http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57896
--- Comment #30 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Jakub Jelinek from comment #25)
> Created attachment 32180 [details]
> gcc48-pr57896.patch
>
> So like this (4.8 version, untested)? Uros, can you please double-check it?
> The point is, if d->testing_p is true, we should do as few work as possible,
> certainly not call any emit_insn or gen_reg_rtx, and as soon as we know
> we'll certainly eventually return true, return it.
I have:
- eyeballed the code a bit and found no issues (this is sorta unreliable test).
- bootstrapped and regression tested on x86_64-pc-linux-gnu {,-m32} with the
gen_reg_rtx patch from Comment #27 applied.
- additionaly tested with
GCC_TEST_RUN_EXPENSIVE=1 make check-gcc
RUNTESTFLAGS='--target_board=unix\{-msse2,-msse4,-mavx\} dg-torture.exp=vshuf*'
on x86_64 AVX target (also with gen_reg_rtx patch applied) and found no issues.
- I don't have AVX2 target to test on, perhaps the above test should be also
run on AVX2 capable target with -mavx2 option added, preferably with
gen_reg_rtx patch applied.
