mbox series

[0/2] Fix some Neon insns on big-endian hosts

Message ID 20201028191712.4910-1-peter.maydell@linaro.org (mailing list archive)
Headers show
Series Fix some Neon insns on big-endian hosts | expand

Message

Peter Maydell Oct. 28, 2020, 7:17 p.m. UTC 
This patchseries fixes a couple of places where the vector
helpers we use for Neon insns were missing uses of the
address-swizzling macros H1(), H2() or H4(). This is harmless
on little-endian hosts but causes the wrong results to be
generated on big-endian hosts. The affected insns are
VUDOT(scalar), VSDOT(scalar), VPADD, VPMAX and VPMIN.

This series is independent of Richard's recent "target/arm:
Fix neon reg offsets"; it fixes the handful of remaining
risu test failures I see.

thanks
-- PMM

Peter Maydell (2):
  target/arm: Fix float16 pairwise Neon ops on big-endian hosts
  target/arm: Fix VUDOT/VSDOT (scalar) on big-endian hosts

 target/arm/vec_helper.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

Comments

Richard Henderson Oct. 28, 2020, 7:34 p.m. UTC | #1 
On 10/28/20 12:17 PM, Peter Maydell wrote:
> Peter Maydell (2):
>   target/arm: Fix float16 pairwise Neon ops on big-endian hosts
>   target/arm: Fix VUDOT/VSDOT (scalar) on big-endian hosts

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~