Fix missing relocations in NativeAOT x86 output

[filipnavara]

This bug was introduced in #97413.

Output without the fix:

IN0011: 000047 mov      edx, dword ptr [ebp+0x10]
IN0012: 00004A mov      ecx, (reloc 0x40420290)
recordRelocation: 10A4E138 (rw: 10A4E138) => 40420310, type 16 (IMAGE_REL_BASED_REL32), delta 0
IN0013: 00004F call     CORINFO_HELP_NEWARR_1_VC

Output with the fix:

IN0011: 000047 mov      edx, dword ptr [ebp+0x10]
recordRelocation: 12241DAB (rw: 12241DAB) => 40420290, type 3 (IMAGE_REL_BASED_MOFFSET), delta 0
IN0012: 00004A mov      ecx, (reloc 0x40420290)
recordRelocation: 12241DB0 (rw: 12241DB0) => 40420310, type 16 (IMAGE_REL_BASED_REL32), delta 0
IN0013: 00004F call     CORINFO_HELP_NEWARR_1_VC

[ghost]

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

Issue Details

---

This bug was introduced in #97413.

Output without the fix:

IN0011: 000047 mov      edx, dword ptr [ebp+0x10]
IN0012: 00004A mov      ecx, (reloc 0x40420290)
recordRelocation: 10A4E138 (rw: 10A4E138) => 40420310, type 16 (IMAGE_REL_BASED_REL32), delta 0
IN0013: 00004F call     CORINFO_HELP_NEWARR_1_VC

Output with the fix:

IN0011: 000047 mov      edx, dword ptr [ebp+0x10]
recordRelocation: 12241DAB (rw: 12241DAB) => 40420290, type 3 (IMAGE_REL_BASED_MOFFSET), delta 0
IN0012: 00004A mov      ecx, (reloc 0x40420290)
recordRelocation: 12241DB0 (rw: 12241DB0) => 40420310, type 16 (IMAGE_REL_BASED_REL32), delta 0
IN0013: 00004F call     CORINFO_HELP_NEWARR_1_VC

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

[ryujit-bot]

Diff results for #98740

Throughput diffs

Throughput diffs for linux/arm64 ran on windows/x64

MinOpts (-0.01% to +0.00%)

Collection	PDIFF
libraries.pmi.linux.arm64.checked.mch	-0.01%

Throughput diffs for windows/arm64 ran on windows/x64

MinOpts (-0.00% to +0.01%)

Collection	PDIFF
libraries.pmi.windows.arm64.checked.mch	+0.01%

Details here

---

[BruceForstall]

cc @dotnet/jit-contrib

[MichalStrehovsky]

@kunalspathak could you have a look? Looks like this was broken by your recent TLS change.

[kunalspathak]

/azp run runtime-nativeaot-outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[kunalspathak]

LGTM, I don't recall why I to do this change during linux/x64 support, probably because there was a case where TLS related constant that was wrongly getting IMAGE_REL_BASED_MOFFSET reloc and should have not get it. I am running outer loop test to make sure.

[kunalspathak]

@filipnavara - where was this issue hitting you? I am surprised why our outerloop didn't catch from past few weeks.

[filipnavara]

win-x86 target which I am bringing up. It’s not part of the tests yet.