( ESNUG 568 Item 2 ) -------------------------------------------- [03/17/17]
Subject: SCOOP -- leaked HiSilicon 16FF+ IC Compiler II slides from DAC'16
2.) My true #2 "must see" is about an IC Compiler II possibility.
Synopsys IC Compiler II -- ever since ICC2's launch in March 2014,
Aart's marketing machine has been projecting that ICC2 works and
was a "game changer". The problem is again and again I found that
IC Compiler II had tech issues from Day One (ESNUG 537 #10, 538 #1)
and onwards in (ESNUG 547 #7, 548 #1, 550 #1, 552 #6, 554 #3) that
looked especially bad since Aart was up on stage during those 27
months saying "ICC2 works! Honest!". On DAC Monday they're doing
an ICC2 lunch. I'd like to attend to see if Aart's R&D has finally
fixed ICC2's multi-databases/capacity/runtime/coloring problems.
This could be a major tipping point for IC Compiler II. Or not.
(booth 149) Ask for Antun Domic. Freebie: lunch
- "My Cheesy Must See List for DAC 2016"
From: [ John Cooley of DeepChip.com ]
Enclosed are 8 slides from the Synopsys IC Compiler II DAC'16 lunch. It
appears to be a short talk by Huawei (HiSilicon) engineers about how they
designed their Kirin 955 processor in TSMC 16FF+ quad ARM Cortex A72's and
quad ARM Cortex A53's, at 2.5Ghz/1.8Ghz.
I have more detailed slides on this same design in ESNUG 568 #3. Enjoy.
- John Cooley
DeepChip.com Holliston, MA
P.S. And a BIG thank you to the spies who sourced me these slides!!! :)
P.P.S. If *anyone* has anon comments they want to add to (or explain) any
of these individual slides, CONTACT ME. I'll edit in your insights.
(And say exactly which specific slide you're commenting on, please.)
---- ---- ---- ---- ---- ---- ----
Slide 1 -- (click on pic to enlarge image)
|
---- ---- ---- ---- ---- ---- ----
Slide 2 -- (click on pic to enlarge image)
|
---- ---- ---- ---- ---- ---- ----
Slide 3 -- (click on pic to enlarge image)
|
---- ---- ---- ---- ---- ---- ----
Slide 4 -- (click on pic to enlarge image)
|
---- ---- ---- ---- ---- ---- ----
Slide 5 -- (click on pic to enlarge image)
|
READER COMMENTS: The HiSilcon engineers themselves ran ICC, it was the
Synopsys AE's who ran ICC II on this chip.
ICC2 is still a taxicab mode tool.
---- ---- ---- ---- ---- ---- ----
Slide 6 -- (click on pic to enlarge image)
|
READER COMMENTS: Much better placement in same runtime!
What's CCD? Is it a type of CTS?
---- ---- ---- ---- ---- ---- ----
Slide 7 -- (click on pic to enlarge image)
|
---- ---- ---- ---- ---- ---- ----
Slide 8 -- (click on pic to enlarge image)
|
---- ---- ---- ---- ---- ---- ----
Related Articles
Anirudh and Sawicki on IC Compiler II, Innovus, Nitro-SoC, Antun
The untold parts of that IMEC "world's first 5nm tapeout" story
ICC2 patch rev, Innovus penetration, and the 10nm layout problem
Join
Index
Next->Item
|
