JIT: Fix delegate GDV for loops with multiple cloning options

[jakobbotsch]

For loops that we consider candidates for cloning with multiple cloning options there was a logic bug that caused cloning to exit early. This affected even canonical loops over arrays if they also had a delegate GDV cloning option.

Fix #109379

New codegen:

; Assembly listing for method C:AggregateDelegate[int](int[],System.Func`3[int,int,int],int):int (Tier1)
; Emitting BLENDED_CODE for X64 with AVX - Windows
; Tier1 code
; optimized code
; optimized using Dynamic PGO
; rsp based frame
; fully interruptible
; with Dynamic PGO: fgCalledCount is 100
; 0 inlinees with PGO data; 1 single block inlinees; 0 inlinees without PGO data
; Final local variable assignments
;
;  V00 arg0         [V00,T06] (  5,  4.00)     ref  ->  rcx         class-hnd single-def <int[]>
;  V01 arg1         [V01,T05] (  7,  5.66)     ref  ->  rbx         class-hnd single-def <System.Func`3[int,int,int]>
;  V02 arg2         [V02,T03] (  8,337.13)     int  ->   r8
;* V03 loc0         [V03,T10] (  0,  0   )     int  ->  zero-ref
;  V04 OutArgs      [V04    ] (  1,  1   )  struct (32) [rsp+0x00]  do-not-enreg[XS] addr-exposed "OutgoingArgSpace"
;  V05 tmp1         [V05,T04] (  5,334.13)     int  ->   r8         "guarded devirt return temp"
;  V06 tmp2         [V06,T00] (  5,668.27)     int  ->  rax         "guarded devirt arg temp"
;* V07 tmp3         [V07    ] (  0,  0   )     ref  ->  zero-ref    class-hnd "guarded devirt this exact temp" <C+<>c>
;  V08 cse0         [V08,T09] (  4,  3.00)     int  ->  rdx         "CSE #01: aggressive"
;  V09 rat0         [V09,T07] (  4,  5.02)   byref  ->  rsi         "Strength reduced derived IV"
;  V10 rat1         [V10,T08] (  4,  5.02)     int  ->  rdi         "Trip count IV"
;  V11 rat2         [V11,T01] (  4,497.18)   byref  ->  rcx         "Strength reduced derived IV"
;  V12 rat3         [V12,T02] (  4,497.18)     int  ->  rdx         "Trip count IV"
;
; Lcl frame size = 32

G_M53109_IG01:  ;; offset=0x0000
       push     rdi
       push     rsi
       push     rbx
       sub      rsp, 32
       mov      rbx, rdx
                                                ;; size=10 bbWeight=1 PerfScore 3.50
G_M53109_IG02:  ;; offset=0x000A
       mov      edx, dword ptr [rcx+0x08]
       test     edx, edx
       jle      SHORT G_M53109_IG05
                                                ;; size=7 bbWeight=1 PerfScore 3.25
G_M53109_IG03:  ;; offset=0x0011
       test     rbx, rbx
       je       SHORT G_M53109_IG07
       mov      rax, 0x7FFA12D60828      ; code for C+<>c:<Main>b__0_0(int,int):int:this
       cmp      qword ptr [rbx+0x18], rax
       jne      SHORT G_M53109_IG07
       add      rcx, 16
       align    [0 bytes for IG04]
                                                ;; size=25 bbWeight=0.99 PerfScore 5.72
G_M53109_IG04:  ;; offset=0x002A
       mov      eax, dword ptr [rcx]
       add      r8d, eax
       add      rcx, 4
       dec      edx
       jne      SHORT G_M53109_IG04
                                                ;; size=13 bbWeight=165.40 PerfScore 620.24
G_M53109_IG05:  ;; offset=0x0037
       mov      eax, r8d
                                                ;; size=3 bbWeight=1.00 PerfScore 0.25
G_M53109_IG06:  ;; offset=0x003A
       add      rsp, 32
       pop      rbx
       pop      rsi
       pop      rdi
       ret
                                                ;; size=8 bbWeight=1.00 PerfScore 2.75
G_M53109_IG07:  ;; offset=0x0042
       lea      rsi, bword ptr [rcx+0x10]
       mov      edi, edx
                                                ;; size=6 bbWeight=0.01 PerfScore 0.01
G_M53109_IG08:  ;; offset=0x0048
       mov      eax, dword ptr [rsi]
       mov      rdx, 0x7FFA12D60828      ; function address
       cmp      qword ptr [rbx+0x18], rdx
       jne      SHORT G_M53109_IG10
       add      r8d, eax
                                                ;; size=21 bbWeight=1.67 PerfScore 10.86
G_M53109_IG09:  ;; offset=0x005D
       add      rsi, 4
       dec      edi
       jne      SHORT G_M53109_IG08
       jmp      SHORT G_M53109_IG05
                                                ;; size=10 bbWeight=1.67 PerfScore 5.85
G_M53109_IG10:  ;; offset=0x0067
       mov      edx, r8d
       mov      r8d, eax
       mov      rcx, gword ptr [rbx+0x08]
       call     [rbx+0x18]System.Func`3[int,int,int]:Invoke(int,int):int:this
       mov      r8d, eax
       jmp      SHORT G_M53109_IG09
                                                ;; size=18 bbWeight=0 PerfScore 0.00

; Total bytes of code 121, prolog size 10, PerfScore 652.42, instruction count 43, allocated bytes for code 121 (MethodHash=0e81308a) for method C:AggregateDelegate[int](int[],System.Func`3[int,int,int],int):int (Tier1)
; ============================================================

[hez2010]

Can we make it into .NET 9 servicing as well? It seems to be a bug instead of an optimization.

[jakobbotsch]

Can we make it into .NET 9 servicing as well? It seems to be a bug instead of an optimization.

This is unlikely to meet the bar for servicing since it is just a missing optimization and no bad codegen.

[jakobbotsch]

cc @dotnet/jit-contrib PTAL @AndyAyersMS

Diffs. For win-x64: 34 affected contexts in benchmarks.run_pgo, and 182 (diffs) + 90 (misses) affected contexts in libraries_tests.run. Sadly we have no aspnet collection at the moment, so can't see how this affects it.

[AndyAyersMS]

Sadly we have no aspnet collection at the moment, so can't see how this affects it

I'll have one up later today...

[jakobbotsch]

/azp run runtime-coreclr superpmi-diffs

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jakobbotsch]

26 hits in the asp.net collection it appears.