This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR82255 (vectorizer cost model overcounts some vector load costs)
On Sep 19, 2017, at 3:58 PM, Richard Sandiford <richard.sandiford@linaro.org> wrote:
>
> Bill Schmidt <wschmidt@linux.vnet.ibm.com> writes:
>> Index: gcc/tree-vect-stmts.c
>> ===================================================================
>> --- gcc/tree-vect-stmts.c (revision 252760)
>> +++ gcc/tree-vect-stmts.c (working copy)
>> @@ -1091,8 +1091,19 @@ vect_model_load_cost (stmt_vec_info stmt_info, int
>> prologue_cost_vec, body_cost_vec, true);
>> if (memory_access_type == VMAT_ELEMENTWISE
>> || memory_access_type == VMAT_STRIDED_SLP)
>> - inside_cost += record_stmt_cost (body_cost_vec, ncopies, vec_construct,
>> - stmt_info, 0, vect_body);
>> + {
>> + int group_size = GROUP_SIZE (stmt_info);
>> + int nunits = TYPE_VECTOR_SUBPARTS (STMT_VINFO_VECTYPE (stmt_info));
>> + if (group_size < nunits)
>> + {
>> + if (dump_enabled_p ())
>> + dump_printf_loc (MSG_NOTE, vect_location,
>> + "vect_model_load_cost: vec_construct required");
>> + inside_cost += record_stmt_cost (body_cost_vec, ncopies,
>> + vec_construct, stmt_info, 0,
>> + vect_body);
>> + }
>> + }
>>
>> if (dump_enabled_p ())
>> dump_printf_loc (MSG_NOTE, vect_location,
>
> This feels like we've probably got the wrong memory_access_type.
> If it's a just a contiguous load then it should be VMAT_CONTIGUOUS
> instead.
I tend to agree -- that will take more surgery to the code that detects strided loads in
both the cost model analysis and in vectorizable_load.
Bill
>
> Thanks,
> Richard
>
>