JIT: Allow more permissive forward subbing into call arguments

[jakobbotsch]

After #70893 and #70931 we should be able to allow this.

Diffs

[ghost]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

Issue Details

---

After #70893 and #70931 we should be able to allow this.

Author:	jakobbotsch
Assignees:	-
Labels:	area-CodeGen-coreclr
Milestone:	-

[jakobbotsch]

/azp run Fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[jakobbotsch]

The failures are #71193 and #71200.

cc @dotnet/jit-contrib PTAL @AndyAyersMS

[AndyAyersMS]

Have you looked into some of these large percentage regressions (x64 win)?

Top method regressions (percentages):
           8 (61.54 % of base) : 120885.dasm - System.Drawing.CharacterRange:GetHashCode():int:this
           8 (61.54 % of base) : 198086.dasm - System.Drawing.Point:GetHashCode():int:this
           8 (61.54 % of base) : 198218.dasm - System.Drawing.Size:GetHashCode():int:this

[jakobbotsch]

Have you looked into some of these large percentage regressions (x64 win)?

Spot checking the regressions they all look like the following:

@@ -7,9 +7,9 @@
 ; 0 inlinees with PGO data; 2 single block inlinees; 0 inlinees without PGO data
 ; Final local variable assignments
 ;
-;  V00 this         [V00,T00] (  4,  4   )   byref  ->  rcx         this single-def
+;  V00 this         [V00,T00] (  4,  4   )   byref  ->  [rsp+08H]   this single-def
 ;# V01 OutArgs      [V01    ] (  1,  1   )  lclBlk ( 0) [rsp+00H]   "OutgoingArgSpace"
-;  V02 tmp1         [V02,T01] (  2,  4   )     int  ->  rax         "impAppendStmt"
+;* V02 tmp1         [V02    ] (  0,  0   )     int  ->  zero-ref    "impAppendStmt"
 ;
 ; Lcl frame size = 0
 
@@ -17,17 +17,21 @@ G_M4318_IG01:        ; gcrefRegs=00000000 {}, byrefRegs=00000000 {}, byref, nogc
 						;; size=0 bbWeight=1    PerfScore 0.00
 G_M4318_IG02:        ; gcrefRegs=00000000 {}, byrefRegs=00000002 {rcx}, byref
        ; byrRegs +[rcx]
-       mov      eax, dword ptr [rcx]
-       mov      edx, dword ptr [rcx+4]
-       mov      ecx, eax
+       mov      bword ptr [rsp+08H], rcx
+       ; GC ptr vars +{V00}
+       mov      ecx, dword ptr [rcx]
        ; byrRegs -[rcx]
-						;; size=7 bbWeight=1    PerfScore 4.25
+       mov      rdx, bword ptr [rsp+08H]
+       ; byrRegs +[rdx]
+       mov      edx, dword ptr [rdx+4]
+       ; byrRegs -[rdx]
+						;; size=15 bbWeight=1    PerfScore 6.00
 G_M4318_IG03:        ; , epilog, nogc, extend
        tail.jmp [System.HashCode:Combine(int,int):int]
        ; gcr arg pop 0
 						;; size=6 bbWeight=1    PerfScore 2.00
 
-; Total bytes of code 13, prolog size 0, PerfScore 7.55, instruction count 4, allocated bytes for code 13 (MethodHash=68b0ef21) for method System.Drawing.CharacterRange:GetHashCode():int:this
+; Total bytes of code 21, prolog size 0, PerfScore 10.10, instruction count 5, allocated bytes for code 21 (MethodHash=68b0ef21) for method System.Drawing.CharacterRange:GetHashCode():int:this
 ; ============================================================
 
 Unwind Info:

This is one of the cases I think we have talked about before where we could make the sort better by taking into account that placing an argument may clobber a register used by a future argument. I hope to look into that in the future.

[jakobbotsch]

Is there anything else you think I should do here @AndyAyersMS?

[AndyAyersMS]

Is there anything else you think I should do here @AndyAyersMS?

No, this looks good.