RFR: 8261235: C1 compilation fails with assert(res->vreg_number() == index) failed: conversion check

[Eirik Bjorsnos]

On Fri, 12 Feb 2021 10:03:25 GMT, Christian Hagedorn <chagedorn at openjdk.org> wrote:

> The assertion is hit because we run out of virtual registers in the linear scan in C1 and do not handle it. I fixed it by applying the same bailout as in `LIRGenerator::new_register()`. 
> 
> There is also a second issue that `LIR_OprDesc::vreg_max` is too big. It is only used in this bailout code. `OprBits::vreg_max` is defined over `OprBits::data_bits` which uses `OprBits::non_data_bits`. But `OprBits::non_data_bits` does not consider `OprBits::pointer_bits` which results in a too large value for `LIR_OprDesc::vreg_max` and the assertion is hit because we don't bail out, yet. This needs to be fixed as well.
> 
> Thanks,
> Christian

I tested this branch on my reproducer and it works like a charm:

    776  302       3       org.jaxen.saxpath.base.Verifier::isXMLLetter (6130 bytes)
    [...]
    compilation bailout: out of virtual registers in linear scan
    4717  302       3       org.jaxen.saxpath.base.Verifier::isXMLLetter (6130 bytes)   COMPILE SKIPPED: out of virtual registers in linear scan (retry at different tier)
    4718  334       4       org.jaxen.saxpath.base.Verifier::isXMLLetter (6130 bytes)

Is it expected that the bailout is logged twice?

First, a log line says: "compilation bailout: out of virtual registers in linear scan", then at the end of the next line: "COMPILE SKIPPED: out of virtual registers in linear scan (retry at different tier)" ?

-------------

PR: https://git.openjdk.java.net/jdk/pull/2543