Re: Re: [PATCH] VECT: Enhance SLP of MASK_LEN_GATHER_LOAD[PR111721]

[Richard Biener <rguenther@suse.de>]

On Thu, 12 Oct 2023, juzhe.zhong@rivai.ai wrote:

> Hi, Richi.
> 
> I restrict as you said into vect_external_def.
> 
> Then this condition made SLP failed:
> 
> -      if (mask_index >= 0
> +      if (mask_index >= 0 && internal_fn_len_index (ifn) < 0
>           && !vect_check_scalar_mask (vinfo, stmt_info, slp_node, mask_index,
>                                       &mask, NULL, &mask_dt, &mask_vectype))
>         return false;
>
> So I add 'internal_fn_len_index (ifn) < 0' for MASK_LEN_GATHER_LOAD does not check scalar mask.

Rather figure why.
 
> Then ICE here:
> 
> vect_slp_analyze_node_operations
> if (child
>           && (SLP_TREE_DEF_TYPE (child) == vect_constant_def
>               || SLP_TREE_DEF_TYPE (child) == vect_external_def)
>           /* Perform usual caching, note code-generation still
>              code-gens these nodes multiple times but we expect
>              to CSE them later.  */
>           && !visited_set.add (child))
>         {
>           visited_vec.safe_push (child);
>           /* ???  After auditing more code paths make a "default"
>              and push the vector type from NODE to all children
>              if it is not already set.  */
>           /* Compute the number of vectors to be generated.  */
>           tree vector_type = SLP_TREE_VECTYPE (child);
>           if (!vector_type)
>             {
>               /* For shifts with a scalar argument we don't need
>                  to cost or code-generate anything.
>                  ???  Represent this more explicitely.  */
>               gcc_assert ((STMT_VINFO_TYPE (SLP_TREE_REPRESENTATIVE (node)) ----> assert FAILed.
>                            == shift_vec_info_type)
>                           && j == 1);
>               continue;
>             }
> 
> Could you help me with that?
> 
> 
> juzhe.zhong@rivai.ai
>  
> From: Richard Biener
> Date: 2023-10-12 17:55
> To: juzhe.zhong@rivai.ai
> CC: gcc-patches; richard.sandiford
> Subject: Re: Re: [PATCH] VECT: Enhance SLP of MASK_LEN_GATHER_LOAD[PR111721]
> On Thu, 12 Oct 2023, juzhe.zhong@rivai.ai wrote:
>  
> > I tree-vect-slp.cc:
> > vect_get_and_check_slp_defs
> > 711: 
> > 
> >           tree type = TREE_TYPE (oprnd);
> >           dt = dts[i];
> >           if ((dt == vect_constant_def
> >                || dt == vect_external_def)
> >               && !GET_MODE_SIZE (vinfo->vector_mode).is_constant ()
> >               && (TREE_CODE (type) == BOOLEAN_TYPE
> >                   || !can_duplicate_and_interleave_p (vinfo, stmts.length (),
> >                                                       type)))
> >             {
> >               if (dump_enabled_p ())
> >                 dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
> >                                  "Build SLP failed: invalid type of def "
> >                                  "for variable-length SLP %T\n", oprnd);
> >               return -1;
> >             }
> > 
> > Here mask = -1 is BOOLEAN type in tree-vect-patterns.cc reaching this condition, then SLP failed:
> > Build SLP failed: invalid type of def
>  
> I think this can be restricted to vect_external_def, but some history
> might reveal the cases we put this code in for (we should be able to
> materialize all constants?).  At least uniform boolean constants
> should be fine.
> >
> > 
> > 
> > juzhe.zhong@rivai.ai
> >  
> > From: Richard Biener
> > Date: 2023-10-12 17:44
> > To: ???
> > CC: gcc-patches; richard.sandiford
> > Subject: Re: Re: [PATCH] VECT: Enhance SLP of MASK_LEN_GATHER_LOAD[PR111721]
> > On Thu, 12 Oct 2023, ??? wrote:
> >  
> > > Thanks Richi point it out.
> > > 
> > > I found this patch can't make conditional gather load succeed on SLP.
> > > 
> > > I am considering change MASK_LEN_GATHER_LOAD in pattern recognization:
> > > 
> > > If no condition mask, in tree-vect-patterns.cc,  I build MASK_LEN_GATHER_LOAD (ptr, offset, scale, 0) -> 4 arguments same as GATHER_LOAD.
> > > In this situation, MASK_LEN_GATHER_LOAD can resue the GATHER_LOAD SLP flow naturally.
> > > 
> > > If has condition mask, in tree-vect-patterns.cc,  I build MASK_LEN_GATHER_LOAD (ptr, offset, scale, 0, condition) -> 5 arguments same as MASK_GATHER_LOAD.
> > > In this situation, MASK_LEN_GATHER_LOAD can resue the MASK_GATHER_LOAD SLP flow naturally.
> > > 
> > > Is it reasonable ?
> >  
> > What's wrong with handling MASK_LEN_GATHER_LOAD with all arguments
> > even when the mask is -1?
> >  
> > > 
> > > juzhe.zhong@rivai.ai
> > >  
> > > From: Richard Biener
> > > Date: 2023-10-11 20:50
> > > To: Juzhe-Zhong
> > > CC: gcc-patches; richard.sandiford
> > > Subject: Re: [PATCH] VECT: Enhance SLP of MASK_LEN_GATHER_LOAD[PR111721]
> > > On Wed, 11 Oct 2023, Juzhe-Zhong wrote:
> > >  
> > > > This patch fixes this following FAILs in RISC-V regression:
> > > > 
> > > > FAIL: gcc.dg/vect/vect-gather-1.c -flto -ffat-lto-objects  scan-tree-dump vect "Loop contains only SLP stmts"
> > > > FAIL: gcc.dg/vect/vect-gather-1.c scan-tree-dump vect "Loop contains only SLP stmts"
> > > > FAIL: gcc.dg/vect/vect-gather-3.c -flto -ffat-lto-objects  scan-tree-dump vect "Loop contains only SLP stmts"
> > > > FAIL: gcc.dg/vect/vect-gather-3.c scan-tree-dump vect "Loop contains only SLP stmts"
> > > > 
> > > > The root cause of these FAIL is that GCC SLP failed on MASK_LEN_GATHER_LOAD.
> > > > 
> > > > Since for RVV, we build MASK_LEN_GATHER_LOAD with dummy mask (-1) in tree-vect-patterns.cc if it is same
> > > > situation as GATHER_LOAD (no conditional mask).
> > > > 
> > > > So we make MASK_LEN_GATHER_LOAD leverage the flow of GATHER_LOAD if mask argument is a dummy mask.
> > > > 
> > > > gcc/ChangeLog:
> > > > 
> > > > * tree-vect-slp.cc (vect_get_operand_map):
> > > > (vect_build_slp_tree_1):
> > > > (vect_build_slp_tree_2):
> > > > * tree-vect-stmts.cc (vectorizable_load):
> > > > 
> > > > ---
> > > >  gcc/tree-vect-slp.cc   | 18 ++++++++++++++++--
> > > >  gcc/tree-vect-stmts.cc |  4 ++--
> > > >  2 files changed, 18 insertions(+), 4 deletions(-)
> > > > 
> > > > diff --git a/gcc/tree-vect-slp.cc b/gcc/tree-vect-slp.cc
> > > > index fa098f9ff4e..712c04ec278 100644
> > > > --- a/gcc/tree-vect-slp.cc
> > > > +++ b/gcc/tree-vect-slp.cc
> > > > @@ -544,6 +544,17 @@ vect_get_operand_map (const gimple *stmt, unsigned char swap = 0)
> > > >    case IFN_MASK_GATHER_LOAD:
> > > >      return arg1_arg4_map;
> > > >  
> > > > +   case IFN_MASK_LEN_GATHER_LOAD:
> > > > +     /* In tree-vect-patterns.cc, we will have these 2 situations:
> > > > +
> > > > + - Unconditional gather load transforms
> > > > +   into MASK_LEN_GATHER_LOAD with dummy mask which is -1.
> > > > +
> > > > + - Conditional gather load transforms
> > > > +   into MASK_LEN_GATHER_LOAD with real conditional mask.*/
> > > > +     return integer_minus_onep (gimple_call_arg (call, 4)) ? arg1_map
> > > > +   : nullptr;
> > > > +
> > > >    case IFN_MASK_STORE:
> > > >      return arg3_arg2_map;
> > > >  
> > > > @@ -1077,7 +1088,8 @@ vect_build_slp_tree_1 (vec_info *vinfo, unsigned char *swap,
> > > >  
> > > >    if (cfn == CFN_MASK_LOAD
> > > >        || cfn == CFN_GATHER_LOAD
> > > > -       || cfn == CFN_MASK_GATHER_LOAD)
> > > > +       || cfn == CFN_MASK_GATHER_LOAD
> > > > +       || cfn == CFN_MASK_LEN_GATHER_LOAD)
> > > >      ldst_p = true;
> > > >    else if (cfn == CFN_MASK_STORE)
> > > >      {
> > > > @@ -1337,6 +1349,7 @@ vect_build_slp_tree_1 (vec_info *vinfo, unsigned char *swap,
> > > >    if (DR_IS_READ (STMT_VINFO_DATA_REF (stmt_info))
> > > >        && rhs_code != CFN_GATHER_LOAD
> > > >        && rhs_code != CFN_MASK_GATHER_LOAD
> > > > +       && rhs_code != CFN_MASK_LEN_GATHER_LOAD
> > > >        /* Not grouped loads are handled as externals for BB
> > > >  vectorization.  For loop vectorization we can handle
> > > >  splats the same we handle single element interleaving.  */
> > > > @@ -1837,7 +1850,8 @@ vect_build_slp_tree_2 (vec_info *vinfo, slp_tree node,
> > > >        if (gcall *stmt = dyn_cast <gcall *> (stmt_info->stmt))
> > > >  gcc_assert (gimple_call_internal_p (stmt, IFN_MASK_LOAD)
> > > >      || gimple_call_internal_p (stmt, IFN_GATHER_LOAD)
> > > > -     || gimple_call_internal_p (stmt, IFN_MASK_GATHER_LOAD));
> > > > +     || gimple_call_internal_p (stmt, IFN_MASK_GATHER_LOAD)
> > > > +     || gimple_call_internal_p (stmt, IFN_MASK_LEN_GATHER_LOAD));
> > > >        else
> > > >  {
> > > >    *max_nunits = this_max_nunits;
> > > > diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
> > > > index cd7c1090d88..263acf5d3cd 100644
> > > > --- a/gcc/tree-vect-stmts.cc
> > > > +++ b/gcc/tree-vect-stmts.cc
> > > > @@ -9575,9 +9575,9 @@ vectorizable_load (vec_info *vinfo,
> > > >  return false;
> > > >  
> > > >        mask_index = internal_fn_mask_index (ifn);
> > > > -      if (mask_index >= 0 && slp_node)
> > > > +      if (mask_index >= 0 && slp_node && internal_fn_len_index (ifn) < 0)
> > > >  mask_index = vect_slp_child_index_for_operand (call, mask_index);
> > > > -      if (mask_index >= 0
> > > > +      if (mask_index >= 0 && internal_fn_len_index (ifn) < 0
> > > >    && !vect_check_scalar_mask (vinfo, stmt_info, slp_node, mask_index,
> > > >        &mask, NULL, &mask_dt, &mask_vectype))
> > > >  return false;
> > >  
> > > You are ignoring the mask argument and here only handle it when the
> > > IFN doesn't have a _LEN.  This doesn't seem to be forward looking
> > > to the point where you want to actually handle masked (aka conditional)
> > > gather.
> > >  
> > > Did you check that SLP is actually used to vectorize this?
> > >  
> > > Richard.
> > >  
> > > 
> >  
> > 
>  
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)