Added "MERGE_LH" micro-operation. To be able to do Garner's formula we need

regular (not modular) multiplication. We're doing this by telling the modular
multiplier to stop after the "square" step, which computes A*B. The problem is
that the multiplier stores the lower part of the product in the internal bank L
and the upper part in the internal bank H, but we need to be able to do
operations on the product as a whole. MERGE_LH that combines the two halves of
the product into one bank.

1 files changed, 39 insertions, 2 deletions

diff --git a/rtl/modexpng_core_top.v b/rtl/modexpng_core_top.v
index dea7f0a..4c1f065 100644
--- a/rtl/modexpng_core_top.v
+++ b/rtl/modexpng_core_top.v
@@ -83,13 +83,15 @@ module modexpng_core_top
     wire uop_opcode_is_ladder = (uop_data_opcode == UOP_OPCODE_LADDER_INIT         ) ||
                                 (uop_data_opcode == UOP_OPCODE_LADDER_STEP         ) ;
     wire uop_opcode_is_mmm    = (uop_data_opcode == UOP_OPCODE_MODULAR_MULTIPLY    ) ||
-                                (uop_data_opcode == UOP_OPCODE_MODULAR_REDUCE_PROC ) ;
+                                (uop_data_opcode == UOP_OPCODE_MODULAR_REDUCE_PROC ) ||
+                                (uop_data_opcode == UOP_OPCODE_REGULAR_MULTIPLY    ) ;
     wire uop_opcode_is_wrk    = (uop_data_opcode == UOP_OPCODE_PROPAGATE_CARRIES   ) ||
                                 (uop_data_opcode == UOP_OPCODE_COPY_CRT_Y2X        ) ||
                                 (uop_data_opcode == UOP_OPCODE_MODULAR_REDUCE_INIT ) ||
                                 (uop_data_opcode == UOP_OPCODE_COPY_LADDERS_X2Y    ) ||
                                 (uop_data_opcode == UOP_OPCODE_CROSS_LADDERS_X2Y   ) ||
-                                (uop_data_opcode == UOP_OPCODE_MODULAR_SUBTRACT    ) ;
+                                (uop_data_opcode == UOP_OPCODE_MODULAR_SUBTRACT    ) ||
+                                (uop_data_opcode == UOP_OPCODE_MERGE_LH            ) ;
                                 
     wire uop_loop_now;
     
@@ -716,6 +718,9 @@ module modexpng_core_top
     reg                     mmm_only_reduce_x;
     reg                     mmm_only_reduce_y;
     
+    reg                     mmm_just_multiply_x;
+    reg                     mmm_just_multiply_y;
+    
     wire                    rdct_ena_x;
     wire                    rdct_ena_y;
     wire                    rdct_rdy_x;
@@ -734,6 +739,7 @@ module modexpng_core_top
         .word_index_last_minus1     (mmm_word_index_last_minus1_x),
         .force_unity_b              (mmm_force_unity_b_x),
         .only_reduce                (mmm_only_reduce_x),
+        .just_multiply              (mmm_just_multiply_x),
         
         .sel_wide_in                (mmm_sel_wide_in_x),
         .sel_narrow_in              (mmm_sel_narrow_in_x),
@@ -790,6 +796,7 @@ module modexpng_core_top
         .word_index_last_minus1     (mmm_word_index_last_minus1_y),
         .force_unity_b              (mmm_force_unity_b_y),
         .only_reduce                (mmm_only_reduce_y),
+        .just_multiply              (mmm_just_multiply_y),
         
         .sel_wide_in                (mmm_sel_wide_in_y),
         .sel_narrow_in              (mmm_sel_narrow_in_y),
@@ -1088,6 +1095,7 @@ module modexpng_core_top
                         UOP_LADDER_PQ: {mmm_ladder_mode_x, mmm_ladder_mode_y} <= {io_mgr_ladder_p, io_mgr_ladder_q};
                     endcase
                     //
+                    {mmm_just_multiply_x,   mmm_just_multiply_y  } <= {2{1'b0}};
                     {mmm_only_reduce_x,     mmm_only_reduce_y    } <= {2{1'b0}};
                     {mmm_force_unity_b_x,   mmm_force_unity_b_y  } <= {2{uop_aux_is_1 ? 1'b0 : 1'b1}};
                     {mmm_sel_wide_in_x,     mmm_sel_wide_in_y    } <= {2{uop_data_sel_wide_in      }};
@@ -1110,6 +1118,20 @@ module modexpng_core_top
                     //
                 end
                 //
+                UOP_OPCODE_REGULAR_MULTIPLY: begin
+                    //
+                    {mmm_ladder_mode_x,     mmm_ladder_mode_y   }  <= {2{1'b1}};
+                    //
+                    {mmm_just_multiply_x,   mmm_just_multiply_y  } <= {2{1'b1}};
+                    {mmm_only_reduce_x,     mmm_only_reduce_y    } <= {2{1'b0}};
+                    {mmm_force_unity_b_x,   mmm_force_unity_b_y  } <= {2{uop_aux_is_1 ? 1'b0 : 1'b1}};
+                    {mmm_sel_wide_in_x,     mmm_sel_wide_in_y    } <= {2{uop_data_sel_wide_in      }};
+                    {mmm_sel_narrow_in_x,   mmm_sel_narrow_in_y  } <= {2{uop_data_sel_narrow_in    }};
+                    {rdct_sel_wide_out_x,   rdct_sel_wide_out_y  } <= {2{uop_data_sel_wide_out     }}; 
+                    {rdct_sel_narrow_out_x, rdct_sel_narrow_out_y} <= {2{uop_data_sel_narrow_out   }};
+                    //
+                end
+                //
                 UOP_OPCODE_PROPAGATE_CARRIES: begin
                     wrk_sel_narrow_in   <= uop_data_sel_narrow_in;
                     wrk_sel_narrow_out  <= uop_data_sel_narrow_out;
@@ -1121,6 +1143,10 @@ module modexpng_core_top
                     wrk_sel_narrow_out <= uop_data_sel_narrow_out;
                 end
                 //
+                UOP_OPCODE_MERGE_LH: begin
+                    wrk_sel_narrow_out <= uop_data_sel_narrow_out;
+                end
+                //
                 UOP_OPCODE_COPY_CRT_Y2X,
                 UOP_OPCODE_COPY_LADDERS_X2Y,
                 UOP_OPCODE_CROSS_LADDERS_X2Y: begin
@@ -1181,10 +1207,21 @@ module modexpng_core_top
                     {rdct_word_index_last_x,       rdct_word_index_last_y      } <= {2{word_index_last_pq       }};
                 end
                 //
+                UOP_OPCODE_REGULAR_MULTIPLY: begin
+                    {mmm_word_index_last_x,        mmm_word_index_last_y       } <= {2{word_index_last_pq       }};  
+                    {mmm_word_index_last_minus1_x, mmm_word_index_last_minus1_y} <= {2{word_index_last_pq_minus1}};
+                    {rdct_word_index_last_x,       rdct_word_index_last_y      } <= {2{word_index_last_pq       }};
+                end
+                //
                 UOP_OPCODE_MODULAR_SUBTRACT: begin
                     wrk_word_index_last <= uop_npq_is_n ? word_index_last_n : word_index_last_pq;
                 end
                 //
+                UOP_OPCODE_MERGE_LH: begin
+                    wrk_word_index_last      <= word_index_last_n;
+                    wrk_word_index_last_half <= word_index_last_pq;
+                end
+                //
                 UOP_OPCODE_LADDER_INIT: begin
                     io_mgr_word_index_last <= OP_ADDR_LADDER_LAST;
                     io_mgr_ladder_steps    <= crt_mode ? bit_index_last_pq : bit_index_last_n;