RAM syntheisis

I almost completed synthesizing all the blocks exept SRAMS.

Memory compiler generatd RAM has following ports

output [127:0]QA;

inputCLKA;

inputCENA;

input [3:0] AA;

inputCLKB;

inputCENB;

input [3:0] AB;

input [127:0]DB;

input [2:0] EMAA

whereas OpensparcT1 rtl has the followig ports

input [127:0] din; // data input--> DB

input [15:0]rd_wl;// read addr

input [15:0]wr_wl; // write addr

input read_en;>CENA

input wr_en;//used in conjunction with->CENB

// word_wen and byte_wen

input rst_tri_en ; // gates off writes during SCAN.

input rclk;->CLKA,CLKB

input se, si ;

input reset_l;

input sehold; // hold scan in data.

output [127:0] dout;-->QA

output so;

Now, the problem is with the extra pins that RTL is having compared

to memory compiler o/p..

I tried to put a wrapper around the actual .But there is some glue logic

also to be implemented because ofrst_l,andsehold.Of course ,se,so

are not used.

And the address byte is 16 bit in the actual RTLcompared to 4-bit.

A 16 to 4 decoder is used.

I need somebody who can help me in this regard, Plz.

Most of my syntheisis work will be finished if I can close this issue.

If it is possible , I'm expecing a detailed reply .

best regards

nag

[1450 byte] By [nagvarre] at [2008-1-19]
# 1

SRAM models in OpenSPARC T1 RTL are generic behavioral models and you

need to match those with your physical implementation of the SRAM.

For each SRAM, you need to match the port names and functionality of the SRAM

using wrapper.

In the example SRAM in your question, there is 16 to 4 decoder in the

OpenSPARC T1 RTL (bw_r_rf16x128d.v) to convert addresses. You can

include that decoder and other glue logic for reset, etc. in your wrapper.

You need to develop your own wrapper to match the port names

and functionality.

OpenSparc at 2007-7-6 > top of java,Open Source Technologies,OpenSPARC...
# 2

the closing remark, regarding the sram, se,so may not totally closing the session,however the sytheisis flow of the internal compressor can change the module scdn and sieh. But who are we to determine if the opening bypass the spram dcnp or remains stable. Ha! tech problems are like math, you see it as problem, try it as the dec. solution

Krazy,

highprofiler at 2007-7-6 > top of java,Open Source Technologies,OpenSPARC...