1. Topics to be covered : How to model memory in Verilog RAM modeling Register Bank

3.  In digital simulation, one often needs to model register files, RAMs, and ROMs. Memories are modeled in Verilog simply as an array of registers.  Each element of the array is known as a word. Each word can be one or more bits. It is important to differentiate between n 1-bit registers and one n-bit register.  A particular word in memory is obtained by using the address as a memory array subscript.

4.  Different types and sizes of memory, register file, stack, etc., can be formed by extending the vector concept. Thus the decleration Reg [15:0] memory[511:0];  Declared an array called memory; it has 512 locations. Each location is 16 bits wide. The value of any chosen location can be assigned to a selected register or vice-versa; this constitutes memory reading or writing. The index used to refer a memory location can be a number or an algebraic expression which reduces to an integral value – positive, zero, or negative.  B = mem[3] //data stored at mem[3] is assigned to B

5.  Reg mem1bit[0:1023] //Memory mem1bit 1k 1-bit words.  Reg [7:0] membyte[0:1023]; //Memory membyte with 1k 8-bit words (bytes)  Membyte [511] //Fetches 1 byte word address is 511.

6.  Verilog provides a very useful system task to initialize memories from a data file. Two tasks are provided to read numbers in binary or hexadecimal format. Keywords $readmemb and $readmemh are used to initialize memories :  Usage: ◦ $readmemb(“ ”, ); ◦ $readmemb(“ ”,,, ); ◦ NOTE :, are optional.

12. Solution 1

14. Basic DRAM Architecture

15. Fast Page Mode DRAM (FP DRAM) A row is selected and the col. addresses are sequenced. A row is considered a page, consisting of multiple words. Each word has a sep. col. address. The sense amplifier buffers a page.

16. EDO DRAM (Extended Data Out DRAM) -- Extra output latch between the sense ampl. and output buffer -- allows overlap bet. Col. Select and previous data out -- saves one cycle over FP DRAM

17. FPM and EDO RAM controlled asynchronously by the processor or the memory controller. A synchronous DRAM interface will eliminate a small amount of time (thus latency) that is needed by the DRAM to detect the ras/cas and rd/wr signals. DRAM latches information to and from the controller on the active edge of the clock signal In addition to a lower latency I/O, after a proper page and column setup, an SDRAM may store the starting address internally and output new data on each active edge of the clock signal, as long as the requested data are consecutive memory locations. This is accomplished by adding a column address counter to the base DRAM architecture. This counter is seeded with a starting column address strobed in by the processor (or memory controller) and is thereafter incremented internally by the DRAM on each clock cycle. SDRAM

19. signalnameactiveI/Odescription CLKclockN/Ainputsystem clock nRSTresetlowinputsystem reset ADDR(20:0)memory addressN/Ainputmemory address for r/w access WnRaccess typeN/Ainput when low read transfer, when high write tran. nASaddress and data strobelowinputstarts transfer nLBE(3:0)input mask for datalowinputinput enable/disable for data DIN(31:0)data inputN/Ainputdata to be written into sdram A(10:0)address busN/Aoutput address or control signals into sdram BS(1:0)bank selectN/Aoutput determines bank to which commands are executed CKEclock enablehighoutputsdram CKE input DQM(3:0)data maskhighoutput sdram data masks, mask individual bytes during data write. nCAScolumn address strobelowoutputsdram nCAS input nCSchip selectlowoutputsdram chip select nRASrow address strobelowoutputsdram nRAS input nWEwrite enablelowoutputsdram nWE input nDTACKtransmission acknowledgelowoutput acknowledges data transfer, strobe for data output from sdram DATA_OUT_SDRdata busN/Aoutputsdram data in bus

22. ARM MIPS

25.  32 registers, each 32-bit wide. 30 are general purpose, R30(Hi) and R31(Low) are reserved for the results of long multiplication (64-bit result).

26.  The register file has three read ports. Reading is done asynchronously and the addresses of the registers to be read are provided by RA1, RA2 and RA3 (each 5-bit wide) while the data of these addressed registers are reflected on RD1, RD2 and RD3 (each 32-bit wide).

27.  Writing into the registers is done at the negative edge of a clock and is enabled by the write control signal. The address of the register to be written is supplied by WA (5-bit wide) and the data is provided by WD (32-bit wide).

29.  There are 32 32-bit registers in a MIPS datapath.  The register bank module includes ◦ two read register address inputs (5-bits each), ◦ one write register address input (5- bits), ◦ a write-enable signal input (1-bit), ◦ a write data input (32-bits), and ◦ two read data outputs (32-bits).

30.  On the positive edge of the write-enable input, the register bank will update the contents of the register at the write address with whatever data is on the write data bus.  The read data outputs are output asynchronously – i.e. they are a function of the read register addresses.