Bug 28436 - accessing an element via a "pointer" on a vector does not cause vec_extract to be used
Summary: accessing an element via a "pointer" on a vector does not cause vec_extract ...
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: middle-end (show other bugs)
Version: 4.2.0
: P3 enhancement
Target Milestone: 4.3.0
Assignee: Andrew Pinski
URL: http://gcc.gnu.org/ml/gcc-patches/200...
Keywords: missed-optimization, patch
Depends on:
Blocks:
 
Reported: 2006-07-19 14:16 UTC by Andrew Pinski
Modified: 2006-12-12 22:34 UTC (History)
3 users (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2006-09-03 21:51:31


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Andrew Pinski 2006-07-19 14:16:31 UTC
Just like PR 28367:
#define vector __attribute__((vector_size(16)))
float f(vector float t)
{
  return *(float*)&t;
}

We get a full vector store which we can get around with a stvewx which just stores that one element and we can get that via a vec_extract pattern.

On x86_64 has the same issue, maybe even worse as there is no need to go through memory there.
X86_64 produces:
f:
.LFB2:
        movaps  %xmm0, -24(%rsp)
        movss   -24(%rsp), %xmm0
        ret
When really it should produce just a ret.
Comment 1 Andrew Pinski 2006-07-19 14:17:26 UTC
I have a patch to fix this and other bugs that show up because of the more uses of vec_set/vec_extract which I will submit when I get back in the US.
Comment 2 Andrew Pinski 2006-11-09 17:24:05 UTC
Testing a fix for this one, though it only does the simple *(int*)&vector_int_var case.
Comment 3 Andrew Pinski 2006-11-21 03:32:45 UTC
Note I Noticed with vector int on x86, it incorrectly uses SF mode because vec_extract is done wrongly.
Comment 4 Andrew Pinski 2006-12-12 22:33:23 UTC
Subject: Bug 28436

Author: pinskia
Date: Tue Dec 12 22:33:06 2006
New Revision: 119801

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=119801
Log:
2006-12-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>

        PR tree-opt/28436
        * tree.h (DECL_COMPLEX_GIMPLE_REG_P): Rename to ...
        (DECL_GIMPLE_REG_P): This.
        * fold-const.c (fold_indirect_ref_1): Fold *(foo *)&vectorfoo into
        using BIT_FIELD_REF.
        * omp-low.c (omp_copy_decl_2): Use the renamed DECL_GIMPLE_REG_P.
        * tree-gimple.c (is_gimple_reg): Use the renamed DECL_GIMPLE_REG_P
        and check for VECTOR_TYPE.
        * expr.c (get_inner_reference): Set the mode for BIT_FIELD_REF with
        vector types.
        * tree-flow-inline.h (var_can_have_subvars): Use the renamed
        DECL_GIMPLE_REG_P.
        * gimplify.c (internal_get_tmp_var): Use the renamed DECL_GIMPLE_REG_P
        and check for VECTOR_TYPE.
        (gimplify_bind_expr): Likewise.
        (gimplify_function_tree): Likewise.
        * expmed.c: Include target.h.
        (extract_bit_field): For vector mode, try find a better mode first.
        If that fails use gen_lowpart (for vectors only).
        * tree-dfa.c (make_rename_temp): Use the renamed DECL_GIMPLE_REG_P
        and check for VECTOR_TYPE.
        * tree-ssa-pre.c (create_expressions_by_pieces): Likewise.
        (insert_into_preds_of_block): Likewise.
        (insert_fake_stores): Create gimple register store_tmps for
        vector types.
        * tree-sra.c (sra_elt): New field, is_vector_lhs.
        (sra_walk_expr <case BIT_FIELD_REF>): For vector types that
        are the left hand side, set the element's is_vector_lhs to true.
        (instantiate_element): For vector types which were on the left         hand size, set DECL_GIMPLE_REG_P to false.
        * tree-nested.c (create_tmp_var_for): Use the renamed DECL_GIMPLE_REG_P.        * tree-inline.c (declare_return_variable):  Use the renamed
        DECL_GIMPLE_REG_P
        and check for VECTOR_TYPE.         (copy_decl_to_var):  Use the renamed DECL_GIMPLE_REG_P.
        (copy_result_decl_to_var): Likewise.
        * tree-vect-transform.c (vect_get_new_vect_var): For vector types,         create a gimple register variable.
        (vect_permute_store_chain): Set DECL_GIMPLE_REG_P to true for the
        vect_inter_* temp variables.
        * Makefile.in (expmed.o): Update dependencies.

2006-12-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>

        PR tree-opt/28436
        * gcc.c-torture/compile/vector-1.c: New test.
        * gcc.c-torture/compile/vector-2.c: New test.
        * gcc.c-torture/compile/vector-3.c: New test.



Added:
    trunk/gcc/testsuite/gcc.c-torture/compile/vector-1.c
    trunk/gcc/testsuite/gcc.c-torture/compile/vector-2.c
    trunk/gcc/testsuite/gcc.c-torture/compile/vector-3.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/Makefile.in
    trunk/gcc/expmed.c
    trunk/gcc/expr.c
    trunk/gcc/fold-const.c
    trunk/gcc/gimplify.c
    trunk/gcc/omp-low.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-dfa.c
    trunk/gcc/tree-flow-inline.h
    trunk/gcc/tree-gimple.c
    trunk/gcc/tree-inline.c
    trunk/gcc/tree-nested.c
    trunk/gcc/tree-sra.c
    trunk/gcc/tree-ssa-pre.c
    trunk/gcc/tree-vect-transform.c
    trunk/gcc/tree.h

Comment 5 Andrew Pinski 2006-12-12 22:34:13 UTC
Fixed mostly, I will file a new bug for the other issues, in that a nonzero index and a non constant index.