This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[trans-mem] PR 68964 fallout -- bootstrap 69343+69339


These two bootstrap PRs, caused by old binutils on s390x and embedded powerpc
that doesn't support altivec, could have been worked around with yet more
configure checks.

But after discussion on IRC, we decided that to revert the arm+ppc+s390 changes
entirely and keep the libitm ABI the same.  Instead of calling new functions
with vector parameters (which internally wound up calling memcpy), we will
instead let the compiler call memcpy directly.  This solves all of the
platform-specific configury wrt when vector instructions are available and
avoids a bit of indirection in the end.

The portion of the patch for 68964 that *isn't* reverted by this does solve the
original ICE by deferring otherwise unhandled types to memcpy.

(Re-)tested on x86_64, s390x, and aarch64, and committed.


r~

Attachment: z
Description: Text document


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]