[ASM] New marshalling system for Waf.Run calls to improve speed and reduce allocations #4302

tonyredondo · 2023-06-23T12:05:56Z

Summary of changes

Replace the marshalling system using in AppSec Context.Run waf call with a new one to avoid allocations and improve performance.
Basically, allocations are made on the tracer side and we don't rely on the waf's helpers method to create the structs.
Introduction of a mechanism to recycle memory faster.

Reason for change

Current Waf calls allocates a lot due to the marshalling of strings and conversion from UTF16 to UTF8.

Benchmarks are much improved (see below)
Reducing a lot of pressure to the garbage collector.

Implementation details

Not using the waf helpers anymore
Allocating our own memory
Using an unmanaged memory pool to recycle memory

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs

tracer/src/Datadog.Trace/Util/UnmanagedMemoryPool.cs

anna-git · 2023-06-26T16:01:56Z

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs

        private static readonly IDatadogLogger Log = DatadogLogging.GetLoggerFor(typeof(Encoder));
        private static readonly int ObjectStructSize = Marshal.SizeOf(typeof(DdwafObjectStruct));

+        [ThreadStatic]


for my understanding, why is 'ThreadStatic' necessary? 😅

The Unmanaged Pool is not thread safe for performance reasons, so the best way to avoid leaking memory in a concurrent scenario like ASP.NET is to create and mantain a pool for each thread (and just dispose the pool if the thread get's killed), this pool instance is copied to the Encoder stack so we avoid accessing a thread static twice in the same processing.

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs

anna-git

thanks a lot!

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs

andrewlock · 2023-07-18T09:01:32Z

Execution-Time Benchmarks Report ⏱️

Execution-time results for samples comparing the following branches/commits:

Execution-time benchmarks measure the whole time it takes to execute a program. And are intended to measure the one-off costs. Cases where the execution time results for the PR are worse than latest master results are shown in red. The following thresholds were used for comparing the execution times:

Welch test with statistical test for significance of 5%
Only results indicating a difference greater than 5% and 5 ms are considered.

Note that these results are based on a single point-in-time result for each branch. For full results, see the dashboard.

Graphs show the p99 interval based on the mean and StdDev of the test run, as well as the mean value of the run (shown as a diamond below the graph).

gantt
    title Execution time (ms) FakeDbCommand (.NET Framework 4.6.2) 
    dateFormat  X
    axisFormat %s
    todayMarker off
    section Baseline
    This PR (4302) - mean (69ms)  : 64, 73
     .   : milestone, 69,
    master - mean (69ms)  : 63, 74
     .   : milestone, 69,

    section CallTarget+Inlining+NGEN
    This PR (4302) - mean (1,037ms)  : 1005, 1070
     .   : milestone, 1037,
    master - mean (1,028ms)  : 1011, 1045
     .   : milestone, 1028,

gantt
    title Execution time (ms) FakeDbCommand (.NET Core 3.1) 
    dateFormat  X
    axisFormat %s
    todayMarker off
    section Baseline
    This PR (4302) - mean (106ms)  : 101, 112
     .   : milestone, 106,
    master - mean (107ms)  : 94, 120
     .   : milestone, 107,

    section CallTarget+Inlining+NGEN
    This PR (4302) - mean (749ms)  : 734, 764
     .   : milestone, 749,
    master - mean (746ms)  : 732, 760
     .   : milestone, 746,

gantt
    title Execution time (ms) FakeDbCommand (.NET 6) 
    dateFormat  X
    axisFormat %s
    todayMarker off
    section Baseline
    This PR (4302) - mean (91ms)  : 84, 97
     .   : milestone, 91,
    master - mean (90ms)  : 83, 96
     .   : milestone, 90,

    section CallTarget+Inlining+NGEN
    This PR (4302) - mean (709ms)  : 693, 726
     .   : milestone, 709,
    master - mean (710ms)  : 694, 727
     .   : milestone, 710,

gantt
    title Execution time (ms) HttpMessageHandler (.NET Framework 4.6.2) 
    dateFormat  X
    axisFormat %s
    todayMarker off
    section Baseline
    This PR (4302) - mean (188ms)  : 182, 195
     .   : milestone, 188,
    master - mean (187ms)  : 179, 196
     .   : milestone, 187,

    section CallTarget+Inlining+NGEN
    This PR (4302) - mean (1,141ms)  : 1095, 1187
     .   : milestone, 1141,
    master - mean (1,124ms)  : 1098, 1149
     .   : milestone, 1124,

gantt
    title Execution time (ms) HttpMessageHandler (.NET Core 3.1) 
    dateFormat  X
    axisFormat %s
    todayMarker off
    section Baseline
    This PR (4302) - mean (276ms)  : 266, 285
     .   : milestone, 276,
    master - mean (270ms)  : 261, 280
     .   : milestone, 270,

    section CallTarget+Inlining+NGEN
    This PR (4302) - mean (1,099ms)  : 1065, 1132
     .   : milestone, 1099,
    master - mean (1,104ms)  : 1084, 1125
     .   : milestone, 1104,

gantt
    title Execution time (ms) HttpMessageHandler (.NET 6) 
    dateFormat  X
    axisFormat %s
    todayMarker off
    section Baseline
    This PR (4302) - mean (261ms)  : 250, 272
     .   : milestone, 261,
    master - mean (260ms)  : 247, 273
     .   : milestone, 260,

    section CallTarget+Inlining+NGEN
    This PR (4302) - mean (1,054ms)  : 1034, 1074
     .   : milestone, 1054,
    master - mean (1,053ms)  : 1024, 1081
     .   : milestone, 1053,

andrewlock · 2023-07-18T09:06:08Z

Benchmarks Report 🐌

Benchmarks for #4302 compared to master:

20 benchmarks are faster, with geometric mean 2.095
1 benchmarks are slower, with geometric mean 1.134
21 benchmarks have fewer allocations
3 benchmarks have more allocations

The following thresholds were used for comparing the benchmark speeds:

Mann–Whitney U test with statistical test for significance of 5%
Only results indicating a difference greater than 10% and 0.3 ns are considered.

Allocation changes below 0.5% are ignored.

Benchmark details

Benchmarks.Trace.ActivityBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Gen 2	Allocated
master	`StartStopWithChild`	net6.0	8.21μs	41.7ns	225ns	0.0237	0.0119	0	7.29 KB
master	`StartStopWithChild`	netcoreapp3.1	10.3μs	58ns	376ns	0.0258	0.0103	0	7.39 KB
master	`StartStopWithChild`	net472	15.6μs	55.7ns	216ns	1.28	0.308	0.0923	7.66 KB
#4302	`StartStopWithChild`	net6.0	8.29μs	47.4ns	345ns	0.0252	0.0126	0	7.29 KB
#4302	`StartStopWithChild`	netcoreapp3.1	9.91μs	49.9ns	269ns	0.0404	0.0202	0.00505	7.38 KB
#4302	`StartStopWithChild`	net472	15.6μs	32.9ns	123ns	1.3	0.341	0.116	7.66 KB

Benchmarks.Trace.AgentWriterBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`WriteAndFlushEnrichedTraces`	net6.0	470μs	279ns	1.08μs	0	2.7 KB
master	`WriteAndFlushEnrichedTraces`	netcoreapp3.1	660μs	254ns	951ns	0	2.7 KB
master	`WriteAndFlushEnrichedTraces`	net472	803μs	373ns	1.34μs	0.401	3.3 KB
#4302	`WriteAndFlushEnrichedTraces`	net6.0	504μs	197ns	763ns	0	2.7 KB
#4302	`WriteAndFlushEnrichedTraces`	netcoreapp3.1	633μs	273ns	1.02μs	0	2.7 KB
#4302	`WriteAndFlushEnrichedTraces`	net472	794μs	217ns	782ns	0.396	3.3 KB

Benchmarks.Trace.Asm.AppSecBodyBenchmark - Same speed ✔️ More allocations ⚠️

More allocations ⚠️ in #4302

Benchmark	Base Allocated	Diff Allocated	Change	Change %
Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleSimpleBody‑netcoreapp3.1	1.75 KB	2.01 KB	256 B	14.61%
Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleSimpleBody‑net6.0	1.78 KB	2.03 KB	256 B	14.41%
Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleSimpleBody‑net472	1.82 KB	2.08 KB	256 B	14.05%

Fewer allocations 🎉 in #4302

Benchmark	Base Allocated	Diff Allocated	Change	Change %
Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleMoreComplexBody‑net472	9.42 KB	8.7 KB	-723 B	-7.67%
Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleMoreComplexBody‑net6.0	9.35 KB	8.63 KB	-720 B	-7.70%
Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleMoreComplexBody‑netcoreapp3.1	9.24 KB	8.52 KB	-721 B	-7.80%

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Allocated
master	`AllCycleSimpleBody`	net6.0	42.8μs	79.7ns	287ns	0.0215	0	1.78 KB
master	`AllCycleSimpleBody`	netcoreapp3.1	45.6μs	84.1ns	315ns	0	0	1.75 KB
master	`AllCycleSimpleBody`	net472	47.9μs	15.7ns	60.7ns	0.283	0	1.82 KB
master	`AllCycleMoreComplexBody`	net6.0	235μs	88.1ns	341ns	0.118	0	9.35 KB
master	`AllCycleMoreComplexBody`	netcoreapp3.1	251μs	368ns	1.43μs	0	0	9.24 KB
master	`AllCycleMoreComplexBody`	net472	261μs	282ns	1.09μs	1.44	0	9.42 KB
master	`ObjectExtractorSimpleBody`	net6.0	116ns	0.0364ns	0.131ns	0.00395	0	280 B
master	`ObjectExtractorSimpleBody`	netcoreapp3.1	175ns	0.178ns	0.691ns	0.00367	0	272 B
master	`ObjectExtractorSimpleBody`	net472	144ns	0.128ns	0.496ns	0.0446	0	281 B
master	`ObjectExtractorMoreComplexBody`	net6.0	2.94μs	0.56ns	2.1ns	0.0543	0	3.88 KB
master	`ObjectExtractorMoreComplexBody`	netcoreapp3.1	4.06μs	1.71ns	6.4ns	0.0508	0	3.78 KB
master	`ObjectExtractorMoreComplexBody`	net472	4.15μs	4.27ns	16ns	0.617	0.00615	3.89 KB
#4302	`AllCycleSimpleBody`	net6.0	42.3μs	15.9ns	59.6ns	0.021	0	2.03 KB
#4302	`AllCycleSimpleBody`	netcoreapp3.1	44.1μs	65ns	234ns	0.022	0	2.01 KB
#4302	`AllCycleSimpleBody`	net472	45.4μs	62.8ns	243ns	0.316	0	2.08 KB
#4302	`AllCycleMoreComplexBody`	net6.0	228μs	56.2ns	210ns	0	0	8.63 KB
#4302	`AllCycleMoreComplexBody`	netcoreapp3.1	233μs	142ns	550ns	0	0	8.52 KB
#4302	`AllCycleMoreComplexBody`	net472	236μs	140ns	542ns	1.3	0	8.7 KB
#4302	`ObjectExtractorSimpleBody`	net6.0	128ns	0.0487ns	0.182ns	0.00396	0	280 B
#4302	`ObjectExtractorSimpleBody`	netcoreapp3.1	174ns	0.0408ns	0.158ns	0.00374	0	272 B
#4302	`ObjectExtractorSimpleBody`	net472	146ns	0.0931ns	0.361ns	0.0446	0	281 B
#4302	`ObjectExtractorMoreComplexBody`	net6.0	2.98μs	0.785ns	2.94ns	0.0551	0	3.88 KB
#4302	`ObjectExtractorMoreComplexBody`	netcoreapp3.1	4.05μs	2.01ns	7.8ns	0.0506	0	3.78 KB
#4302	`ObjectExtractorMoreComplexBody`	net472	4.13μs	1.35ns	5.23ns	0.617	0.00619	3.89 KB

Benchmarks.Trace.Asm.AppSecWafBenchmark - Faster 🎉 Fewer allocations 🎉

Faster 🎉 in #4302

Benchmark	base/diff	Base Median (ns)	Diff Median (ns)	Modality
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (10))‑net6.0	3.884	49,313.75	12,695.19
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (10))‑net472	3.434	96,104.83	27,989.58
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (10))‑netcoreapp3.1	3.403	67,385.13	19,800.35	bimodal
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (1000))‑net6.0	2.918	130,238.34	44,640.39
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (100))‑net6.0	2.801	126,958.86	45,321.28
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (100))‑net472	2.729	222,391.87	81,492.96
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (1000))‑netcoreapp3.1	2.723	171,701.07	63,055.73
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (100))‑netcoreapp3.1	2.690	167,989.34	62,448.69
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (1000))‑net472	2.557	213,375.31	83,437.91
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (100, attack))‑net472	1.803	311,935.10	172,964.97

Fewer allocations 🎉 in #4302

Benchmark	Base Allocated	Diff Allocated	Change	Change %
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (10, attack))‑net472	22.71 KB	16.04 KB	-6.67 KB	-29.37%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (10, attack))‑net6.0	22.42 KB	15.77 KB	-6.65 KB	-29.66%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (10, attack))‑netcoreapp3.1	22.37 KB	15.72 KB	-6.65 KB	-29.72%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (100, attack))‑net472	42.89 KB	29.84 KB	-13.05 KB	-30.42%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (1000, attack))‑net472	42.9 KB	29.85 KB	-13.05 KB	-30.42%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (1000, attack))‑netcoreapp3.1	41.69 KB	28.77 KB	-12.91 KB	-30.98%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (100, attack))‑netcoreapp3.1	41.7 KB	28.77 KB	-12.93 KB	-31.00%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (100, attack))‑net6.0	41.09 KB	27.99 KB	-13.1 KB	-31.89%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWafWithAttack(args: NestedMap (1000, attack))‑net6.0	41.09 KB	27.98 KB	-13.11 KB	-31.90%
Benchmarks.Trace.Asm.AppSecWafBenchmark.RunWaf(args: NestedMap (100))‑net472	36.33 KB	23.29 KB	-13.05 KB	-35.91%

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Allocated
master	`RunWaf(args=NestedMap (10))`	net6.0	49.5μs	182ns	629ns	0.219	0	16.07 KB
master	`RunWaf(args=NestedMap (10))`	netcoreapp3.1	67.7μs	381ns	2.64μs	0.2	0	16.06 KB
master	`RunWaf(args=NestedMap (10))`	net472	96.1μs	35ns	121ns	2.54	0.0957	16.14 KB
master	`RunWafWithAttack(args=Neste(...)tack) [22])`	net6.0	116μs	38.8ns	145ns	0.291	0	22.42 KB
master	`RunWafWithAttack(args=Neste(...)tack) [22])`	netcoreapp3.1	138μs	527ns	1.9μs	0.276	0	22.37 KB
master	`RunWafWithAttack(args=Neste(...)tack) [22])`	net472	166μs	75ns	291ns	3.54	0.165	22.71 KB
master	`RunWaf(args=NestedMap (100))`	net6.0	126μs	604ns	2.34μs	0.447	0	34.74 KB
master	`RunWaf(args=NestedMap (100))`	netcoreapp3.1	165μs	864ns	4.14μs	0.469	0	35.39 KB
master	`RunWaf(args=NestedMap (100))`	net472	222μs	154ns	596ns	5.73	0.424	36.33 KB
master	`RunWafWithAttack(args=Neste(...)tack) [23])`	net6.0	207μs	1.12μs	6.33μs	0.492	0	41.09 KB
master	`RunWafWithAttack(args=Neste(...)tack) [23])`	netcoreapp3.1	247μs	955ns	3.7μs	0.497	0	41.7 KB
master	`RunWafWithAttack(args=Neste(...)tack) [23])`	net472	311μs	538ns	2.08μs	6.76	0.601	42.89 KB
master	`RunWaf(args=NestedMap (1000))`	net6.0	130μs	110ns	426ns	0.471	0	34.74 KB
master	`RunWaf(args=NestedMap (1000))`	netcoreapp3.1	170μs	613ns	2.38μs	0.414	0	35.39 KB
master	`RunWaf(args=NestedMap (1000))`	net472	213μs	65ns	252ns	5.76	0.427	36.34 KB
master	`RunWafWithAttack(args=Neste(...)tack) [24])`	net6.0	204μs	185ns	718ns	0.52	0	41.09 KB
master	`RunWafWithAttack(args=Neste(...)tack) [24])`	netcoreapp3.1	250μs	975ns	3.78μs	0.488	0	41.69 KB
master	`RunWafWithAttack(args=Neste(...)tack) [24])`	net472	306μs	1.04μs	4.03μs	6.79	0.453	42.9 KB
#4302	`RunWaf(args=NestedMap (10))`	net6.0	12.7μs	3.02ns	10.9ns	0.128	0	9.42 KB
#4302	`RunWaf(args=NestedMap (10))`	netcoreapp3.1	19.8μs	7.04ns	25.4ns	0.128	0	9.42 KB
#4302	`RunWaf(args=NestedMap (10))`	net472	28μs	21.6ns	80.8ns	1.5	0	9.48 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [22])`	net6.0	77.4μs	46.9ns	181ns	0.195	0	15.77 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [22])`	netcoreapp3.1	84.9μs	132ns	513ns	0.211	0	15.72 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [22])`	net472	98.7μs	56.6ns	212ns	2.52	0	16.04 KB
#4302	`RunWaf(args=NestedMap (100))`	net6.0	45.4μs	57.1ns	221ns	0.294	0	21.64 KB
#4302	`RunWaf(args=NestedMap (100))`	netcoreapp3.1	62.5μs	64.3ns	240ns	0.31	0	22.47 KB
#4302	`RunWaf(args=NestedMap (100))`	net472	81.5μs	51.7ns	193ns	3.7	0.0406	23.29 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [23])`	net6.0	125μs	123ns	460ns	0.371	0	27.99 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [23])`	netcoreapp3.1	146μs	266ns	1.03μs	0.367	0	28.77 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [23])`	net472	173μs	153ns	593ns	4.74	0.0862	29.84 KB
#4302	`RunWaf(args=NestedMap (1000))`	net6.0	44.7μs	47.4ns	184ns	0.289	0	21.64 KB
#4302	`RunWaf(args=NestedMap (1000))`	netcoreapp3.1	63μs	71.7ns	278ns	0.284	0	22.47 KB
#4302	`RunWaf(args=NestedMap (1000))`	net472	83.5μs	53.5ns	207ns	3.66	0.0416	23.28 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [24])`	net6.0	127μs	157ns	608ns	0.384	0	27.98 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [24])`	netcoreapp3.1	146μs	292ns	1.09μs	0.367	0	28.77 KB
#4302	`RunWafWithAttack(args=Neste(...)tack) [24])`	net472	173μs	185ns	718ns	4.73	0.086	29.85 KB

Benchmarks.Trace.AspNetCoreBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`SendRequest`	net6.0	169μs	88.3ns	342ns	0.254	18.09 KB
master	`SendRequest`	netcoreapp3.1	189μs	327ns	1.27μs	0.189	20.25 KB
master	`SendRequest`	net472	0.000182ns	0.000101ns	0.000391ns	0	0 b
#4302	`SendRequest`	net6.0	168μs	158ns	611ns	0.249	18.09 KB
#4302	`SendRequest`	netcoreapp3.1	192μs	311ns	1.2μs	0.193	20.25 KB
#4302	`SendRequest`	net472	0.00243ns	0.000367ns	0.00137ns	0	0 b

Benchmarks.Trace.CIVisibilityProtocolWriterBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Gen 2	Allocated
master	`WriteAndFlushEnrichedTraces`	net6.0	537μs	2.4μs	9.28μs	0.519	0	0	41.46 KB
master	`WriteAndFlushEnrichedTraces`	netcoreapp3.1	624μs	1.7μs	5.9μs	0.316	0	0	41.77 KB
master	`WriteAndFlushEnrichedTraces`	net472	815μs	3.81μs	17.9μs	8.58	2.45	0.408	53.26 KB
#4302	`WriteAndFlushEnrichedTraces`	net6.0	532μs	1.42μs	5.12μs	0.506	0	0	41.48 KB
#4302	`WriteAndFlushEnrichedTraces`	netcoreapp3.1	613μs	388ns	1.5μs	0.312	0	0	41.64 KB
#4302	`WriteAndFlushEnrichedTraces`	net472	799μs	2.35μs	8.78μs	8.31	2.37	0.396	53.24 KB

Benchmarks.Trace.DbCommandBenchmark - Faster 🎉 Same allocations ✔️

Faster 🎉 in #4302

Benchmark	base/diff	Base Median (ns)	Diff Median (ns)	Modality
Benchmarks.Trace.DbCommandBenchmark.ExecuteNonQuery‑net6.0	1.198	1,174.96	981.11

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`ExecuteNonQuery`	net6.0	1.18μs	0.715ns	2.77ns	0.0106	768 B
master	`ExecuteNonQuery`	netcoreapp3.1	1.35μs	0.616ns	2.39ns	0.0101	768 B
master	`ExecuteNonQuery`	net472	1.63μs	0.372ns	1.44ns	0.116	730 B
#4302	`ExecuteNonQuery`	net6.0	981ns	0.369ns	1.43ns	0.0104	768 B
#4302	`ExecuteNonQuery`	netcoreapp3.1	1.42μs	0.453ns	1.7ns	0.00997	768 B
#4302	`ExecuteNonQuery`	net472	1.58μs	0.447ns	1.67ns	0.116	730 B

Benchmarks.Trace.ElasticsearchBenchmark - Slower ⚠️ Same allocations ✔️

Slower ⚠️ in #4302

Benchmark	diff/base	Base Median (ns)	Diff Median (ns)	Modality
Benchmarks.Trace.ElasticsearchBenchmark.CallElasticsearch‑net472	1.134	2,231.82	2,530.40

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Allocated
master	`CallElasticsearch`	net6.0	1.2μs	1.37ns	5.32ns	0.0138	0	992 B
master	`CallElasticsearch`	netcoreapp3.1	1.46μs	0.463ns	1.79ns	0.0131	0	992 B
master	`CallElasticsearch`	net472	2.23μs	0.631ns	2.36ns	0.159	0.00112	1 KB
master	`CallElasticsearchAsync`	net6.0	1.31μs	1.64ns	6.12ns	0.0137	0	968 B
master	`CallElasticsearchAsync`	netcoreapp3.1	1.51μs	0.815ns	3.05ns	0.0143	0	1.04 KB
master	`CallElasticsearchAsync`	net472	2.49μs	0.776ns	3ns	0.167	0	1.06 KB
#4302	`CallElasticsearch`	net6.0	1.19μs	1.47ns	5.52ns	0.0137	0	992 B
#4302	`CallElasticsearch`	netcoreapp3.1	1.41μs	0.795ns	2.98ns	0.0135	0	992 B
#4302	`CallElasticsearch`	net472	2.53μs	0.522ns	2.02ns	0.158	0	1 KB
#4302	`CallElasticsearchAsync`	net6.0	1.34μs	0.947ns	3.67ns	0.0134	0	968 B
#4302	`CallElasticsearchAsync`	netcoreapp3.1	1.57μs	0.73ns	2.83ns	0.014	0	1.04 KB
#4302	`CallElasticsearchAsync`	net472	2.52μs	0.571ns	2.14ns	0.168	0.00126	1.06 KB

Benchmarks.Trace.GraphQLBenchmark - Faster 🎉 Same allocations ✔️

Faster 🎉 in #4302

Benchmark	base/diff	Base Median (ns)	Diff Median (ns)	Modality
Benchmarks.Trace.GraphQLBenchmark.ExecuteAsync‑net6.0	1.134	1,335.74	1,177.50

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Allocated
master	`ExecuteAsync`	net6.0	1.34μs	0.695ns	2.6ns	0.0129	0	912 B
master	`ExecuteAsync`	netcoreapp3.1	1.41μs	0.595ns	2.22ns	0.0121	0	912 B
master	`ExecuteAsync`	net472	1.73μs	1.34ns	5.19ns	0.139	0.000862	875 B
#4302	`ExecuteAsync`	net6.0	1.18μs	0.634ns	2.37ns	0.0124	0	912 B
#4302	`ExecuteAsync`	netcoreapp3.1	1.4μs	0.473ns	1.77ns	0.0119	0	912 B
#4302	`ExecuteAsync`	net472	1.7μs	0.448ns	1.62ns	0.138	0.000853	875 B

Benchmarks.Trace.HttpClientBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`SendAsync`	net6.0	3.8μs	1.7ns	6.35ns	0.0267	1.94 KB
master	`SendAsync`	netcoreapp3.1	4.33μs	2.17ns	8.39ns	0.0325	2.48 KB
master	`SendAsync`	net472	7.23μs	4.65ns	18ns	0.482	3.05 KB
#4302	`SendAsync`	net6.0	3.77μs	1.54ns	5.77ns	0.0264	1.94 KB
#4302	`SendAsync`	netcoreapp3.1	4.48μs	1.85ns	6.94ns	0.0336	2.48 KB
#4302	`SendAsync`	net472	7.08μs	2.28ns	8.53ns	0.481	3.05 KB

Benchmarks.Trace.ILoggerBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`EnrichedLog`	net6.0	1.34μs	3.19ns	12.4ns	0.0231	1.62 KB
master	`EnrichedLog`	netcoreapp3.1	2.01μs	0.971ns	3.63ns	0.0221	1.62 KB
master	`EnrichedLog`	net472	2.38μs	4.33ns	16.8ns	0.244	1.54 KB
#4302	`EnrichedLog`	net6.0	1.32μs	0.569ns	2.13ns	0.0226	1.62 KB
#4302	`EnrichedLog`	netcoreapp3.1	1.94μs	0.675ns	2.43ns	0.0218	1.62 KB
#4302	`EnrichedLog`	net472	2.42μs	2.59ns	9.33ns	0.245	1.54 KB

Benchmarks.Trace.Log4netBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Gen 1	Allocated
master	`EnrichedLog`	net6.0	114μs	192ns	742ns	0.0571	0	4.21 KB
master	`EnrichedLog`	netcoreapp3.1	118μs	179ns	669ns	0	0	4.21 KB
master	`EnrichedLog`	net472	150μs	159ns	594ns	0.671	0.224	4.38 KB
#4302	`EnrichedLog`	net6.0	115μs	117ns	423ns	0.0582	0	4.21 KB
#4302	`EnrichedLog`	netcoreapp3.1	119μs	273ns	1.02μs	0.0594	0	4.21 KB
#4302	`EnrichedLog`	net472	150μs	96ns	346ns	0.674	0.225	4.38 KB

Benchmarks.Trace.NLogBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`EnrichedLog`	net6.0	2.96μs	1ns	3.75ns	0.0298	2.18 KB
master	`EnrichedLog`	netcoreapp3.1	3.77μs	1.09ns	4.07ns	0.0282	2.18 KB
master	`EnrichedLog`	net472	4.59μs	2.28ns	8.84ns	0.314	1.99 KB
#4302	`EnrichedLog`	net6.0	2.89μs	1.41ns	5.29ns	0.03	2.18 KB
#4302	`EnrichedLog`	netcoreapp3.1	3.94μs	1.4ns	5.24ns	0.0295	2.18 KB
#4302	`EnrichedLog`	net472	4.59μs	3.4ns	12.7ns	0.316	1.99 KB

Benchmarks.Trace.RedisBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`SendReceive`	net6.0	1.24μs	0.844ns	3.16ns	0.0162	1.16 KB
master	`SendReceive`	netcoreapp3.1	1.62μs	2.34ns	8.1ns	0.0155	1.16 KB
master	`SendReceive`	net472	2.12μs	2.2ns	8.53ns	0.185	1.16 KB
#4302	`SendReceive`	net6.0	1.27μs	1.1ns	4.26ns	0.0159	1.16 KB
#4302	`SendReceive`	netcoreapp3.1	1.49μs	0.771ns	2.88ns	0.0159	1.16 KB
#4302	`SendReceive`	net472	2.14μs	2.57ns	9.95ns	0.184	1.16 KB

Benchmarks.Trace.SerilogBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`EnrichedLog`	net6.0	2.62μs	1.77ns	6.86ns	0.0209	1.53 KB
master	`EnrichedLog`	netcoreapp3.1	3.66μs	1.16ns	4.33ns	0.0201	1.58 KB
master	`EnrichedLog`	net472	3.97μs	3.62ns	13.5ns	0.309	1.96 KB
#4302	`EnrichedLog`	net6.0	2.5μs	0.846ns	3.28ns	0.0213	1.53 KB
#4302	`EnrichedLog`	netcoreapp3.1	3.61μs	1.76ns	6.81ns	0.0216	1.58 KB
#4302	`EnrichedLog`	net472	4.11μs	1.79ns	6.95ns	0.31	1.96 KB

Benchmarks.Trace.SpanBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`StartFinishSpan`	net6.0	441ns	0.694ns	2.69ns	0.00751	536 B
master	`StartFinishSpan`	netcoreapp3.1	575ns	0.192ns	0.693ns	0.00721	536 B
master	`StartFinishSpan`	net472	621ns	0.285ns	1.07ns	0.0852	538 B
master	`StartFinishScope`	net6.0	551ns	0.176ns	0.657ns	0.00916	656 B
master	`StartFinishScope`	netcoreapp3.1	773ns	0.459ns	1.78ns	0.00888	656 B
master	`StartFinishScope`	net472	847ns	0.321ns	1.24ns	0.098	618 B
#4302	`StartFinishSpan`	net6.0	455ns	0.304ns	1.18ns	0.00747	536 B
#4302	`StartFinishSpan`	netcoreapp3.1	579ns	0.308ns	1.19ns	0.00712	536 B
#4302	`StartFinishSpan`	net472	597ns	0.136ns	0.528ns	0.0851	538 B
#4302	`StartFinishScope`	net6.0	546ns	0.301ns	1.17ns	0.00931	656 B
#4302	`StartFinishScope`	netcoreapp3.1	751ns	0.324ns	1.26ns	0.00868	656 B
#4302	`StartFinishScope`	net472	868ns	0.341ns	1.32ns	0.0978	618 B

Benchmarks.Trace.TraceAnnotationsBenchmark - Same speed ✔️ Same allocations ✔️

Raw results

Branch	Method	Toolchain	Mean	StdError	StdDev	Gen 0	Allocated
master	`RunOnMethodBegin`	net6.0	563ns	0.244ns	0.946ns	0.00937	656 B
master	`RunOnMethodBegin`	netcoreapp3.1	811ns	0.538ns	2.01ns	0.00888	656 B
master	`RunOnMethodBegin`	net472	887ns	0.283ns	1.02ns	0.098	618 B
#4302	`RunOnMethodBegin`	net6.0	558ns	0.173ns	0.649ns	0.0093	656 B
#4302	`RunOnMethodBegin`	netcoreapp3.1	787ns	0.605ns	2.26ns	0.00876	656 B
#4302	`RunOnMethodBegin`	net472	926ns	0.233ns	0.903ns	0.0979	618 B

daniel-romano-DD · 2023-07-18T09:11:33Z

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs

+
+                if (enumerableDic is IDictionary)
+                {
+                    var typeKVP = typeof(KeyValuePair<TKey, TValue>);


I'd cache the typeofs in static fields to avoid the Reflection slugginess

typeof is already performant, in IL this is transformed into a ldtoken and then a GetTypeFromHandle that is really fast because is only a getter to the runtime handle field. https://source.dot.net/#System.Private.CoreLib/src/System/Type.CoreCLR.cs,97

So I don't think caching is required here.

andrewlock · 2023-07-18T09:39:52Z

Throughput/Crank Report:zap:

Throughput results for AspNetCoreSimpleController comparing the following branches/commits:

Cases where throughput results for the PR are worse than latest master (5% drop or greater), results are shown in red.

Note that these results are based on a single point-in-time result for each branch. For full results, see one of the many, many dashboards!

gantt
    title Throughput Linux x64 (Total requests) 
    dateFormat  X
    axisFormat %s
    section Baseline
    This PR (4302) (11.636M)   : 0, 11635977
    master (11.433M)   : 0, 11433406
    benchmarks/2.37.0 (11.517M)   : 0, 11517370
    benchmarks/2.9.0 (11.506M)   : 0, 11506345

    section Automatic
    This PR (4302) (8.180M)   : 0, 8180319
    master (8.114M)   : 0, 8114165
    benchmarks/2.37.0 (8.143M)   : 0, 8142506
    benchmarks/2.9.0 (8.280M)   : 0, 8279564

    section Trace stats
    master (8.118M)   : 0, 8118262
    benchmarks/2.37.0 (8.044M)   : 0, 8044129

    section Manual
    This PR (4302) (10.253M)   : 0, 10253315
    master (10.339M)   : 0, 10338672
    benchmarks/2.37.0 (10.278M)   : 0, 10278226

    section Manual + Automatic
    This PR (4302) (7.828M)   : 0, 7827523
    master (7.711M)   : 0, 7710676
    benchmarks/2.37.0 (7.786M)   : 0, 7785678

    section Version Conflict
    master (7.078M)   : 0, 7077875
    benchmarks/2.37.0 (7.067M)   : 0, 7066760

gantt
    title Throughput Linux arm64 (Total requests) 
    dateFormat  X
    axisFormat %s
    section Baseline
    This PR (4302) (9.637M)   : 0, 9636754
    master (9.576M)   : 0, 9576012
    benchmarks/2.37.0 (9.619M)   : 0, 9618848
    benchmarks/2.9.0 (9.542M)   : 0, 9542061

    section Automatic
    This PR (4302) (6.699M)   : 0, 6698584
    master (6.687M)   : 0, 6687360
    benchmarks/2.37.0 (6.844M)   : 0, 6843850

    section Trace stats
    master (6.893M)   : 0, 6893377
    benchmarks/2.37.0 (6.801M)   : 0, 6801072

    section Manual
    This PR (4302) (8.542M)   : 0, 8541678
    master (8.538M)   : 0, 8538336
    benchmarks/2.37.0 (8.587M)   : 0, 8587283

    section Manual + Automatic
    This PR (4302) (6.617M)   : 0, 6616821
    master (6.448M)   : 0, 6448420
    benchmarks/2.37.0 (6.537M)   : 0, 6536939

    section Version Conflict
    master (5.788M)   : 0, 5788108
    benchmarks/2.37.0 (5.769M)   : 0, 5768881

gantt
    title Throughput Windows x64 (Total requests) 
    dateFormat  X
    axisFormat %s
    section Baseline
    This PR (4302) (9.126M)   : 0, 9125559
    master (9.569M)   : 0, 9569061
    benchmarks/2.37.0 (9.309M)   : 0, 9308797
    benchmarks/2.9.0 (9.366M)   : 0, 9366476

    section Automatic
    This PR (4302) (6.739M)   : 0, 6739382
    master (6.663M)   : 0, 6662702
    benchmarks/2.37.0 (6.552M)   : 0, 6552099
    benchmarks/2.9.0 (6.864M)   : 0, 6864235

    section Trace stats
    master (6.622M)   : 0, 6622086
    benchmarks/2.37.0 (6.685M)   : 0, 6685340

    section Manual
    This PR (4302) (8.200M)   : 0, 8199967
    master (8.316M)   : 0, 8316117
    benchmarks/2.37.0 (8.493M)   : 0, 8492505

    section Manual + Automatic
    This PR (4302) (6.391M)   : 0, 6391465
    master (6.415M)   : 0, 6414915
    benchmarks/2.37.0 (6.373M)   : 0, 6372588

    section Version Conflict
    master (5.818M)   : 0, 5817640
    benchmarks/2.37.0 (5.938M)   : 0, 5937795

gantt
    title Throughput Linux x64 (ASM) (Total requests) 
    dateFormat  X
    axisFormat %s
    section Baseline
    This PR (4302) (7.369M)   : 0, 7369404
    master (7.389M)   : 0, 7388844
    benchmarks/2.37.0 (7.239M)   : 0, 7239363
    benchmarks/2.9.0 (7.508M)   : 0, 7508023

    section No attack
    This PR (4302) (2.140M)   : 0, 2139796
    master (2.036M)   : 0, 2036034
    benchmarks/2.37.0 (2.105M)   : 0, 2105389
    benchmarks/2.9.0 (3.275M)   : 0, 3275098

    section Attack
    This PR (4302) (1.762M)   : 0, 1762035
    master (1.644M)   : 0, 1643534
    benchmarks/2.37.0 (1.686M)   : 0, 1685550
    benchmarks/2.9.0 (2.570M)   : 0, 2570112

    section Blocking
    This PR (4302) (3.456M)   : 0, 3456302
    master (3.180M)   : 0, 3179669
    benchmarks/2.37.0 (3.048M)   : 0, 3048473

daniel-romano-DD · 2023-07-18T11:07:31Z

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs

+                    void EnumerateItems<TKeySource, TValueSource>()
+                        where TKeySource : notnull
+                    {
+                        var itemData = childrenData;


Maybe it's better to pass captured vars as params in order to avoid extra heap allocations?
https://learn.microsoft.com/en-us/dotnet/csharp/programming-guide/classes-and-structs/local-functions#heap-allocations

good point, but if we want to avoid any captured vars in this one, it really damages readability, we would have to pass childrenData, enumerableDic, getKey,getValue, argToFree, applySafetyLimits, remainingDepth, pool`...
and wouldnt the copy of all these arguments balance out the capture? I dont know..
but as we're already gaining a lot in allocations/speed in this PR (see waf benchmark above), maybe this could be kept as is, for the sake of readability?

daniel-romano-DD

Nice work

datadog-ddstaging · 2023-09-14T10:47:10Z

Datadog Report

Branch report: tony/improve-marshalling-for-waf
Commit report: a6c44b7

✅ dd-trace-dotnet: 0 Failed, 0 New Flaky, 302174 Passed, 1154 Skipped, 29m 25.6s Wall Time

andrewlock

LGTM on the tracer side, but I didn't look at the ASM side

andrewlock · 2023-09-15T11:43:24Z

tracer/src/Datadog.Trace/AppSec/Waf/Context.cs

+                var pool = Encoder.Pool;
+                var pwArgs = Encoder.Encode(addresses, applySafetyLimits: true, argToFree: _argCache, pool: pool);
+                code = _waf.Run(_contextHandle, ref pwArgs, ref retNative, timeoutMicroSeconds);
+                pool.Return(_argCache);


Should pool.Return() (and _argCache.Clear();) be in a try{}finally{} to be on the safe side?

fixed in a6c44b7

andrewlock · 2023-09-15T11:52:15Z

tracer/src/Datadog.Trace/Util/UnmanagedMemoryPool.cs

+        var items = _items;
+        var length = _length;


Just wondering, given these variables are both readonly, is their much benefit to the local copy here?

yes thanks!, Tony checked, and removing the copy makes the generated code shorter, fixed in a6c44b7

andrewlock · 2023-09-15T11:56:47Z

tracer/src/Datadog.Trace/Util/UnmanagedMemoryPool.cs

+
+namespace Datadog.Trace.Util;
+
+internal unsafe class UnmanagedMemoryPool : IDisposable


I think we should add comments to this type and the Rent() methods etc (at the very least) about the fact that this type is not thread safe, and should only be used as such with [ThreadStatic]. Because the equivalent BCL types are thread safe, so this could cause some interesting issues if used incorrectly 😄

yes! added in a6c44b7

DataDog#4891) * Revert "[ASM] New marshalling system for Waf.Run calls to improve speed and reduce allocations (DataDog#4302)" This reverts commit 096f106. * fix schema test by syncing with master

tonyredondo self-assigned this Jun 23, 2023

github-advanced-security bot found potential problems Jun 23, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Fixed Show fixed Hide fixed

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Fixed Show fixed Hide fixed

tonyredondo changed the title ~~New marshalling system for Waf.Run call to improve speed and reduce allocations~~ New marshalling system for Waf.Run calls to improve speed and reduce allocations Jun 23, 2023

tonyredondo marked this pull request as ready for review June 26, 2023 13:18

tonyredondo requested review from a team as code owners June 26, 2023 13:18

github-advanced-security bot found potential problems Jun 26, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Fixed Show fixed Hide fixed

DataDog deleted a comment from andrewlock Jun 26, 2023

DataDog deleted a comment from datadog-ddstaging bot Jun 26, 2023

anna-git reviewed Jun 26, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Outdated Show resolved Hide resolved

anna-git reviewed Jun 26, 2023

View reviewed changes

tracer/src/Datadog.Trace/Util/UnmanagedMemoryPool.cs Show resolved Hide resolved

anna-git reviewed Jun 26, 2023

View reviewed changes

anna-git reviewed Jun 27, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Outdated Show resolved Hide resolved

anna-git reviewed Jun 27, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Outdated Show resolved Hide resolved

anna-git reviewed Jun 27, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Outdated Show resolved Hide resolved

anna-git added the area:asm label Jun 27, 2023

anna-git reviewed Jun 27, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Outdated Show resolved Hide resolved

anna-git approved these changes Jun 28, 2023

View reviewed changes

github-advanced-security bot found potential problems Jun 29, 2023

View reviewed changes

tracer/src/Datadog.Trace/AppSec/Waf/Encoder.cs Fixed Show fixed Hide fixed

DataDog deleted a comment from andrewlock Jun 29, 2023

This comment has been minimized.

Sign in to view

daniel-romano-DD reviewed Jul 18, 2023

View reviewed changes

robertpi approved these changes Jul 20, 2023

View reviewed changes

daniel-romano-DD approved these changes Jul 20, 2023

View reviewed changes

anna-git force-pushed the tony/improve-marshalling-for-waf branch from c3a2411 to 7260172 Compare September 14, 2023 10:21

anna-git force-pushed the tony/improve-marshalling-for-waf branch 3 times, most recently from 345f76f to c5e5822 Compare September 14, 2023 13:10

anna-git added 2 commits September 14, 2023 17:36

Change encoding way, create objects managed side

88c1e13

Update appsec classes to use with new encoder

2253f18

anna-git force-pushed the tony/improve-marshalling-for-waf branch from c5e5822 to 97b27b5 Compare September 14, 2023 15:36

Adapt tests

51b4a45

anna-git force-pushed the tony/improve-marshalling-for-waf branch from 97b27b5 to 51b4a45 Compare September 15, 2023 06:20

andrewlock approved these changes Sep 15, 2023

View reviewed changes

Response to comments

a6c44b7

anna-git changed the title ~~New marshalling system for Waf.Run calls to improve speed and reduce allocations~~ [ASM] New marshalling system for Waf.Run calls to improve speed and reduce allocations Sep 16, 2023

anna-git merged commit 096f106 into master Sep 16, 2023
11 checks passed

anna-git deleted the tony/improve-marshalling-for-waf branch September 16, 2023 11:45

github-actions bot added this to the vNext milestone Sep 16, 2023

anna-git mentioned this pull request Sep 18, 2023

[ASM] Fix waf memory tests flakiness #4636

Merged

anna-git mentioned this pull request Oct 4, 2023

[ASM] Api security foundation for dotnet core #4654

Merged


		namespace Datadog.Trace.Util;

		internal unsafe class UnmanagedMemoryPool : IDisposable

[ASM] New marshalling system for Waf.Run calls to improve speed and reduce allocations #4302

[ASM] New marshalling system for Waf.Run calls to improve speed and reduce allocations #4302

Conversation

tonyredondo commented Jun 23, 2023 • edited by anna-git Loading

Summary of changes

Reason for change

Implementation details

anna-git Jun 26, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

anna-git left a comment

Choose a reason for hiding this comment

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

andrewlock commented Jul 18, 2023 • edited Loading

Execution-Time Benchmarks Report ⏱️

andrewlock commented Jul 18, 2023 • edited Loading

Benchmarks Report 🐌

Benchmark details

Raw results

Raw results

More allocations ⚠️ in #4302

Fewer allocations 🎉 in #4302

Raw results

Faster 🎉 in #4302

Fewer allocations 🎉 in #4302

Raw results

Raw results

Raw results

Faster 🎉 in #4302

Raw results

Slower ⚠️ in #4302

Raw results

Faster 🎉 in #4302

Raw results

Raw results

Raw results

Raw results

Raw results

Raw results

Raw results

Raw results

Raw results

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andrewlock commented Jul 18, 2023 • edited Loading

Throughput/Crank Report:zap:

Choose a reason for hiding this comment

Choose a reason for hiding this comment

daniel-romano-DD left a comment

Choose a reason for hiding this comment

datadog-ddstaging bot commented Sep 14, 2023 • edited Loading

Datadog Report

andrewlock left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tonyredondo commented Jun 23, 2023 •

edited by anna-git

Loading

anna-git Jun 26, 2023 •

edited

Loading

andrewlock commented Jul 18, 2023 •

edited

Loading

andrewlock commented Jul 18, 2023 •

edited

Loading

andrewlock commented Jul 18, 2023 •

edited

Loading

datadog-ddstaging bot commented Sep 14, 2023 •

edited

Loading