| Message ID | 8736vybfzp.fsf@arm.com |
|---|---|
| State | New |
| Headers | show |
| Series | Fix remove_stmt in vectorizable_simd_clone_call (PR 86758) | expand |
On Wed, Aug 1, 2018 at 11:20 AM Richard Sandiford <richard.sandiford@arm.com> wrote: > > vectorizable_simd_clone_call was trying to remove a pattern statement > instead of the original statement, Fixes existing tests > gcc.dg/pr84452.c and gcc.target/i386/pr84309.c on x86. > > This relies on a function added by: > https://gcc.gnu.org/ml/gcc-patches/2018-07/msg01825.html > which can be applied independently of the rest of the series. > > Sorry for the breakage. It looks like I compared the x86 results > with the wrong baseline. :-( > > Tested on aarch64-linux-gnu (with and without SVE), aarch64_be-elf > and x86_64-linux-gnu. OK to install? OK. Richard. > Richard > > > 2018-08-01 Richard Sandiford <richard.sandiford@arm.com> > > gcc/ > PR tree-optimization/86748 > * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try > to remove pattern statements. > > Index: gcc/tree-vect-stmts.c > =================================================================== > --- gcc/tree-vect-stmts.c 2018-08-01 10:12:54.801835439 +0100 > +++ gcc/tree-vect-stmts.c 2018-08-01 10:13:17.629637260 +0100 > @@ -4368,7 +4368,7 @@ vectorizable_simd_clone_call (stmt_vec_i > } > else > new_stmt = gimple_build_nop (); > - vinfo->replace_stmt (gsi, stmt_info, new_stmt); > + vinfo->replace_stmt (gsi, vect_orig_stmt (stmt_info), new_stmt); > unlink_stmt_vdef (stmt); > > return true;
Index: gcc/tree-vect-stmts.c =================================================================== --- gcc/tree-vect-stmts.c 2018-08-01 10:12:54.801835439 +0100 +++ gcc/tree-vect-stmts.c 2018-08-01 10:13:17.629637260 +0100 @@ -4368,7 +4368,7 @@ vectorizable_simd_clone_call (stmt_vec_i } else new_stmt = gimple_build_nop (); - vinfo->replace_stmt (gsi, stmt_info, new_stmt); + vinfo->replace_stmt (gsi, vect_orig_stmt (stmt_info), new_stmt); unlink_stmt_vdef (stmt); return true;