URL
https://opencores.org/ocsvn/openrisc/openrisc/trunk
Subversion Repositories openrisc
[/] [openrisc/] [trunk/] [or1200/] [rtl/] [verilog/] [or1200_fpu_post_norm_div.v] - Rev 602
Go to most recent revision | Compare with Previous | Blame | View Log
////////////////////////////////////////////////////////////////////// //// //// //// or1200_fpu_post_norm_div //// //// //// //// This file is part of the OpenRISC 1200 project //// //// http://opencores.org/project,or1k //// //// //// //// Description //// //// post-normalization entity for the division unit //// //// //// //// To Do: //// //// //// //// //// //// Author(s): //// //// - Original design (FPU100) - //// //// Jidan Al-eryani, jidan@gmx.net //// //// - Conv. to Verilog and inclusion in OR1200 - //// //// Julius Baxter, julius@opencores.org //// //// //// ////////////////////////////////////////////////////////////////////// // // Copyright (C) 2006, 2010 // // This source file may be used and distributed without // restriction provided that this copyright statement is not // removed from the file and that any derivative work contains // the original copyright notice and the associated disclaimer. // // THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY // EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED // TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS // FOR A PARTICULAR PURPOSE. IN NO EVENT SHALL THE AUTHOR // OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, // INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES // (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE // GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR // BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT // OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE // POSSIBILITY OF SUCH DAMAGE. // module or1200_fpu_post_norm_div ( clk_i, opa_i, opb_i, qutnt_i, rmndr_i, exp_10_i, sign_i, rmode_i, output_o, ine_o ); parameter FP_WIDTH = 32; parameter MUL_SERIAL = 0; // 0 for parallel multiplier, 1 for serial parameter MUL_COUNT = 11; //11 for parallel multiplier, 34 for serial parameter FRAC_WIDTH = 23; parameter EXP_WIDTH = 8; parameter ZERO_VECTOR = 31'd0; parameter INF = 31'b1111111100000000000000000000000; parameter QNAN = 31'b1111111110000000000000000000000; parameter SNAN = 31'b1111111100000000000000000000001; input clk_i; input [FP_WIDTH-1:0] opa_i; input [FP_WIDTH-1:0] opb_i; input [FRAC_WIDTH+3:0] qutnt_i; input [FRAC_WIDTH+3:0] rmndr_i; input [EXP_WIDTH+1:0] exp_10_i; input sign_i; input [1:0] rmode_i; output reg [FP_WIDTH-1:0] output_o; output reg ine_o; // input&output register wires reg [FP_WIDTH-1:0] s_opa_i; reg [FP_WIDTH-1:0] s_opb_i; reg [EXP_WIDTH-1:0] s_expa; reg [EXP_WIDTH-1:0] s_expb; reg [FRAC_WIDTH+3:0] s_qutnt_i; reg [FRAC_WIDTH+3:0] s_rmndr_i; reg [5:0] s_r_zeros; reg [EXP_WIDTH+1:0] s_exp_10_i; reg s_sign_i; reg [1:0] s_rmode_i; wire [FP_WIDTH-1:0] s_output_o; wire s_ine_o, s_overflow; wire s_opa_dn, s_opb_dn; wire s_qutdn; wire [9:0] s_exp_10b; reg [5:0] s_shr1; reg [5:0] s_shl1; wire s_shr2; reg [8:0] s_expo1; wire [8:0] s_expo2; reg [8:0] s_expo3; reg [26:0] s_fraco1; wire [24:0] s_frac_rnd; reg [24:0] s_fraco2; wire s_guard, s_round, s_sticky, s_roundup; wire s_lost; wire s_op_0, s_opab_0, s_opb_0; wire s_infa, s_infb; wire s_nan_in, s_nan_op, s_nan_a, s_nan_b; wire s_inf_result; always @(posedge clk_i) begin s_opa_i <= opa_i; s_opb_i <= opb_i; s_expa <= opa_i[30:23]; s_expb <= opb_i[30:23]; s_qutnt_i <= qutnt_i; s_rmndr_i <= rmndr_i; s_exp_10_i <= exp_10_i; s_sign_i <= sign_i; s_rmode_i <= rmode_i; end // Output Register always @(posedge clk_i) begin output_o <= s_output_o; ine_o <= s_ine_o; end // qutnt_i // 26 25 3 // | | | // h fffffffffffffffffffffff grs //*** Stage 1 **** // figure out the exponent and how far the fraction has to be shifted // right or left assign s_opa_dn = !(|s_expa) & (|opa_i[22:0]); assign s_opb_dn = !(|s_expb) & (|opb_i[22:0]); assign s_qutdn = !s_qutnt_i[26]; assign s_exp_10b = s_exp_10_i - {9'd0,s_qutdn}; wire [9:0] v_shr; wire [9:0] v_shl; assign v_shr = (s_exp_10b[9] | !(|s_exp_10b)) ? (10'd1 - s_exp_10b) - {9'd0,s_qutdn} : 0; assign v_shl = (s_exp_10b[9] | !(|s_exp_10b)) ? 0 : s_exp_10b[8] ? 0 : {9'd0,s_qutdn}; always @(posedge clk_i) if (s_exp_10b[9] | !(|s_exp_10b)) s_expo1 <= 9'd1; else s_expo1 <= s_exp_10b[8:0]; always @(posedge clk_i) s_shr1 <= v_shr[6] ? 6'b111111 : v_shr[5:0]; always @(posedge clk_i) s_shl1 <= v_shl[5:0]; // *** Stage 2 *** // Shifting the fraction and rounding // shift the fraction always @(posedge clk_i) if (|s_shr1) s_fraco1 <= s_qutnt_i >> s_shr1; else s_fraco1 <= s_qutnt_i << s_shl1; assign s_expo2 = s_fraco1[26] ? s_expo1 : s_expo1 - 9'd1; //s_r_zeros <= count_r_zeros(s_qutnt_i); always @(s_qutnt_i) casez(s_qutnt_i) // synopsys full_case parallel_case 27'b??????????????????????????1: s_r_zeros = 0; 27'b?????????????????????????10: s_r_zeros = 1; 27'b????????????????????????100: s_r_zeros = 2; 27'b???????????????????????1000: s_r_zeros = 3; 27'b??????????????????????10000: s_r_zeros = 4; 27'b?????????????????????100000: s_r_zeros = 5; 27'b????????????????????1000000: s_r_zeros = 6; 27'b???????????????????10000000: s_r_zeros = 7; 27'b??????????????????100000000: s_r_zeros = 8; 27'b?????????????????1000000000: s_r_zeros = 9; 27'b????????????????10000000000: s_r_zeros = 10; 27'b???????????????100000000000: s_r_zeros = 11; 27'b??????????????1000000000000: s_r_zeros = 12; 27'b?????????????10000000000000: s_r_zeros = 13; 27'b????????????100000000000000: s_r_zeros = 14; 27'b???????????1000000000000000: s_r_zeros = 15; 27'b??????????10000000000000000: s_r_zeros = 16; 27'b?????????100000000000000000: s_r_zeros = 17; 27'b????????1000000000000000000: s_r_zeros = 18; 27'b???????10000000000000000000: s_r_zeros = 19; 27'b??????100000000000000000000: s_r_zeros = 20; 27'b?????1000000000000000000000: s_r_zeros = 21; 27'b????10000000000000000000000: s_r_zeros = 22; 27'b???100000000000000000000000: s_r_zeros = 23; 27'b??1000000000000000000000000: s_r_zeros = 24; 27'b?10000000000000000000000000: s_r_zeros = 25; 27'b100000000000000000000000000: s_r_zeros = 26; 27'b000000000000000000000000000: s_r_zeros = 27; endcase // casex (s_qutnt_i) assign s_lost = (s_shr1+{5'd0,s_shr2}) > s_r_zeros; // ***Stage 3*** // Rounding assign s_guard = s_fraco1[2]; assign s_round = s_fraco1[1]; assign s_sticky = s_fraco1[0] | (|s_rmndr_i); assign s_roundup = s_rmode_i==2'b00 ? // round to nearest even s_guard & ((s_round | s_sticky) | s_fraco1[3]) : s_rmode_i==2'b10 ? // round up (s_guard | s_round | s_sticky) & !s_sign_i : s_rmode_i==2'b11 ? // round down (s_guard | s_round | s_sticky) & s_sign_i : 0; // round to zero(truncate = no rounding) assign s_frac_rnd = s_roundup ?{1'b0,s_fraco1[26:3]} + 1 : {1'b0,s_fraco1[26:3]}; assign s_shr2 = s_frac_rnd[24]; always @(posedge clk_i) begin s_expo3 <= s_shr2 ? s_expo2 + "1" : s_expo2; s_fraco2 <= s_shr2 ? {1'b0,s_frac_rnd[24:1]} : s_frac_rnd; end // // ***Stage 4**** // Output assign s_op_0 = !((|s_opa_i[30:0]) & (|s_opb_i[30:0])); assign s_opab_0 = !((|s_opa_i[30:0]) | (|s_opb_i[30:0])); assign s_opb_0 = !(|s_opb_i[30:0]); assign s_infa = &s_expa; assign s_infb = &s_expb; assign s_nan_a = s_infa & (|s_opa_i[22:0]); assign s_nan_b = s_infb & (|s_opb_i[22:0]); assign s_nan_in = s_nan_a | s_nan_b; assign s_nan_op = (s_infa & s_infb) | s_opab_0; // 0 / 0, inf / inf assign s_inf_result = (&s_expo3[7:0]) | s_expo3[8] | s_opb_0; assign s_overflow = s_inf_result & !(s_infa) & !s_opb_0; assign s_ine_o = !s_op_0 & (s_lost | (|s_fraco1[2:0]) | s_overflow | (|s_rmndr_i)); assign s_output_o = (s_nan_in | s_nan_op) ? {s_sign_i,QNAN} : s_infa | s_overflow | s_inf_result ? {s_sign_i,INF} : s_op_0 | s_infb ? {s_sign_i,ZERO_VECTOR} : {s_sign_i,s_expo3[7:0],s_fraco2[22:0]}; endmodule // or1200_fpu_post_norm_div
Go to most recent revision | Compare with Previous | Blame | View Log