JIT: remove bound check for negated ranges

[EgorBo]

Closes #96046

byte Test(int a) => rva[8 - (a & 7)];

static ReadOnlySpan<byte> rva => [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];

Was:

; Method Program:Test(int):ubyte:this (FullOpts)
       sub      rsp, 40
       and      ecx, 7
       mov      eax, ecx
       neg      eax
       add      eax, 8
       cmp      eax, 10
       jae      SHORT G_M949_IG04
       mov      rcx, 0x1735CBE29D0
       movzx    rax, byte  ptr [rax+rcx]
       add      rsp, 40
       ret
G_M949_IG04:
       call     CORINFO_HELP_RNGCHKFAIL
       int3
; Total bytes of code 44

Now:

; Method Program:Test(int):ubyte:this (FullOpts)
       sub      rsp, 40
       and      edx, 7
       mov      eax, edx
       neg      eax
       add      eax, 8
       mov      rcx, 0x228C6572A20
       movzx    rax, byte  ptr [rax+rcx]
       add      rsp, 40
       ret      
; Total bytes of code: 33

[ghost]

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

Issue Details

---

Closes #96046

byte Test(int a) => rva[8 - (a & 7)];

static ReadOnlySpan<byte> rva => [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];

Was:

; Method Program:Test(int):ubyte:this (FullOpts)
       sub      rsp, 40
       and      ecx, 7
       mov      eax, ecx
       neg      eax
       add      eax, 8
       cmp      eax, 10
       jae      SHORT G_M949_IG04
       mov      rcx, 0x1735CBE29D0
       movzx    rax, byte  ptr [rax+rcx]
       add      rsp, 40
       ret
G_M949_IG04:
       call     CORINFO_HELP_RNGCHKFAIL
       int3
; Total bytes of code 44

Now:

; Method Program:Test(int):ubyte:this (FullOpts)
       sub      rsp, 40
       and      edx, 7
       mov      eax, edx
       neg      eax
       add      eax, 8
       mov      rcx, 0x228C6572A20
       movzx    rax, byte  ptr [rax+rcx]
       add      rsp, 40
       ret      
; Total bytes of code: 33

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

[EgorBo]

@dotnet/jit-contrib PTAL, small change, diffs aren't too big, but it closes #96046

[BruceForstall]

Can either the existing upper or lower limit be INT_MIN, in which case -INT_MIN == INT_MIN?

[EgorBo]

I wasn't able to repro such a case, but added a check just in case 👍