_s #(4) oddwin_dff(.din(oddwin_m[3:0]), .clk(clk), .q(exu_ifu_oddwin_s[3:0]), .se(se), .si(), .so()); integer i; wire [11:0] next_cwp; always @(posedge clk) begin current_cwp[2:0]<=(cwp_wen_m & ecl_rml_thr_m[0])? next_cwp_m: next_cwp[2:0]; current_cwp[5:3]<=(cwp_wen_m & ecl_rml_thr_m[1])? next_cwp_m: next_cwp[5:3]; current_cwp[8:6]<=(cwp_wen_m & ecl_rml_thr_m[2])? next_cwp_m: next_cwp[8:6]; current_cwp[11:9]<=(cwp_wen_m & ecl_rml_thr_m[3])? next_cwp_m: next_cwp[11:9]; end sparc_exu_rml_cwp cwp( .swap_outs (swap_outs), .swap_locals_ins(swap_locals_ins), .rml_ecl_cwp_e (rml_ecl_cwp_e[2:0]), .old_cwp_e (old_cwp_e[2:0]), .new_cwp_e (new_cwp_e[2:0]), .oddwin_w (oddwin_w[3:0]), .next_cwp (next_cwp), /*AUTOINST*/ // Outputs .rml_ecl_cwp_d (rml_ecl_cwp_d[2:0]), .exu_tlu_cwp0_w(exu_tlu_cwp0_w[2:0]), .exu_tlu_cwp1_w(exu_tlu_cwp1_w[2:0]), .exu_tlu_cwp2_w(exu_tlu_cwp2_w[2:0]), .exu_tlu_cwp3_w(exu_tlu_cwp3_w[2:0]), .rml_irf_cwpswap_tid_e(rml_irf_cwpswap_tid_e[1:0]), .exu_tlu_spill (exu_tlu_spill), .exu_tlu_spill_wtype(exu_tlu_spill_wtype[2:0]), .exu_tlu_spill_other(exu_tlu_spill_other), .exu_tlu_spill_tid(exu_tlu_spill_tid[1:0]), .rml_ecl_swap_done(rml_ecl_swap_done[3:0]), .exu_tlu_cwp_cmplt(exu_tlu_cwp_cmplt), .exu_tlu_cwp_cmplt_tid(exu_tlu_cwp_cmplt_tid[1:0]), .exu_tlu_cwp_retry(exu_tlu_cwp_retry), // Inputs .clk (clk), .se (se), .reset (reset), .rst_tri_en (rst_tri_en), .rml_ecl_wtype_e(rml_ecl_wtype_e[2:0]), .rml_ecl_other_e(rml_ecl_other_e), .exu_tlu_spill_e(exu_tlu_spill_e), .tlu_exu_cwpccr_update_m(tlu_exu_cwpccr_update_m), .tlu_exu_cwp_retry_m(tlu_exu_cwp_retry_m), .tlu_exu_cwp_m (tlu_exu_cwp_m[2:0]), .thr_d (thr_d[3:0]), .ecl_rml_thr_m (ecl_rml_thr_m[3:0]), .ecl_rml_thr_w (ecl_rml_thr_w[3:0]), .tid_e (tid_e[1:0]), .next_cwp_w (next_cwp_w[2:0]), .next_cwp_e (next_cwp_e[2:0]), .cwp_wen_w (cwp_wen_w), .save_e (save_e), .restore_e (restore_e), .ifu_exu_flushw_e(ifu_exu_flushw_e), .ecl_rml_cwp_wen_e(ecl_rml_cwp_wen_e), .full_swap_e (full_swap_e), .rml_kill_w (rml_kill_w)); /////////////////////////////// // Cansave logic /////////////////////////////// assign cansave_wen_e = ((save_e & ~cansave_is0_e & ~rml_ecl_clean_window_e) | ifu_exu_saved_e | (restore_e & ~canrestore_is0_e) | (ifu_exu_restored_e & otherwin_is0_e)); sparc_exu_rml_inc3 cansave_inc(.dout(rml_next_cansave_e[2:0]), .din(rml_ecl_cansave_e[2:0]), .inc(cansave_inc_e)); assign cansave_inc_e = restore_e | ifu_exu_saved_e; mux2ds #(3) next_cansave_mux(.dout(next_cansave_e[2:0]), .in0(ecl_rml_xor_data_e[2:0]), .in1(rml_next_cansave_e[2:0]), .sel0(~cansave_wen_e), .sel1(cansave_wen_e)); dff_s cansave_wen_e2m(.din(cansave_wen_e), .clk(clk), .q(cansave_wen_m), .se(se), .si(), .so()); dff_s #(3) next_cansave_e2m(.din(next_cansave_e[2:0]), .clk(clk), .q(next_cansave_m[2:0]), .se(se), .si(), .so()); assign cansave_wen_valid_m = cansave_wen_m; dff_s cansave_wen_m2w(.din(cansave_wen_valid_m), .clk(clk), .q(rml_cansave_wen_w), .se(se), .si(), .so()); dff_s #(3) next_cansave_m2w(.din(next_cansave_m[2:0]), .clk(clk), .q(next_cansave_w[2:0]), .se(se), .si(), .so()); assign cansave_wen_w = (rml_cansave_wen_w | ecl_rml_cansave_wen_w) & ~rml_kill_w; /////////////////////////////// // Canrestore logic /////////////////////////////// assign canrestore_wen_e = ((save_e & ~cansave_is0_e & ~rml_ecl_clean_window_e) | ifu_exu_restored_e | (restore_e & ~canrestore_is0_e) | (ifu_exu_saved_e & otherwin_is0_e)); sparc_exu_rml_inc3 canrestore_inc(.dout(rml_next_canrestore_e[2:0]), .din(rml_ecl_canrestore_e[2:0]), .inc(canrestore_inc_e)); assign canrestore_inc_e = ifu_exu_restored_e | save_e; mux2ds #(3) next_canrestore_mux(.dout(next_canrestore_e[2:0]), .in0(ecl_rml_xor_data_e[2:0]), .in1(rml_next_canrestore_e[2:0]), .sel0(~canrestore_wen_e), .sel1(canrestore_wen_e)); dff_s canrestore_wen_e2m(.din(canrestore_wen_e), .clk(clk), .q(canrestore_wen_m), .se(se), .si(), .so()); dff_s #(3) next_canrestore_e2m(.din(next_canrestore_e[2:0]), .clk(clk), .q(next_canrestore_m[2:0]), .se(se), .si(), .so()); assign canrestore_wen_valid_m = canrestore_wen_m; dff_s canrestore_wen_m2w(.din(canrestore_wen_valid_m), .clk(clk), .q(rml_canrestore_wen_w), .se(se), .si(), .so()); dff_s #(3) next_canrestore_m2w(.din(next_canrestore_m[2:0]), .clk(clk), .q(next_canrestore_w[2:0]), .se(se), .si(), .so()); assign canrestore_wen_w = (rml_canrestore_wen_w | ecl_rml_canrestore_wen_w) & ~rml_kill_w; /////////////////////////////// // Otherwin logic /////////////////////////////// // Decrements on saved or restored if otherwin != 0 assign otherwin_wen_e = ((ifu_exu_saved_e | ifu_exu_restored_e) & ~otherwin_is0_e); assign rml_next_otherwin_e[2] = ((rml_ecl_otherwin_e[2] & rml_ecl_otherwin_e[1]) | (rml_ecl_otherwin_e[2] & rml_ecl_otherwin_e[0])); assign rml_next_otherwin_e[1] = rml_ecl_otherwin_e[1] ^ ~rml_ecl_otherwin_e[0]; assign rml_next_otherwin_e[0] = ~rml_ecl_otherwin_e[0]; mux2ds #(3) next_otherwin_mux(.dout(next_otherwin_e[2:0]), .in0(ecl_rml_xor_data_e[2:0]), .in1(rml_next_otherwin_e[2:0]), .sel0(~otherwin_wen_e), .sel1(otherwin_wen_e)); dff_s otherwin_wen_e2m(.din(otherwin_wen_e), .clk(clk), .q(otherwin_wen_m), .se(se), .si(), .so()); dff_s #(3) next_otherwin_e2m(.din(next_otherwin_e[2:0]), .clk(clk), .q(next_otherwin_m[2:0]), .se(se), .si(), .so()); assign otherwin_wen_valid_m = otherwin_wen_m; dff_s otherwin_wen_m2w(.din(otherwin_wen_valid_m), .clk(clk), .q(rml_otherwin_wen_w), .se(se), .si(), .so()); dff_s #(3) next_otherwin_m2w(.din(next_otherwin_m[2:0]), .clk(clk), .q(next_otherwin_w[2:0]), .se(se), .si(), .so()); assign otherwin_wen_w = (rml_otherwin_wen_w | ecl_rml_otherwin_wen_w) & ~rml_kill_w; /////////////////////////////// // Cleanwin logic /////////////////////////////// // increments on restored if cleanwin != 7 assign cleanwin_wen_e = (ifu_exu_restored_e & ~(rml_ecl_cleanwin_e[2] & rml_ecl_cleanwin_e[1] & rml_ecl_cleanwin_e[0])); assign rml_next_cleanwin_e[2] = ((~rml_ecl_cleanwin_e[2] & rml_ecl_cleanwin_e[1] & rml_ecl_cleanwin_e[0]) | rml_ecl_cleanwin_e[2]); assign rml_next_cleanwin_e[1] = rml_ecl_cleanwin_e[1] ^ rml_ecl_cleanwin_e[0]; assign rml_next_cleanwin_e[0] = ~rml_ecl_cleanwin_e[0]; mux2ds #(3) next_cleanwin_mux(.dout(next_cleanwin_e[2:0]), .in0(ecl_rml_xor_data_e[2:0]), .in1(rml_next_cleanwin_e[2:0]), .sel0(~cleanwin_wen_e), .sel1(cleanwin_wen_e)); dff_s cleanwin_wen_e2m(.din(cleanwin_wen_e), .clk(clk), .q(cleanwin_wen_m), .se(se), .si(), .so()); dff_s #(3) next_cleanwin_e2m(.din(next_cleanwin_e[2:0]), .clk(clk), .q(next_cleanwin_m[2:0]), .se(se), .si(), .so()); assign cleanwin_wen_valid_m = cleanwin_wen_m; dff_s cleanwin_wen_m2w(.din(cleanwin_wen_valid_m), .clk(clk), .q(rml_cleanwin_wen_w), .se(se), .si(), .so()); dff_s #(3) next_cleanwin_m2w(.din(next_cleanwin_m[2:0]), .clk(clk), .q(next_cleanwin_w[2:0]), .se(se), .si(), .so()); assign cleanwin_wen_w = (rml_cleanwin_wen_w | ecl_rml_cleanwin_wen_w) & ~rml_kill_w; /////////////////////////////// // WSTATE logic /////////////////////////////// assign wstate_wen_w = ecl_rml_wstate_wen_w & ~rml_kill_w; /////////////////////////////// // Storage of other WMRs /////////////////////////////// sparc_exu_reg cansave_reg(.clk(clk), .se(se), .data_out(rml_ecl_cansave_d[2:0]), .thr_out(thr_d[3:0]), .thr_w(ecl_rml_thr_w[3:0]), .wen_w(cansave_wen_w), .data_in_w(next_cansave_w[2:0])); dff_s #(3) cansave_d2e(.din(rml_ecl_cansave_d[2:0]), .clk(clk), .q(rml_ecl_cansave_e[2:0]), .se(se), .si(), .so()); sparc_exu_reg canrestore_reg(.clk(clk), .se(se), .data_out(rml_ecl_canrestore_d[2:0]), .thr_out(thr_d[3:0]), .thr_w(ecl_rml_thr_w[3:0]), .wen_w(canrestore_wen_w), .data_in_w(next_canrestore_w[2:0])); dff_s #(3) canrestore_d2e(.din(rml_ecl_canrestore_d[2:0]), .clk(clk), .q(rml_ecl_canrestore_e[2:0]), .se(se), .si(), .so()); sparc_exu_reg otherwin_reg(.clk(clk), .se(se), .data_out(rml_ecl_otherwin_d[2:0]), .thr_out(thr_d[3:0]), .thr_w(ecl_rml_thr_w[3:0]), .wen_w(otherwin_wen_w), .data_in_w(next_otherwin_w[2:0])); dff_s #(3) otherwin_d2e(.din(rml_ecl_otherwin_d[2:0]), .clk(clk), .q(rml_ecl_otherwin_e[2:0]), .se(se), .si(), .so()); sparc_exu_reg cleanwin_reg(.clk(clk), .se(se), .data_out(rml_ecl_cleanwin_d[2:0]), .thr_out(thr_d[3:0]), .thr_w(ecl_rml_thr_w[3:0]), .wen_w(cleanwin_wen_w), .data_in_w(next_cleanwin_w[2:0])); dff_s #(3) cleanwin_d2e(.din(rml_ecl_cleanwin_d[2:0]), .clk(clk), .q(rml_ecl_cleanwin_e[2:0]), .se(se), .si(), .so()); sparc_exu_reg hi_wstate_reg(.clk(clk), .se(se), .data_out(rml_ecl_wstate_d[5:3]), .thr_out(thr_d[3:0]), .thr_w(ecl_rml_thr_w[3:0]), .wen_w(wstate_wen_w), .data_in_w(exu_tlu_wsr_data_w[5:3])); sparc_exu_reg lo_wstate_reg(.clk(clk), .se(se), .data_out(rml_ecl_wstate_d[2:0]), .thr_out(thr_d[3:0]), .thr_w(ecl_rml_thr_w[3:0]), .wen_w(wstate_wen_w), .data_in_w(exu_tlu_wsr_data_w[2:0])); ///////////////////////////////// // Alternate Globals control //---------------------------- ///////////////////////////////// assign rml_irf_new_agp[1:0] = tlu_exu_agp[1:0]; assign agp_tid[1:0] = tlu_exu_agp_tid[1:0]; `ifdef FPGA_SYN_1THREAD assign rml_irf_old_agp[1:0] = agp_thr0[1:0]; assign agp_wen_thr0_w = (agp_thr[0] & agp_wen) | reset; // mux between new and current value mux2ds #(2) agp_next0_mux(.dout(agp_thr0_next[1:0]), .in0(agp_thr0[1:0]), .in1(new_agp[1:0]), .sel0(~agp_wen_thr0_w), .sel1(agp_wen_thr0_w)); dff_s #(2) dff_agp_thr0(.din(agp_thr0_next[1:0]), .clk(clk), .q(agp_thr0[1:0]), .se(se), .si(), .so()); // generation of controls assign agp_wen = tlu_exu_agp_swap; assign rml_irf_swap_global = agp_wen; assign rml_irf_global_tid[1:0] = agp_tid[1:0]; // decode tids assign agp_thr[0] = ~agp_tid[1] & ~agp_tid[0]; // Decode agp input assign new_agp[1:0] = rml_irf_new_agp[1:0] | {2{reset}}; // send current global level to ecl for error logging assign rml_ecl_gl_e[1:0] = agp_thr0[1:0]; `else // Output selection for current agp mux4ds #(2) mux_agp_out1(.dout(rml_irf_old_agp[1:0]), .sel0(agp_thr[0]), .sel1(agp_thr[1]), .sel2(agp_thr[2]), .sel3(agp_thr[3]), .in0(agp_thr0[1:0]), .in1(agp_thr1[1:0]), .in2(agp_thr2[1:0]), .in3(agp_thr3[1:0])); ////////////////////////////////////// // Storage of agp ////////////////////////////////////// // enable input for each thread assign agp_wen_thr0_w = (agp_thr[0] & agp_wen) | reset; assign agp_wen_thr1_w = (agp_thr[1] & agp_wen) | reset; assign agp_wen_thr2_w = (agp_thr[2] & agp_wen) | reset; assign agp_wen_thr3_w = (agp_thr[3] & agp_wen) | reset; // mux between new and current value mux2ds #(2) agp_next0_mux(.dout(agp_thr0_next[1:0]), .in0(agp_thr0[1:0]), .in1(new_agp[1:0]), .sel0(~agp_wen_thr0_w), .sel1(agp_wen_thr0_w)); mux2ds #(2) agp_next1_mux(.dout(agp_thr1_next[1:0]), .in0(agp_thr1[1:0]), .in1(new_agp[1:0]), .sel0(~agp_wen_thr1_w), .sel1(agp_wen_thr1_w)); mux2ds #(2) agp_next2_mux(.dout(agp_thr2_next[1:0]), .in0(agp_thr2[1:0]), .in1(new_agp[1:0]), .sel0(~agp_wen_thr2_w), .sel1(agp_wen_thr2_w)); mux2ds #(2) agp_next3_mux(.dout(agp_thr3_next[1:0]), .in0(agp_thr3[1:0]), .in1(new_agp[1:0]), .sel0(~agp_wen_thr3_w), .sel1(agp_wen_thr3_w)); // store new value dff_s #(2) dff_agp_thr0(.din(agp_thr0_next[1:0]), .clk(clk), .q(agp_thr0[1:0]), .se(se), .si(), .so()); dff_s #(2) dff_agp_thr1(.din(agp_thr1_next[1:0]), .clk(clk), .q(agp_thr1[1:0]), .se(se), .si(), .so()); dff_s #(2) dff_agp_thr2(.din(agp_thr2_next[1:0]), .clk(clk), .q(agp_thr2[1:0]), .se(se), .si(), .so()); dff_s #(2) dff_agp_thr3(.din(agp_thr3_next[1:0]), .clk(clk), .q(agp_thr3[1:0]), .se(se), .si(), .so()); // generation of controls assign agp_wen = tlu_exu_agp_swap; assign rml_irf_swap_global = agp_wen; assign rml_irf_global_tid[1:0] = agp_tid[1:0]; // decode tids assign agp_thr[0] = ~agp_tid[1] & ~agp_tid[0]; assign agp_thr[1] = ~agp_tid[1] & agp_tid[0]; assign agp_thr[2] = agp_tid[1] & ~agp_tid[0]; assign agp_thr[3] = agp_tid[1] & agp_tid[0]; // Decode agp input assign new_agp[1:0] = rml_irf_new_agp[1:0] | {2{reset}}; // send current global level to ecl for error logging assign rml_ecl_gl_e[1:0] = ((tid_e[1:0] == 2'b00)? agp_thr0[1:0]: (tid_e[1:0] == 2'b01)? agp_thr1[1:0]: (tid_e[1:0] == 2'b10)? agp_thr2[1:0]: agp_thr3[1:0]); `endif // !`ifdef FPGA_SYN_1THREAD endmodule // sparc_exu_rml