Add arm64 benchmarks

bwesterb · Jul 24, 2024 · fd4bd90 · fd4bd90
1 parent 7a469d2
commit fd4bd90
Showing 1 changed file with 28 additions and 0 deletions.
diff --git a/README.md b/README.md
@@ -30,6 +30,7 @@ This shows three benchmarks:
 * `purego` are our slightly optimized versions that work on uint64s instead of bytes.
 * `asm` are the AVX2 implementations and the reason to use this library.
 
+### Intel Core i9-13900
 ```
 goos: linux
 goarch: amd64
@@ -53,3 +54,30 @@ Memset-32   4.126Gi ± 3%   16.444Gi ± 1%  +298.59% (p=0.000 n=10)   59.051Gi
 Popcnt-32   7.251Gi ± 1%   13.428Gi ± 0%   +85.20% (p=0.000 n=10)   29.288Gi ± 3%   +303.94% (p=0.000 n=10)
 geomean     4.042Gi         14.22Gi       +251.80%                   39.49Gi        +876.93%
 ```
+
+### Apple M2 Pro
+
+```
+goos: darwin
+goarch: arm64
+pkg: github.com/bwesterb/go-and
+          │    naive     │               purego                │                 asm                  │
+          │    sec/op    │   sec/op     vs base                │    sec/op     vs base                │
+And-12      354.88µ ± 1%   84.98µ ± 1%  -76.05% (p=0.000 n=10)   26.38µ ±  1%  -92.57% (p=0.000 n=10)
+Or-12       357.74µ ± 2%   84.90µ ± 2%  -76.27% (p=0.000 n=10)   26.23µ ±  1%  -92.67% (p=0.000 n=10)
+Xor-12      365.42µ ± 3%   85.10µ ± 1%  -76.71% (p=0.000 n=10)   26.38µ ±  1%  -92.78% (p=0.000 n=10)
+AndNot-12   358.92µ ± 1%   85.65µ ± 4%  -76.14% (p=0.000 n=10)   26.47µ ±  0%  -92.63% (p=0.000 n=10)
+Memset-12   297.70µ ± 1%   80.65µ ± 1%  -72.91% (p=0.000 n=10)   12.27µ ± 12%  -95.88% (p=0.000 n=10)
+Popcnt-12   293.81µ ± 1%   45.94µ ± 1%  -84.36% (p=0.000 n=10)   11.42µ ±  1%  -96.11% (p=0.000 n=10)
+geomean      336.7µ        76.14µ       -77.38%                  20.19µ        -94.00%
+
+          │    naive     │                 purego                 │                   asm                    │
+          │     B/s      │      B/s       vs base                 │      B/s        vs base                  │
+And-12      2.624Gi ± 1%   10.959Gi ± 1%  +317.58% (p=0.000 n=10)   35.305Gi ±  1%  +1245.28% (p=0.000 n=10)
+Or-12       2.603Gi ± 2%   10.970Gi ± 2%  +321.35% (p=0.000 n=10)   35.502Gi ±  1%  +1263.65% (p=0.000 n=10)
+Xor-12      2.549Gi ± 3%   10.944Gi ± 1%  +329.42% (p=0.000 n=10)   35.309Gi ±  1%  +1285.41% (p=0.000 n=10)
+AndNot-12   2.595Gi ± 1%   10.873Gi ± 4%  +319.04% (p=0.000 n=10)   35.183Gi ±  0%  +1255.91% (p=0.000 n=10)
+Memset-12   3.128Gi ± 1%   11.548Gi ± 1%  +269.12% (p=0.000 n=10)   75.993Gi ± 13%  +2329.13% (p=0.000 n=10)
+Popcnt-12   3.170Gi ± 1%   20.271Gi ± 1%  +539.48% (p=0.000 n=10)   81.560Gi ±  1%  +2472.98% (p=0.000 n=10)
+geomean     2.766Gi         12.23Gi       +342.17%                   46.14Gi        +1568.02%
+```