perf: rework decoder interface (#22)

The updated interface decodes codepoints directly from a reader rather than being implemented as a state machine. This turns out to be considerably more efficient than the previous implementation, with around 25% improvement on the `token_reader` and `reader` benchmarks: ``` Benchmark 1 (27 runs): zig-out/bin-old/token_reader Gtk-4.0.gir measurement mean ± σ min … max outliers delta wall_time 188ms ± 14.5ms 168ms … 205ms 0 ( 0%) 0% peak_rss 7.31MB ± 58.5KB 7.21MB … 7.34MB 0 ( 0%) 0% cpu_cycles 688M ± 4.20M 684M … 706M 1 ( 4%) 0% instructions 1.19G ± 29.4 1.19G … 1.19G 0 ( 0%) 0% cache_references 412K ± 763K 239K … 4.21M 2 ( 7%) 0% cache_misses 10.0K ± 7.40K 7.90K … 46.8K 2 ( 7%) 0% branch_misses 814K ± 1.37K 813K … 821K 1 ( 4%) 0% Benchmark 2 (37 runs): zig-out/bin/token_reader Gtk-4.0.gir measurement mean ± σ min … max outliers delta wall_time 136ms ± 13.8ms 115ms … 147ms 0 ( 0%) ⚡- 27.7% ± 3.8% peak_rss 7.31MB ± 54.7KB 7.21MB … 7.34MB 8 (22%) + 0.1% ± 0.4% cpu_cycles 462M ± 1.87M 459M … 466M 0 ( 0%) ⚡- 32.8% ± 0.2% instructions 1.14G ± 26.6 1.14G … 1.14G 0 ( 0%) ⚡- 4.1% ± 0.0% cache_references 236K ± 4.86K 227K … 244K 0 ( 0%) - 42.7% ± 60.7% cache_misses 9.40K ± 1.25K 7.88K … 11.5K 0 ( 0%) - 6.5% ± 24.6% branch_misses 815K ± 1.01K 813K … 817K 0 ( 0%) + 0.1% ± 0.1% ``` ``` Benchmark 1 (23 runs): zig-out/bin-old/reader Gtk-4.0.gir measurement mean ± σ min … max outliers delta wall_time 225ms ± 14.2ms 199ms … 249ms 0 ( 0%) 0% peak_rss 7.25MB ± 100KB 7.08MB … 7.34MB 0 ( 0%) 0% cpu_cycles 823M ± 12.2M 813M … 847M 0 ( 0%) 0% instructions 1.43G ± 23.0 1.43G … 1.43G 0 ( 0%) 0% cache_references 757K ± 129K 635K … 1.07M 1 ( 4%) 0% cache_misses 13.7K ± 1.18K 12.5K … 17.2K 2 ( 9%) 0% branch_misses 1.43M ± 3.35K 1.42M … 1.43M 0 ( 0%) 0% Benchmark 2 (31 runs): zig-out/bin/reader Gtk-4.0.gir measurement mean ± σ min … max outliers delta wall_time 166ms ± 13.9ms 144ms … 175ms 0 ( 0%) ⚡- 26.5% ± 3.4% peak_rss 7.27MB ± 81.8KB 7.08MB … 7.34MB 0 ( 0%) + 0.3% ± 0.7% cpu_cycles 581M ± 1.54M 579M … 584M 0 ( 0%) ⚡- 29.4% ± 0.5% instructions 1.38G ± 16.0 1.38G … 1.38G 9 (29%) ⚡- 3.8% ± 0.0% cache_references 715K ± 219K 563K … 1.71M 3 (10%) - 5.5% ± 13.6% cache_misses 13.5K ± 1.31K 11.4K … 16.5K 2 ( 6%) - 1.2% ± 5.1% branch_misses 1.07M ± 20.3K 1.05M … 1.11M 5 (16%) ⚡- 25.3% ± 0.6% ```
ianprime0509 · Oct 15, 2023 · 9c6389d · 9c6389d
1 parent ea0fcb1
commit 9c6389d
Show file tree

Hide file tree

Showing 6 changed files with 311 additions and 312 deletions.
diff --git a/README.md b/README.md
@@ -123,23 +123,79 @@ parsing through an entire XML file without doing any additional processing. The
 XML file is loaded completely into memory first, then the parser is executed on
 it until it completes.
 
-Below are some benchmarking results as of commit
-`e9809855f7ee3403efa1fdc5f9010182f47361d0`, as performed on my laptop. The
-results were obtained by executing [poop](https://github.com/andrewrk/poop) on
-the benchmark implementations.
+Below are some benchmarking results as of August 14, 2023, using Zig
+`0.12.0-dev.906+2d7d037c4`, as performed on my laptop. The results were obtained
+by executing [poop](https://github.com/andrewrk/poop) on the benchmark
+implementations.
 
 ### GTK 4 GIR
 
-This is a 7.6MiB XML file containing GObject introspection metadata for GTK 4.
-
-| Implementation             | Execution time  | Memory usage    |
-| -------------------------- | --------------- | --------------- |
-| zig-xml (`Reader`)         | 242ms ± 5.50ms  | 9.12MB ± 66.5KB |
-| zig-xml (`TokenReader`)    | 169ms ± 13.4ms  | 9.07MB ± 97.9KB |
-| zig-xml (`Scanner`)        | 40.2ms ± 2.25ms | 9.09MB ± 97.0KB |
-| libxml2 (`xmlreader.h`)    | 74.0ms ± 3.16ms | 10.4MB ± 104KB  |
-| mxml (`mxmlSAXLoadString`) | 97.1ms ± 1.63ms | 9.12MB ± 64.9KB |
-| yxml                       | 36.2ms ± 999us  | 9.09MB ± 92.3KB |
+This is a 5.7MB XML file containing GObject introspection metadata for GTK 4. In
+the output below, libxml2 is used as the baseline. The three benchmarks
+`reader`, `token_reader`, and `scanner` test the three APIs provided by this
+library, and the mxml and yxml libraries are also included for comparison.
+
+```
+Benchmark 1 (78 runs): zig-out/bin/libxml2 Gtk-4.0.gir
+  measurement          mean ± σ            min … max           outliers         delta
+  wall_time          64.2ms ± 1.87ms    55.5ms … 70.1ms          4 ( 5%)        0%
+  peak_rss           14.6MB ± 76.4KB    14.4MB … 14.7MB          0 ( 0%)        0%
+  cpu_cycles          196M  ± 1.03M      194M  …  200M           3 ( 4%)        0%
+  instructions        409M  ± 43.1       409M  …  409M           0 ( 0%)        0%
+  cache_references   5.44M  ±  325K     5.08M  … 6.97M           5 ( 6%)        0%
+  cache_misses       66.0K  ± 5.36K     55.0K  … 91.0K           3 ( 4%)        0%
+  branch_misses       874K  ± 3.80K      868K  …  890K           1 ( 1%)        0%
+
+Benchmark 2 (30 runs): zig-out/bin/reader Gtk-4.0.gir
+  measurement          mean ± σ            min … max           outliers         delta
+  wall_time           170ms ± 1.59ms     167ms …  173ms          0 ( 0%)        💩+164.2% ±  1.2%
+  peak_rss           7.29MB ± 73.8KB    7.08MB … 7.34MB          0 ( 0%)        ⚡- 50.0% ±  0.2%
+  cpu_cycles          583M  ± 2.88M      579M  …  590M           0 ( 0%)        💩+196.9% ±  0.4%
+  instructions       1.38G  ± 32.2      1.38G  … 1.38G           0 ( 0%)        💩+237.2% ±  0.0%
+  cache_references    751K  ±  135K      580K  … 1.12M           0 ( 0%)        ⚡- 86.2% ±  2.2%
+  cache_misses       17.5K  ± 5.41K     12.9K  … 34.5K           3 (10%)        ⚡- 73.5% ±  3.5%
+  branch_misses      1.06M  ± 10.9K     1.05M  … 1.11M           2 ( 7%)        💩+ 21.5% ±  0.3%
+
+Benchmark 3 (38 runs): zig-out/bin/token_reader Gtk-4.0.gir
+  measurement          mean ± σ            min … max           outliers         delta
+  wall_time           135ms ± 1.59ms     132ms …  138ms          0 ( 0%)        💩+110.4% ±  1.1%
+  peak_rss           7.31MB ± 54.2KB    7.21MB … 7.34MB          8 (21%)        ⚡- 49.8% ±  0.2%
+  cpu_cycles          462M  ± 2.20M      459M  …  467M           0 ( 0%)        💩+135.5% ±  0.3%
+  instructions       1.14G  ± 21.0      1.14G  … 1.14G           0 ( 0%)        💩+179.9% ±  0.0%
+  cache_references    237K  ± 7.40K      225K  …  255K           0 ( 0%)        ⚡- 95.6% ±  1.9%
+  cache_misses       10.1K  ± 1.29K     8.16K  … 13.2K           0 ( 0%)        ⚡- 84.8% ±  2.7%
+  branch_misses       815K  ±  919       813K  …  816K           3 ( 8%)        ⚡-  6.8% ±  0.1%
+
+Benchmark 4 (103 runs): zig-out/bin/scanner Gtk-4.0.gir
+  measurement          mean ± σ            min … max           outliers         delta
+  wall_time          48.6ms ± 1.82ms    45.8ms … 55.2ms          4 ( 4%)        ⚡- 24.3% ±  0.8%
+  peak_rss           7.27MB ± 87.8KB    7.08MB … 7.34MB          0 ( 0%)        ⚡- 50.1% ±  0.2%
+  cpu_cycles          152M  ± 3.48M      151M  …  177M           5 ( 5%)        ⚡- 22.4% ±  0.4%
+  instructions        472M  ± 19.9       472M  …  472M           0 ( 0%)        💩+ 15.6% ±  0.0%
+  cache_references    209K  ± 1.80K      207K  …  222K           4 ( 4%)        ⚡- 96.2% ±  1.2%
+  cache_misses       7.95K  ±  179      7.59K  … 8.50K           0 ( 0%)        ⚡- 88.0% ±  1.6%
+  branch_misses       511K  ±  874       510K  …  518K          13 (13%)        ⚡- 41.6% ±  0.1%
+
+Benchmark 5 (63 runs): zig-out/bin/mxml Gtk-4.0.gir
+  measurement          mean ± σ            min … max           outliers         delta
+  wall_time          80.2ms ± 2.44ms    76.0ms … 87.9ms          3 ( 5%)        💩+ 24.9% ±  1.1%
+  peak_rss           7.44MB ± 56.3KB    7.34MB … 7.47MB         15 (24%)        ⚡- 48.9% ±  0.2%
+  cpu_cycles          262M  ± 2.95M      258M  …  281M           1 ( 2%)        💩+ 33.4% ±  0.4%
+  instructions        762M  ± 56.7K      762M  …  762M           3 ( 5%)        💩+ 86.4% ±  0.0%
+  cache_references    401K  ±  473K      272K  … 3.08M          10 (16%)        ⚡- 92.6% ±  2.4%
+  cache_misses       14.2K  ± 2.62K     12.0K  … 31.1K           2 ( 3%)        ⚡- 78.5% ±  2.2%
+  branch_misses      1.02M  ± 99.5K      998K  … 1.79M           4 ( 6%)        💩+ 16.3% ±  2.5%
+
+Benchmark 6 (196 runs): zig-out/bin/yxml Gtk-4.0.gir
+  measurement          mean ± σ            min … max           outliers         delta
+  wall_time          25.4ms ± 1.03ms    23.9ms … 34.3ms          3 ( 2%)        ⚡- 60.4% ±  0.5%
+  peak_rss           7.29MB ± 77.0KB    7.08MB … 7.34MB          0 ( 0%)        ⚡- 50.0% ±  0.1%
+  cpu_cycles         71.0M  ± 1.03M     70.5M  … 84.2M           5 ( 3%)        ⚡- 63.8% ±  0.1%
+  instructions        236M  ± 20.1       236M  …  236M           0 ( 0%)        ⚡- 42.2% ±  0.0%
+  cache_references    202K  ±  805       201K  …  210K           7 ( 4%)        ⚡- 96.3% ±  0.8%
+  cache_misses       8.00K  ±  215      7.64K  … 9.57K           4 ( 2%)        ⚡- 87.9% ±  1.1%
+  branch_misses       239K  ±  787       238K  …  248K          21 (11%)        ⚡- 72.7% ±  0.1%
+```
 
 ## License
 

diff --git a/bench/build.zig b/bench/build.zig
@@ -18,6 +18,7 @@ pub fn build(b: *Build) !void {
     bench_reader.linkLibC();
 
     const libxml2 = b.dependency("libxml2", .{
+        .optimize = .ReleaseFast,
         .iconv = false,
         .lzma = false,
         .zlib = false,

diff --git a/bench/src/scanner.zig b/bench/src/scanner.zig
@@ -1,13 +1,16 @@
+const std = @import("std");
 const xml = @import("xml");
 
 pub const main = @import("common.zig").main;
 
 pub fn runBench(data: []const u8) !void {
     var scanner = xml.Scanner{};
+    var data_stream = std.io.fixedBufferStream(data);
     var decoder = xml.encoding.Utf8Decoder{};
-    for (data) |b| {
-        if (try decoder.next(b)) |c| {
-            _ = try scanner.next(c, 1);
-        }
+    var buf: [4]u8 = undefined;
+    while (true) {
+        const c = try decoder.readCodepoint(data_stream.reader(), &buf);
+        if (!c.present) break;
+        _ = try scanner.next(c.codepoint, c.byte_length);
     }
 }
diff --git a/examples/scan.zig b/examples/scan.zig
@@ -27,27 +27,19 @@ pub fn main() !void {
 
     var line: usize = 1;
     var column: usize = 1;
-    read: while (true) {
-        var codepoint_bytes: usize = 0;
-        const c = while (true) {
-            const b = input_reader.readByte() catch |e| switch (e) {
-                error.EndOfStream => break :read,
-                else => |other| return other,
-            };
-            codepoint_bytes += 1;
-            if (try decoder.next(b)) |codepoint| {
-                break codepoint;
-            }
-        };
-        const token = scanner.next(c, codepoint_bytes) catch |e| {
+    while (true) {
+        var buf: [4]u8 = undefined;
+        const c = try decoder.readCodepoint(input_reader, &buf);
+        if (!c.present) break;
+        const token = scanner.next(c.codepoint, c.byte_length) catch |e| {
             try stdout_buffered_writer.flush();
             try stderr.print("error: {} ({}:{}): {}\n", .{ scanner.pos, line, column, e });
             return;
         };
         if (token != .ok) {
             try stdout.print("{} ({}:{}): {}\n", .{ scanner.pos, line, column, token });
         }
-        if (c == '\n') {
+        if (c.codepoint == '\n') {
             line += 1;
             column = 1;
         } else {