Re: Ping: Extra subreg fold for variable-length CONST_VECTORs

2018-01-13 Thread Jeff Law
On 01/12/2018 03:00 AM, Richard Sandiford wrote: > Ping. > > FWIW, the SLP test failures it fixes were ICEs rather than code-quality > tests, so this is a correctness fix rather than an optimisation. > > Thanks, > Richard > > Richard Sandiford writes: >> The SVE support for the new CONST_VECTOR

Ping: Extra subreg fold for variable-length CONST_VECTORs

2018-01-12 Thread Richard Sandiford
Ping. FWIW, the SLP test failures it fixes were ICEs rather than code-quality tests, so this is a correctness fix rather than an optimisation. Thanks, Richard Richard Sandiford writes: > The SVE support for the new CONST_VECTOR encoding needs to be able > to extract the first N bits of the vect

Extra subreg fold for variable-length CONST_VECTORs

2018-01-04 Thread Richard Sandiford
The SVE support for the new CONST_VECTOR encoding needs to be able to extract the first N bits of the vector and duplicate it. This patch adds a simplify_subreg rule for that. The code is covered by the gcc.target/aarch64/sve_slp_*.c tests. Tested on aarch64-linux-gnu, x86_64-linux-gnu and power