Eliminate redundant residual re-read in optimized Ok() switch arms (#272)

Lever A of the Ok() rework: in a conditional-switch arm whose field carries
a residual predicate beyond the discriminant equality (e.g.
`if tag == K && outer == 1`), gate the field's Ok() check on the residual
alone rather than re-deriving the full has_<field>() existence condition.

Inside `case K:` the switch has already decoded and (provably, or via the
emitted Known() guard) established the discriminant, so
`has_<field>() <=> residual`. The prior form re-invoked has_<field>(), which
redundantly re-reads and re-compares the discriminant on every arm; the
residual (a single hoisted, shared subexpression) drops that per-arm work.

Implementation: preserve the residual conjunct list through
_generate_optimized_ok_method_body instead of collapsing it to a bool, thread
(ir, subexpressions) into _emit_switch_block/_render_case_body, and render the
residual at emit time (so demoted groups leak no dead subexpression). Multiple
conjuncts fold into one synthesized boolean AND so the residual renders once.
The gate is required for correctness, not just size: <field>() returns a null
view when absent, so an unguarded <field>().Ok() would wrongly fail when the
residual is false; we also bail when the residual is not Known.

Adds testdata OuterGuardedUnion (100 tag arms each guarded by a shared
`outer == 1` residual) to exercise and measure the change; flat
LargeConditionals (all bare arms) is the unaffected control.

Measured OuterGuardedUnion::Ok() reader symbol @ -Os vs the pre-Lever-A
generator (schema held fixed): MicroBlaze LE/Vitis (authoritative) -18.7%,
MicroBlaze BE -21.5%, ARM Cortex-M4 -17.0%, x86-64 gcc -9.5% / clang -50.7%.
Retired guest instructions (OuterGuardedUnion hot loop): MicroBlaze BE -3.2%,
MicroBlaze LE -1.6%, ARM -1.1%. Flat LargeConditionals codegen is
byte-identical (0 symbol delta on every target); x86 wall-clock unchanged.
Golden diff is confined to the residual-arm schemas (condition.emb.h,
many_conditionals.emb.h); every pre-existing struct is byte-identical. Full
compiler+runtime suite (205 tests) and the #266 condition_test oracle stay green.
4 files changed
tree: 25cf264d75103741ce609060fef5f43c2c43cc82
  1. .github/
  2. compiler/
  3. doc/
  4. gen/
  5. integration/
  6. runtime/
  7. scripts/
  8. testdata/
  9. .bazelrc
  10. .gitignore
  11. BUILD
  12. build.json
  13. build_defs.bzl
  14. CONTRIBUTING.md
  15. emboss-format
  16. embossc
  17. LICENSE
  18. license_header
  19. MODULE.bazel
  20. pyproject.toml
  21. README.md
  22. requirements.txt
README.md

Emboss

Emboss is a tool for generating code that reads and writes binary data structures. It is designed to help write code that communicates with hardware devices such as GPS receivers, LIDAR scanners, or actuators.

What does Emboss do?

Emboss takes specifications of binary data structures, and produces code that will efficiently and safely read and write those structures.

Currently, Emboss only generates C++ code, but the compiler is structured so that writing new back ends is relatively easy -- contact emboss-dev@google.com if you think Emboss would be useful, but your project uses a different language.

When should I use Emboss?

If you're sitting down with a manual that looks something like this or this, Emboss is meant for you.

When should I not use Emboss?

Emboss is not designed to handle text-based protocols; if you can use minicom or telnet to connect to your device, and manually enter commands and see responses, Emboss probably won't help you.

Emboss is intended for cases where you do not control the data format. If you are defining your own format, you may be better off using Protocol Buffers or Cap'n Proto or BSON or some similar system.

Why not just use packed structs?

In C++, packed structs are most common method of dealing with these kinds of structures; however, they have a number of drawbacks compared to Emboss views:

  1. Access to packed structs is not checked. Emboss (by default) ensures that you do not read or write out of bounds.
  2. It is easy to accidentally trigger C++ undefined behavior using packed structs, for example by not respecting the struct's alignment restrictions or by running afoul of strict aliasing rules. Emboss is designed to work with misaligned data, and is careful to use strict-aliasing-safe constructs.
  3. Packed structs do not handle variable-size arrays, nor arrays of sub-byte-size fields, such as boolean flags.
  4. Packed structs do not handle endianness; your code must be very careful to correctly convert stored endianness to native.
  5. Packed structs do not handle variable-sized fields, such as embedded substructs with variable length.
  6. Although unions can sometimes help, packed structs do not handle overlapping fields well.
  7. Although unions can sometimes help, packed structs do not handle optional fields well.
  8. Certain aspects of bitfields in C++, such as their exact placement within the larger containing block, are implementation-defined. Emboss always reads and writes bitfields in a portable way.
  9. Packed structs do not have support for conversion to human-readable text format.
  10. It is difficult to read the definition of a packed struct in order to generate documentation, alternate representations, or support in languages other than C and C++.

What does Emboss not do?

Emboss does not help you transmit data over a wire -- you must use something else to actually transmit bytes back and forth. This is partly because there are too many possible ways of communicating with devices, but also because it allows you to manipulate structures independently of where they came from or where they are going.

Emboss does not help you interpret your data, or implement any kind of higher-level logic. It is strictly meant to help you turn bit patterns into something suitable for your programming language to handle.

What state is Emboss in?

Emboss is currently under development. While it should be entirely ready for many data formats, it may still be missing features. If you find something that Emboss can't handle, please contact emboss-dev@google.com to see if and when support can be added.

Emboss is not an officially supported Google product: while the Emboss authors will try to answer feature requests, bug reports, and questions, there is no SLA (service level agreement).

Getting Started

Head over to the User Guide to get started.