mirror of
https://github.com/carbon-language/carbon-lang.git
synced 2026-09-24 22:02:23 +01:00
ca8df34d0d495de29cb97f979daae2d006742ad1
52
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
bf32da8dad |
Add missing standard library header inclusions (#5316)
Discovered by clang-tidy. |
||
|
|
acbe6530c3 |
Move diagnostics into a namespace (#5173)
What this really does is avoids shadowing names, so that we can comfortable have things like `Check::DiagnosticEmitter` or `Check::DiagnosticLoc` without shadowing being a concern. Note, down this path I'm also thinking about: - Renaming misc DiagnosticConsumer/DiagnosticEmitter classes, possibly just to DiagnosticConsumer/DiagnosticEmitter (so `Check::DiagnosticEmitter` instead of `SemIRLocDiagnosticEmitter`). - Dropping `Diagnostic` from `Emitter::DiagnosticBuilder`. - But not for `Check::DiagnosticBuilder`, because `Check::Builder` would be ambiguous. - Renaming diagnostics/diagnostic_* to drop "diagnostic". [Discussion about SemIRLoc -> DiagnosticLoc](https://discord.com/channels/655572317891461132/655578254970716160/1353771570463768698) reminded me of this (in particular the older [Check::DiagnosticBuilder discussion](https://discord.com/channels/655572317891461132/655578254970716160/1344363562608627763)), but I'd only do that rename if there's matching consensus about a path forward where we keep SemIRLoc, and in a way that it's only ever used for diagnostics (the divergence from which is at the root of current LocId discussion). I'm trying to keep that separate from a namespace addition for clarity. |
||
|
|
e79d3be5bd |
Combine DiagnosticConverter into DiagnosticEmitter (#4878)
At present, we typically define a DiagnosticConverter, then store an instance of it and a DiagnosticEmitter that wraps it. This is relatively minor in general, but I've been trying to create more self-contained DiagnosticEmitter classes (which hold their own DiagnosticConverter, similar to NullDiagnosticEmitter), and there it just gets in the way. Since we don't reuse DiagnosticConverter instances, this combines the definition into DiagnosticEmitter. Mainly this means we don't have a separate object in play, and less to carry around. The most impact is probably to SemIRDiagnosticConverter, which was also the most complex. Now `SemIRLocDiagnosticEmitter`, this gets some different construction flow. Note in the PR I've split the file rename to its own commit, to try to help delta views. However, the most substantial parts of the refactoring are split into #4876, which this depends upon. |
||
|
|
133717cd7e |
Eliminate NodeLocConverter (#4870)
I'm looking at eliminating `DiagnosticConverter`. This change removes `NodeLocConverter` (albeit adding `UnitAndImportsDiagnosticConverter`), and in doing so, refactors lex conversion functions to extract them out from the `DiagnosticConverter` functions. I'll be following up with changes that collapse `DiagnosticConverter` logic into `DiagnosticEmitter` locations. The intent is that we shouldn't need separate ownership of both types. |
||
|
|
b06fcc97f6 |
Clean up a few details of lex yaml printing (#4845)
- Escape dumped token strings (what got me here) - Change the quoting from backticks to quotes - Also add a `FormatEscaped` helper function for this, updating other `.write_escaped` uses |
||
|
|
8f685b6953 |
Change how diagnostics are ordered (#4778)
This change deliberately breaks away from the line/column ordering, and instead focuses on a last byte offset corresponding to the final token processed as part of producing the message. Where that's equal, this maintains stable ordering in order to reflect the order that diagnostics were produced. The intent of this approach is that lex, parse, and check diagnostics are interleaved based on where they are produced, but that subexpressions still have diagnostics emitted prior to containing expressions. In particular, the prior line/column sort essentially sorted on the _start_ of where a diagnostic was associated, and this is closer to sorting based on the _end_. As a consequence, something like `F(1 2)` will have the error for `1 2` emitted _before_ a diagnostic for `F(1 2)` not matching parameters, instead of _after_. In check, we track the last handled node. This provides a last_byte_offset _separate_ from where a diagnostic is associated. The intent is that this creates an ordering of diagnostics which may be associated with earlier code, to cause the diagnostics to be emitted later. An example consequence of this is the change in ordering of modifier diagnostics: we are diagnosing those from the same place, but they have the same last_byte_offset, so we print them out in the order produced. I've added similar tracking to parse, but cannot identify any test which is affected by it (note the separate commit, I thought about this late). I'm not sure whether we have good out-of-order errors we could produce for this. A significant number of tests have reordered diagnostics as a consequence of this change, so this change does not add further testing. |
||
|
|
28602a87c2 |
Fix handling of repeated tuple indexing. (#4733)
Per [the design](https://docs.carbon-lang.dev/docs/design/lexical_conventions/), `x.1.2` should lex as `(x.1).2`, not as `x.(1.2)`. |
||
|
|
3ce0df67bb |
Add Dump functions to Check, Parse, and Lex (#4669)
- Provide `Check::Dump(context, arg)` and similar. - gdb and lldb should do contextual lookup, and `call Dump(*this, Lex::TokenIndex::Invalid)` has been tested with gdb. - Since this is only for debug, keeps the functions fully separated from code. - Uses alwayslink to ensure objects are correctly linked, even though there are no calls. - `-Wno-missing-prototypes` is needed when we don't have forward declarations. - Code is not linked in opt builds, using `#ifndef NDEBUG`. - This probably could be doing something in BUILD files with a `select()`, but the `#ifndef` seemed easier. This is based on #4620, but uses free functions instead of member functions. Co-authored-by: Dana Jansens <danakj@orodu.net> --------- Co-authored-by: danakj <danakj@orodu.net> |
||
|
|
61c0a8b676 |
Make more use of llvm STLExtras (#4668)
This is essentially the result of looking at `.begin()` uses. We also frequently do `std::shuffle`, but unfortunately STLExtras doesn't provide a wrapper for that. |
||
|
|
361efa90a8 |
Always call MemUsage::Collect to collect metrics from a field (#4480)
Previously Collect() was used for types that implemented CollectMemUsage() but otherwise Add() was used. This required the caller to think about the type of the field and know/decide which method to use. Now, the caller always uses Collect() unless they are adding specific byte values, in which case Add is used. Typically then, Add will only be used to implement the CollectMemUsage() function. To do this we require all Collect() methods to be templates so that they all be a single overload set. The Collect on BumpPtrAllocator is converted to a template that checks `std::same_as<llvm::BumpPtrAllocator, T>`. |
||
|
|
4148161e24 |
Refactor value store code to use separate files. (#4477)
This is in anticipation of making the integer value store be customized heavily. I'd like to extract it from the common code when doing that, so first disentangling them here without any intended change in functionality or behavior to enable that. I've tried to update `#include`s to be as minimal as I can and added a few missing includes spotted in the process. I've split the test for value store to include what was easy focused on just the value store templates rather than the unified shared value stores. This might surface some opportunities for adding more tests, but for this PR, just doing the minimal restructuring. |
||
|
|
06f4eec91e |
Modify lex yaml output to elide FileStart/End in tests. (#4433)
Trying to make split file tests of lex functionality shorter and easier to read. numeric_literals.carbon in particular has an example of why I'm interested in this (at the bottom). This also switches from `[]` list format to `-` list format so that the trailing `]` is removed. Trimming comments in tokenized_buffer.h because (1) it feels like it's giving too much detail about what's printed, which has drifted slightly and (2) it also feels like it's trying to justify YAML output, when that's just what we're doing in general. --------- Co-authored-by: Geoff Romer <gromer@google.com> |
||
|
|
0db96ebc52 |
Stitch together adjacent comments using the indent. (#4397)
This is improving the comment production to produce fewer distinct comments. At present, comment processing uses strict prefix matching. It either expects `// ` (with a space) for valid comments, or just `//` (without a space) for invalid comments that lacked the space. As a consequence, the following would be three comments: ``` // Comment 1 // // // Comment 4 ``` This is because a 3-character prefix is used for valid comments. The prefix switches between lines 1 and 2, and again between lines 3 and 4, each resulting in a separate comment. For contrast, this is one comment because only a 2-character prefix is used: ``` //Comment 1 // // //Comment 4 ``` That's because all lines lack a suffix space. Additionally, with SIMD 16-byte boundaries, further splits can occur if processing needs to transition to non-SIMD. Here, I'm trying to just address all of this by: 1. Stitching together adjacent comments. Since a lexed comment starts at the `//` excluding the indent, the delta from the prior comment must be precisely the indent. 2. Adding support for switching from SIMD to non-SIMD on file boundaries. I considered trying to have a separate `//\n` prefix for SIMD processing of `// `, but I wasn't sure about the tradeoff of doing both at the same time (in particular, it'd require constructing a string for the different prefix), thus this stitch approach. This does mean multiple passes will be required for a typical long comment structure using blank comment lines to separate paragraphs (for performance reasons, I will recommend engineers not write comm... nevermind). --------- Co-authored-by: Chandler Carruth <chandlerc@gmail.com> |
||
|
|
1338f9e0ad |
Add tracking of lexed comments, with skeletal formatting. (#4385)
In order to format comments, it's helpful if they're tracked. This tracks them separately from tokens in order to avoid interfering with parse; it'd be inconvenient if comment tokens could show up in arbitrary locations, albeit possible to support. This additionally extracts out the TokenIterator support into a template in order to generally have it available for IndexBase types. I'm only adding it for CommentInfo, not sure if we'll want it elsewhere, but this structure still felt like a good fit. |
||
|
|
06344aeb7c |
Do some tactical inlining across lexer and parser. (#4307)
These are based on looking at our compilation benchmark and looking at function bodies that seem surprising to not get inlined. Note that this will have a bit more impact on x86 where function call overhead (especially due to pushing and popping registers) is a bit higher than Arm. For a recent AMD server, this makes parsing around 15% faster, and full "check" phase 5% faster. Benchmark results: ``` name old cpu/op new cpu/op delta BM_CompileAPIFileDenseDecls<Phase::Lex>/256 40.2µs ± 2% 37.8µs ± 1% -5.89% (p=0.000 n=19+17) BM_CompileAPIFileDenseDecls<Phase::Lex>/1024 190µs ± 2% 181µs ± 2% -4.93% (p=0.000 n=19+18) BM_CompileAPIFileDenseDecls<Phase::Lex>/4096 779µs ± 1% 745µs ± 2% -4.29% (p=0.000 n=19+19) BM_CompileAPIFileDenseDecls<Phase::Lex>/16384 3.44ms ± 1% 3.32ms ± 3% -3.32% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Lex>/65536 14.6ms ± 2% 14.3ms ± 3% -2.46% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Lex>/262144 66.7ms ± 2% 65.0ms ± 4% -2.52% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/256 85.7µs ± 2% 71.3µs ± 2% -16.77% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/1024 421µs ± 2% 352µs ± 2% -16.38% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/4096 1.71ms ± 2% 1.44ms ± 2% -15.89% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/16384 7.19ms ± 2% 6.10ms ± 2% -15.24% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/65536 29.8ms ± 2% 25.3ms ± 2% -14.91% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/262144 127ms ± 2% 109ms ± 2% -14.28% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Check>/256 785µs ± 1% 752µs ± 1% -4.13% (p=0.000 n=20+18) BM_CompileAPIFileDenseDecls<Phase::Check>/1024 1.71ms ± 1% 1.62ms ± 1% -5.17% (p=0.000 n=20+18) BM_CompileAPIFileDenseDecls<Phase::Check>/4096 5.28ms ± 1% 4.97ms ± 1% -6.04% (p=0.000 n=20+19) BM_CompileAPIFileDenseDecls<Phase::Check>/16384 20.2ms ± 1% 19.0ms ± 2% -5.98% (p=0.000 n=20+20) BM_CompileAPIFileDenseDecls<Phase::Check>/65536 83.8ms ± 1% 78.9ms ± 2% -5.84% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Check>/262144 354ms ± 1% 335ms ± 1% -5.41% (p=0.000 n=19+20) ``` |
||
|
|
4845f40dff |
Switch CARBON_CHECK to a format string API (#4285)
This switches `DCHECK` and `FATAL` as well. The goal is to reduce the code size impact of these assertions so that we can keep more of them enabled. Currently, the largest cost I see from `CHECK` is not the actual check or the cold code itself, but actually the failure to inline trivial functions due to the presence of the cold code. This means that our goal isn't to reduce apparent code size in the final binary but the LLVM IR cost assessed for these routines in the inliner, which closely correlates with code size but is a bit different. As discussed in #4283, experimentation shows that a single function call with a minimal number of arguments is the lowest cost model for these. This is easily achieved with a format-string API that internally uses `llvm::formatv`. This PR is essentially the `CHECK` version of #4283. However, the check macros are substantially harder to make work with both format strings and streaming because they also take a condition. Also, unexpectedly, I was very successful at devising a regular expression based automated rewrite from the streaming to the format string form with only low 10s of manual fixes. This includes compacting strings broken up across lines, etc. Given how well that went, I've prepared this PR which just directly switches to the format string API and migrate everything to use it. One nice side-effect is that the format string approach ends up greatly simplifying the implementation here as well. This is ... *shockingly* effective. Parsing speeds up by more than 3% with just this change. And checking speeds up by **8%** with this change alone: ``` BM_CompileAPIFileDenseDecls<Phase::Parse>/256 86.3µs ± 1% 82.9µs ± 1% -3.94% (p=0.000 n=17+19) BM_CompileAPIFileDenseDecls<Phase::Parse>/1024 431µs ± 1% 415µs ± 1% -3.76% (p=0.000 n=18+19) BM_CompileAPIFileDenseDecls<Phase::Parse>/4096 1.77ms ± 1% 1.71ms ± 1% -3.18% (p=0.000 n=18+19) BM_CompileAPIFileDenseDecls<Phase::Parse>/16384 7.44ms ± 1% 7.17ms ± 2% -3.56% (p=0.000 n=18+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/65536 30.7ms ± 1% 29.7ms ± 1% -3.15% (p=0.000 n=18+20) BM_CompileAPIFileDenseDecls<Phase::Parse>/262144 131ms ± 1% 127ms ± 1% -2.81% (p=0.000 n=18+18) BM_CompileAPIFileDenseDecls<Phase::Check>/256 878µs ± 2% 800µs ± 1% -8.91% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Check>/1024 1.88ms ± 2% 1.72ms ± 1% -8.56% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Check>/4096 5.78ms ± 2% 5.28ms ± 1% -8.70% (p=0.000 n=20+18) BM_CompileAPIFileDenseDecls<Phase::Check>/16384 21.9ms ± 1% 20.1ms ± 1% -8.02% (p=0.000 n=18+20) BM_CompileAPIFileDenseDecls<Phase::Check>/65536 90.4ms ± 2% 83.1ms ± 1% -8.04% (p=0.000 n=19+20) BM_CompileAPIFileDenseDecls<Phase::Check>/262144 381ms ± 2% 352ms ± 1% -7.79% (p=0.000 n=19+19) ``` --------- Co-authored-by: Richard Smith <richard@metafoo.co.uk> Co-authored-by: josh11b <15258583+josh11b@users.noreply.github.com> |
||
|
|
d6b2fb1736 |
Add parse support for multiple requirements after where separated by and (#4298)
Follow on to #4275 that added `where` parse support. --------- Co-authored-by: Josh L <josh11b@users.noreply.github.com> Co-authored-by: Richard Smith <richard@metafoo.co.uk> |
||
|
|
c43fa3a8a5 |
Bit-pack the lexer's token info (#4270)
This makes each token info consist of 8 bytes of data: - 1 byte of the kind - 1 bit for whitespace tracking - 23 bits of payload - 32 bits for byte offset in the file This builds directly on representing the location of the token as a single 32-bit offset, now compressing the rest of the data into a single 32-bit bitfield. This adds some implementation limits: we can no longer lex more than 2^23 tokens in a single source file. Nor can we have more than 2^23 string literals, integer literals, real literals, or identifiers. Only the first of these is even close to an issue, and even then seems unlikely to ever be a problem in practice. The memory efficiency here is great and the motivating goal. But to make this work well, we also need to streamline how we create the tokens. Otherwise, all the bit fiddling can end up erasing our gains. This PR adds a number of APIs to manage creating and accessing the now significantly more complex storage of token infos to try and help with this. One big change required to simplify the writes here is to switch from computing whether a token has trailing space after-the-fact to pre-computing whether a token will have leading space. That lets us have the leading space information available immediately when forming the token, and avoids doing a single bit flip afterward. Another change that helps with this representation is to minimize the updating of groups after-the-fact. The code now tries to set the opening index directly when creating the closing token and only updates the opening group afterward. Because of the bit packing, this is a reduction of 0.5% of dynamic instructions in the compile benchmark, and has dramatic improvements for the grouping symbol focused benchmarks. All combined, this is a significant improvement on the lexer-focused benchmarks despite the added complexity, and a significant win on our compile time benchmarks due to both the lexer improvements and downstream memory density improvements: 5-12% reduction in lex time, growing larger as files get larger. About a 4.5% reduction in parse time, and even a 1-2% reduction in total check time. =D --------- Co-authored-by: Jon Ross-Perkins <jperkins@google.com> Co-authored-by: Geoff Romer <gromer@google.com> |
||
|
|
97e98bcc5a |
Shrink the lexer's token location and line data structures. (#4269)
First, this replaces the separate line index and column index in the token information with a single 32-bit byte offset of the token. This is then used to compute line and column numbers with a binary search of the line structure and then using that to compute the column within the line. In practice, this is _much_ more efficient: - Smaller token data structure. This will hopefully combine with a subsequent optimization PR that shrinks the token data structure still further. - Fewer stores to form each token's information in the tight hot loop of the lexer. - Less state to maintain while lexing, fewer computations while lexing. We only have to search to build the line and column information off the hot lexing path, and so this ends up being a significant win and shrinks some of the more significant data structures. Second, this shrinks the line start to a 32-bit integer and removes the line length. Our source buffer already ensures we only have 2 GiB of source with a nice diagnostic. I've just added a check to help document this in the lexer. The line length can be avoided in all of the cases it was being used, largely by looking at the next line's start and working from there. This also precipitated cleaning up some code that dated from when lines were only built during lexing rather than being pre-built, which resulted in nice simplifications. With this PR, I think it makes sense to re-name a bunch of methods on `TokenizedBuffer`, but to an extent that was already needed as these methods somewhat predate the more pervasive style conventions. I avoided that here to keep this PR focused on the implementation change, I'll create a subsequent PR to update the API to both better nomenclature and remove deviations from our conventions. There may also be a way to de-duplicate the binary search in the diagnostic location conversion and the main line accessor binary search, but it wasn't obvious to me that it would be a net savings, so left it alone for now. The performance impact of this varies quite a bit... The lexer's benchmark improves pretty consistent across the board on both x86 and Arm. For x86, where I have nice comparison tools, it appears 3% to 20% faster depending on the specific pattern. For Arm server CPUs at least it seems a much smaller but still an improvement. The overall compilation benchmarks however don't improve much with these changes alone on x86. Significant reduction in instruction count required for lexing, but the overall performance is bottlenecked elsewhere in the overall compilation it seems. However, on Arm, despite the more modest gains in special cases of lexing, this shows fairly consistent 1-2% improvements in overall lexing performance on our compilation benchmark. And the expectaiton is these improvements will compound with subsequent work to further compact our representation. --------- Co-authored-by: Jon Ross-Perkins <jperkins@google.com> |
||
|
|
f1190a4792 |
Add basic output of where memory is stored after a compile. (#4136)
The output is really basic, I'm just adding this to help track how memory is allocated. ``` --- filename: 'check/testdata/expr_category/in_place_tuple_init.carbon' source_: used_bytes: 8057 reserved_bytes: 8057 tokens_.allocator_: used_bytes: 0 reserved_bytes: 0 tokens_.token_infos_: used_bytes: 1040 reserved_bytes: 2032 (eliding) value_stores_.string_literals_.set_: used_bytes: 320 reserved_bytes: 320 Total: used_bytes: 20609 reserved_bytes: 29437 ... ``` --------- Co-authored-by: josh11b <15258583+josh11b@users.noreply.github.com> |
||
|
|
bb117aea3a |
Add support for iN and uN for all suitable N. (#3868)
`i32` is retained as a special case for now, for bootstrapping purposes, and maps to `BuiltinIntType`, which is distinct from `Core.Int(32)`. This will be removed later once we support `Core.BigInt`. For now this provides both the `iN` types and also the builtins to support `Core.Int(N)`. The intent is that we'll change the `iN` support to rewrite to calls here when we do that for the other type literals and type keywords. No conversions between integer types are supported yet, and all literals are of type `i32`, so we can't actually form values of any of these new types. --------- Co-authored-by: Jon Ross-Perkins <jperkins@google.com> Co-authored-by: Carbon Infra Bot <carbon-external-infra@google.com> |
||
|
|
b5d28f2c4b | location -> loc abbreviation (#3826) | ||
|
|
6c458ffe7e |
Add import context for locations. (#3807)
As discussed around #3792, identify the import a diagnostic message came from prior to the diagnostic message itself. This occurs during location translation so that the logic can be central. I'd considered associating the parse node with ImportRef instructions, but I realized about halfway through that because I need to store the ImportDirectiveId on the ImportIR for cross-package imports, it's there for use in location translation without extra work. That saves a fair amount of stringing it through declarations, as well as an oddity where ImportRef instructions would have a node that didn't really represent them. |
||
|
|
0bd45f0d6b |
Rename DiagnosticLocationTranslator -> DiagnosticConverter (#3804)
Since the addition of TranslateArg, I don't think this type is going to go away (cutting a TODO). Refactoring names slightly to fit the current role, and adding const to ConvertLocation. |
||
|
|
4421a75c36 |
file_name -> filename (#3791)
I wanted to choose one or the other. I think some code has been using each from early on. We're predominately using `filename`, so consolidating on that. This conveniently matches the [Google dev doc style guide](https://developers.google.com/style/word-list#filename) (which we use for docs) which says "filename: Not file name". In toolchain: ``` ╚╡git grep file_name . | wc -l 35 ╚╡git grep filename . | wc -l 489 ``` |
||
|
|
fea2651e7c |
chore: fix typos (#3738)
Signed-off-by: cui fliter <imcusg@gmail.com> |
||
|
|
0a06fceb5f |
Improve diagnosis of mismatched brackets. (#3282)
Move handling of mismatched brackets out of the main lexing loop into a separate pass that is only run if there are mismatched brackets This is done in preparation for using both lookahead and lookbehind to work out how to match brackets, and to get this code far away from the hot lexing loop. Fix bracket insertion location to be immediately after the token that we're inserting the bracket after, rather than potentially at the end of a comment. When there are open brackets at the end of the file, say that there are open brackets, not that there's a closing bracket without a matching opening bracket. |
||
|
|
d87fe8b532 |
Rename Carbon::StringLiteralId -> Carbon::StringLiteralValueId. (#3522)
We have `StringLiteral`s in multiple other `Carbon` sub-namespaces. Rename to a more specific name to avoid collisions. We should likely also rename `Carbon::IntId` -> `Carbon::IntValueId` and `Carbon::RealId` -> `Carbon::RealValueId`, but this collision is prioritized because it was blocking work on typed parse nodes which introduces a `Carbon::Parse::StringLiteralId`. |
||
|
|
0db63ff17a |
Abbreviate Integer and FloatingPoint (#3435)
I was suggesting this because `FloatingPoint` is pretty long. `int` and `float` should be familiar abbreviations. `unsigned` should be familiar to developers too, but `UnsignedInt` still feels usefully clearer for the additional chars. |
||
|
|
eae630a3db |
Rename Lex::{Token,Line} -> Lex::{Token,Line}Index. (#3433)
As discussed [on discord](https://discord.com/channels/655572317891461132/655578254970716160/1178878128714678282) and today's toolchain discussion. |
||
|
|
3f208e27f9 |
Align on FileStart/FileEnd for naming. (#3428)
The lexer has been using EndOfFile form (stemming from EOF), parser went to FileEnd form. This consolidates on FileEnd form. |
||
|
|
1d443a3617 |
Underline the entire token when producing diagnostic messages. (#3413)
This is an incremental improvement on our diagnostic messages that simply underlines an entire token if the token is larger than 1 char (else it points to the single char with a caret like it used to). |
||
|
|
cafcd88882 |
Split lexing logic and storage to separate files. (#3365)
Just reorganizing logic a little, trying to mirror the direction we've gone with check, lower, etc. That is, lex.h contains a function `Lex` that is used directly. Note, I'm avoiding making meaningful changes here. It could in theory still affect inlining in benchmarks, but I'm not seeing an impact. Before: ``` ------------------------------------------------------------------------------------------------------ Benchmark Time CPU Iterations UserCounters... ------------------------------------------------------------------------------------------------------ BM_ValidKeywords 2784949 ns 2784867 ns 249 bytes_per_second=214.452M/s tokens_per_second=35.9084M/s BM_ValidKeywordsAsRawIdentifiers 3222597 ns 3222551 ns 210 bytes_per_second=244.513M/s tokens_per_second=31.0313M/s BM_RawIdentifierFocus 5907836 ns 5907518 ns 103 bytes_per_second=264.873M/s tokens_per_second=16.9276M/s BM_ValidIdentifiers<1, 64, false> 6255128 ns 6254297 ns 105 bytes_per_second=235.488M/s tokens_per_second=15.989M/s BM_ValidIdentifiers<1, 1, true> |
||
|
|
3401eed8d8 |
Split IdentifierId and StringLiteralId from StringId (#3352)
Following up on discussion yesterday regarding this split. Note, I'm expecting #3341 to do IdentifierId -> NameId in SemIR. It might be worth adding NameId creation directly to StringStore if you're content with this setup though. |
||
|
|
6742d0d048 |
Add partial raw identifier support. (#3344)
I'm looking at this due to the conversation on #3341. Although diagnostics aren't where they should be, I thought it may help to start adding raw identifier support (which may also help show how I was thinking about this). Note regarding the TODO on how to form the token, `GetTokenText` returns the `string_id`'s reference value for an `Identifier`. So to make `GetTokenText` work in a way that returns `r#foo` for a raw identifier, I think there are a few options: 1. Add additional data indicating the end of the identifier. 2. Add `RawIdentifier` as a token kind to indicate that it's raw and should be prefixed with `r#` (but also giving later stages one more token kind to handle) 3. Make the `string_id` correspond to `r#foo`, and have later stages add `foo` to the strings table whenever `r#foo` is encountered (with map lookups leading to deduplication). 4. Add `StringId::RawKeyword` special values for each keyword. - This would mean `self` prints as `self`, `r#self` prints as `r#self`, but `r#foo` is not a keyword so prints as `foo`. - This means keywords would need to be listed in a place `StringId` can depend on them, one way or the other (e.g., a `keywords.def` file in `base/` should work). 5. Say that it _is_ an `Identifier`, and if it's a keyword spelling, it must have been a raw identifier. - Same limitation as above: This would mean `self` prints as `self`, `r#self` prints as `r#self`, but `r#foo` is not a keyword so prints as `foo`. I'm hoping to resolve this issue separately though. :) |
||
|
|
b2cfd5a8a8 |
Change StringLiteral to less frequently allocate a new string. (#3314)
Building on #3311, which started moving the result string into a `unique_ptr`, instead have `StringLiteral` use a `BumpPtrAllocator` to manage memory. But also, detect when a string is really trivial during `Lex` and, if so, return `contents_` directly. |
||
|
|
1b0e2d3a4b |
Cleanups of SIMD code and document no Arm port. (#3325)
I spent (a lot) of time working to see if there was any profitable way to port the SIMD code that scans for identifier length to Arm. There isn't really. =/ While working on these, I made some cleanups to the SIMD code that seemed worth landing, and added some benchmarks. All this PR does is the cleanups, benchmarks, and documents that Arm isn't just waiting to get attention but doesn't really have good options (so far). For posterity, here are the core techniques I tried: 1) Direct 32-byte SIMD scanning using pair-wise add trees to build a 32-bit mask of valid identifier and then `clz` to compute the distance. This is a very good analog to the 16-byte SIMD structure used on x86-64. The pair-wise summing technique is the one used in simdjson for similar purposes. 2) A 16-byte SIMD scanning similar to the x86 version but using `shrn` to produce a 64-bit scalar bitmask with 4 bits per byte, and then scaling the bit-count distance. 3) Various hybrid versions of (1) and (2) with short scalar scans to identify short identifiers before paying the SIMD start-up cost. 4) A much fancier version of (1) that scanned 64-bytes at a time, but cached the resulting 64-bit mask and re-used it until exhausted. Some good background on these techniques on Arm CPUs is in this blog post: https://community.arm.com/arm-community-blogs/b/infrastructure-solutions-blog/posts/porting-x86-vector-bitmask-optimizations-to-arm-neon Sadly, both (1) and (2) were significantly slower than a scalar loop over the bytes. Even (3) was consistently slower. The only approach that came close was (4) and it was very *slightly* slower in typical examples and very *slightly* faster in extremely difficult cases like huge identifiers. Ultimately, the only path I see (suggested by Dougall on a Mastodon discussion of this whole problem space) is to take (4) to the limit of computing an identifier-or-not bitmask *for the entire source file* using a deeply throughput optimized routine (maybe as part of the line scanning). That should be able to manage the high latency you end up with when handling these patterns in SIMD on Arm. The good news is that at least the M1 is *so* fast in the byte-scanning loop that this isn't hurting nearly as much as I feared. --------- Co-authored-by: Jon Ross-Perkins <jperkins@google.com> |
||
|
|
1b55ad86dd |
Extend SharedValueStores to SemIR (#3313)
Building on #3311, change SemIR to use the SharedValueStore. Since this removes hermeticity, raw output no longer prints ints, reals, and strings. TokenizedBuffer accessors are modified to return IDs because values are often passed through in semantics without needing to read them. I would've put SharedValueStores on Context, except for the GetArrayBoundValue convenience method. I felt awkward removing that, so it's on File, at least for now. That's then used by the formatter and Lower too. The flipside of this is that TokenizedBuffer has a SharedValueStores only for printing, so maybe that's similar enough to what File is doing. This doesn't start shifting other SemIR members to ValueStore, but that seems like a next step. |
||
|
|
d13f76e001 |
Add value store to be shared across compile stages. (#3311)
This updates lexing to use the data. I'll do checking separately, just to split changes. Note the ValueStore structure is also set up such that SemIR::File can use it for other fields. --------- Co-authored-by: Richard Smith <richard@metafoo.co.uk> |
||
|
|
629c63c7d1 |
Port the comment block scanning SIMD to Arm. (#3300)
This adds an Arm Neon code path for comment block scanning. It also tries to tightened up the exact architecture-specific coding pattern for SIMD code here. Notably, moving to always having architecture-specific code inside a macro that can be used to globally disable SIMD, but any place where *all* architectures will require some custom code, an `else` branch with an error. Overall, this makes large-block lexing >5% faster on an ARM server I have access to, but that's a bit misleading. The Neon performance there doesn't seem very good. On my M1 laptop the difference is *much* larger. There, even 4-line comments show a noticable improvement and the block speed looks well over 20%. Sadly, I don't have the same nice scripts to generate good statistical data. Raw benchmark data from an ARM sever for reference: ``` BM_CommentLines/1/0/0 19.4ms ± 7% 19.7ms ± 6% ~ (p=0.121 n=20+20) BM_CommentLines/4/0/0 24.8ms ± 7% 24.8ms ± 6% ~ (p=0.904 n=20+20) BM_CommentLines/128/0/0 251ms ± 2% 234ms ± 3% -6.70% (p=0.000 n=20+20) BM_CommentLines/1/30/0 21.6ms ±10% 21.9ms ±10% ~ (p=0.157 n=20+20) BM_CommentLines/4/30/0 28.8ms ± 9% 28.8ms ±11% ~ (p=0.779 n=20+20) BM_CommentLines/128/30/0 248ms ± 2% 233ms ± 2% -5.72% (p=0.000 n=20+20) BM_CommentLines/1/70/0 23.4ms ±12% 23.7ms ±12% ~ (p=0.341 n=20+20) BM_CommentLines/4/70/0 30.8ms ± 9% 31.1ms ±11% ~ (p=0.602 n=20+20) BM_CommentLines/128/70/0 302ms ± 4% 292ms ± 4% -3.46% (p=0.000 n=18+20) BM_CommentLines/1/0/2 19.8ms ± 7% 20.0ms ± 6% ~ (p=0.149 n=20+20) BM_CommentLines/4/0/2 25.1ms ± 7% 25.3ms ± 8% ~ (p=0.659 n=20+20) BM_CommentLines/128/0/2 225ms ± 2% 212ms ± 2% -5.88% (p=0.000 n=20+20) BM_CommentLines/1/30/2 22.0ms ± 9% 22.2ms ±10% ~ (p=0.289 n=20+20) BM_CommentLines/4/30/2 29.0ms ±11% 29.1ms ±12% ~ (p=0.738 n=20+20) BM_CommentLines/128/30/2 261ms ±10% 243ms ± 3% -6.85% (p=0.000 n=20+20) BM_CommentLines/1/70/2 23.5ms ±11% 23.8ms ±15% ~ (p=0.429 n=20+20) BM_CommentLines/4/70/2 31.3ms ±10% 31.5ms ±11% ~ (p=0.478 n=20+20) BM_CommentLines/128/70/2 306ms ± 4% 292ms ± 4% -4.52% (p=0.000 n=18+19) BM_CommentLines/1/0/8 20.9ms ± 8% 21.2ms ± 7% ~ (p=0.127 n=20+20) BM_CommentLines/4/0/8 27.3ms ± 9% 27.5ms ±12% ~ (p=0.678 n=20+20) BM_CommentLines/128/0/8 227ms ± 2% 210ms ± 2% -7.35% (p=0.000 n=19+20) BM_CommentLines/1/30/8 22.6ms ±11% 23.0ms ±10% ~ (p=0.114 n=20+20) BM_CommentLines/4/30/8 29.4ms ±10% 29.4ms ±12% ~ (p=0.947 n=20+20) BM_CommentLines/128/30/8 275ms ± 4% 257ms ± 7% -6.59% (p=0.000 n=19+20) BM_CommentLines/1/70/8 23.9ms ±13% 24.3ms ±14% ~ (p=0.265 n=20+20) BM_CommentLines/4/70/8 32.3ms ±11% 32.4ms ± 9% ~ (p=0.478 n=20+20) BM_CommentLines/128/70/8 319ms ± 4% 307ms ± 4% -3.83% (p=0.000 n=18+19) ``` --------- Co-authored-by: Richard Smith <richard@metafoo.co.uk> Co-authored-by: Jon Ross-Perkins <jperkins@google.com> |
||
|
|
3015135a52 |
Skip blocks of comments with identical prefixes. (#3299)
Specifically, after lexing a comment line, look at the next line and see if it starts with an identical sequence of indent, comment '/'s and character after the '/'s. If so, skip it as part of a block of comments. This skips repeatedly diagnosing the same erroneous comment introducer after the first one in a block, but that seems like a feature rather than a bug. The big motivation is to make sure the lexer is minimally impacted by the length of comment blocks and skips them as efficiently as possible. While they aren't exactly common, large block comments do come up and it'd be unfortunate for those to actually slow down the toolchain. It also happens that this is particularly easy to do because we're just looking to see if we see the same prefix byte sequence. With SIMD we can typically handle the most common indents with just a few instructions. Because of the diagnostic differences, I've included a scalar fallback that replicates the functionality but has no limit on indent size or CPU features. I've also added testing to cover this behavior. The only non-noise benchmark changes are as expected the comment ones, with a nice improvement across the board: ``` BM_CommentLines/1/0/0 15.8ms ± 2% 15.6ms ± 2% -0.87% (p=0.004 n=19+19) BM_CommentLines/4/0/0 20.2ms ± 1% 18.8ms ± 1% -6.75% (p=0.000 n=18+18) BM_CommentLines/128/0/0 221ms ± 1% 167ms ± 1% -24.44% (p=0.000 n=20+19) BM_CommentLines/1/30/0 16.6ms ± 3% 16.5ms ± 3% ~ (p=0.175 n=19+20) BM_CommentLines/4/30/0 26.1ms ± 1% 24.8ms ± 2% -5.05% (p=0.000 n=18+19) BM_CommentLines/128/30/0 233ms ± 1% 185ms ± 1% -20.38% (p=0.000 n=19+20) BM_CommentLines/1/70/0 19.2ms ± 1% 19.0ms ± 2% -0.66% (p=0.016 n=19+20) BM_CommentLines/4/70/0 27.9ms ± 1% 26.6ms ± 1% -4.63% (p=0.000 n=19+19) BM_CommentLines/128/70/0 251ms ± 1% 213ms ± 1% -15.18% (p=0.000 n=20+18) BM_CommentLines/1/0/2 15.9ms ± 1% 15.8ms ± 2% ~ (p=0.061 n=19+19) BM_CommentLines/4/0/2 20.5ms ± 2% 19.0ms ± 2% -7.53% (p=0.000 n=20+20) BM_CommentLines/128/0/2 213ms ± 1% 153ms ± 1% -28.18% (p=0.000 n=19+20) BM_CommentLines/1/30/2 16.8ms ± 2% 16.7ms ± 3% ~ (p=0.134 n=20+20) BM_CommentLines/4/30/2 26.6ms ± 1% 25.2ms ± 3% -5.50% (p=0.000 n=20+20) BM_CommentLines/128/30/2 238ms ± 1% 187ms ± 2% -21.49% (p=0.000 n=17+19) BM_CommentLines/1/70/2 19.3ms ± 1% 19.4ms ± 3% ~ (p=0.407 n=17+20) BM_CommentLines/4/70/2 28.2ms ± 1% 26.9ms ± 2% -4.70% (p=0.000 n=19+19) BM_CommentLines/128/70/2 257ms ± 2% 214ms ± 1% -16.52% (p=0.000 n=20+18) BM_CommentLines/1/0/8 16.3ms ± 2% 16.1ms ± 2% -1.22% (p=0.001 n=20+20) BM_CommentLines/4/0/8 22.7ms ± 2% 20.4ms ± 2% -10.20% (p=0.000 n=20+20) BM_CommentLines/128/0/8 244ms ± 1% 153ms ± 1% -37.26% (p=0.000 n=20+18) BM_CommentLines/1/30/8 17.3ms ± 2% 17.2ms ± 3% ~ (p=0.192 n=20+20) BM_CommentLines/4/30/8 28.0ms ± 2% 25.6ms ± 3% -8.46% (p=0.000 n=19+18) BM_CommentLines/128/30/8 272ms ± 1% 196ms ± 2% -27.90% (p=0.000 n=18+20) BM_CommentLines/1/70/8 19.9ms ± 2% 19.9ms ± 2% ~ (p=0.531 n=20+19) BM_CommentLines/4/70/8 29.3ms ± 1% 27.3ms ± 1% -6.87% (p=0.000 n=19+19) BM_CommentLines/128/70/8 292ms ± 1% 228ms ± 1% -21.97% (p=0.000 n=20+19) ``` --------- Co-authored-by: Richard Smith <richard@metafoo.co.uk> |
||
|
|
7371354dc7 |
Consolidate indent handling to following newlines. (#3298)
The biggest advantage of this is reducing the repeated code in every non-whitespace code path of the lexer to set indent appropriately. Now we handle it cleanly at the start and after vertical whitespace. While this makes the generated code for all the other paths through the lexer quite a bit nicer, it doesn't actually move performance in interesting ways outside of making large blocks of blank lines slightly slower. While there are lots of fluctuations in the benchmark data, they mostly seem to be either noise or artifacts of loop alignment and not really due to an important change here. Raw benchmark data: ``` BM_ValidKeywords 3.10ms ± 1% 3.13ms ± 1% +1.02% (p=0.000 n=20+19) BM_ValidIdentifiers<1, 64, false> 10.8ms ± 3% 10.7ms ± 3% ~ (p=0.383 n=20+20) BM_ValidIdentifiers<1, 1, true> 3.71ms ± 1% 3.67ms ± 2% -1.05% (p=0.000 n=19+19) BM_ValidIdentifiers<3, 5, true> 13.1ms ± 2% 13.0ms ± 2% ~ (p=0.120 n=20+19) BM_ValidIdentifiers<3, 16, true> 13.1ms ± 2% 13.2ms ± 2% ~ (p=0.091 n=20+20) BM_ValidIdentifiers<12, 64, true> 15.1ms ± 1% 15.1ms ± 1% ~ (p=0.138 n=19+19) BM_HorizontalWhitespace/1 13.2ms ± 3% 13.2ms ± 0% ~ (p=0.458 n=20+15) BM_HorizontalWhitespace/4 13.4ms ± 3% 13.3ms ± 1% -1.12% (p=0.000 n=20+19) BM_HorizontalWhitespace/16 14.1ms ± 2% 14.0ms ± 2% -0.89% (p=0.010 n=20+20) BM_HorizontalWhitespace/64 17.9ms ± 2% 17.8ms ± 1% -0.73% (p=0.002 n=20+16) BM_HorizontalWhitespace/128 24.2ms ± 2% 24.1ms ± 1% ~ (p=0.346 n=19+17) BM_RandomSource 7.88ms ± 2% 7.83ms ± 3% ~ (p=0.052 n=20+20) BM_GroupingSymbols/1/0/0 6.25ms ± 2% 6.38ms ± 1% +1.99% (p=0.000 n=20+19) BM_GroupingSymbols/2/0/0 5.24ms ± 2% 5.27ms ± 2% ~ (p=0.065 n=20+19) BM_GroupingSymbols/3/0/0 4.09ms ± 1% 4.07ms ± 1% ~ (p=0.063 n=20+20) BM_GroupingSymbols/4/0/0 3.80ms ± 1% 3.79ms ± 2% ~ (p=0.820 n=20+20) BM_GroupingSymbols/8/0/0 3.05ms ± 1% 3.09ms ± 1% +1.41% (p=0.000 n=20+20) BM_GroupingSymbols/16/0/0 2.95ms ± 1% 3.01ms ± 1% +1.79% (p=0.000 n=20+20) BM_GroupingSymbols/32/0/0 3.67ms ± 1% 3.77ms ± 1% +2.75% (p=0.000 n=20+20) BM_GroupingSymbols/0/1/0 5.66ms ± 1% 5.61ms ± 1% -0.88% (p=0.000 n=20+18) BM_GroupingSymbols/0/2/0 4.43ms ± 2% 4.40ms ± 1% -0.74% (p=0.005 n=20+17) BM_GroupingSymbols/0/3/0 3.15ms ± 2% 3.12ms ± 2% -0.96% (p=0.002 n=20+19) BM_GroupingSymbols/0/4/0 2.79ms ± 2% 2.77ms ± 2% -0.88% (p=0.005 n=20+20) BM_GroupingSymbols/0/8/0 1.81ms ± 2% 1.79ms ± 2% ~ (p=0.056 n=20+20) BM_GroupingSymbols/0/16/0 1.26ms ± 2% 1.26ms ± 2% ~ (p=0.547 n=20+20) BM_GroupingSymbols/0/32/0 1.05ms ± 1% 0.96ms ± 1% -8.52% (p=0.000 n=19+19) BM_GroupingSymbols/0/0/1 5.68ms ± 2% 5.65ms ± 2% ~ (p=0.126 n=20+18) BM_GroupingSymbols/0/0/2 4.44ms ± 2% 4.40ms ± 2% -0.99% (p=0.001 n=20+18) BM_GroupingSymbols/0/0/3 3.15ms ± 1% 3.13ms ± 1% -0.74% (p=0.005 n=20+18) BM_GroupingSymbols/0/0/4 2.80ms ± 1% 2.77ms ± 2% -1.01% (p=0.000 n=20+19) BM_GroupingSymbols/0/0/8 1.81ms ± 2% 1.79ms ± 2% -0.97% (p=0.006 n=20+20) BM_GroupingSymbols/0/0/16 1.26ms ± 1% 1.26ms ± 2% ~ (p=0.678 n=20+20) BM_GroupingSymbols/0/0/32 1.05ms ± 1% 0.96ms ± 1% -8.59% (p=0.000 n=20+19) BM_GroupingSymbols/32/1/0 3.57ms ± 1% 3.69ms ± 1% +3.39% (p=0.000 n=19+20) BM_GroupingSymbols/32/2/0 3.49ms ± 1% 3.60ms ± 1% +3.16% (p=0.000 n=20+20) BM_GroupingSymbols/32/3/0 3.41ms ± 1% 3.52ms ± 1% +3.49% (p=0.000 n=19+20) BM_GroupingSymbols/32/4/0 3.35ms ± 1% 3.45ms ± 1% +3.13% (p=0.000 n=20+20) BM_GroupingSymbols/32/8/0 3.12ms ± 1% 3.22ms ± 1% +3.09% (p=0.000 n=19+20) BM_GroupingSymbols/32/16/0 2.75ms ± 1% 2.82ms ± 1% +2.71% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/0 2.24ms ± 2% 2.24ms ± 1% ~ (p=0.311 n=20+17) BM_GroupingSymbols/32/32/1 2.20ms ± 1% 2.23ms ± 1% +1.23% (p=0.000 n=20+20) BM_GroupingSymbols/32/32/2 2.19ms ± 1% 2.21ms ± 1% +1.08% (p=0.000 n=20+20) BM_GroupingSymbols/32/32/3 2.16ms ± 0% 2.19ms ± 1% +1.09% (p=0.000 n=18+20) BM_GroupingSymbols/32/32/4 2.14ms ± 1% 2.19ms ± 1% +2.25% (p=0.000 n=20+20) BM_GroupingSymbols/32/32/8 2.07ms ± 1% 2.11ms ± 1% +2.04% (p=0.000 n=17+20) BM_GroupingSymbols/32/32/16 1.94ms ± 1% 1.98ms ± 1% +1.95% (p=0.000 n=20+20) BM_GroupingSymbols/32/32/32 1.74ms ± 1% 1.77ms ± 1% +1.88% (p=0.000 n=18+20) BM_BlankLines/1 14.2ms ± 1% 14.4ms ± 1% +1.47% (p=0.000 n=18+16) BM_BlankLines/4 17.3ms ± 1% 17.5ms ± 2% +1.32% (p=0.000 n=17+20) BM_BlankLines/16 31.3ms ± 1% 33.0ms ± 2% +5.35% (p=0.000 n=19+20) BM_BlankLines/64 89.6ms ± 1% 101.3ms ± 1% +12.98% (p=0.000 n=20+20) BM_BlankLines/128 167ms ± 3% 185ms ± 1% +11.11% (p=0.000 n=19+20) BM_CommentLines/1/0/0 15.7ms ± 1% 15.8ms ± 1% ~ (p=0.109 n=19+16) BM_CommentLines/4/0/0 19.8ms ± 1% 20.2ms ± 1% +2.12% (p=0.000 n=20+19) BM_CommentLines/128/0/0 208ms ± 1% 221ms ± 1% +6.29% (p=0.000 n=20+19) BM_CommentLines/1/30/0 16.6ms ± 1% 16.6ms ± 1% ~ (p=0.354 n=19+19) BM_CommentLines/4/30/0 25.8ms ± 2% 26.1ms ± 1% +1.39% (p=0.000 n=20+19) BM_CommentLines/128/30/0 222ms ± 2% 232ms ± 1% +4.46% (p=0.000 n=20+18) BM_CommentLines/1/70/0 19.2ms ± 2% 19.1ms ± 1% ~ (p=0.478 n=20+19) BM_CommentLines/4/70/0 27.6ms ± 2% 27.9ms ± 1% +0.95% (p=0.001 n=20+18) BM_CommentLines/128/70/0 245ms ± 2% 250ms ± 1% +1.96% (p=0.000 n=20+18) BM_CommentLines/1/0/2 16.0ms ± 1% 15.9ms ± 2% -0.68% (p=0.015 n=15+18) BM_CommentLines/4/0/2 20.6ms ± 1% 20.5ms ± 1% -0.53% (p=0.024 n=18+19) BM_CommentLines/128/0/2 216ms ± 2% 213ms ± 1% -1.70% (p=0.000 n=17+19) BM_CommentLines/1/30/2 16.9ms ± 2% 16.8ms ± 3% ~ (p=0.072 n=20+20) BM_CommentLines/4/30/2 26.8ms ± 2% 26.5ms ± 1% -0.91% (p=0.000 n=20+19) BM_CommentLines/128/30/2 244ms ± 2% 238ms ± 1% -2.30% (p=0.000 n=19+17) BM_CommentLines/1/70/2 19.4ms ± 2% 19.4ms ± 2% ~ (p=0.059 n=18+20) BM_CommentLines/4/70/2 28.4ms ± 2% 28.2ms ± 1% -0.81% (p=0.003 n=20+19) BM_CommentLines/128/70/2 261ms ± 2% 255ms ± 1% -2.33% (p=0.000 n=20+18) BM_CommentLines/1/0/8 16.2ms ± 2% 16.2ms ± 2% ~ (p=0.904 n=20+20) BM_CommentLines/4/0/8 25.0ms ± 1% 22.6ms ± 1% -9.40% (p=0.000 n=20+19) BM_CommentLines/128/0/8 246ms ± 2% 244ms ± 1% -1.14% (p=0.000 n=20+19) BM_CommentLines/1/30/8 17.3ms ± 2% 17.3ms ± 1% ~ (p=1.000 n=19+18) BM_CommentLines/4/30/8 30.3ms ± 2% 27.9ms ± 1% -7.83% (p=0.000 n=20+20) BM_CommentLines/128/30/8 277ms ± 3% 271ms ± 1% -2.26% (p=0.000 n=20+17) BM_CommentLines/1/70/8 19.9ms ± 2% 19.9ms ± 2% ~ (p=0.687 n=19+20) BM_CommentLines/4/70/8 31.7ms ± 2% 29.3ms ± 1% -7.55% (p=0.000 n=20+18) BM_CommentLines/128/70/8 296ms ± 2% 291ms ± 1% -1.50% (p=0.000 n=20+18) ``` |
||
|
|
95a1cc8cba |
Tweak lexer to improve generated code. (#3296)
This is a collection of tweaks and they are ones I'm less confident in FWIW. I set out to make the generated code cleaner and reduce loading pointers through pointers in a bunch of cases. It also works to reduce the working-set-size, and reduce the set of mutated values on each iteration. But I didn't benchmark at each step and it's not clear that incremental benchmarks will even be meaningful, so its hard to say which tweaks were load bearing and which weren't. For example, with this, the core lexer dispatch loop doesn't mutate anything in memory -- the position is passed in register and incremented in register throughout. The pointer to the source text, the size, and the pointer to the lexer are also passed in registers. By making the tokenized buffer a direct member of the lexer, all of its members can be accessed at a constant offset from the lexer pointer itself. The cost of this is a move at the end of lexing, but especially as the buffer size gets large this seems like a trivial cost compared to the previous double-indirection. This uses a (significantly) cheaper representation for the line index -- the `Line` type is optimized for dense *storage*, and is not a great type for using in a tight loop like the lexer. Once using a good index and once the line table is readily available from the above, we can simply index the line table rather than store (and update) a separate pointer. Last but not least, this removes the column as suggested in a previous review. Now it is computed from the position and the line information. My macro benchmark looks like it improves in the 2% - 5% range, but its getting into the noise sadly (or maybe this is good?). On a quiet AMD server with 20 runs before/after I get reasonably compelling across-the-board improvements on all the microbenchmarks, including 5% on `RandomSource` which is key: ``` BM_ValidKeywords 3.44ms ± 1% 3.11ms ± 1% -9.52% (p=0.000 n=19+20) BM_ValidIdentifiers<1, 64, false> 11.2ms ± 1% 10.8ms ± 2% -3.23% (p=0.000 n=19+20) BM_ValidIdentifiers<1, 1, true> 3.96ms ± 1% 3.72ms ± 1% -6.10% (p=0.000 n=18+20) BM_ValidIdentifiers<3, 5, true> 13.5ms ± 1% 13.2ms ± 3% -2.45% (p=0.000 n=19+20) BM_ValidIdentifiers<3, 16, true> 13.5ms ± 2% 13.1ms ± 2% -2.81% (p=0.000 n=19+20) BM_ValidIdentifiers<12, 64, true> 15.6ms ± 1% 15.2ms ± 3% -2.41% (p=0.000 n=19+20) BM_HorizontalWhitespace/1 13.6ms ± 2% 13.3ms ± 3% -2.34% (p=0.000 n=19+20) BM_HorizontalWhitespace/4 13.9ms ± 2% 13.6ms ± 3% -2.06% (p=0.000 n=19+20) BM_HorizontalWhitespace/16 14.5ms ± 2% 14.3ms ± 2% -1.17% (p=0.002 n=19+20) BM_HorizontalWhitespace/64 18.5ms ± 3% 18.1ms ± 2% -2.17% (p=0.000 n=19+19) BM_HorizontalWhitespace/128 24.7ms ± 1% 24.4ms ± 2% -1.29% (p=0.000 n=18+20) BM_RandomSource 8.38ms ± 2% 7.92ms ± 2% -5.50% (p=0.000 n=19+20) BM_GroupingSymbols/1/0/0 6.63ms ± 2% 6.30ms ± 2% -4.97% (p=0.000 n=19+20) BM_GroupingSymbols/2/0/0 5.65ms ± 2% 5.27ms ± 2% -6.75% (p=0.000 n=19+20) BM_GroupingSymbols/3/0/0 4.51ms ± 1% 4.12ms ± 1% -8.62% (p=0.000 n=19+18) BM_GroupingSymbols/4/0/0 4.23ms ± 1% 3.82ms ± 1% -9.65% (p=0.000 n=19+18) BM_GroupingSymbols/8/0/0 3.50ms ± 1% 3.08ms ± 1% -12.21% (p=0.000 n=19+19) BM_GroupingSymbols/16/0/0 3.38ms ± 1% 2.97ms ± 1% -11.98% (p=0.000 n=19+19) BM_GroupingSymbols/32/0/0 4.10ms ± 1% 3.69ms ± 1% -10.02% (p=0.000 n=17+19) BM_GroupingSymbols/0/1/0 5.85ms ± 2% 5.70ms ± 1% -2.54% (p=0.000 n=19+20) BM_GroupingSymbols/0/2/0 4.60ms ± 2% 4.46ms ± 1% -3.12% (p=0.000 n=19+19) BM_GroupingSymbols/0/3/0 3.32ms ± 2% 3.18ms ± 1% -4.21% (p=0.000 n=18+20) BM_GroupingSymbols/0/4/0 2.94ms ± 2% 2.82ms ± 1% -4.22% (p=0.000 n=19+20) BM_GroupingSymbols/0/8/0 1.95ms ± 1% 1.83ms ± 1% -6.47% (p=0.000 n=17+19) BM_GroupingSymbols/0/16/0 1.40ms ± 1% 1.27ms ± 1% -9.12% (p=0.000 n=19+20) BM_GroupingSymbols/0/32/0 1.19ms ± 1% 1.05ms ± 1% -11.50% (p=0.000 n=19+20) BM_GroupingSymbols/0/0/1 5.88ms ± 2% 5.71ms ± 1% -2.87% (p=0.000 n=19+20) BM_GroupingSymbols/0/0/2 4.62ms ± 1% 4.45ms ± 2% -3.58% (p=0.000 n=18+20) BM_GroupingSymbols/0/0/3 3.32ms ± 1% 3.17ms ± 1% -4.72% (p=0.000 n=16+17) BM_GroupingSymbols/0/0/4 2.95ms ± 1% 2.81ms ± 2% -4.50% (p=0.000 n=16+19) BM_GroupingSymbols/0/0/8 1.96ms ± 1% 1.82ms ± 2% -6.80% (p=0.000 n=19+20) BM_GroupingSymbols/0/0/16 1.40ms ± 1% 1.27ms ± 1% -8.96% (p=0.000 n=19+20) BM_GroupingSymbols/0/0/32 1.19ms ± 1% 1.05ms ± 1% -11.64% (p=0.000 n=19+20) BM_GroupingSymbols/32/1/0 4.04ms ± 1% 3.60ms ± 1% -10.81% (p=0.000 n=17+20) BM_GroupingSymbols/32/2/0 3.93ms ± 1% 3.51ms ± 1% -10.55% (p=0.000 n=18+20) BM_GroupingSymbols/32/3/0 3.84ms ± 1% 3.44ms ± 1% -10.38% (p=0.000 n=18+19) BM_GroupingSymbols/32/4/0 3.76ms ± 2% 3.38ms ± 1% -10.21% (p=0.000 n=19+20) BM_GroupingSymbols/32/8/0 3.51ms ± 1% 3.14ms ± 1% -10.61% (p=0.000 n=19+19) BM_GroupingSymbols/32/16/0 3.11ms ± 1% 2.77ms ± 1% -10.95% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/0 2.54ms ± 2% 2.25ms ± 1% -11.77% (p=0.000 n=19+16) BM_GroupingSymbols/32/32/1 2.50ms ± 1% 2.22ms ± 2% -11.19% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/2 2.48ms ± 1% 2.21ms ± 1% -11.15% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/3 2.46ms ± 1% 2.18ms ± 1% -11.11% (p=0.000 n=18+20) BM_GroupingSymbols/32/32/4 2.43ms ± 1% 2.16ms ± 1% -11.04% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/8 2.35ms ± 1% 2.09ms ± 1% -11.11% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/16 2.20ms ± 1% 1.96ms ± 1% -11.30% (p=0.000 n=19+18) BM_GroupingSymbols/32/32/32 1.98ms ± 1% 1.75ms ± 1% -11.54% (p=0.000 n=19+19) BM_BlankLines/1 14.6ms ± 1% 14.2ms ± 3% -2.53% (p=0.000 n=19+19) BM_BlankLines/4 18.3ms ± 1% 17.4ms ± 2% -4.68% (p=0.000 n=19+20) BM_BlankLines/16 34.9ms ± 2% 31.6ms ± 3% -9.42% (p=0.000 n=19+20) BM_BlankLines/64 102ms ± 2% 91ms ± 3% -10.75% (p=0.000 n=19+20) BM_BlankLines/128 190ms ± 2% 169ms ± 3% -10.82% (p=0.000 n=19+20) BM_CommentLines/1/0/0 16.2ms ± 1% 15.8ms ± 2% -2.29% (p=0.000 n=19+19) BM_CommentLines/4/0/0 21.0ms ± 1% 19.9ms ± 2% -5.17% (p=0.000 n=19+19) BM_CommentLines/128/0/0 237ms ± 1% 211ms ± 2% -10.93% (p=0.000 n=19+20) BM_CommentLines/1/30/0 17.0ms ± 1% 16.8ms ± 3% -1.36% (p=0.001 n=17+20) BM_CommentLines/4/30/0 27.2ms ± 2% 26.0ms ± 2% -4.24% (p=0.000 n=19+20) BM_CommentLines/128/30/0 255ms ± 2% 223ms ± 2% -12.58% (p=0.000 n=19+20) BM_CommentLines/1/70/0 19.7ms ± 1% 19.2ms ± 1% -2.35% (p=0.000 n=18+18) BM_CommentLines/4/70/0 29.0ms ± 2% 27.6ms ± 1% -4.81% (p=0.000 n=18+17) BM_CommentLines/128/70/0 273ms ± 7% 249ms ± 3% -9.05% (p=0.000 n=20+20) BM_CommentLines/1/0/2 16.4ms ± 2% 16.0ms ± 1% -2.49% (p=0.000 n=19+15) BM_CommentLines/4/0/2 22.0ms ± 2% 20.7ms ± 1% -5.84% (p=0.000 n=19+17) BM_CommentLines/128/0/2 256ms ± 2% 218ms ± 1% -14.88% (p=0.000 n=19+17) BM_CommentLines/1/30/2 17.4ms ± 1% 17.0ms ± 2% -2.26% (p=0.000 n=19+19) BM_CommentLines/4/30/2 28.5ms ± 1% 26.9ms ± 2% -5.67% (p=0.000 n=19+18) BM_CommentLines/128/30/2 288ms ± 5% 246ms ± 3% -14.70% (p=0.000 n=19+20) BM_CommentLines/1/70/2 19.9ms ± 1% 19.4ms ± 1% -2.56% (p=0.000 n=19+18) BM_CommentLines/4/70/2 30.0ms ± 2% 28.5ms ± 2% -5.20% (p=0.000 n=19+19) BM_CommentLines/128/70/2 300ms ± 3% 265ms ± 3% -11.69% (p=0.000 n=18+20) BM_CommentLines/1/0/8 16.7ms ± 2% 16.3ms ± 2% -2.22% (p=0.000 n=19+19) BM_CommentLines/4/0/8 26.4ms ± 1% 25.2ms ± 1% -4.42% (p=0.000 n=19+18) BM_CommentLines/128/0/8 285ms ± 2% 248ms ± 1% -13.16% (p=0.000 n=19+18) BM_CommentLines/1/30/8 17.9ms ± 2% 17.4ms ± 1% -2.99% (p=0.000 n=19+17) BM_CommentLines/4/30/8 32.0ms ± 1% 30.5ms ± 2% -4.73% (p=0.000 n=19+20) BM_CommentLines/128/30/8 320ms ± 2% 280ms ± 3% -12.52% (p=0.000 n=18+20) BM_CommentLines/1/70/8 20.7ms ± 2% 20.0ms ± 2% -3.22% (p=0.000 n=19+19) BM_CommentLines/4/70/8 33.5ms ± 2% 31.8ms ± 1% -5.23% (p=0.000 n=19+19) BM_CommentLines/128/70/8 338ms ± 3% 298ms ± 2% -11.64% (p=0.000 n=19+20) ``` |
||
|
|
0e2b6c7f1a |
Optimize runs of horizontal whitespace. (#3288)
So, this is a somewhat fun, simple improvement. =] Just use a loop and count runs of whitespace. I didn't even work all that hard to make the loop fast, but it seems great. Makes long runs of indentation more than 2x faster at basically no code complexity. I thought about doing this for vertical whitespace as well but it's not easy to do, and didn't seem worth adding complexity. Huge runs of blank lines aren't nearly as common as lots of indentation. I do have a plan for an analogous optimization for comment blocks, but want to simplify some other code first. We could also make this (hilariously) faster with some judicious use of SIMD or clever use of a string function, but it doesn't seem worth it given how fast the simple loop is already. I didn't work to get a high N count and so there's plenty of noise here, but the benchmark data speaks for itself: ``` BM_ValidKeywords 2.71ms ± 1% 2.75ms ± 1% +1.32% (p=0.016 n=5+5) BM_ValidIdentifiers<1, 64, false> 9.71ms ± 2% 9.74ms ± 1% ~ (p=1.000 n=5+5) BM_ValidIdentifiers<1, 1, true> 3.17ms ± 1% 3.22ms ± 2% ~ (p=0.151 n=5+5) BM_ValidIdentifiers<3, 5, true> 11.5ms ± 2% 11.6ms ± 3% ~ (p=0.548 n=5+5) BM_ValidIdentifiers<3, 16, true> 11.3ms ± 3% 11.6ms ± 3% ~ (p=0.095 n=5+5) BM_ValidIdentifiers<12, 64, true> 12.8ms ± 1% 12.8ms ± 1% ~ (p=1.000 n=5+5) BM_HorizontalWhitespace/1 11.6ms ± 1% 11.7ms ± 2% ~ (p=0.310 n=5+5) BM_HorizontalWhitespace/4 12.9ms ± 1% 11.8ms ± 0% -8.50% (p=0.008 n=5+5) BM_HorizontalWhitespace/16 17.3ms ± 4% 12.5ms ± 1% -27.69% (p=0.008 n=5+5) BM_HorizontalWhitespace/64 28.9ms ± 3% 16.0ms ± 2% -44.88% (p=0.008 n=5+5) BM_HorizontalWhitespace/128 47.6ms ± 3% 21.0ms ± 0% -55.88% (p=0.016 n=5+4) BM_RandomSource 7.92ms ± 1% 7.70ms ± 3% -2.74% (p=0.016 n=5+5) BM_GroupingSymbols/1/0/0 5.92ms ± 2% 5.86ms ± 1% ~ (p=0.310 n=5+5) BM_GroupingSymbols/2/0/0 5.30ms ± 1% 5.01ms ± 1% -5.52% (p=0.008 n=5+5) BM_GroupingSymbols/3/0/0 4.48ms ± 0% 3.95ms ± 1% -11.69% (p=0.008 n=5+5) BM_GroupingSymbols/4/0/0 4.61ms ± 1% 3.73ms ± 1% -19.12% (p=0.008 n=5+5) BM_GroupingSymbols/8/0/0 5.34ms ± 6% 3.05ms ± 1% -42.82% (p=0.008 n=5+5) BM_GroupingSymbols/16/0/0 6.44ms ± 1% 3.20ms ± 2% -50.24% (p=0.008 n=5+5) BM_GroupingSymbols/32/0/0 10.3ms ± 5% 4.2ms ± 1% -59.81% (p=0.008 n=5+5) BM_GroupingSymbols/0/1/0 5.17ms ± 2% 5.17ms ± 1% ~ (p=0.690 n=5+5) BM_GroupingSymbols/0/2/0 4.04ms ± 1% 4.04ms ± 2% ~ (p=1.000 n=5+5) BM_GroupingSymbols/0/3/0 2.89ms ± 1% 2.89ms ± 1% ~ (p=1.000 n=5+5) BM_GroupingSymbols/0/4/0 2.54ms ± 1% 2.55ms ± 1% ~ (p=0.421 n=5+5) BM_GroupingSymbols/0/8/0 1.65ms ± 2% 1.67ms ± 1% ~ (p=0.310 n=5+5) BM_GroupingSymbols/0/16/0 1.18ms ± 1% 1.19ms ± 2% ~ (p=0.222 n=5+5) BM_GroupingSymbols/0/32/0 948µs ± 1% 957µs ± 2% ~ (p=0.310 n=5+5) BM_GroupingSymbols/0/0/1 5.17ms ± 1% 5.16ms ± 2% ~ (p=0.841 n=5+5) BM_GroupingSymbols/0/0/2 4.03ms ± 2% 4.04ms ± 2% ~ (p=0.548 n=5+5) BM_GroupingSymbols/0/0/3 2.89ms ± 1% 2.88ms ± 1% ~ (p=0.841 n=5+5) BM_GroupingSymbols/0/0/4 2.54ms ± 1% 2.54ms ± 2% ~ (p=0.548 n=5+5) BM_GroupingSymbols/0/0/8 1.66ms ± 3% 1.67ms ± 2% ~ (p=0.690 n=5+5) BM_GroupingSymbols/0/0/16 1.19ms ± 1% 1.18ms ± 2% ~ (p=0.548 n=5+5) BM_GroupingSymbols/0/0/32 949µs ± 2% 955µs ± 1% ~ (p=0.310 n=5+5) BM_GroupingSymbols/32/1/0 10.0ms ± 1% 4.0ms ± 1% -59.76% (p=0.008 n=5+5) BM_GroupingSymbols/32/2/0 9.73ms ± 2% 3.92ms ± 2% -59.73% (p=0.008 n=5+5) BM_GroupingSymbols/32/3/0 9.40ms ± 2% 3.84ms ± 3% -59.10% (p=0.008 n=5+5) BM_GroupingSymbols/32/4/0 9.20ms ± 2% 3.74ms ± 1% -59.35% (p=0.008 n=5+5) BM_GroupingSymbols/32/8/0 8.38ms ± 2% 3.50ms ± 1% -58.22% (p=0.008 n=5+5) BM_GroupingSymbols/32/16/0 7.18ms ± 2% 3.06ms ± 3% -57.41% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/0 5.53ms ± 1% 2.44ms ± 2% -55.99% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/1 5.47ms ± 2% 2.40ms ± 2% -56.08% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/2 5.41ms ± 3% 2.40ms ± 2% -55.66% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/3 5.28ms ± 2% 2.37ms ± 2% -55.18% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/4 5.25ms ± 2% 2.34ms ± 2% -55.52% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/8 5.03ms ± 3% 2.25ms ± 1% -55.29% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/16 4.62ms ± 1% 2.15ms ± 3% -53.41% (p=0.008 n=5+5) BM_GroupingSymbols/32/32/32 3.99ms ± 1% 1.86ms ± 1% -53.24% (p=0.008 n=5+5) BM_BlankLines/1 12.8ms ± 1% 12.6ms ± 2% ~ (p=0.310 n=5+5) BM_BlankLines/4 16.0ms ± 3% 15.8ms ± 1% ~ (p=0.310 n=5+5) BM_BlankLines/16 33.1ms ± 2% 32.1ms ± 1% -3.29% (p=0.008 n=5+5) BM_BlankLines/64 84.6ms ± 3% 84.2ms ± 3% ~ (p=0.690 n=5+5) BM_BlankLines/128 159ms ± 4% 155ms ± 3% ~ (p=0.310 n=5+5) BM_CommentLines/1/0/0 14.5ms ± 2% 14.5ms ± 2% ~ (p=0.690 n=5+5) BM_CommentLines/4/0/0 19.0ms ± 1% 19.0ms ± 1% ~ (p=0.548 n=5+5) BM_CommentLines/128/0/0 188ms ± 2% 186ms ± 1% ~ (p=0.151 n=5+5) BM_CommentLines/1/30/0 14.8ms ± 3% 14.7ms ± 2% ~ (p=0.421 n=5+5) BM_CommentLines/4/30/0 21.8ms ± 1% 21.3ms ± 4% ~ (p=0.095 n=5+5) BM_CommentLines/128/30/0 201ms ± 1% 200ms ± 2% ~ (p=0.421 n=5+5) BM_CommentLines/1/70/0 15.5ms ± 3% 15.2ms ± 4% ~ (p=0.095 n=5+5) BM_CommentLines/4/70/0 23.2ms ± 1% 22.5ms ± 2% -2.76% (p=0.008 n=5+5) BM_CommentLines/128/70/0 213ms ± 1% 209ms ± 1% -1.54% (p=0.016 n=5+5) BM_CommentLines/1/0/2 15.3ms ± 1% 14.5ms ± 1% -4.99% (p=0.008 n=5+5) BM_CommentLines/4/0/2 21.4ms ± 1% 20.2ms ± 2% -5.48% (p=0.008 n=5+5) BM_CommentLines/128/0/2 242ms ± 2% 218ms ± 5% -10.13% (p=0.008 n=5+5) BM_CommentLines/1/30/2 15.7ms ± 4% 15.1ms ± 2% -3.37% (p=0.008 n=5+5) BM_CommentLines/4/30/2 24.3ms ± 1% 22.5ms ± 3% -7.42% (p=0.008 n=5+5) BM_CommentLines/128/30/2 268ms ± 2% 240ms ± 3% -10.22% (p=0.008 n=5+5) BM_CommentLines/1/70/2 16.1ms ± 2% 15.3ms ± 3% -5.24% (p=0.008 n=5+5) BM_CommentLines/4/70/2 25.7ms ± 3% 24.0ms ± 3% -6.69% (p=0.008 n=5+5) BM_CommentLines/128/70/2 272ms ± 1% 247ms ± 1% -9.21% (p=0.008 n=5+5) BM_CommentLines/1/0/8 17.2ms ± 5% 14.8ms ± 2% -14.32% (p=0.008 n=5+5) BM_CommentLines/4/0/8 30.4ms ± 2% 20.8ms ± 2% -31.47% (p=0.008 n=5+5) BM_CommentLines/128/0/8 463ms ± 1% 260ms ± 1% -43.89% (p=0.008 n=5+5) BM_CommentLines/1/30/8 17.2ms ± 2% 15.2ms ± 2% -12.01% (p=0.008 n=5+5) BM_CommentLines/4/30/8 32.4ms ± 1% 23.2ms ± 3% -28.61% (p=0.008 n=5+5) BM_CommentLines/128/30/8 498ms ± 3% 287ms ± 2% -42.32% (p=0.008 n=5+5) BM_CommentLines/1/70/8 17.6ms ± 2% 15.5ms ± 3% -11.61% (p=0.008 n=5+5) BM_CommentLines/4/70/8 34.0ms ± 4% 24.7ms ± 3% -27.31% (p=0.008 n=5+5) BM_CommentLines/128/70/8 497ms ± 2% 297ms ± 4% -40.29% (p=0.008 n=5+5) ``` Stacked on top of #3287 -- only the last commit should be reviewed here. |
||
|
|
a79ea4b28d |
Do more precise table dispatch for symbols. (#3287)
When originally switching to the table dispatch approach we discussed that it'd be nice to disentangle the monolithic symbol lexing routine with this as we'll typically have fairly precise dispatch. This is especially true for grouping symbols, which in Carbon are all constructively one-character (at this point). I think this provides a substantial improvement to the clarity of the code by disentangling the different paths. It also allowed a bunch of simplifications / clarifications to exactly what the behavior with closing invalid groups actually involves currently. This was initially motivated by code organization improvements, and any performance wins were speculative. However, when benchmarking it surfaced a problem that hadn't been clear -- we're generating too many distinct functions here, and the table-based dispatch slows down in the face of that. So this PR also includes a fix for that, removing the template-generated fan-out of dispatch functions for distinct symbols. Instead, we have a dedicated table to translate one character into the token kinds. This seems to work quite well, avoiding the huge branch-y structure and just do fairly cheap table translation & dispatch for all one-character symbols. Building the table requires the token kinds to be default constructable, so this also enables that and arranges for the zero-value kind to be the error kind. Combined, this is a modest speedup for *non* grouping symbols (3-4%, a bit noisy). And in some cases it is a huge speedup for grouping symbols (>10%). Raw benchmark data with 20 runs before/after -- despite the # of runs, the grouping symbols benchmarks were frustratingly noisy in non-uniform ways that couldn't fully be accounted for here. Still, this seems like an overall improvement. ``` BM_RandomSource 7.98ms ± 2% 7.73ms ± 3% -3.12% (p=0.000 n=18+19) BM_GroupingSymbols/1/0/0 5.90ms ± 2% 5.82ms ± 4% -1.38% (p=0.001 n=20+20) BM_GroupingSymbols/2/0/0 5.21ms ± 2% 5.15ms ± 2% -1.14% (p=0.002 n=20+18) BM_GroupingSymbols/3/0/0 4.42ms ± 2% 4.34ms ± 2% -1.87% (p=0.000 n=19+18) BM_GroupingSymbols/4/0/0 4.29ms ± 2% 4.38ms ± 5% ~ (p=0.297 n=17+20) BM_GroupingSymbols/8/0/0 5.09ms ±10% 5.10ms ± 7% ~ (p=0.919 n=18+20) BM_GroupingSymbols/16/0/0 6.35ms ± 8% 6.29ms ± 6% ~ (p=0.201 n=20+20) BM_GroupingSymbols/32/0/0 9.88ms ± 2% 9.83ms ± 1% ~ (p=0.167 n=18+20) BM_GroupingSymbols/0/1/0 5.12ms ± 2% 5.01ms ± 2% -2.14% (p=0.000 n=20+19) BM_GroupingSymbols/0/2/0 4.01ms ± 2% 3.93ms ± 4% -2.03% (p=0.000 n=20+19) BM_GroupingSymbols/0/3/0 2.92ms ± 3% 2.81ms ± 2% -3.87% (p=0.000 n=20+19) BM_GroupingSymbols/0/4/0 2.61ms ± 3% 2.47ms ± 2% -5.30% (p=0.000 n=20+18) BM_GroupingSymbols/0/8/0 1.77ms ± 3% 1.61ms ± 2% -8.91% (p=0.000 n=18+19) BM_GroupingSymbols/0/16/0 1.41ms ± 3% 1.16ms ± 4% -17.66% (p=0.000 n=20+20) BM_GroupingSymbols/0/32/0 1.10ms ± 2% 0.92ms ± 3% -16.36% (p=0.000 n=20+17) BM_GroupingSymbols/0/0/1 5.09ms ± 2% 5.03ms ± 3% -1.11% (p=0.001 n=20+18) BM_GroupingSymbols/0/0/2 4.01ms ± 2% 3.91ms ± 2% -2.67% (p=0.000 n=20+18) BM_GroupingSymbols/0/0/3 2.93ms ± 3% 2.81ms ± 2% -4.23% (p=0.000 n=20+19) BM_GroupingSymbols/0/0/4 2.59ms ± 2% 2.48ms ± 3% -4.48% (p=0.000 n=20+19) BM_GroupingSymbols/0/0/8 1.75ms ± 1% 1.62ms ± 3% -7.65% (p=0.000 n=17+19) BM_GroupingSymbols/0/0/16 1.40ms ± 2% 1.15ms ± 3% -17.67% (p=0.000 n=19+20) BM_GroupingSymbols/0/0/32 1.10ms ± 2% 0.92ms ± 3% -15.91% (p=0.000 n=20+19) BM_GroupingSymbols/32/1/0 9.62ms ± 2% 9.65ms ± 2% ~ (p=0.654 n=18+20) BM_GroupingSymbols/32/2/0 9.41ms ± 2% 9.37ms ± 2% ~ (p=0.095 n=20+19) BM_GroupingSymbols/32/3/0 9.13ms ± 2% 9.13ms ± 3% ~ (p=0.687 n=19+20) BM_GroupingSymbols/32/4/0 8.93ms ± 1% 8.87ms ± 2% -0.69% (p=0.010 n=20+18) BM_GroupingSymbols/32/8/0 8.15ms ± 2% 8.14ms ± 3% ~ (p=0.729 n=19+19) BM_GroupingSymbols/32/16/0 7.04ms ± 3% 6.92ms ± 1% -1.71% (p=0.000 n=20+18) BM_GroupingSymbols/32/32/0 5.48ms ± 2% 5.38ms ± 3% -1.81% (p=0.000 n=20+20) BM_GroupingSymbols/32/32/1 5.39ms ± 2% 5.29ms ± 2% -1.87% (p=0.000 n=19+19) BM_GroupingSymbols/32/32/2 5.34ms ± 2% 5.21ms ± 1% -2.45% (p=0.000 n=20+18) BM_GroupingSymbols/32/32/3 5.27ms ± 3% 5.16ms ± 2% -2.18% (p=0.000 n=20+19) BM_GroupingSymbols/32/32/4 5.21ms ± 2% 5.10ms ± 3% -2.11% (p=0.000 n=19+20) BM_GroupingSymbols/32/32/8 4.98ms ± 2% 4.83ms ± 2% -2.85% (p=0.000 n=19+19) BM_GroupingSymbols/32/32/16 4.55ms ± 2% 4.45ms ± 2% -2.25% (p=0.000 n=18+20) BM_GroupingSymbols/32/32/32 3.95ms ± 2% 3.84ms ± 2% -2.98% (p=0.000 n=19+20) ``` |
||
|
|
6ba8712fbd |
Predetermine all the line splits in the lexer. (#3278)
## Summary ## Restructures the lexer to first scan the entire source text for newlines and create all the line structures needed. Doing this up-front makes it easy to produce an optimized version with minimal complexity. Currently, it leverages the system `memchr`, but even when expanded to handle more complex cases like CR+LF line endings, being isolated in this way will result in a significantly simpler implementation. This change improves the lexing of comment lines significantly by skipping their contents immediately. The overhead of the pre-scan is unmeasurable in all realistic benchmarks, and 10-30% in benchmarks consisting almost entirely of blank lines or comments. The improvement of comment lexing with average length comment lines mixed with code starts at 20% and goes up. Regressing blank line handling for non-empty comments seems like the right tradeoff (by far). ## Background and details ## One weak point in the lexer implementation were large runs of comments. While those aren't terribly common, they shouldn't present a hazard to the lexer performance. A bit more common is a pattern of comments like the following: ```carbon // Some method comment here. fn SomeMethodName(...) -> ...; // Some other method comment here. fn SomeOtherMethodName(...) -> ...; ``` Here, the lexer spends an inordinate amount of time getting from the `\n` after the first semicolon to the `fn` token. It has to skip a blank line, scan a line, find the `//` comment start, then scan to find the next `\n`, and then scan horizontal whitespace, etc. It is tempting to build a scanner *exactly* for this. In fact, I built one, and I can publish it in a PR if folks are interested in what it looks like. For x86-64, the PSHUFB trick used for scanning identifiers technically works. But it is *complicated*. Amazingly so. 150 lines of very subtle code with subtle performance pitfalls at ever turn. I felt very uncomfortable submitting it, but we can always go back to it. Nothing I've come up with quite matches it for sheer speed. However, most of the complexity and time is spent walking from a `//` to the end of the line. And *that* is something we can do very simply. In fact, there is a tuned function for that in libc: `memchr`. Using this we can build a very fast and much simpler scanner to split lines up-front. This PR uses that and a carefully crafted fast loop to first build up all the line info we need. Getting this to be as fast as possible required some other subtle changes, for example always creating a line structure that goes from the last `\n` and the end of the file. We then back up the EOF token to avoid surfacing this to users. The nice thing is the EOF token isn't part of any hot loop, and so this removes branches everywhere else at modest complexity. Once we have that, the rest of the lexer just needs to keep track of its current line in order to record column offsets. I've taken some care to try and optimize the lexer's usage of the line structures but there are more opportunities here I suspect. Combined, this gets much but not all of the performance of a huge SIMD scanner for newline-through-to-next-token. For extreme cases (100s of blank lines or empty comment lines between tokens) the holistic scanner is of course still much faster, but those don't seem nearly worth the cost. I was initially worried about the overhead of taking two passes over the source text, but in practice I've not been able to measure any appreciable cost to this with realistic source files. In some cases benchmarks with no newlines get *faster* because we use a much more efficient approach to fetch the source text into cache as a happenstance. And that in turn makes the byte-wise dispatched loop run faster as it stalls less. I'm particularly happy with this approach because it seems very clear how to extend this to support CR+LF, bare CR, and even complex mixtures without any significant speed cost. That wasn't at all true for the other approaches explored. I may try some further PRs to smooth out the last bits of slowness here, but already this is working excellent for me in practice. My 10mloc test case is down to 2.3s to lex. ## Raw benchmark data Using a tool that runs benchmarks before and after and analyzes the results, the following summarizes the CPU-time impact, each of these for lexing 100k tokens: ``` BM_ValidKeywords 2.57ms ± 1% 2.58ms ± 0% ~ (p=0.190 n=5+4) BM_ValidIdentifiers<1, 64, false> 9.24ms ± 4% 9.31ms ± 4% ~ (p=0.421 n=5+5) BM_ValidIdentifiers<1, 1, true> 3.05ms ± 4% 3.11ms ± 4% ~ (p=0.222 n=5+5) BM_ValidIdentifiers<3, 5, true> 10.9ms ± 0% 11.1ms ± 1% +1.76% (p=0.016 n=4+5) BM_ValidIdentifiers<3, 16, true> 11.1ms ± 7% 11.0ms ± 1% ~ (p=0.310 n=5+5) BM_ValidIdentifiers<12, 64, true> 12.2ms ± 1% 12.3ms ± 2% ~ (p=0.111 n=4+5) BM_HorizontalWhitespace/1 11.2ms ± 6% 11.1ms ± 2% ~ (p=0.841 n=5+5) BM_HorizontalWhitespace/4 12.0ms ± 3% 12.0ms ± 2% ~ (p=0.548 n=5+5) BM_HorizontalWhitespace/16 16.2ms ± 6% 15.9ms ± 8% ~ (p=0.690 n=5+5) BM_HorizontalWhitespace/64 27.7ms ± 3% 28.4ms ± 3% ~ (p=0.151 n=5+5) BM_HorizontalWhitespace/128 44.3ms ± 1% 45.6ms ± 6% +3.15% (p=0.032 n=5+5) BM_RandomSource 7.75ms ± 2% 7.72ms ± 1% ~ (p=1.000 n=5+5) BM_BlankLines/1 11.7ms ± 1% 12.1ms ± 1% +3.46% (p=0.008 n=5+5) BM_BlankLines/4 14.0ms ± 2% 15.2ms ± 3% +8.12% (p=0.008 n=5+5) BM_BlankLines/16 23.5ms ± 2% 31.1ms ± 4% +32.26% (p=0.008 n=5+5) BM_BlankLines/64 75.3ms ± 1% 81.2ms ± 3% +7.83% (p=0.008 n=5+5) BM_BlankLines/128 133ms ± 3% 150ms ± 2% +12.74% (p=0.008 n=5+5) BM_CommentLines/1/0/0 13.1ms ± 0% 13.7ms ± 1% +5.11% (p=0.008 n=5+5) BM_CommentLines/4/0/0 16.6ms ± 1% 18.2ms ± 4% +9.56% (p=0.008 n=5+5) BM_CommentLines/128/0/0 169ms ± 4% 182ms ± 1% +7.24% (p=0.008 n=5+5) BM_CommentLines/1/30/0 18.7ms ± 5% 14.1ms ± 0% -24.84% (p=0.008 n=5+5) BM_CommentLines/4/30/0 36.5ms ± 6% 20.6ms ± 3% -43.59% (p=0.008 n=5+5) BM_CommentLines/128/30/0 525ms ± 4% 198ms ± 1% -62.38% (p=0.008 n=5+5) BM_CommentLines/1/70/0 23.4ms ± 6% 14.7ms ± 2% -37.15% (p=0.008 n=5+5) BM_CommentLines/4/70/0 53.3ms ± 7% 22.4ms ± 4% -57.99% (p=0.008 n=5+5) BM_CommentLines/128/70/0 1.05s ± 4% 0.21s ± 2% -80.31% (p=0.008 n=5+5) BM_CommentLines/1/0/2 14.1ms ± 6% 14.3ms ± 1% ~ (p=0.151 n=5+5) BM_CommentLines/4/0/2 19.4ms ± 5% 20.1ms ± 1% ~ (p=0.151 n=5+5) BM_CommentLines/128/0/2 238ms ± 8% 229ms ± 0% ~ (p=0.151 n=5+5) BM_CommentLines/1/30/2 19.2ms ± 7% 14.6ms ± 1% -23.87% (p=0.008 n=5+5) BM_CommentLines/4/30/2 40.3ms ±13% 22.3ms ± 4% -44.63% (p=0.008 n=5+5) BM_CommentLines/128/30/2 568ms ± 7% 254ms ± 3% -55.28% (p=0.008 n=5+5) BM_CommentLines/1/70/2 23.3ms ± 1% 15.0ms ± 3% -35.61% (p=0.016 n=4+5) BM_CommentLines/4/70/2 57.2ms ± 9% 24.1ms ± 2% -57.81% (p=0.008 n=5+5) BM_CommentLines/128/70/2 1.07s ± 0% 0.26s ± 2% -75.51% (p=0.016 n=4+5) BM_CommentLines/1/0/8 15.9ms ± 7% 16.0ms ± 1% ~ (p=0.151 n=5+5) BM_CommentLines/4/0/8 24.2ms ± 6% 27.9ms ± 2% +15.36% (p=0.008 n=5+5) BM_CommentLines/128/0/8 386ms ± 4% 445ms ± 1% +15.28% (p=0.008 n=5+5) BM_CommentLines/1/30/8 20.6ms ± 5% 16.3ms ± 1% -20.95% (p=0.008 n=5+5) BM_CommentLines/4/30/8 45.3ms ± 6% 30.3ms ± 3% -32.98% (p=0.008 n=5+5) BM_CommentLines/128/30/8 699ms ± 3% 477ms ± 3% -31.83% (p=0.008 n=5+5) BM_CommentLines/1/70/8 25.7ms ± 5% 16.8ms ± 2% -34.67% (p=0.008 n=5+5) BM_CommentLines/4/70/8 62.0ms ± 4% 31.6ms ± 2% -49.10% (p=0.008 n=5+5) BM_CommentLines/128/70/8 1.20s ± 2% 0.48s ± 4% -59.60% (p=0.008 n=5+5) ``` The horizontal whitespace benchmark (and all of the non-line-oriented ones) are noisier than they appear here but do show some improvements (surprisingly). My guess is that it has a lot to do with system load, as the advantage is that we're using a vectorized loop to scan the text first and then doing the byte-dispatched loop. So when the cache is a bit slower to populate, the vectorized version starts to be faster. --------- Co-authored-by: Richard Smith <richard@metafoo.co.uk> Co-authored-by: josh11b <josh11b@users.noreply.github.com> |
||
|
|
6f5934a505 |
Unblock more lexer inlining. (#3274)
The big change is to make the lexer helpers have internal linkage, making all of them easy to inline into single call sites. Looking at the profile showed several other cases of unfortunate out-of-line functions. Two were due to the code size produced for checks -- those are switched to `DCHECK`s to remove that code from optimized builds. The loss of coverage seems minor. A last one was closing open groups. This was a surprising routine to be hot, but it the paths to discover "nothing to do here" were intertwined into the code. This PR extracts this common trace into a separate function that delegates to the looping recovery path. This lets the hot path inline easily. At this point, for a large lexing benchmark I'm using, 50% of the time is in the identifier hash table at this point. The remaining improvements are to actually make some of the hot routines like symbol lexing and comment lexing faster. |
||
|
|
03c3b86758 |
Switch lexer to fully table-driven design. (#3273)
This uses the musttail dispatched table approach to drive the entire lexing. The result is that there is no main lexer loop at all in a traditional sense, now everything is driven through tail recursive dispatch on the next byte of the source text. This should be easy to extend still -- the design pattern is to add lexer methods for handling specific cases, and then add a dispatch function to dispatch to them from the table. For example, we can add a method that handles decoding UTF-8 outside of the ASCII subset and set the table entries used by non-ASCII initial bytes to dispatch to it. The performance is already surprisingly good, benchmarks show a modest improvement across the board. That's despite there still being some *serious* performance issues that I'll fix in a separate patch. There are also opportunities to leverage this structure more heavily as needed by putting more specialized dispatch targets in for specific bytes. A follow-up PR will re-organize the functions here, as almost all of the methods on the `Lexer` should become private, but I wanted to keep that a separate change since it will probably render the diff even more hard to read than it already is. |
||
|
|
d552545c6d |
Move dispatch routines to be static member functions. (#3265)
These routines have a regular signature and are used to build a table of function pointers for fast dispatch. However, the previous approach relied on lambdas to build these functions which resulted in very hard to read functions in the profile and backtrace. In preparation for expanding dispatch to handle (many) more cases in the lexer and also enabling more aggressive inlining into the dispatch routines, I wanted to tidy up how they appear. This PR alone shouldn't have any interesting functional effect, it's just re-organizing the code. Co-authored-by: Richard Smith <richard@metafoo.co.uk> |
||
|
|
a46ca6bf7a |
Add a start-of-file token and parse node. (#3263)
This removes a (very) hot branch in the lexer where we need to special case when a token is the first token and can't look at its previous token. It also seems like a generally nice change to the structure of both the token buffer and parse tree as there are now bracketing elements for both ends and we should be able to avoid similar branching in the future. Mostly mechanical updates to the lexer and parser code to handle this, but also needed to special case the location information in the autoupdate code. And then the usual large body of auto-updated tests. No benchmark data for this change alone as in isolation and in the current lexer structure it doesn't make a big difference. But this branch was particularly difficult to handle when trying to update the whitespace skipping code to be faster, and so I think it is worth systematically avoiding the special case here. |