Write barrier elision on object reference assignment to a return buffer?

[neon-sunset]

Description

Consider the method static (object, object) GetObj() => (new(), new());.

On win-x64 JIT emits the following codegen for it:

G_M000_IG01:                ;; offset=0x0000
       push     rdi
       push     rsi
       push     rbx
       sub      rsp, 32
       mov      rbx, rcx

G_M000_IG02:                ;; offset=0x000A
       mov      rsi, 0x7FFCD3835FA0
       mov      rcx, rsi
       call     CORINFO_HELP_NEWSFAST
       mov      rdi, rax
       mov      rcx, rsi
       call     CORINFO_HELP_NEWSFAST
       mov      rsi, rax
       mov      rdx, rdi
       mov      rcx, rbx
       call     CORINFO_HELP_CHECKED_ASSIGN_REF
       lea      rcx, bword ptr [rbx+0x08]
       mov      rdx, rsi
       call     CORINFO_HELP_CHECKED_ASSIGN_REF
       mov      rax, rbx

G_M000_IG03:                ;; offset=0x0044
       add      rsp, 32
       pop      rbx
       pop      rsi
       pop      rdi
       ret      
; Total bytes of code: 76

Are write barriers necessary given the fact that it is known that object references are being assigned to a return buffer? Is there something I am missing?

Configuration

.NET SDK:
 Version:   8.0.100-rtm.23471.13
 Commit:    1de2c23f77

Runtime Environment:
 OS Name:     Windows
 OS Version:  10.0.22621
 OS Platform: Windows
 RID:         win-x64
 Base Path:   C:\Program Files\dotnet\sdk\8.0.100-rtm.23471.13\

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

Issue Details

---

Description

Consider the method static (object, object) GetObj() => (new(), new());.

On win-x64 JIT emits the following codegen for it:

G_M000_IG01:                ;; offset=0x0000
       push     rdi
       push     rsi
       push     rbx
       sub      rsp, 32
       mov      rbx, rcx

G_M000_IG02:                ;; offset=0x000A
       mov      rsi, 0x7FFCD3835FA0
       mov      rcx, rsi
       call     CORINFO_HELP_NEWSFAST
       mov      rdi, rax
       mov      rcx, rsi
       call     CORINFO_HELP_NEWSFAST
       mov      rsi, rax
       mov      rdx, rdi
       mov      rcx, rbx
       call     CORINFO_HELP_CHECKED_ASSIGN_REF
       lea      rcx, bword ptr [rbx+0x08]
       mov      rdx, rsi
       call     CORINFO_HELP_CHECKED_ASSIGN_REF
       mov      rax, rbx

G_M000_IG03:                ;; offset=0x0044
       add      rsp, 32
       pop      rbx
       pop      rsi
       pop      rdi
       ret      
; Total bytes of code: 76

Are write barriers necessary given the fact that it is known that object references are being assigned to a return buffer which should not require write barriers? Is there something I am missing?

Configuration

.NET SDK:
 Version:   8.0.100-rtm.23471.13
 Commit:    1de2c23f77

Runtime Environment:
 OS Name:     Windows
 OS Version:  10.0.22621
 OS Platform: Windows
 RID:         win-x64
 Base Path:   C:\Program Files\dotnet\sdk\8.0.100-rtm.23471.13\

Author:	neon-sunset
Assignees:	-
Labels:	tenet-performance, area-CodeGen-coreclr, untriaged
Milestone:	-

[jakobbotsch]

They are necessary because the JIT can optimize passed return buffers to point to heap. For example, check the codegen for something like c.obj = GetObj() where c is a class.

[neon-sunset]

Today I learned, thanks!