From: Richard Biener Date: Wed, 25 Sep 2024 11:15:42 +0000 (+0200) Subject: Speed up wide_int_storage::operator=(wide_int_storage const&) X-Git-Tag: basepoints/gcc-16~5689 X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=6efc770a71b7227cdbdc24c947ce1fef10794f4c;p=thirdparty%2Fgcc.git Speed up wide_int_storage::operator=(wide_int_storage const&) wide_int_storage shows up high in the profile for the testcase in PR114855 where the apparent issue is that the conditional jump on 'precision' after the (inlined) memcpy stalls the pipeline due to the data dependence and required store-to-load forwarding. We can add scheduling freedom by instead testing precision as from the source which speeds up the function by 30%. I've applied the same logic to the copy CTOR. * wide-int.h (wide_int_storage::wide_int_storage): Branch on source precision to avoid data dependence on memcpy destination. (wide_int_storage::operator=): Likewise. --- diff --git a/gcc/wide-int.h b/gcc/wide-int.h index 64b8bf2040c..777f017f5ae 100644 --- a/gcc/wide-int.h +++ b/gcc/wide-int.h @@ -1196,7 +1196,7 @@ inline wide_int_storage::wide_int_storage (const T &x) inline wide_int_storage::wide_int_storage (const wide_int_storage &x) { memcpy (this, &x, sizeof (wide_int_storage)); - if (UNLIKELY (precision > WIDE_INT_MAX_INL_PRECISION)) + if (UNLIKELY (x.precision > WIDE_INT_MAX_INL_PRECISION)) { u.valp = XNEWVEC (HOST_WIDE_INT, CEIL (precision, HOST_BITS_PER_WIDE_INT)); memcpy (u.valp, x.u.valp, len * sizeof (HOST_WIDE_INT)); @@ -1219,9 +1219,9 @@ wide_int_storage::operator = (const wide_int_storage &x) XDELETEVEC (u.valp); } memcpy (this, &x, sizeof (wide_int_storage)); - if (UNLIKELY (precision > WIDE_INT_MAX_INL_PRECISION)) + if (UNLIKELY (x.precision > WIDE_INT_MAX_INL_PRECISION)) { - u.valp = XNEWVEC (HOST_WIDE_INT, CEIL (precision, HOST_BITS_PER_WIDE_INT)); + u.valp = XNEWVEC (HOST_WIDE_INT, CEIL (x.precision, HOST_BITS_PER_WIDE_INT)); memcpy (u.valp, x.u.valp, len * sizeof (HOST_WIDE_INT)); } return *this;