[Perf -97%] System.Collections.IterateForEach<String>.ImmutableArray

[performanceautofiler]

Run Information

Architecture	x64
OS	ubuntu 18.04
Changes	diff

Regressions in System.Collections.IterateForEach

Benchmark	Baseline	Test	Test/Base	Modality	Baseline Outlier	Baseline ETL	Comapre ETL
ImmutableArray	229.85 ns	452.97 ns	1.97		True

Historical Data in Reporting System

Repro

git clone https://github.com/dotnet/performance.git
python3 .\performance\scripts\benchmarks_ci.py -f netcoreapp5.0 --filter 'System.Collections.IterateForEach<String>*'

Histogram

System.Collections.IterateForEach.ImmutableArray(Size: 512)

[191.001 ; 241.559) | @@@@@@@@@@@@@@@@@@@@@@
[241.559 ; 292.117) | 
[292.117 ; 342.675) | 
[342.675 ; 393.232) | 
[393.232 ; 428.361) | 
[428.361 ; 480.151) | @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@

Docs

Profiling workflow for dotnet/runtime repository
Benchmarking workflow for dotnet/runtime repository

[kunalspathak]

I am not sure if this is alignment related issue or the loop alignment would help here. The loop in the benchmark fits in 2 32B chunks as seen below and we don't need to add extra alignment here.

; =========================== 32B boundary ===========================
 00007ffb`9af2e863        EB3E                 jmp      SHORT G_M56622_IG07
; ~~~~~~~~~~~~~~~~~~~~~~ Skipping because (currentOffset < extraBytesNotInLoop) ~~~~~~~~~~~~~~~~~~~~~~
 00007ffb`9af2e865                             align
 00007ffb`9af2e865                             align
 00007ffb`9af2e865                             align
                                                ;; bbWeight=1    PerfScore 4.75
G_M56622_IG06:              ;; offset=0065H
 00007ffb`9af2e865        488B442428           mov      rax, gword ptr [rsp+28H]
 00007ffb`9af2e86a        8B4008               mov      eax, dword ptr [rax+8]
 00007ffb`9af2e86d        39442430             cmp      dword ptr [rsp+30H], eax
 00007ffb`9af2e871        732F                 jae      SHORT G_M56622_IG10
 00007ffb`9af2e873        488B442428           mov      rax, gword ptr [rsp+28H]
 00007ffb`9af2e878        8B542430             mov      edx, dword ptr [rsp+30H]
 00007ffb`9af2e87c        4863D2               movsxd   rdx, edx
 00007ffb`9af2e87f        488B7CD010           mov      rdi, gword ptr [rax+8*rdx+16]
; =========================== 32B boundary ===========================
                                                ;; bbWeight=2    PerfScore 18.50
G_M56622_IG07:              ;; offset=0084H
 00007ffb`9af2e884        8B442430             mov      eax, dword ptr [rsp+30H]
 00007ffb`9af2e888        FFC0                 inc      eax
 00007ffb`9af2e88a        89442430             mov      dword ptr [rsp+30H], eax
 00007ffb`9af2e88e        488B542428           mov      rdx, gword ptr [rsp+28H]
 00007ffb`9af2e893        394208               cmp      dword ptr [rdx+8], eax
 00007ffb`9af2e896        7FCD                 jg       SHORT G_M56622_IG06
                                                ;; bbWeight=8    PerfScore 50.00

CC: @adamsitnik , @AndyAyersMS