On Sat, 2017-04-22 at 16:28 -0700, Kenneth Graunke wrote: > Curro pointed out that I should not just check for MACH, but use > the reads_accumulator_implicitly() helper, which would also prevent > the same bug with MAC and SADA2 (if we ever decide to use them). >
This seems to rely on commit 2faf227ec2e (i965/vec4: Avoid reswizzling MACH instructions in opt_register_coalesce()), which is not available in stable. But taking a look at it, I think it could be good candidate to be also in stable. WDYT? Can we propose 2faf227ec2e to stable? J.A. > Cc: Francisco Jerez <curroje...@riseup.net> > Cc: mesa-sta...@lists.freedesktop.org > --- > src/intel/compiler/brw_vec4.cpp | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > Curro - sorry, I already pushed the MACH patch...but I agree with your > feedback, so here it is as a follow-up patch :) > > Emil - you'll need to cherry-pick 9347acac440190af67fce7b5f28e8eee7245fb6d > before picking this patch. I forgot to nominate it for stable, sorry! > > diff --git a/src/intel/compiler/brw_vec4.cpp b/src/intel/compiler/brw_vec4.cpp > index 4bb774bf10e..0909ddb5861 100644 > --- a/src/intel/compiler/brw_vec4.cpp > +++ b/src/intel/compiler/brw_vec4.cpp > @@ -1071,11 +1071,11 @@ vec4_instruction::can_reswizzle(const struct > gen_device_info *devinfo, > if (devinfo->gen == 6 && is_math() && swizzle != BRW_SWIZZLE_XYZW) > return false; > > - /* Don't touch MACH - it uses the accumulator results from an earlier > - * MUL - so we'd need to reswizzle both. We don't do that, so just > - * avoid it entirely. > + /* We can't swizzle implicit accumulator access. We'd have to > + * reswizzle the producer of the accumulator value in addition > + * to the consumer (i.e. both MUL and MACH). Just skip this. > */ > - if (opcode == BRW_OPCODE_MACH) > + if (reads_accumulator_implicitly()) > return false; > > if (!can_do_writemask(devinfo) && dst_writemask != WRITEMASK_XYZW) _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev