Sugar 2.0

1
Sugar 2.0 Proposal Presented to Accellera FVTC

• Cindy Eisner
• Joint Work with Dana Fisman

IBM Research Laboratory in Haifa
2
Overview

• Whats new in Sugar 2.0 (reminder)
• Whats new since February presentation
• Sugar with respect to requirements
• Sugar advantages
• Summary

3
Whats new in Sugar 2.0 (reminder)

• Sugar Foundation Language
• Three flavors
• Verilog, VHDL, EDL
• Four layers
• boolean, temporal, modeling, verification
• Verification layer
• verification directives , statement grouping
• Support for multiple clocks (strong/weak)
• Abort operator
• Formal definition of finite semantics

4
Whats new since February

• No major changes since draft presentation of last
  month
• Document proposal has been completed and fleshed
  out with more details
• Added a discussion of Sugar formulas which are
  checked on-the-fly
• VHDL flavor is left undefined at this time (but
  syntax has been checked so as not to conflict
  with VHDL boolean expressions)
• Two minor changes see next slides

5
Minor change 1 from February draft

• Simplified use of multiple clocks doesnt require
  as many parentheses

• always (p -gt X q_at_clkq) _at_clkp

6
Minor change 2 from February draft

• A module can be bound to a verification unit
• Meaning is that signals of the verification unit
  map by name to signals in the bound module
• Allows same specification to be used no matter
  how block is embedded in design
• Vunit can also be bound to a specific instance
  (requirement R50a)
• See example on next slide

7
Binding a vunit

• vunit default
• property pulsed(s) always (s -gt next !s)
• vunit myunit(blocka)
• assume pulsed(req_in)
• assert pulsed(ack_out)
• vunit yourunit(blockb)
• assert pulsed(req_out)
• assume pulsed (ack_in)

8
Sugar 2.0 with respect to requirements

• Sugar 2.0 proposal fulfills all requirements
• Highlights
• Layered definition
• Three flavors
• Powerful sequences
• Multiple clocks
• Reset support (through abort operator)
• Named properties, sequences, and endpoints
• Formal semantics both finite and infinite
• Can specify any omega-regular language

9
Advantage 1 History and Track Record

• Mature eight years of practical experience
• Extensive user base
• Contrast with CBV which has recently undergone a
  major revolution

10
Advantage 2 Easy to learn, use

• A Sugar specification says what it means
• always ((snoop hitm) gt
• !trans_start trans_start writeback)
• Contrast to somewhat convoluted formulation in
  CBV using not and fail
• always
• if (snoop hitm)
• (1) not fail !trans_start trans_start
  writeback

11
Advantage 3 Succinct

• A Sugar specification is concise
• forall N 1..8
• always within(write_command_start sizeN
  LAST)gx_startN
• Contrast with verbosity of CBV
• task sugar_within (r regexp, b bool, s
  regexp)
• if (r)
• not fail !b s b
• endtask
• function \sugar_.. (b bool, n nat)
  regexp
• return !bbn !b
• endfunction
• begin_and
• for (N nat 1 N lt 8 N N 1)
• always
• sugar_within
• (
• write_command_start (size N),
• LAST,

12
Advantage 4 Elegant formal semantics

• Semantics are
• well defined formal semantics appear in Sugar
  2.0 proposal
• elegant formal semantics are only 3 pages long
• Contrast with lengthy semantics document of CBV

13
Advantage 5 Known efficient model checking
algorithms

• Known model checking algorithms
• LTL and CTL model checking well-known
• Translation of regular expressions to automata is
  well-known
• Contrast with specialized algorithms of CBV

14
Advantage 6 Known efficient simulation
algorithms

• Known algorithms for finite time reasoning
• as implemented in FoCs
• implementation sketched in Sugar 2.0 proposal
• Contrast with CBV, whose implementation of the
  newer parts of the language is untried

15
Advantage 7 Declarative language

• Temporal layer is declarative
• Specification can be easily read
• Contrast with CBV, which mixes declarative and
  procedural code

16
Advantage 8 Expressive power

• Sugar Foundation Language
• allows the expression of any omega-regular
  language
• Optional Branching Extension
• allows specification of deadlock-freedom
• Contrast with less expressive CBV

17
Advantage 9 Built-in syntactic sugar

• Specifications are unambiguous, so can be easily
  read
• always ((snoop hitm) -gt next next_event(trans_st
  art)(writeback))
• Contrast with CBV, which requires the user to do
  a lot of the work
• function cbv_next_event(e bool) regexp
• return first_match 1e
• endfunction
• always
• if (snoop hitm)
• (1) if (cbv_next_event(trans_start))
• writeback

18
Summary

• Sugar is an elegant, powerful specification
  language with a strong track record both within
  and outside of IBM
• Sugar has existing implementations in both model
  checking (RuleBase) and simulation (Focs)
• Implementations are sketched in Sugar 2.0 proposal