Fix zero initialization use of "initReg"

[BruceForstall]

If the only local is an enregistered must-init struct,
we were setting initReg to its register (in this case, xmm0).
However, initReg is expected to be an integer register. In the
test case, with GS cookie stress, the GS cookie code asserted
that initReg was an integer register, but it wasn't.

The fix is the change the condition to use the actual register
assigned to the variable (in this case, xmm0), not the variable
type (here, TYP_STRUCT).

No spmi asm diffs.

Fixes #57911

[ghost]

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

Issue Details

---

If the only local is an enregistered must-init struct,
we were setting initReg to its register (in this case, xmm0).
However, initReg is expected to be an integer register. In the
test case, with GS cookie stress, the GS cookie code asserted
that initReg was an integer register, but it wasn't.

The fix is the change the condition to use the actual register
assigned to the variable (in this case, xmm0), not the variable
type (here, TYP_STRUCT).

No spmi asm diffs.

Fixes #57911

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

[BruceForstall]

/azp run runtime-coreclr jitstress

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[BruceForstall]

@sandreenko This looks like fallout from enregistered struct locals. Is the condition appropriate, or is there some other standard condition that should be used? PTAL.

[sandreenko]

LGTM, I think the fix is correct.
Looks like LclVarDsc::IsFloatRegType is not prepared for enreg structs, we can change the function or we can delete its usage, there are only 3 callers, 1 you have fixed here, the other 2 are:

runtime/src/coreclr/jit/codegenlinear.cpp

Lines 54 to 55 in de6b1e9

	// Is this a floating-point argument?
	if (varDsc->IsFloatRegType())

runtime/src/coreclr/jit/compiler.cpp

Line 2173 in de6b1e9

(varDsc->IsFloatRegType() || !isFloatReg) && (varDsc->lvSlotNum < info.compVarScopesCount))

could you please look at them as well?

[BruceForstall]

How about usage of varTypeUsesFloatReg when applied to LclVarDsc::TypeGet()? Is that also problematic? If we have LclVarDsc vars with TYP_STRUCT, we can't tell whether they use integer registers or floating-point registers just by looking at their type?

[sandreenko]

How about usage of varTypeUsesFloatReg when applied to LclVarDsc::TypeGet()? Is that also problematic? If we have LclVarDsc vars with TYP_STRUCT, we can't tell whether they use integer registers or floating-point registers just by looking at their type?

You need to use LclVarDsc->GetRegisterType() instead if TypeGet().

[BruceForstall]

/azp run runtime-coreclr jitstress

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[BruceForstall]

@sandreenko Can you look at the additional changes in 16a9f78?

[BruceForstall]

cc @dotnet/jit-contrib