diff options
| -rw-r--r-- | rtl/gfx/gfx_fixed_fma.sv | 91 |
1 files changed, 43 insertions, 48 deletions
diff --git a/rtl/gfx/gfx_fixed_fma.sv b/rtl/gfx/gfx_fixed_fma.sv index e1dd6bb..b458010 100644 --- a/rtl/gfx/gfx_fixed_fma.sv +++ b/rtl/gfx/gfx_fixed_fma.sv @@ -1,15 +1,5 @@ `include "gfx/gfx_defs.sv" -/* Operación a * b + c en punto fijo, diseñada para cerrar timing fácilmente - * en Cyclone V donde los bloques de DSP soportan 18x18. Las etapas son: - * - fma_ab: Productos de permutaciones a_lo/hi con b_lo/hi - * - fma_pp: Recombinación en FMAs parciales - * - fma_lo: Mitad baja del resultado y mitad alta pre-carry - * - fma_hi: Mitad alta post-carry - * - * Nótese que esto toma exactamente el mismo trabajo que a * b - * (ver rtl/core/mul.sv en proyecto 2 de arqui 1). - */ module gfx_fixed_fma ( input logic clk, @@ -22,51 +12,56 @@ module gfx_fixed_fma output fixed q ); - fixed a_ab, b_ab, c_ab, c_pp; - logic[1:0] carry; - logic[16:0] lo_left, lo_right; - logic[17:0] lo_with_carry; - logic[35:0] ab_ll, ab_lh, ab_hl, ab_hh; + logic[$bits(fixed) + `FIXED_FRAC - 1:0] q_ext; - logic[15:0] a_lo, a_hi, b_lo, b_hi, ab_ll_hi, ab_ll_lo, ab_hl_hi, ab_hl_lo, - ab_lh_hi, ab_lh_lo, ab_hh_hi, ab_hh_lo, c_hi, c_lo, hi, hi_left, hi_right, lo; +`ifndef VERILATOR + assign q = q_ext[$bits(fixed) + `FIXED_FRAC - 1:`FIXED_FRAC]; - assign {a_hi, a_lo} = a_ab; - assign {b_hi, b_lo} = b_ab; - assign {c_hi, c_lo} = c_pp; + lpm_mult mult + ( + .aclr(0), + .clock(clk), + .clken(!stall), - assign {ab_ll_hi, ab_ll_lo} = ab_ll[31:0]; - assign {ab_lh_hi, ab_lh_lo} = ab_lh[31:0]; - assign {ab_hl_hi, ab_hl_lo} = ab_hl[31:0]; - assign {ab_hh_hi, ab_hh_lo} = ab_hh[31:0]; + .sum({c, {`FIXED_FRAC{1'b0}}}), + .dataa({{`FIXED_FRAC{a[$bits(a) - 1]}}, a}), + .datab({{`FIXED_FRAC{b[$bits(b) - 1]}}, b}), + .result(q_ext) + ); - assign {carry, lo} = lo_with_carry; + defparam + mult.lpm_widtha = $bits(fixed) + `FIXED_FRAC, + mult.lpm_widthb = $bits(fixed) + `FIXED_FRAC, + mult.lpm_widths = $bits(fixed) + `FIXED_FRAC, + mult.lpm_widthp = $bits(fixed) + `FIXED_FRAC, + mult.lpm_representation = "SIGNED", + mult.lpm_pipeline = `FIXED_FMA_STAGES; +`else + fixed a_hold, b_hold, c_hold; - always @(posedge clk) - if (!stall) begin - a_ab <= a; - b_ab <= b; - c_ab <= c; + assign q = q_ext[$bits(fixed) + `FIXED_FRAC - 1:`FIXED_FRAC] + c_hold; + assign q_ext = a_hold * b_hold; - /* Como los operandos son pequeños (16 bits), esto no se sintetiza, - * sino que se enruta a través de los bloques de DSP más cercanos - */ - ab_ll <= {2'd0, a_lo} * {2'd0, b_lo}; - ab_lh <= {2'd0, a_lo} * {2'd0, b_hi}; - ab_hl <= {2'd0, a_hi} * {2'd0, b_lo}; - ab_hh <= {2'd0, a_hi} * {2'd0, b_hi}; + gfx_pipes #(.WIDTH($bits(a)), .DEPTH(`FIXED_FMA_STAGES)) a_pipes + ( + .in(a), + .out(a_hold), + .* + ); - c_pp <= c_ab; + gfx_pipes #(.WIDTH($bits(b)), .DEPTH(`FIXED_FMA_STAGES)) b_pipes + ( + .in(b), + .out(b_hold), + .* + ); - hi_left <= ab_hh_lo + ab_lh_hi; - lo_left <= {1'd0, ab_lh_lo} + {1'd0, ab_hl_lo}; - hi_right <= ab_hl_hi + c_hi; - lo_right <= {1'd0, ab_ll_hi} + {1'd0, c_lo}; - - hi <= hi_left + hi_right; - lo_with_carry <= {1'd0, lo_left} + {1'd0, lo_right}; - - q <= {hi + {14'd0, carry}, lo}; - end + gfx_pipes #(.WIDTH($bits(c)), .DEPTH(`FIXED_FMA_STAGES)) c_pipes + ( + .in(c), + .out(c_hold), + .* + ); +`endif endmodule |