JIT: fix another case of early flow graph divergence #85873
Conversation
The JIT will sometimes decide to instrument a Tier0 method even if `BBINSTR` is not passed by the VM (this is enabled when the VM passes `BBINSTR_IF_LOOPS` so that we can provide some PGO data to OSR methods). In such cases we build the flow graph and then decide to instrument, so the flow graph shape may differ from the case where we know up front that we are going to instrument or optimize. Remedy this by also running the early branch to next flow graph opt when a Tier0 JIT is passed `BBINSTR_IF_LOOPS`. Addresses another case of dotnet#85856.
ghost
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch Issue DetailsThe JIT will sometimes decide to instrument a Tier0 method even if In such cases we build the flow graph and then decide to instrument, so the flow graph shape may differ from the case where we know up front that we are going to instrument or optimize. Remedy this by also running the early branch to next flow graph opt when a Tier0 JIT is passed Addresses another case of #85856.
|
|
@EgorBo PTAL Relatively small number of diffs, but this impacts some important methods.... |
We can always change it later.... |
|
Diffs these are all in Tier0 instrumentation strategy / schema. |
ghost
locked as resolved and limited conversation to collaborators
The JIT will sometimes decide to instrument a Tier0 method even if
BBINSTRis not passed by the VM (this is enabled when the VM passesBBINSTR_IF_LOOPSso that we can provide some PGO data to OSR methods).In such cases we build the flow graph and then decide to instrument, so the flow graph shape may differ from the case where we know up front that we are going to instrument or optimize.
Remedy this by also running the early branch to next flow graph opt when a Tier0 JIT is passed
BBINSTR_IF_LOOPS.Addresses another case of #85856.