> Thansk for review.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
> Ok for trunk?
>
> n some benchmark, I notice stv failed due to cost unprofitable, but the igain
> is inside the loop, but sse<->integer conversion is outside the loop, current
> cost
> model doesn't consider the
Update in V3
> > > + cost_sse_integer = 0;
> > > + weighted_cost_sse_integer = 0 ;
> Extra space here.
Changed.
> > > + : ix86_size_cost.sse_to_integer;
>
> Please be sure to not revert the changes from my patch adding
> COSTS_N_INSNS (...) / 2
> here and some other places.
Yes, keep the
> > gcc/ChangeLog:
> >
> > * config/i386/i386-features.cc
> > (scalar_chain::mark_dual_mode_def): Weight
> > n_integer_to_sse/n_sse_to_integer with bb frequency.
> > (general_scalar_chain::compute_convert_gain): Ditto, and
> > adjust function prototype to ret
On Thu, May 8, 2025 at 2:40 PM liuhongt wrote:
>
> The only part I changed is related to size_cost of sse_to_ineteger, as below
>
> 114+ /* Under TARGET_SSE4_1, it's vmovd + vpextrd/vpinsrd.
> 115+ W/o it, it's movd + psrlq/unpckldq + movd. */
> 116+ else if (!TARGET_64BIT && smode != SImod
The only part I changed is related to size_cost of sse_to_ineteger, as below
114+ /* Under TARGET_SSE4_1, it's vmovd + vpextrd/vpinsrd.
115+ W/o it, it's movd + psrlq/unpckldq + movd. */
116+ else if (!TARGET_64BIT && smode != SImode)
117+cost *= TARGET_SSE4_1 ? 2 : 3;
118+
Ok for trun
