| Message ID | 87a7zwwwa1.fsf@linaro.org |
|---|---|
| State | Accepted |
| Commit | 3ea518f6f63e66e48f2d41cfa41e1efae653a484 |
| Headers | show |
| Series | Protect against min_profitable_iters going negative | expand |
On 11/08/2017 09:49 AM, Richard Sandiford wrote: > We had: > > if (vec_outside_cost <= 0) > min_profitable_iters = 0; > else > { > min_profitable_iters = ((vec_outside_cost - scalar_outside_cost) > * assumed_vf > - vec_inside_cost * peel_iters_prologue > - vec_inside_cost * peel_iters_epilogue) > / ((scalar_single_iter_cost * assumed_vf) > - vec_inside_cost); > > which can lead to negative min_profitable_iters when the *_outside_costs > are the same and peel_iters_epilogue is nonzero (e.g. if we're peeling > for gaps). > > This is tested as part of the patch that adds support for fully-predicated > loops. > > Tested on aarch64-linux-gnu (both with and without SVE), x86_64-linux-gnu > and powerpc64le-linux-gnu. OK to install? > > Thanks, > Richard > > > 2017-11-08 Richard Sandiford <richard.sandiford@linaro.org> > Alan Hayward <alan.hayward@arm.com> > David Sherwood <david.sherwood@arm.com> > > gcc/ > * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure > min_profitable_iters doesn't go negative. OK. jeff
Index: gcc/tree-vect-loop.c =================================================================== --- gcc/tree-vect-loop.c 2017-11-08 16:35:04.770241799 +0000 +++ gcc/tree-vect-loop.c 2017-11-08 16:47:06.386264822 +0000 @@ -3651,23 +3651,24 @@ vect_estimate_min_profitable_iters (loop if ((scalar_single_iter_cost * assumed_vf) > (int) vec_inside_cost) { - if (vec_outside_cost <= 0) + min_profitable_iters = ((vec_outside_cost - scalar_outside_cost) + * assumed_vf + - vec_inside_cost * peel_iters_prologue + - vec_inside_cost * peel_iters_epilogue); + + if (min_profitable_iters <= 0) min_profitable_iters = 0; else - { - min_profitable_iters = ((vec_outside_cost - scalar_outside_cost) - * assumed_vf - - vec_inside_cost * peel_iters_prologue - - vec_inside_cost * peel_iters_epilogue) - / ((scalar_single_iter_cost * assumed_vf) - - vec_inside_cost); + { + min_profitable_iters /= ((scalar_single_iter_cost * assumed_vf) + - vec_inside_cost); if ((scalar_single_iter_cost * assumed_vf * min_profitable_iters) <= (((int) vec_inside_cost * min_profitable_iters) + (((int) vec_outside_cost - scalar_outside_cost) * assumed_vf))) min_profitable_iters++; - } + } } /* vector version will never be profitable. */ else
We had: if (vec_outside_cost <= 0) min_profitable_iters = 0; else { min_profitable_iters = ((vec_outside_cost - scalar_outside_cost) * assumed_vf - vec_inside_cost * peel_iters_prologue - vec_inside_cost * peel_iters_epilogue) / ((scalar_single_iter_cost * assumed_vf) - vec_inside_cost); which can lead to negative min_profitable_iters when the *_outside_costs are the same and peel_iters_epilogue is nonzero (e.g. if we're peeling for gaps). This is tested as part of the patch that adds support for fully-predicated loops. Tested on aarch64-linux-gnu (both with and without SVE), x86_64-linux-gnu and powerpc64le-linux-gnu. OK to install? Thanks, Richard 2017-11-08 Richard Sandiford <richard.sandiford@linaro.org> Alan Hayward <alan.hayward@arm.com> David Sherwood <david.sherwood@arm.com> gcc/ * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure min_profitable_iters doesn't go negative.