open-source-silicon.dev

Has anyone used any of the `gf180mcu_fd_ip_sram` modules, without modification, in an openlane config file? `caravel_core` uses one here, but it’s in a <https://github.com/efabless/caravel-gf180mcu/blob/main/openlane/caravel_core/config.tcl|wrapper>. Is that the best way?

<@U017X0NM2E7>, I am also looking for 512x32 bit SRAM,   I don't able see lef/gds corresponds to 512x32 wrapper inside the efabless git repo: <https://github.com/efabless/caravel_mgmt_soc_gf180mcu.git> .
I am bit worried about the memory area needed for 512x32 memory in GF180nm
As per generated .lef in PDK:
64Byte - 431.86 BY 232.88
128Byte - 431.86 BY 268.88
258Byte - 431.86 BY 340.88 ;
512Byte - 431.86 BY 484.88 ;
i.e 512 * 4 Byte = 431x484x4*1.2 = 1mm2 (I have assumed 20% free space penalty)
Which is quite large compare to Sky130nm area of 683.1 BY 416.54  = 0.283mm2.
<@U0175T39732> Any comment

If I recall, these aren't really 180nm memories since this is the 5V process. It's been a while since I looked

I see four SRAM are part  GF180nm pdk Folder ?
-rw-r--r-- 1 dinesha pyfive 49733 Nov 17 16:27 gf180mcu_fd_ip_sram__sram64x8m8wm1.lef
-rw-r--r-- 1 dinesha pyfive 79888 Nov 17 16:27 gf180mcu_fd_ip_sram__sram512x8m8wm1.lef
-rw-r--r-- 1 dinesha pyfive 62309 Nov 17 16:27 gf180mcu_fd_ip_sram__sram256x8m8wm1.lef
-rw-r--r-- 1 dinesha pyfive 52789 Nov 17 16:27 gf180mcu_fd_ip_sram__sram128x8m8wm1.lef

<@U016EM8L91B> Are we allowed use these Memory in GFMPW-1 tapeout ?

<@U025YC3GX9R>:  Yes, you may use any of those SRAM blocks.

The GF180MCU process is a special version of GF180 which uses a thick oxide as the base gate thickness, then has a "thicker" oxide in addition to that.  All of the 3.3V IP was (as far as I can tell) farmed out by GF to others (probably ARM), and so they had only 5V IP to release to open source.  All that means that the process itself is more like 0.25um or 0.35um than 0.18um, and if you have only 5V IP, it's more like a 0.5um process.  So yes, it looks _very_ unfavorable when compared to SkyWater.

And Skywater itself is technically a 150nm process which looks less unfavorable when compared to other 130nm processes.

At least SkyWater does advertise itself (somewhere) as a "hybrid" process.  But for both of them, their advertised "feature size" is more marketing reality than physical reality.  I assume that GF180MCU is run on the same equipment as other GF180 processes, just with different specifications for the front-end.