Pipeline matmul multiplier: 2-stage to fix WNS

Split compute into stage1 (compute → partial) and
stage2 (partial → c_data). Halves CARRY4 chain depth
from 16 to ~8 levels. Compute cycles: 16 → 32.

rtl/ip/matmul/matmul_axi_slave.v

@@ -71,6 +71,8 @@ reg        busy;
 reg        done;
 reg        error;
 reg [3:0]  calc_index;
+reg [65:0] partial;       // pipeline register for 2-stage multiplier
+reg        pipe_valid;     // partial has valid data
 
 wire       write_fire;
 wire       start_fire;
@@ -288,14 +290,25 @@ always @(posedge clk) begin
             done       <= 1'b0;
             error      <= 1'b0;
             calc_index <= 4'd0;
+            pipe_valid <= 1'b0;
         end else if (busy) begin
-            c_data[calc_index] <= calc_c_element(calc_index);
+            if (!pipe_valid) begin
-            if (calc_index == 4'd15) begin
+                // Stage 1: compute element, store in partial
-                busy       <= 1'b0;
+                partial    <= calc_c_element(calc_index);
-                done       <= 1'b1;
+                pipe_valid <= 1'b1;
-                calc_index <= 4'd0;
             end else begin
-                calc_index <= calc_index + 4'd1;
+                // Stage 2: store partial into c_data, advance
+                c_data[calc_index] <= partial;
+                pipe_valid <= 1'b0;
+                if (calc_index == 4'd15) begin
+                    busy       <= 1'b0;
+                    done       <= 1'b1;
+        calc_index    <= 4'd0;
+        partial       <= 66'b0;
+        pipe_valid    <= 1'b0;
+                end else begin
+                    calc_index <= calc_index + 4'd1;
+                end
             end
         end
     end