The old program took to make Logic elements to fit. Now it uses a 2-port RAM element to replace the huge registers. The protocol for shifting in is a bit different now. Instead of shifting in all the data at once and then Latching the user shifts in the data 64 bits at a time then latches.

module LED_Matrix_32x128
(
clk,e_cnt,row_cnt,col,SDRAM_CS,SRAM_CS,LAT,INPUT,SCLK
);

input  wire  clk;
input  wire  LAT;
input  wire  [7:0]  INPUT;
input  wire  SCLK;

output  reg [63:0]  col;
output  reg [0:0]   SDRAM_CS;
output  reg [0:0]   SRAM_CS;
output  reg [5:0]  row_cnt;
output  reg [4:0]  e_cnt;

reg [24:0]  clk_slow;
reg [63:0]  col_buffer;

    reg  [5:0]  read_addr;
reg  [5:0]  write_addr;
reg  [63:0]  disp_mem_data;
reg      disp_mem_wren;
wire  [63:0]  disp_mem_q;
initial
begin
SDRAM_CS <= 1'b0;
    SRAM_CS <= 1'b0;
    clk_slow <= 16'b0000000000000000;
    row_cnt <= 5'b00000;
    e_cnt <= 4'b1000;
    write_addr <= 5'b00000;
    disp_mem_wren <= 0;
  end

always @(posedge clk) 
begin
  clk_slow <= clk_slow + 1'b1;
end

always @(posedge clk_slow[10]) 
begin
  if(row_cnt > 5'b01111)
begin
row_cnt <= 5'b00000;
    e_cnt <= {e_cnt[2:0],e_cnt[3]};
  end

  col <= disp_mem_q;  
  row_cnt <= row_cnt + 1'b1;
  read_addr <= read_addr + 1'b1;

end

always @(posedge SCLK)
begin
  if(LAT)
  begin
    col_buffer <= {col_buffer[62:0],INPUT[0]};
  end
  else
  begin
    disp_mem_data <= col_buffer;
    disp_mem_wren <= 1;
    disp_mem_wren <= 0;
    write_addr <= write_addr + 1'b1;
  end
end

display_memory disp_mem_inst (
        .rdaddress (read_addr),
        .wraddress (write_addr),
        .clock  (clk),
        .data   (disp_mem_data),
        .wren   (disp_mem_wren),
        .q      (disp_mem_q)
);

endmodule