The fields of upb_MiniTableField are intended to be internal-only, accessed only through public functions like `upb_MiniTable_GetSubMessageTable()`. But over time, clients have started accessing many of these fields directly. This is an easy mistake to make, as there is no clear signal that the fields should not be used in applications. This makes the implementation difficult to change without breaking users.
The new `UPB_PRIVATE()` macro appends an unpredictable string to each private symbol. This makes it very difficult to accidentally use a private symbol, since users would need to write something like `field->submsg_index_dont_copy_me__upb_internal_use_only`. This is still possible to do, but it leaves a clear wart in the code showing that an an encapsulation break has occurred. The `UPB_PRIVATE()` macro itself is defined in `port/def.inc`, which users cannot include directly.
Once we land this, more such CLs will follow for the other fields of `upb_MiniTable*`. We will add inline functions as needed to provide the semantic functionality needed by users.
PiperOrigin-RevId: 523166901
The offsetof trick is not compatible with an incoming Clang change.
Clang has started to enforce that it is UB to declare types inside offsetof. See e327b52766.
```
third_party/upb/upb/mem/arena.c:166:25: error: 'struct (unnamed at third_party/upb/upb/mem/arena.c:166:7)' cannot be defined in '__builtin_offsetof'
n = UPB_ALIGN_DOWN(n, UPB_ALIGN_OF(upb_Arena));
^
./third_party/upb/upb/port/def.inc:113:38: note: expanded from macro 'UPB_ALIGN_OF'
#define UPB_ALIGN_OF(type) offsetof (struct { char c; type member; }, member)
^
third_party/upb/upb/mem/arena.c:166:25: error: 'struct (unnamed at third_party/upb/upb/mem/arena.c:166:7)' cannot be defined in '__builtin_offsetof'
./third_party/upb/upb/port/def.inc:113:38: note: expanded from macro 'UPB_ALIGN_OF'
#define UPB_ALIGN_OF(type) offsetof (struct { char c; type member; }, member)
```
PiperOrigin-RevId: 501872556
This CL changes the upb compiler to no longer depend on C++ protobuf libraries. upb now uses its own reflection libraries to implement its code generator.
# Key Benefits
1. upb can now use its own reflection libraries throughout the compiler. This makes upb more consistent and principled, and gives us more chances to dogfood our own C++ reflection API. This highlighted several parts of the C++ reflection API that were incomplete.
2. This CL removes code duplication that previously existed in the compiler. The upb reflection library has code to build MiniDescriptors and MiniTables out of descriptors, but prior to this CL the upb compiler could not use it. The upb compiler had a separate copy of this logic, and the compiler's copy of this logic was especially tricky and hard to maintain. This CL removes the separate copy of that logic.
3. This CL (mostly) removes upb's dependency on the C++ protobuf library. We still depend on `protoc` (the binary), but the runtime and compiler no longer link against C++'s libraries. This opens up the possibility of speeding up some builds significantly if we can use a prebuilt `protoc` binary.
# Bootstrap Stages
To bootstrap, we check in a copy of our generated code for `descriptor.proto` and `plugin.proto`. This allows the compiler to depend on the generated code for these two protos without creating a circular dependency. This code is checked in to the `stage0` directory.
The bootstrapping process is divided into a few stages. All `cc_library()`, `upb_proto_library()`, and `cc_binary()` targets that would otherwise be circular participate in this staging process. That currently includes:
* `//third_party/upb:descriptor_upb_proto`
* `//third_party/upb:plugin_upb_proto`
* `//third_party/upb:reflection`
* `//third_party/upb:reflection_internal`
* `//third_party/upbc:common`
* `//third_party/upbc:file_layout`
* `//third_party/upbc:plugin`
* `//third_party/upbc:protoc-gen-upb`
For each of these targets, we produce a rule for each stage (the logic for this is nicely encapsulated in Blaze/Bazel macros like `bootstrap_cc_library()` and `bootstrap_upb_proto_library()`, so the `BUILD` file remains readable). For example:
* `//third_party/upb:descriptor_upb_proto_stage0`
* `//third_party/upb:descriptor_upb_proto_stage1`
* `//third_party/upb:descriptor_upb_proto`
The stages are:
1. `stage0`: This uses the checked-in version of the generated code. The stage0 compiler is correct and outputs the same code as all other compilers, but it is unnecessarily slow because its protos were compiled in bootstrap mode. The stage0 compiler is used to generate protos for stage1.
2. `stage1`: The stage1 compiler is correct and fast, and therefore we use it in almost all cases (eg. `upb_proto_library()`). However its own protos were not generated using `upb_proto_library()`, so its `cc_library()` targets cannot be safely mixed with `upb_proto_library()`, as this would lead to duplicate symbols.
3. final (no stage): The final compiler is identical to the `stage1` compiler. The only difference is that its protos were built with `upb_proto_library()`. This doesn't matter very much for the compiler binary, but for the `cc_library()` targets like `//third_party/upb:reflection`, only the final targets can be safely linked in by other applications.
# "Bootstrap Mode" Protos
The checked-in generated code is generated in a special "bootstrap" mode that is a bit different than normal generated code. Bootstrap mode avoids depending on the internal representation of MiniTables or the messages, at the cost of slower runtime performance.
Bootstrap mode only interacts with MiniTables and messages using public APIs such as `upb_MiniTable_Build()`, `upb_Message_GetInt32()`, etc. This is very important as it allows us to change the internal representation without needing to regenerate our bootstrap protos. This will make it far easier to write CLs that change the internal representation, because it avoids the awkward dance of trying to regenerate the bootstrap protos when the compiler itself is broken due to bootstrap protos being out of date.
The bootstrap generated code does have two downsides:
1. The accessors are less efficient, because they look up MiniTable fields by number instead of hard-coding the MiniTableField into the generated code.
2. It requires runtime initialization of the MiniTables, which costs CPU cycles at startup, and also allocates memory which is never freed. Per google3 rules this is not really a leak, since this memory is still reachable via static variables, but it is undesirable in many contexts. We could fix this part by introducing the equivalent of `google::protobuf::ShutdownProtobufLibrary()`).
These downsides are fine for the bootstrapping process, but they are reason enough not to enable bootstrap mode in general for all protos.
# Bootstrapping Always Uses OSS Protos
To enable smooth syncing between Google3 and OSS, we always use an OSS version of the checked in generated code for `stage0`, even in google3.
This requires that the google3 code can be switched to reference the OSS proto names using a preprocessor define. We introduce the `UPB_DESC(xyz)` macro for this, which will expand into either `proto2_xyz` or `google_protobuf_xyz`. Any libraries used in `stage0` must use `UPB_DESC(xyz)` rather than refer to the symbol names directly.
PiperOrigin-RevId: 501458451
We have previously been using Copybara to rewrite these names, but for bootstrapping we will want to be able to sometimes use OSS names inside google3.
PiperOrigin-RevId: 500294974
This simplifies the code generation by making output agnostic to whether fasttables will be used or not.
This grows the generated code in the common case, but when fasttables are not being used the preprocessor will strip away the unused tables.
PiperOrigin-RevId: 499340805
Prior to this CL, there were several different code paths for reading/writing message data. Generated code, MiniTable accessors, and reflection all performed direct manipulation of the bits and bytes in a message, but they all had distinct implementations that did not share much of any code. This divergence meant that they could easily have different behavior, bugs could creep into one but not another, and we would need three different sets of tests to get full test coverage. This also made it very difficult to change the internal representation in any way, since it would require updating many places in the code.
With this CL, the three different APIs for accessing message data now all share a common set of functions. The common functions all take a `upb_MiniTableField` as the canonical description of a field's type and layout. The lowest-level functions are very branchy, as they must test for every possible variation in the field type (field vs oneof, hasbit vs no-hasbit, different field sizes, whether a nonzero default value exists, extension vs. regular field), however these functions are declared inline and designed to be very optimizable when values are known at compile time.
In generated accessors, for example, we can declare constant `upb_MiniTableField` instances so that all values can constant-propagate, and we can get fully specialized code even though we are calling a generic function. On the other hand, when we use the generic functions from reflection, we get runtime branches since values are not known at compile time. But even the function is written to still be as efficient as possible even when used from reflection. For example, we use memcpy() calls with constant length so that the compiler can optimize these into inline loads/stores without having to make an out-of-line call to memcpy().
In this way, this CL should be a benefit to both correctness and performance. It will also make it easier to change the message representation, for example to optimize the encoder by giving hasbits to all fields.
Note that we have not completely consolidated all access in this CL:
1. Some functions outside of get/set such as clear and hazzers are not yet unified.
2. The encoder and decoder still touch the message without going through the common functions. The encoder and decoder require a bit more specialized code to get good performance when reading/writing fields en masse.
PiperOrigin-RevId: 490016095
upb has traditionally returned 16-byte-aligned pointers from arena allocation. This was out of an abundance of caution, since users could theoretically be using upb arenas to allocate memory that is then used for SSE/AVX values (eg. [`__m128`](https://docs.microsoft.com/en-us/cpp/cpp/m128?view=msvc-170), which require 16-byte alignment.
In practice, the protobuf C++ arena has used 8-byte alignment for 8 years with no significant problems I know of arising from SSE etc.
Reducing the alignment requirement to 8 will save memory. It will also help with compatibility on 32-bit architectures where `malloc()` only returns 8-byte aligned memory. The immediate motivation is to fix the win32 build for Python protobuf.
PiperOrigin-RevId: 448331777
Unfortunately a few of the Clang warnings did not have easy fixes:
../../../../ext/google/protobuf_c/ruby-upb.c: In function ‘fastdecode_err’:
../../../../ext/google/protobuf_c/ruby-upb.c:353:13: warning: function might be candidate for attribute ‘noreturn’ [-Wsuggest-attribute=noreturn]
353 | const char *fastdecode_err(upb_decstate *d) {
| ^~~~~~~~~~~~~~
../../../../ext/google/protobuf_c/ruby-upb.c: In function ‘_upb_decode’:
../../../../ext/google/protobuf_c/ruby-upb.c:867:30: warning: argument ‘buf’ might be clobbered by ‘longjmp’ or ‘vfork’ [-Wclobbered]
867 | bool _upb_decode(const char *buf, size_t size, void *msg,
I even tried to suppress the first error, but it still shows up.
upb previously attempted to support C89 and pre-2015 versions
of Visual Studio. This was to support older compilers with
limited C99 support (particularly MSVC). But as of last August,
even gRPC has dropped support for MSVC prior to 2015
c87276d058
Therefore it seems safe for upb to no longer attempt C89 support
(we were already not truly C89 compliant, with our use of "bool").
We now explicitly require C99 or greater and MSVC 2015 or greater.
This cleaned up port_def.inc a fair bit. I took the chance to
also remove some obsolete macros.
* resolvename is declared to return a bool value, but instead can return NULL. MSVC 2019 does not like that an throws a compile error. Fixed by returning false instead of NULL.
* When compiling with MSVC 2019, the UPB_ASSUME macro expands out to:
do {} if (false && (ok))
That isn't valid C code. Fixed by adding an elif for MSVC that uses __assume(0), which is similar to gcc's __builtin_unreachable according to http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2017/p0627r0.pdf.
* WIP.
* WIP.
* Tests are passing.
* Recover some perf: LIKELY doesn't propagate through functions. :(
* Added some more benchmarks.
* Simplify & optimize upb_arena_realloc().
* Only add owned blocks to the freelist.
* More optimization/simplification.
* Re-fixed the bug.
* Revert unintentional changes to parser.rl.
* Revert Lua changes for now.
* Revert the arena fuse changes for now.
* Added last_size to the arena representation.
* Fixed compile errors.
* Fixed compile error and changed benchmarks to do one allocation.
New code is smaller (in both source size and compiled size) and faster.
# Speed
The decoder speeds up on all machines I tested, though the amount of speedup varies. I was only able to test Intel CPUs.
### Linux Desktop
```
CPU: Intel(R) Core(TM) i7-8700K CPU @ 3.70GHz
OS: Linux
name old time/op new time/op delta
CreateArena 4.72ns ± 0% 4.93ns ± 0% +4.47% (p=0.000 n=11+11)
ParseDescriptor 12.4µs ± 1% 9.1µs ± 1% -26.65% (p=0.000 n=11+11)
```
### Mac Laptop
```
CPU: Intel(R) Core(TM) i7-8850H CPU @ 2.60GHz
OS: macOS
name old time/op new time/op delta
CreateArena 5.33ns ± 3% 5.58ns ± 2% +4.69% (p=0.000 n=12+12)
ParseDescriptor 15.0µs ± 2% 11.9µs ± 2% -20.20% (p=0.000 n=12+12)
```
### Linux Workstation
```
CPU: Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz
OS: Linux
name old time/op new time/op delta
CreateArena 5.29ns ± 0% 5.52ns ± 0% +4.37% (p=0.000 n=10+12)
ParseDescriptor 18.6µs ± 0% 16.4µs ± 0% -11.54% (p=0.000 n=12+12)
```
# Size
A few source files grow marginally because of some arena functionality moved inline. But `upb/decode.c` shrinks by 30% on Linux:
```
VM SIZE
--------------
+2.1% +283 upb/json_decode.c
+24% +205 upb/msg.c
+8.4% +115 upb/upb.c
+0.9% +28 upb/reflection.c
[ = ] 0 upb/def.c
[ = ] 0 upb/encode.c
[ = ] 0 upb/json_encode.c
[ = ] 0 upb/table.c
-30.3% -1.51Ki upb/decode.c
-0.7% -738 TOTAL
```
upb/json/parser.rl: In function 'end_member.isra.150':
bazel-out/k8-opt/bin/upb.c:5536:13: error: 'sel' may be used uninitialized in this function [-Werror=maybe-uninitialized]
upb_func *ret = (upb_func *)h->table[s].func;
Joshua Haberman
Dmitri Gribenko
Protobuf Team Bot
Eric Salo
Joshua Haberman
Esun Kim
Igor Kostenko
Charlie Savage
Anna R