Hello, @cincuranet , I was fiddling with @fdcastel `s PR (#1203) when I have noticed that between used in his PR nuget 10.3.1 and master there was an update (10.3.3) which significantly affected performance and allocation rates when woring with UTF8 strings.
| Method |
Job |
BuildConfiguration |
DataType |
Mean |
Error |
StdDev |
Ratio |
RatioSD |
Gen0 |
Gen1 |
Allocated |
Alloc Ratio |
| Fetch |
NuGet100 |
10.3.2 |
CHAR((...) UTF8 [28] |
1.007 s |
0.0261 s |
0.0770 s |
1.01 |
0.11 |
46000.0000 |
7000.0000 |
369.79 MB |
1.00 |
| Fetch |
Core100 |
10.3.3 |
CHAR((...) UTF8 [28] |
2.131 s |
0.0420 s |
0.0810 s |
2.13 |
0.18 |
543000.0000 |
112000.0000 |
4339.37 MB |
11.73 |
Interesting fact is that the update was fixing an issue around the same place where one of my old PRs (#1252) was implementing optimizations directly related to affected methods. Related issue resulting in problematic fix bd0db88: #1213
Table relaed to my PR vs 10.3.1
| Method |
Job |
BuildConfiguration |
DataType |
Mean |
Error |
StdDev |
Ratio |
RatioSD |
Gen0 |
Gen1 |
Allocated |
Alloc Ratio |
| Fetch |
NuGet100 |
ReleaseNuGet |
CHAR((...) UTF8 [28] |
1.066 s |
0.0398 s |
0.1175 s |
1.01 |
0.16 |
46000.0000 |
7000.0000 |
369.79 MB |
1.00 |
| Fetch |
Core100 |
Release |
CHAR((...) UTF8 [28] |
1.161 s |
0.0232 s |
0.0683 s |
1.10 |
0.14 |
46000.0000 |
7000.0000 |
369.79 MB |
1.00 |
Can you, please, review my PR changes and check if they are correct (aka they won't reintroduce the fixed issue, tests are passing successfully meanwhile)?
Hello, @cincuranet , I was fiddling with @fdcastel `s PR (#1203) when I have noticed that between used in his PR nuget 10.3.1 and master there was an update (10.3.3) which significantly affected performance and allocation rates when woring with UTF8 strings.
Interesting fact is that the update was fixing an issue around the same place where one of my old PRs (#1252) was implementing optimizations directly related to affected methods. Related issue resulting in problematic fix bd0db88: #1213
Table relaed to my PR vs 10.3.1
Can you, please, review my PR changes and check if they are correct (aka they won't reintroduce the fixed issue, tests are passing successfully meanwhile)?