E203数据冲突处理OITF


      流水线的数据冲突分为三类:WAR,RAW,WAW

       https://wenku.baidu.com/view/e066926d48d7c1c708a14508.html

  • WAR: write after read 相关性,又称先读后写相关性。比如下面的指令序列,第一条指令会读取x4,第二条指令会写x4。在流水线中,如果第二条指令比第一条指令先写x4,则第一条指令就会读出错误的值。

           add x5, x4,x6

           add x4, x3, x2

  • WAW: write after write 相关性,又称先写后写相关性。比如下面的指令序列,两条指令都会写x5。在流水线中,如果第二条指令比第一条指令先写x5,就会引起逻辑错误。

           add x5, x4,x6

           add x5, x3, x2

  • RAW:read after write相关性,又称先写后读相关性。比如下面指令序列,如果第二条指令,在第一条指令写x5之前,第二条指令先读x4,就会引起逻辑错误。

           add x5, x4,x6

           add x4, x5, x2


      由于蜂鸟E200系列是按序派遣,按顺序写回的微架构,在指令派遣时候就已经从通用寄存器数组中读取了源操作数。后续执行的指令写回regfile的操作不可能影响到前面指令的读取,所以不可能发生WAR相关性造成的数据冲突。

      正在派遣的指令处在流水线的第二级,假设之前派遣的指令是单周期指令,则前序指令肯定已经完成了执行且将结果写回了Regfile。因此正在派遣的指令不可能会发生RAW数据冲突。但是假设之前派遣的指令是多周期指令(长指令),由于指令需要多个周期才能写回结果。因此正在派遣的指令可能会产生前序相关的RAW相关性。

     正在派遣的指令处在流水线的第二级,假设之前派遣的指令是单周期指令,则前序指令肯定已经完成了执行且将结果写回了Regfile。因此正在派遣的指令不可能会发生WAW数据冲突。但是假设之前派遣的指令是多周期指令(长指令),由于指令需要多个周期才能写回结果。因此正在派遣的指令可能会产生前序相关的WAW相关性。

    为了能检测出长指令的RAW和WAW相关性,蜂鸟E200使用了一个outstanding instruction track fifo(OITF)模块。在流水线的派遣(Dispatch)点,每一次派遣一个长指令,则会在OITF中分配一个表项(Entry),在这个表项中会存储该长指令的结果寄存器索引。在流水线的写回(Write-back)点,每次按顺序写回一个长指令之后,就会将此指令在OITF中的表项移除。

      每条指令派遣时,都会将本指令的源操作数和目的操作数寄存器索引和OITF中的各个表项进行比对,从而判断本指令是否与已经被派遣出,且尚未写回的长指令产生RAW和WAW相关性。如果产生相关性,则stall住当前指令的派遣。如果没有RAW和WAW相关性,且该指令为多周期长指令,把该指令写入OITF,如果OITF是full,则仍要stall住管线,等待OITF释放空间后,再写入并派遣。

   在writeback模块,会进行长指令写回仲裁,长指令写回regfile后,会释放OITF中相应的表项。

image

     



OITF代码如下,如果fifo full,则dis_ready=0, 与dispatch模块握手失败,不会发送新的dispatch进来。如果不为空,会发送新的指令进来进行判断。

`include "e203_defines.v"

module e203_exu_oitf (
  output dis_ready,

  input  dis_ena, //dispatch a long instruction enable signal
  input  ret_ena, //write back a long instruction enable signal

  output [`E203_ITAG_WIDTH-1:0] dis_ptr,  //write pointer
  output [`E203_ITAG_WIDTH-1:0] ret_ptr,  //read pointer

  output [`E203_RFIDX_WIDTH-1:0] ret_rdidx,
  output ret_rdwen,
  output ret_rdfpu,
  output [`E203_PC_SIZE-1:0] ret_pc,

  input  disp_i_rs1en, // enable if current dispatch instruction fetch first source operand  
  input  disp_i_rs2en, // ...
  input  disp_i_rs3en, // ...
  input  disp_i_rdwen, // enable if current dispatch instruction write back to register
  input  disp_i_rs1fpu, // enable if current dispath instruction need to read float gpr
  input  disp_i_rs2fpu, //...
  input  disp_i_rs3fpu, //...
  input  disp_i_rdfpu,  //enable if current dipatch instruction need to write back to float register files.
  //register index
  input  [`E203_RFIDX_WIDTH-1:0] disp_i_rs1idx,
  input  [`E203_RFIDX_WIDTH-1:0] disp_i_rs2idx,
  input  [`E203_RFIDX_WIDTH-1:0] disp_i_rs3idx,
  input  [`E203_RFIDX_WIDTH-1:0] disp_i_rdidx,
  input  [`E203_PC_SIZE    -1:0] disp_i_pc, //pc of current dispatch instruciotn 

  output oitfrd_match_disprs1, //dispatch instruction rs1 is same as any item of result register in oitf
  output oitfrd_match_disprs2, //...
  output oitfrd_match_disprs3, //...
  output oitfrd_match_disprd,  //dispatch instruction rd is same as any item of result register in oitf.
  //if empty, no conflict
  output oitf_empty,
  input  clk,
  input  rst_n
);

  wire [`E203_OITF_DEPTH-1:0] vld_set;
  wire [`E203_OITF_DEPTH-1:0] vld_clr;
  wire [`E203_OITF_DEPTH-1:0] vld_ena;
  wire [`E203_OITF_DEPTH-1:0] vld_nxt;
  wire [`E203_OITF_DEPTH-1:0] vld_r; //if it is valid signal in all item
  wire [`E203_OITF_DEPTH-1:0] rdwen_r;// if it is write back register in all item
  wire [`E203_OITF_DEPTH-1:0] rdfpu_r; //result register in all item if are float
  wire [`E203_RFIDX_WIDTH-1:0] rdidx_r[`E203_OITF_DEPTH-1:0]; //register index in all items
  // The PC here is to be used at wback stage to track out the
  //  PC of exception of long-pipe instruction
  wire [`E203_PC_SIZE-1:0] pc_r[`E203_OITF_DEPTH-1:0];

  wire alc_ptr_ena = dis_ena;  //dispatch a long instruction enable signal, as write pointer enable signal
  wire ret_ptr_ena = ret_ena;  //write back a long instruction enable signal, as read pointer enable signal

  wire oitf_full ;

  wire [`E203_ITAG_WIDTH-1:0] alc_ptr_r; //write pointer, long instruction dispatch
  wire [`E203_ITAG_WIDTH-1:0] ret_ptr_r; //read pointer, long instruction write back

  generate
  if(`E203_OITF_DEPTH > 1) begin: depth_gt1//{
      //extra mark bit for write full
      wire alc_ptr_flg_r;
      wire alc_ptr_flg_nxt = ~alc_ptr_flg_r;
      wire alc_ptr_flg_ena = (alc_ptr_r == ($unsigned(`E203_OITF_DEPTH-1))) & alc_ptr_ena;

      sirv_gnrl_dfflr #(1) alc_ptr_flg_dfflrs(alc_ptr_flg_ena, alc_ptr_flg_nxt, alc_ptr_flg_r, clk, rst_n);

      wire [`E203_ITAG_WIDTH-1:0] alc_ptr_nxt;
      //if write to fifo depth, write ptr = 0,otherwise write ptr = write ptr + 1 
      assign alc_ptr_nxt = alc_ptr_flg_ena ? `E203_ITAG_WIDTH'b0 : (alc_ptr_r + 1'b1);

      sirv_gnrl_dfflr #(`E203_ITAG_WIDTH) alc_ptr_dfflrs(alc_ptr_ena, alc_ptr_nxt, alc_ptr_r, clk, rst_n);

      //extra mark bit for read empty 
      wire ret_ptr_flg_r;
      wire ret_ptr_flg_nxt = ~ret_ptr_flg_r;
      wire ret_ptr_flg_ena = (ret_ptr_r == ($unsigned(`E203_OITF_DEPTH-1))) & ret_ptr_ena;

      sirv_gnrl_dfflr #(1) ret_ptr_flg_dfflrs(ret_ptr_flg_ena, ret_ptr_flg_nxt, ret_ptr_flg_r, clk, rst_n);

      wire [`E203_ITAG_WIDTH-1:0] ret_ptr_nxt;
      //if read to fifo depth, read ptr = 0, otherwise read prt = read prt + 1
      assign ret_ptr_nxt = ret_ptr_flg_ena ? `E203_ITAG_WIDTH'b0 : (ret_ptr_r + 1'b1);

      sirv_gnrl_dfflr #(`E203_ITAG_WIDTH) ret_ptr_dfflrs(ret_ptr_ena, ret_ptr_nxt, ret_ptr_r, clk, rst_n);
      //empty, full mark
      assign oitf_empty = (ret_ptr_r == alc_ptr_r) &   (ret_ptr_flg_r == alc_ptr_flg_r);
      assign oitf_full  = (ret_ptr_r == alc_ptr_r) & (~(ret_ptr_flg_r == alc_ptr_flg_r));
  end//}
  else begin: depth_eq1//}{
      assign alc_ptr_r =1'b0;
      assign ret_ptr_r =1'b0;
      assign oitf_empty = ~vld_r[0];
      assign oitf_full  = vld_r[0];
  end//}
  endgenerate//}

  assign ret_ptr = ret_ptr_r;
  assign dis_ptr = alc_ptr_r;

 //// 
 //// // If the OITF is not full, or it is under retiring, then it is ready to accept new dispatch
 //// assign dis_ready = (~oitf_full) | ret_ena;
 // To cut down the loop between ALU write-back valid --> oitf_ret_ena --> oitf_ready ---> dispatch_ready --- > alu_i_valid
 //   we exclude the ret_ena from the ready signal
 assign dis_ready = (~oitf_full);

  wire [`E203_OITF_DEPTH-1:0] rd_match_rs1idx;
  wire [`E203_OITF_DEPTH-1:0] rd_match_rs2idx;
  wire [`E203_OITF_DEPTH-1:0] rd_match_rs3idx;
  wire [`E203_OITF_DEPTH-1:0] rd_match_rdidx;

  genvar i;
  generate //{
      for (i=0; i<`E203_OITF_DEPTH; i=i+1) begin:oitf_entries//{
        //every time, assign a item and write pointer same as current i, then
	//valid set is high
        assign vld_set[i] = alc_ptr_ena & (alc_ptr_r == i);
        //every time, assign a item and read pointer same as current i, then
	//valid clr is high
        assign vld_clr[i] = ret_ptr_ena & (ret_ptr_r == i);
        assign vld_ena[i] = vld_set[i] |   vld_clr[i];
        assign vld_nxt[i] = vld_set[i] | (~vld_clr[i]);

        sirv_gnrl_dfflr #(1) vld_dfflrs(vld_ena[i], vld_nxt[i], vld_r[i], clk, rst_n);
        //Payload only set, no need to clear
        sirv_gnrl_dffl #(`E203_RFIDX_WIDTH) rdidx_dfflrs(vld_set[i], disp_i_rdidx, rdidx_r[i], clk);
        sirv_gnrl_dffl #(`E203_PC_SIZE    ) pc_dfflrs   (vld_set[i], disp_i_pc   , pc_r[i]   , clk);
        sirv_gnrl_dffl #(1)                 rdwen_dfflrs(vld_set[i], disp_i_rdwen, rdwen_r[i], clk);
        sirv_gnrl_dffl #(1)                 rdfpu_dfflrs(vld_set[i], disp_i_rdfpu, rdfpu_r[i], clk);
        //compare dispatch source operand with result register in fifo
        assign rd_match_rs1idx[i] = vld_r[i] & rdwen_r[i] & disp_i_rs1en & (rdfpu_r[i] == disp_i_rs1fpu) & (rdidx_r[i] == disp_i_rs1idx);
        assign rd_match_rs2idx[i] = vld_r[i] & rdwen_r[i] & disp_i_rs2en & (rdfpu_r[i] == disp_i_rs2fpu) & (rdidx_r[i] == disp_i_rs2idx);
        assign rd_match_rs3idx[i] = vld_r[i] & rdwen_r[i] & disp_i_rs3en & (rdfpu_r[i] == disp_i_rs3fpu) & (rdidx_r[i] == disp_i_rs3idx);
        assign rd_match_rdidx [i] = vld_r[i] & rdwen_r[i] & disp_i_rdwen & (rdfpu_r[i] == disp_i_rdfpu ) & (rdidx_r[i] == disp_i_rdidx );

      end//}
  endgenerate//}
  //rs1 in fifo, so RAW relative
  assign oitfrd_match_disprs1 = |rd_match_rs1idx;
  //rs2 in fifo, so RAW relative
  assign oitfrd_match_disprs2 = |rd_match_rs2idx;
  //rs3 in fifo, so RAW relative
  assign oitfrd_match_disprs3 = |rd_match_rs3idx;
  //rd in fifo, so WAW relative
  assign oitfrd_match_disprd  = |rd_match_rdidx ;

  assign ret_rdidx = rdidx_r[ret_ptr];
  assign ret_pc    = pc_r [ret_ptr];
  assign ret_rdwen = rdwen_r[ret_ptr];
  assign ret_rdfpu = rdfpu_r[ret_ptr];

endmodule










posted on 2019-08-14 09:37  迈克老狼2012  阅读(1973)  评论(0编辑  收藏  举报

导航