Skip to content

Commit debf55a

Browse files
committed
Update xxhash3 performance tables and graphs
1 parent 4158077 commit debf55a

13 files changed

Lines changed: 1696 additions & 1314 deletions

‎comparison/README.md‎

Lines changed: 41 additions & 42 deletions
Original file line numberDiff line numberDiff line change
@@ -86,22 +86,21 @@ graphs are boring flat lines, so a table is used instead.
8686
### aarch64
8787

8888
| Implementation | Throughput (GiB/s) |
89-
|----------------|--------------------|
90-
| Rust | 35.0 |
91-
| C | 35.0 |
92-
| C (scalar) | 21.2 |
93-
| C (NEON) | 35.0 |
89+
|:---------------|-------------------:|
90+
| Rust | 35.2 |
91+
| C | 35.3 |
92+
| C (scalar) | 21.9 |
93+
| C (NEON) | 35.2 |
9494

9595
### x86_64
9696

9797
| Implementation | Throughput (GiB/s) |
98-
|----------------|--------------------|
99-
| Rust | 58.9 |
100-
| C | 25.1 |
101-
| C (scalar) | 7.6 |
102-
| C (SSE2) | 25.1 |
103-
| C (AVX2) | 58.4 |
104-
98+
|:---------------|-------------------:|
99+
| Rust | 57.7 |
100+
| C | 25.8 |
101+
| C (scalar) | 14.4 |
102+
| C (SSE2) | 26.0 |
103+
| C (AVX2) | 57.2 |
105104

106105
## Streaming data
107106

@@ -128,24 +127,24 @@ various chunk sizes.
128127

129128
## Small amounts of data
130129

131-
Compares the **time taken** to hash 0 to 230 bytes of
132-
data. Representative samples are taken from similar times to avoid
133-
cluttering the graph and wasting benchmarking time.
130+
Compares the **time taken** to hash 0 to 230 bytes of data using
131+
`oneshot_with_seed`. Representative samples are taken from similar
132+
times to avoid cluttering the graph and wasting benchmarking time.
134133

135134
### aarch64
136135

137-
<a href="./results/xxhash3_64-tiny_data-aarch64.svg">
136+
<a href="./results/xxhash3_64-tiny_data-oneshot_with_seed-aarch64.svg">
138137
<img
139-
src="./results/xxhash3_64-tiny_data-aarch64.svg"
138+
src="./results/xxhash3_64-tiny_data-oneshot_with_seed-aarch64.svg"
140139
alt="xxHash3, 64-bit, small data, on an aarch64 processor"
141140
/>
142141
</a>
143142

144143
### x86_64
145144

146-
<a href="./results/xxhash3_64-tiny_data-x86_64.svg">
145+
<a href="./results/xxhash3_64-tiny_data-oneshot_with_seed-x86_64.svg">
147146
<img
148-
src="./results/xxhash3_64-tiny_data-x86_64.svg"
147+
src="./results/xxhash3_64-tiny_data-oneshot_with_seed-x86_64.svg"
149148
alt="xxHash3, 64-bit, small data, on an x86_64 processor"
150149
/>
151150
</a>
@@ -159,21 +158,21 @@ function call. Data sizes from 256 KiB to 4 MiB are tested. These
159158
graphs are boring flat lines, so a table is used instead.
160159

161160
| Implementation | Throughput (GiB/s) |
162-
|----------------|--------------------|
163-
| Rust | 34.4 |
164-
| C | 34.8 |
165-
| C (scalar) | 21.3 |
166-
| C (NEON) | 34.6 |
161+
|:---------------|-------------------:|
162+
| Rust | 34.9 |
163+
| C | 34.7 |
164+
| C (scalar) | 21.6 |
165+
| C (NEON) | 34.6 |
167166

168167
### x86_64
169168

170169
| Implementation | Throughput (GiB/s) |
171-
|----------------|--------------------|
172-
| Rust | 58.3 |
173-
| C | 25.6 |
174-
| C (scalar) | 7.6 |
175-
| C (SSE2) | 25.5 |
176-
| C (AVX2) | 57.4 |
170+
|:---------------|-------------------:|
171+
| Rust | 57.8 |
172+
| C | 28.0 |
173+
| C (scalar) | 13.1 |
174+
| C (SSE2) | 28.0 |
175+
| C (AVX2) | 57.4 |
177176

178177
## Streaming data
179178

@@ -200,24 +199,24 @@ various chunk sizes.
200199

201200
## Small amounts of data
202201

203-
Compares the **time taken** to hash 0 to 230 bytes of
204-
data. Representative samples are taken from similar times to avoid
205-
cluttering the graph and wasting benchmarking time.
202+
Compares the **time taken** to hash 0 to 230 bytes of data using
203+
`oneshot_with_seed`. Representative samples are taken from similar
204+
times to avoid cluttering the graph and wasting benchmarking time.
206205

207206
### aarch64
208207

209-
<a href="./results/xxhash3_128-tiny_data-aarch64.svg">
208+
<a href="./results/xxhash3_128-tiny_data-oneshot_with_seed-aarch64.svg">
210209
<img
211-
src="./results/xxhash3_128-tiny_data-aarch64.svg"
210+
src="./results/xxhash3_128-tiny_data-oneshot_with_seed-aarch64.svg"
212211
alt="xxHash3, 128-bit, small data, on an aarch64 processor"
213212
/>
214213
</a>
215214

216215
### x86_64
217216

218-
<a href="./results/xxhash3_128-tiny_data-x86_64.svg">
217+
<a href="./results/xxhash3_128-tiny_data-oneshot_with_seed-x86_64.svg">
219218
<img
220-
src="./results/xxhash3_128-tiny_data-x86_64.svg"
219+
src="./results/xxhash3_128-tiny_data-oneshot_with_seed-x86_64.svg"
221220
alt="xxHash3, 128-bit, small data, on an x86_64 processor"
222221
/>
223222
</a>
@@ -228,10 +227,10 @@ cluttering the graph and wasting benchmarking time.
228227

229228
| CPU | Memory | C compiler |
230229
|-------------------|--------|--------------------|
231-
| Apple M1 Max | 64 GiB | clang 16.0.0 |
232-
| AMD Ryzen 9 3950X | 32 GiB | cl.exe 19.41.34120 |
230+
| Apple M1 Max | 64 GiB | clang 21.0.0 |
231+
| AMD Ryzen 9 3950X | 32 GiB | cl.exe 19.51.36260 |
233232

234-
Tests were run with `rustc 1.82.0 (f6e511eec 2024-10-15)`.
233+
Tests were run with `rustc 1.99.0 (b940084d7 2026-09-28)`.
235234

236235
## Details
237236

@@ -250,7 +249,7 @@ Tests were run with `rustc 1.82.0 (f6e511eec 2024-10-15)`.
250249

251250
<tr>
252251
<th>C compiler</th>
253-
<td>Apple clang version 16.0.0 (clang-1600.0.26.3)</td>
252+
<td>Apple clang version 21.0.0 (clang-2100.3.34.2)</td>
254253
</tr>
255254
</table>
256255

@@ -269,6 +268,6 @@ Tests were run with `rustc 1.82.0 (f6e511eec 2024-10-15)`.
269268

270269
<tr>
271270
<th>C compiler</th>
272-
<td>Microsoft (R) C/C++ Optimizing Compiler Version 19.41.34120 for x86</td>
271+
<td>Microsoft (R) C/C++ Optimizing Compiler Version 19.51.36260 for x64</td>
273272
</tr>
274273
</table>

0 commit comments

Comments
 (0)