| Message ID | 20220323145726.98F2F13302@imap2.suse-dmz.suse.de |
|---|---|
| State | New |
| Headers | show |
| Series | target/102125 - alternative memcpy folding improvement | expand |
On 23/03/2022 14:57, Richard Biener via Gcc-patches wrote: > The following extends the heuristical memcpy folding path with the > ability to use misaligned accesses on strict-alignment targets just > like the size-based path does. That avoids regressing the following > testcase on arm > > uint64_t bar64(const uint8_t *rData1) > { > uint64_t buffer; > memcpy(&buffer, rData1, sizeof(buffer)); > return buffer; > } > > when r12-3482-g5f6a6c91d7c592 is reverted. > > Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed to trunk. > > OK to revert r12-3482-g5f6a6c91d7c592? Yes, go ahead. R. > > Thanks, > Richard. > > 2022-03-23 Richard Biener <rguenther@suse.de> > > PR target/102125 > * gimple-fold.cc (gimple_fold_builtin_memory_op): Allow the > use of movmisalign when either the source or destination > decl is properly aligned. > --- > gcc/gimple-fold.cc | 12 ++++++++++-- > 1 file changed, 10 insertions(+), 2 deletions(-) > > diff --git a/gcc/gimple-fold.cc b/gcc/gimple-fold.cc > index c9179abb27e..5eff7d68ac1 100644 > --- a/gcc/gimple-fold.cc > +++ b/gcc/gimple-fold.cc > @@ -1254,7 +1254,11 @@ gimple_fold_builtin_memory_op (gimple_stmt_iterator *gsi, > srcvar = fold_build2 (MEM_REF, desttype, src, off0); > else > { > - if (STRICT_ALIGNMENT) > + enum machine_mode mode = TYPE_MODE (desttype); > + if ((mode == BLKmode && STRICT_ALIGNMENT) > + || (targetm.slow_unaligned_access (mode, src_align) > + && (optab_handler (movmisalign_optab, mode) > + == CODE_FOR_nothing))) > return false; > srctype = build_aligned_type (TYPE_MAIN_VARIANT (desttype), > src_align); > @@ -1267,7 +1271,11 @@ gimple_fold_builtin_memory_op (gimple_stmt_iterator *gsi, > destvar = fold_build2 (MEM_REF, srctype, dest, off0); > else > { > - if (STRICT_ALIGNMENT) > + enum machine_mode mode = TYPE_MODE (srctype); > + if ((mode == BLKmode && STRICT_ALIGNMENT) > + || (targetm.slow_unaligned_access (mode, dest_align) > + && (optab_handler (movmisalign_optab, mode) > + == CODE_FOR_nothing))) > return false; > desttype = build_aligned_type (TYPE_MAIN_VARIANT (srctype), > dest_align);
diff --git a/gcc/gimple-fold.cc b/gcc/gimple-fold.cc index c9179abb27e..5eff7d68ac1 100644 --- a/gcc/gimple-fold.cc +++ b/gcc/gimple-fold.cc @@ -1254,7 +1254,11 @@ gimple_fold_builtin_memory_op (gimple_stmt_iterator *gsi, srcvar = fold_build2 (MEM_REF, desttype, src, off0); else { - if (STRICT_ALIGNMENT) + enum machine_mode mode = TYPE_MODE (desttype); + if ((mode == BLKmode && STRICT_ALIGNMENT) + || (targetm.slow_unaligned_access (mode, src_align) + && (optab_handler (movmisalign_optab, mode) + == CODE_FOR_nothing))) return false; srctype = build_aligned_type (TYPE_MAIN_VARIANT (desttype), src_align); @@ -1267,7 +1271,11 @@ gimple_fold_builtin_memory_op (gimple_stmt_iterator *gsi, destvar = fold_build2 (MEM_REF, srctype, dest, off0); else { - if (STRICT_ALIGNMENT) + enum machine_mode mode = TYPE_MODE (srctype); + if ((mode == BLKmode && STRICT_ALIGNMENT) + || (targetm.slow_unaligned_access (mode, dest_align) + && (optab_handler (movmisalign_optab, mode) + == CODE_FOR_nothing))) return false; desttype = build_aligned_type (TYPE_MAIN_VARIANT (srctype), dest_align);
The following extends the heuristical memcpy folding path with the ability to use misaligned accesses on strict-alignment targets just like the size-based path does. That avoids regressing the following testcase on arm uint64_t bar64(const uint8_t *rData1) { uint64_t buffer; memcpy(&buffer, rData1, sizeof(buffer)); return buffer; } when r12-3482-g5f6a6c91d7c592 is reverted. Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed to trunk. OK to revert r12-3482-g5f6a6c91d7c592? Thanks, Richard. 2022-03-23 Richard Biener <rguenther@suse.de> PR target/102125 * gimple-fold.cc (gimple_fold_builtin_memory_op): Allow the use of movmisalign when either the source or destination decl is properly aligned. --- gcc/gimple-fold.cc | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)