[V3] VECT: Change flow of decrement IV

Message ID 20230601043617.173986-1-juzhe.zhong@rivai.ai
State Committed
Commit bffc52838e393a775e13dc48162669b0f43ebe09
Headers
Series [V3] VECT: Change flow of decrement IV |

Commit Message

钟居哲 June 1, 2023, 4:36 a.m. UTC
  From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>

Follow Richi's suggestion, I change current decrement IV flow from:

do {
   remain -= MIN (vf, remain);
} while (remain != 0);

into:

do {
   old_remain = remain;
   len = MIN (vf, remain);
   remain -= vf;
} while (old_remain >= vf);

to enhance SCEV.

Include fixes from kewen.


This patch will need to wait for Kewen's test feedback.

Testing on X86 is on-going

Co-Authored by: Kewen Lin  <linkw@linux.ibm.com>

  PR tree-optimization/109971

gcc/ChangeLog:

        * tree-vect-loop-manip.cc (vect_set_loop_controls_directly): Change decrement IV flow.
        (vect_set_loop_condition_partial_vectors): Ditto.

---
 gcc/tree-vect-loop-manip.cc | 36 +++++++++++++++++++++++++-----------
 1 file changed, 25 insertions(+), 11 deletions(-)
  

Comments

钟居哲 June 1, 2023, 5 a.m. UTC | #1
This patch is no difference from V2.
Just add PR tree-optimization/109971 as Kewen's suggested.

Already bootstrapped and Regression on X86 no difference.

Ok for trunk ?


juzhe.zhong@rivai.ai
 
From: juzhe.zhong
Date: 2023-06-01 12:36
To: gcc-patches
CC: richard.sandiford; rguenther; linkw; Ju-Zhe Zhong
Subject: [PATCH V3] VECT: Change flow of decrement IV
From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
 
Follow Richi's suggestion, I change current decrement IV flow from:
 
do {
   remain -= MIN (vf, remain);
} while (remain != 0);
 
into:
 
do {
   old_remain = remain;
   len = MIN (vf, remain);
   remain -= vf;
} while (old_remain >= vf);
 
to enhance SCEV.
 
Include fixes from kewen.
 
 
This patch will need to wait for Kewen's test feedback.
 
Testing on X86 is on-going
 
Co-Authored by: Kewen Lin  <linkw@linux.ibm.com>
 
  PR tree-optimization/109971
 
gcc/ChangeLog:
 
        * tree-vect-loop-manip.cc (vect_set_loop_controls_directly): Change decrement IV flow.
        (vect_set_loop_condition_partial_vectors): Ditto.
 
---
gcc/tree-vect-loop-manip.cc | 36 +++++++++++++++++++++++++-----------
1 file changed, 25 insertions(+), 11 deletions(-)
 
diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
index acf3642ceb2..3f735945e67 100644
--- a/gcc/tree-vect-loop-manip.cc
+++ b/gcc/tree-vect-loop-manip.cc
@@ -483,7 +483,7 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
gimple_stmt_iterator loop_cond_gsi,
rgroup_controls *rgc, tree niters,
tree niters_skip, bool might_wrap_p,
- tree *iv_step)
+ tree *iv_step, tree *compare_step)
{
   tree compare_type = LOOP_VINFO_RGROUP_COMPARE_TYPE (loop_vinfo);
   tree iv_type = LOOP_VINFO_RGROUP_IV_TYPE (loop_vinfo);
@@ -538,9 +538,9 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
   ...
   vect__4.8_28 = .LEN_LOAD (_17, 32B, _36, 0);
   ...
-    ivtmp_35 = ivtmp_9 - _36;
+    ivtmp_35 = ivtmp_9 - POLY_INT_CST [4, 4];
   ...
-    if (ivtmp_35 != 0)
+    if (ivtmp_9 > POLY_INT_CST [4, 4])
     goto <bb 4>; [83.33%]
   else
     goto <bb 5>; [16.67%]
@@ -549,13 +549,15 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
       tree step = rgc->controls.length () == 1 ? rgc->controls[0]
       : make_ssa_name (iv_type);
       /* Create decrement IV.  */
-      create_iv (nitems_total, MINUS_EXPR, step, NULL_TREE, loop, &incr_gsi,
- insert_after, &index_before_incr, &index_after_incr);
+      create_iv (nitems_total, MINUS_EXPR, nitems_step, NULL_TREE, loop,
+ &incr_gsi, insert_after, &index_before_incr,
+ &index_after_incr);
       gimple_seq_add_stmt (header_seq, gimple_build_assign (step, MIN_EXPR,
    index_before_incr,
    nitems_step));
       *iv_step = step;
-      return index_after_incr;
+      *compare_step = nitems_step;
+      return index_before_incr;
     }
   /* Create increment IV.  */
@@ -825,6 +827,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
      arbitrarily pick the last.  */
   tree test_ctrl = NULL_TREE;
   tree iv_step = NULL_TREE;
+  tree compare_step = NULL_TREE;
   rgroup_controls *rgc;
   rgroup_controls *iv_rgc = nullptr;
   unsigned int i;
@@ -861,7 +864,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
&preheader_seq, &header_seq,
loop_cond_gsi, rgc, niters,
niters_skip, might_wrap_p,
- &iv_step);
+ &iv_step, &compare_step);
    iv_rgc = rgc;
  }
@@ -884,10 +887,21 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
   /* Get a boolean result that tells us whether to iterate.  */
   edge exit_edge = single_exit (loop);
-  tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
-  tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
-  gcond *cond_stmt = gimple_build_cond (code, test_ctrl, zero_ctrl,
- NULL_TREE, NULL_TREE);
+  gcond *cond_stmt;
+  if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
+    {
+      gcc_assert (compare_step);
+      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? LE_EXPR : GT_EXPR;
+      cond_stmt = gimple_build_cond (code, test_ctrl, compare_step, NULL_TREE,
+      NULL_TREE);
+    }
+  else
+    {
+      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
+      tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
+      cond_stmt
+ = gimple_build_cond (code, test_ctrl, zero_ctrl, NULL_TREE, NULL_TREE);
+    }
   gsi_insert_before (&loop_cond_gsi, cond_stmt, GSI_SAME_STMT);
   /* The loop iterates (NITERS - 1) / VF + 1 times.
-- 
2.36.3
  
Kewen.Lin June 1, 2023, 5:24 a.m. UTC | #2
Hi,

on 2023/6/1 13:00, juzhe.zhong@rivai.ai wrote:
> This patch is no difference from V2.

I support this patch based on the testing and SPEC2017 evaluation
results on Power (see my comments on patch v2).

> Just add PR tree-optimization/109971 as Kewen's suggested.

Thanks for adding that, I was expecting you will add that when you
are committing it, not really requesting one new version. :)  btw,
the PR marker(s) will trigger scripts to comment some commit info
(commit link, commit log) into the specified PR(s), people can
find some connections between PRs and (fixing or progressing forward)
commits easily.

BR,
Kewen

> 
> Already bootstrapped and Regression on X86 no difference.
> 
> Ok for trunk ?
> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> juzhe.zhong@rivai.ai
> 
>      
>     *From:* juzhe.zhong <mailto:juzhe.zhong@rivai.ai>
>     *Date:* 2023-06-01 12:36
>     *To:* gcc-patches <mailto:gcc-patches@gcc.gnu.org>
>     *CC:* richard.sandiford <mailto:richard.sandiford@arm.com>; rguenther <mailto:rguenther@suse.de>; linkw <mailto:linkw@linux.ibm.com>; Ju-Zhe Zhong <mailto:juzhe.zhong@rivai.ai>
>     *Subject:* [PATCH V3] VECT: Change flow of decrement IV
>     From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
>      
>     Follow Richi's suggestion, I change current decrement IV flow from:
>      
>     do {
>        remain -= MIN (vf, remain);
>     } while (remain != 0);
>      
>     into:
>      
>     do {
>        old_remain = remain;
>        len = MIN (vf, remain);
>        remain -= vf;
>     } while (old_remain >= vf);
>      
>     to enhance SCEV.
>      
>     Include fixes from kewen.
>      
>      
>     This patch will need to wait for Kewen's test feedback.
>      
>     Testing on X86 is on-going
>      
>     Co-Authored by: Kewen Lin  <linkw@linux.ibm.com>
>      
>       PR tree-optimization/109971
>      
>     gcc/ChangeLog:
>      
>             * tree-vect-loop-manip.cc (vect_set_loop_controls_directly): Change decrement IV flow.
>             (vect_set_loop_condition_partial_vectors): Ditto.
>      
>     ---
>     gcc/tree-vect-loop-manip.cc | 36 +++++++++++++++++++++++++-----------
>     1 file changed, 25 insertions(+), 11 deletions(-)
>      
>     diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
>     index acf3642ceb2..3f735945e67 100644
>     --- a/gcc/tree-vect-loop-manip.cc
>     +++ b/gcc/tree-vect-loop-manip.cc
>     @@ -483,7 +483,7 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>     gimple_stmt_iterator loop_cond_gsi,
>     rgroup_controls *rgc, tree niters,
>     tree niters_skip, bool might_wrap_p,
>     - tree *iv_step)
>     + tree *iv_step, tree *compare_step)
>     {
>        tree compare_type = LOOP_VINFO_RGROUP_COMPARE_TYPE (loop_vinfo);
>        tree iv_type = LOOP_VINFO_RGROUP_IV_TYPE (loop_vinfo);
>     @@ -538,9 +538,9 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>        ...
>        vect__4.8_28 = .LEN_LOAD (_17, 32B, _36, 0);
>        ...
>     -    ivtmp_35 = ivtmp_9 - _36;
>     +    ivtmp_35 = ivtmp_9 - POLY_INT_CST [4, 4];
>        ...
>     -    if (ivtmp_35 != 0)
>     +    if (ivtmp_9 > POLY_INT_CST [4, 4])
>          goto <bb 4>; [83.33%]
>        else
>          goto <bb 5>; [16.67%]
>     @@ -549,13 +549,15 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>            tree step = rgc->controls.length () == 1 ? rgc->controls[0]
>            : make_ssa_name (iv_type);
>            /* Create decrement IV.  */
>     -      create_iv (nitems_total, MINUS_EXPR, step, NULL_TREE, loop, &incr_gsi,
>     - insert_after, &index_before_incr, &index_after_incr);
>     +      create_iv (nitems_total, MINUS_EXPR, nitems_step, NULL_TREE, loop,
>     + &incr_gsi, insert_after, &index_before_incr,
>     + &index_after_incr);
>            gimple_seq_add_stmt (header_seq, gimple_build_assign (step, MIN_EXPR,
>         index_before_incr,
>         nitems_step));
>            *iv_step = step;
>     -      return index_after_incr;
>     +      *compare_step = nitems_step;
>     +      return index_before_incr;
>          }
>        /* Create increment IV.  */
>     @@ -825,6 +827,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>           arbitrarily pick the last.  */
>        tree test_ctrl = NULL_TREE;
>        tree iv_step = NULL_TREE;
>     +  tree compare_step = NULL_TREE;
>        rgroup_controls *rgc;
>        rgroup_controls *iv_rgc = nullptr;
>        unsigned int i;
>     @@ -861,7 +864,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>     &preheader_seq, &header_seq,
>     loop_cond_gsi, rgc, niters,
>     niters_skip, might_wrap_p,
>     - &iv_step);
>     + &iv_step, &compare_step);
>         iv_rgc = rgc;
>       }
>     @@ -884,10 +887,21 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>        /* Get a boolean result that tells us whether to iterate.  */
>        edge exit_edge = single_exit (loop);
>     -  tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
>     -  tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
>     -  gcond *cond_stmt = gimple_build_cond (code, test_ctrl, zero_ctrl,
>     - NULL_TREE, NULL_TREE);
>     +  gcond *cond_stmt;
>     +  if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
>     +    {
>     +      gcc_assert (compare_step);
>     +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? LE_EXPR : GT_EXPR;
>     +      cond_stmt = gimple_build_cond (code, test_ctrl, compare_step, NULL_TREE,
>     +      NULL_TREE);
>     +    }
>     +  else
>     +    {
>     +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
>     +      tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
>     +      cond_stmt
>     + = gimple_build_cond (code, test_ctrl, zero_ctrl, NULL_TREE, NULL_TREE);
>     +    }
>        gsi_insert_before (&loop_cond_gsi, cond_stmt, GSI_SAME_STMT);
>        /* The loop iterates (NITERS - 1) / VF + 1 times.
>     -- 
>     2.36.3
>      
>
  
钟居哲 June 1, 2023, 8:09 a.m. UTC | #3
Thanks Kewen. Let's wait for Richard and Richi.



juzhe.zhong@rivai.ai
 
From: Kewen.Lin
Date: 2023-06-01 13:24
To: juzhe.zhong@rivai.ai
CC: richard.sandiford; rguenther; gcc-patches
Subject: Re: [PATCH V3] VECT: Change flow of decrement IV
Hi,
 
on 2023/6/1 13:00, juzhe.zhong@rivai.ai wrote:
> This patch is no difference from V2.
 
I support this patch based on the testing and SPEC2017 evaluation
results on Power (see my comments on patch v2).
 
> Just add PR tree-optimization/109971 as Kewen's suggested.
 
Thanks for adding that, I was expecting you will add that when you
are committing it, not really requesting one new version. :)  btw,
the PR marker(s) will trigger scripts to comment some commit info
(commit link, commit log) into the specified PR(s), people can
find some connections between PRs and (fixing or progressing forward)
commits easily.
 
BR,
Kewen
 
> 
> Already bootstrapped and Regression on X86 no difference.
> 
> Ok for trunk ?
> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
> juzhe.zhong@rivai.ai
> 
>      
>     *From:* juzhe.zhong <mailto:juzhe.zhong@rivai.ai>
>     *Date:* 2023-06-01 12:36
>     *To:* gcc-patches <mailto:gcc-patches@gcc.gnu.org>
>     *CC:* richard.sandiford <mailto:richard.sandiford@arm.com>; rguenther <mailto:rguenther@suse.de>; linkw <mailto:linkw@linux.ibm.com>; Ju-Zhe Zhong <mailto:juzhe.zhong@rivai.ai>
>     *Subject:* [PATCH V3] VECT: Change flow of decrement IV
>     From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
>      
>     Follow Richi's suggestion, I change current decrement IV flow from:
>      
>     do {
>        remain -= MIN (vf, remain);
>     } while (remain != 0);
>      
>     into:
>      
>     do {
>        old_remain = remain;
>        len = MIN (vf, remain);
>        remain -= vf;
>     } while (old_remain >= vf);
>      
>     to enhance SCEV.
>      
>     Include fixes from kewen.
>      
>      
>     This patch will need to wait for Kewen's test feedback.
>      
>     Testing on X86 is on-going
>      
>     Co-Authored by: Kewen Lin  <linkw@linux.ibm.com>
>      
>       PR tree-optimization/109971
>      
>     gcc/ChangeLog:
>      
>             * tree-vect-loop-manip.cc (vect_set_loop_controls_directly): Change decrement IV flow.
>             (vect_set_loop_condition_partial_vectors): Ditto.
>      
>     ---
>     gcc/tree-vect-loop-manip.cc | 36 +++++++++++++++++++++++++-----------
>     1 file changed, 25 insertions(+), 11 deletions(-)
>      
>     diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
>     index acf3642ceb2..3f735945e67 100644
>     --- a/gcc/tree-vect-loop-manip.cc
>     +++ b/gcc/tree-vect-loop-manip.cc
>     @@ -483,7 +483,7 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>     gimple_stmt_iterator loop_cond_gsi,
>     rgroup_controls *rgc, tree niters,
>     tree niters_skip, bool might_wrap_p,
>     - tree *iv_step)
>     + tree *iv_step, tree *compare_step)
>     {
>        tree compare_type = LOOP_VINFO_RGROUP_COMPARE_TYPE (loop_vinfo);
>        tree iv_type = LOOP_VINFO_RGROUP_IV_TYPE (loop_vinfo);
>     @@ -538,9 +538,9 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>        ...
>        vect__4.8_28 = .LEN_LOAD (_17, 32B, _36, 0);
>        ...
>     -    ivtmp_35 = ivtmp_9 - _36;
>     +    ivtmp_35 = ivtmp_9 - POLY_INT_CST [4, 4];
>        ...
>     -    if (ivtmp_35 != 0)
>     +    if (ivtmp_9 > POLY_INT_CST [4, 4])
>          goto <bb 4>; [83.33%]
>        else
>          goto <bb 5>; [16.67%]
>     @@ -549,13 +549,15 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>            tree step = rgc->controls.length () == 1 ? rgc->controls[0]
>            : make_ssa_name (iv_type);
>            /* Create decrement IV.  */
>     -      create_iv (nitems_total, MINUS_EXPR, step, NULL_TREE, loop, &incr_gsi,
>     - insert_after, &index_before_incr, &index_after_incr);
>     +      create_iv (nitems_total, MINUS_EXPR, nitems_step, NULL_TREE, loop,
>     + &incr_gsi, insert_after, &index_before_incr,
>     + &index_after_incr);
>            gimple_seq_add_stmt (header_seq, gimple_build_assign (step, MIN_EXPR,
>         index_before_incr,
>         nitems_step));
>            *iv_step = step;
>     -      return index_after_incr;
>     +      *compare_step = nitems_step;
>     +      return index_before_incr;
>          }
>        /* Create increment IV.  */
>     @@ -825,6 +827,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>           arbitrarily pick the last.  */
>        tree test_ctrl = NULL_TREE;
>        tree iv_step = NULL_TREE;
>     +  tree compare_step = NULL_TREE;
>        rgroup_controls *rgc;
>        rgroup_controls *iv_rgc = nullptr;
>        unsigned int i;
>     @@ -861,7 +864,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>     &preheader_seq, &header_seq,
>     loop_cond_gsi, rgc, niters,
>     niters_skip, might_wrap_p,
>     - &iv_step);
>     + &iv_step, &compare_step);
>         iv_rgc = rgc;
>       }
>     @@ -884,10 +887,21 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>        /* Get a boolean result that tells us whether to iterate.  */
>        edge exit_edge = single_exit (loop);
>     -  tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
>     -  tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
>     -  gcond *cond_stmt = gimple_build_cond (code, test_ctrl, zero_ctrl,
>     - NULL_TREE, NULL_TREE);
>     +  gcond *cond_stmt;
>     +  if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
>     +    {
>     +      gcc_assert (compare_step);
>     +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? LE_EXPR : GT_EXPR;
>     +      cond_stmt = gimple_build_cond (code, test_ctrl, compare_step, NULL_TREE,
>     +      NULL_TREE);
>     +    }
>     +  else
>     +    {
>     +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
>     +      tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
>     +      cond_stmt
>     + = gimple_build_cond (code, test_ctrl, zero_ctrl, NULL_TREE, NULL_TREE);
>     +    }
>        gsi_insert_before (&loop_cond_gsi, cond_stmt, GSI_SAME_STMT);
>        /* The loop iterates (NITERS - 1) / VF + 1 times.
>     -- 
>     2.36.3
>      
>
  
Richard Biener June 2, 2023, 8:56 a.m. UTC | #4
On Thu, 1 Jun 2023, juzhe.zhong@rivai.ai wrote:

> This patch is no difference from V2.
> Just add PR tree-optimization/109971 as Kewen's suggested.
> 
> Already bootstrapped and Regression on X86 no difference.
> 
> Ok for trunk ?

OK.

Richard.

> 
> juzhe.zhong@rivai.ai
>  
> From: juzhe.zhong
> Date: 2023-06-01 12:36
> To: gcc-patches
> CC: richard.sandiford; rguenther; linkw; Ju-Zhe Zhong
> Subject: [PATCH V3] VECT: Change flow of decrement IV
> From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
>  
> Follow Richi's suggestion, I change current decrement IV flow from:
>  
> do {
>    remain -= MIN (vf, remain);
> } while (remain != 0);
>  
> into:
>  
> do {
>    old_remain = remain;
>    len = MIN (vf, remain);
>    remain -= vf;
> } while (old_remain >= vf);
>  
> to enhance SCEV.
>  
> Include fixes from kewen.
>  
>  
> This patch will need to wait for Kewen's test feedback.
>  
> Testing on X86 is on-going
>  
> Co-Authored by: Kewen Lin  <linkw@linux.ibm.com>
>  
>   PR tree-optimization/109971
>  
> gcc/ChangeLog:
>  
>         * tree-vect-loop-manip.cc (vect_set_loop_controls_directly): Change decrement IV flow.
>         (vect_set_loop_condition_partial_vectors): Ditto.
>  
> ---
> gcc/tree-vect-loop-manip.cc | 36 +++++++++++++++++++++++++-----------
> 1 file changed, 25 insertions(+), 11 deletions(-)
>  
> diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
> index acf3642ceb2..3f735945e67 100644
> --- a/gcc/tree-vect-loop-manip.cc
> +++ b/gcc/tree-vect-loop-manip.cc
> @@ -483,7 +483,7 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
> gimple_stmt_iterator loop_cond_gsi,
> rgroup_controls *rgc, tree niters,
> tree niters_skip, bool might_wrap_p,
> - tree *iv_step)
> + tree *iv_step, tree *compare_step)
> {
>    tree compare_type = LOOP_VINFO_RGROUP_COMPARE_TYPE (loop_vinfo);
>    tree iv_type = LOOP_VINFO_RGROUP_IV_TYPE (loop_vinfo);
> @@ -538,9 +538,9 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>    ...
>    vect__4.8_28 = .LEN_LOAD (_17, 32B, _36, 0);
>    ...
> -    ivtmp_35 = ivtmp_9 - _36;
> +    ivtmp_35 = ivtmp_9 - POLY_INT_CST [4, 4];
>    ...
> -    if (ivtmp_35 != 0)
> +    if (ivtmp_9 > POLY_INT_CST [4, 4])
>      goto <bb 4>; [83.33%]
>    else
>      goto <bb 5>; [16.67%]
> @@ -549,13 +549,15 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>        tree step = rgc->controls.length () == 1 ? rgc->controls[0]
>        : make_ssa_name (iv_type);
>        /* Create decrement IV.  */
> -      create_iv (nitems_total, MINUS_EXPR, step, NULL_TREE, loop, &incr_gsi,
> - insert_after, &index_before_incr, &index_after_incr);
> +      create_iv (nitems_total, MINUS_EXPR, nitems_step, NULL_TREE, loop,
> + &incr_gsi, insert_after, &index_before_incr,
> + &index_after_incr);
>        gimple_seq_add_stmt (header_seq, gimple_build_assign (step, MIN_EXPR,
>     index_before_incr,
>     nitems_step));
>        *iv_step = step;
> -      return index_after_incr;
> +      *compare_step = nitems_step;
> +      return index_before_incr;
>      }
>    /* Create increment IV.  */
> @@ -825,6 +827,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>       arbitrarily pick the last.  */
>    tree test_ctrl = NULL_TREE;
>    tree iv_step = NULL_TREE;
> +  tree compare_step = NULL_TREE;
>    rgroup_controls *rgc;
>    rgroup_controls *iv_rgc = nullptr;
>    unsigned int i;
> @@ -861,7 +864,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
> &preheader_seq, &header_seq,
> loop_cond_gsi, rgc, niters,
> niters_skip, might_wrap_p,
> - &iv_step);
> + &iv_step, &compare_step);
>     iv_rgc = rgc;
>   }
> @@ -884,10 +887,21 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>    /* Get a boolean result that tells us whether to iterate.  */
>    edge exit_edge = single_exit (loop);
> -  tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
> -  tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
> -  gcond *cond_stmt = gimple_build_cond (code, test_ctrl, zero_ctrl,
> - NULL_TREE, NULL_TREE);
> +  gcond *cond_stmt;
> +  if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
> +    {
> +      gcc_assert (compare_step);
> +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? LE_EXPR : GT_EXPR;
> +      cond_stmt = gimple_build_cond (code, test_ctrl, compare_step, NULL_TREE,
> +      NULL_TREE);
> +    }
> +  else
> +    {
> +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
> +      tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
> +      cond_stmt
> + = gimple_build_cond (code, test_ctrl, zero_ctrl, NULL_TREE, NULL_TREE);
> +    }
>    gsi_insert_before (&loop_cond_gsi, cond_stmt, GSI_SAME_STMT);
>    /* The loop iterates (NITERS - 1) / VF + 1 times.
>
  
钟居哲 June 2, 2023, 8:58 a.m. UTC | #5
Thanks Richi. I am gonna merge it after Richard's final approve.



juzhe.zhong@rivai.ai
 
From: Richard Biener
Date: 2023-06-02 16:56
To: juzhe.zhong@rivai.ai
CC: gcc-patches; richard.sandiford; linkw
Subject: Re: [PATCH V3] VECT: Change flow of decrement IV
On Thu, 1 Jun 2023, juzhe.zhong@rivai.ai wrote:
 
> This patch is no difference from V2.
> Just add PR tree-optimization/109971 as Kewen's suggested.
> 
> Already bootstrapped and Regression on X86 no difference.
> 
> Ok for trunk ?
 
OK.
 
Richard.
 
> 
> juzhe.zhong@rivai.ai
>  
> From: juzhe.zhong
> Date: 2023-06-01 12:36
> To: gcc-patches
> CC: richard.sandiford; rguenther; linkw; Ju-Zhe Zhong
> Subject: [PATCH V3] VECT: Change flow of decrement IV
> From: Ju-Zhe Zhong <juzhe.zhong@rivai.ai>
>  
> Follow Richi's suggestion, I change current decrement IV flow from:
>  
> do {
>    remain -= MIN (vf, remain);
> } while (remain != 0);
>  
> into:
>  
> do {
>    old_remain = remain;
>    len = MIN (vf, remain);
>    remain -= vf;
> } while (old_remain >= vf);
>  
> to enhance SCEV.
>  
> Include fixes from kewen.
>  
>  
> This patch will need to wait for Kewen's test feedback.
>  
> Testing on X86 is on-going
>  
> Co-Authored by: Kewen Lin  <linkw@linux.ibm.com>
>  
>   PR tree-optimization/109971
>  
> gcc/ChangeLog:
>  
>         * tree-vect-loop-manip.cc (vect_set_loop_controls_directly): Change decrement IV flow.
>         (vect_set_loop_condition_partial_vectors): Ditto.
>  
> ---
> gcc/tree-vect-loop-manip.cc | 36 +++++++++++++++++++++++++-----------
> 1 file changed, 25 insertions(+), 11 deletions(-)
>  
> diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
> index acf3642ceb2..3f735945e67 100644
> --- a/gcc/tree-vect-loop-manip.cc
> +++ b/gcc/tree-vect-loop-manip.cc
> @@ -483,7 +483,7 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
> gimple_stmt_iterator loop_cond_gsi,
> rgroup_controls *rgc, tree niters,
> tree niters_skip, bool might_wrap_p,
> - tree *iv_step)
> + tree *iv_step, tree *compare_step)
> {
>    tree compare_type = LOOP_VINFO_RGROUP_COMPARE_TYPE (loop_vinfo);
>    tree iv_type = LOOP_VINFO_RGROUP_IV_TYPE (loop_vinfo);
> @@ -538,9 +538,9 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>    ...
>    vect__4.8_28 = .LEN_LOAD (_17, 32B, _36, 0);
>    ...
> -    ivtmp_35 = ivtmp_9 - _36;
> +    ivtmp_35 = ivtmp_9 - POLY_INT_CST [4, 4];
>    ...
> -    if (ivtmp_35 != 0)
> +    if (ivtmp_9 > POLY_INT_CST [4, 4])
>      goto <bb 4>; [83.33%]
>    else
>      goto <bb 5>; [16.67%]
> @@ -549,13 +549,15 @@ vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
>        tree step = rgc->controls.length () == 1 ? rgc->controls[0]
>        : make_ssa_name (iv_type);
>        /* Create decrement IV.  */
> -      create_iv (nitems_total, MINUS_EXPR, step, NULL_TREE, loop, &incr_gsi,
> - insert_after, &index_before_incr, &index_after_incr);
> +      create_iv (nitems_total, MINUS_EXPR, nitems_step, NULL_TREE, loop,
> + &incr_gsi, insert_after, &index_before_incr,
> + &index_after_incr);
>        gimple_seq_add_stmt (header_seq, gimple_build_assign (step, MIN_EXPR,
>     index_before_incr,
>     nitems_step));
>        *iv_step = step;
> -      return index_after_incr;
> +      *compare_step = nitems_step;
> +      return index_before_incr;
>      }
>    /* Create increment IV.  */
> @@ -825,6 +827,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>       arbitrarily pick the last.  */
>    tree test_ctrl = NULL_TREE;
>    tree iv_step = NULL_TREE;
> +  tree compare_step = NULL_TREE;
>    rgroup_controls *rgc;
>    rgroup_controls *iv_rgc = nullptr;
>    unsigned int i;
> @@ -861,7 +864,7 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
> &preheader_seq, &header_seq,
> loop_cond_gsi, rgc, niters,
> niters_skip, might_wrap_p,
> - &iv_step);
> + &iv_step, &compare_step);
>     iv_rgc = rgc;
>   }
> @@ -884,10 +887,21 @@ vect_set_loop_condition_partial_vectors (class loop *loop,
>    /* Get a boolean result that tells us whether to iterate.  */
>    edge exit_edge = single_exit (loop);
> -  tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
> -  tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
> -  gcond *cond_stmt = gimple_build_cond (code, test_ctrl, zero_ctrl,
> - NULL_TREE, NULL_TREE);
> +  gcond *cond_stmt;
> +  if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
> +    {
> +      gcc_assert (compare_step);
> +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? LE_EXPR : GT_EXPR;
> +      cond_stmt = gimple_build_cond (code, test_ctrl, compare_step, NULL_TREE,
> +      NULL_TREE);
> +    }
> +  else
> +    {
> +      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
> +      tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
> +      cond_stmt
> + = gimple_build_cond (code, test_ctrl, zero_ctrl, NULL_TREE, NULL_TREE);
> +    }
>    gsi_insert_before (&loop_cond_gsi, cond_stmt, GSI_SAME_STMT);
>    /* The loop iterates (NITERS - 1) / VF + 1 times.
> 
 
-- 
Richard Biener <rguenther@suse.de>
SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg,
Germany; GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman;
HRB 36809 (AG Nuernberg)
  
Richard Sandiford June 2, 2023, 11:43 a.m. UTC | #6
"juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai> writes:
> Thanks Richi. I am gonna merge it after Richard's final approve.

Thanks for checking, but no need to wait for a second ack from me!
Please go ahead and commit.

Richard
  
Li, Pan2 via Gcc-patches June 2, 2023, 11:50 a.m. UTC | #7
Committed, thanks all.

Pan

-----Original Message-----
From: Gcc-patches <gcc-patches-bounces+pan2.li=intel.com@gcc.gnu.org> On Behalf Of Richard Sandiford via Gcc-patches
Sent: Friday, June 2, 2023 7:44 PM
To: juzhe.zhong@rivai.ai
Cc: rguenther <rguenther@suse.de>; gcc-patches <gcc-patches@gcc.gnu.org>; linkw <linkw@linux.ibm.com>
Subject: Re: [PATCH V3] VECT: Change flow of decrement IV

"juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai> writes:
> Thanks Richi. I am gonna merge it after Richard's final approve.

Thanks for checking, but no need to wait for a second ack from me!
Please go ahead and commit.

Richard
  
钟居哲 June 6, 2023, 8:02 a.m. UTC | #8
Hi, Richard and Richi.
Recently, we have some tries on our downstream RVV LLVM.
We change "get_vector_length" (same IR like GCC "select_vl") into "umin", turns out LLVM SCEV analysis succeed.
The unroll PASS works in LLVM.

I think Richard's  suggestion is very reasonable.

I didn't have the time take a look at how LLVM do now but I will take a look at it in the future. 

Thanks.


juzhe.zhong@rivai.ai
 
From: Richard Sandiford
Date: 2023-06-02 19:43
To: juzhe.zhong\@rivai.ai
CC: rguenther; gcc-patches; linkw
Subject: Re: [PATCH V3] VECT: Change flow of decrement IV
"juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai> writes:
> Thanks Richi. I am gonna merge it after Richard's final approve.
 
Thanks for checking, but no need to wait for a second ack from me!
Please go ahead and commit.
 
Richard
  

Patch

diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
index acf3642ceb2..3f735945e67 100644
--- a/gcc/tree-vect-loop-manip.cc
+++ b/gcc/tree-vect-loop-manip.cc
@@ -483,7 +483,7 @@  vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
 				 gimple_stmt_iterator loop_cond_gsi,
 				 rgroup_controls *rgc, tree niters,
 				 tree niters_skip, bool might_wrap_p,
-				 tree *iv_step)
+				 tree *iv_step, tree *compare_step)
 {
   tree compare_type = LOOP_VINFO_RGROUP_COMPARE_TYPE (loop_vinfo);
   tree iv_type = LOOP_VINFO_RGROUP_IV_TYPE (loop_vinfo);
@@ -538,9 +538,9 @@  vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
 	   ...
 	   vect__4.8_28 = .LEN_LOAD (_17, 32B, _36, 0);
 	   ...
-	   ivtmp_35 = ivtmp_9 - _36;
+	   ivtmp_35 = ivtmp_9 - POLY_INT_CST [4, 4];
 	   ...
-	   if (ivtmp_35 != 0)
+	   if (ivtmp_9 > POLY_INT_CST [4, 4])
 	     goto <bb 4>; [83.33%]
 	   else
 	     goto <bb 5>; [16.67%]
@@ -549,13 +549,15 @@  vect_set_loop_controls_directly (class loop *loop, loop_vec_info loop_vinfo,
       tree step = rgc->controls.length () == 1 ? rgc->controls[0]
 					       : make_ssa_name (iv_type);
       /* Create decrement IV.  */
-      create_iv (nitems_total, MINUS_EXPR, step, NULL_TREE, loop, &incr_gsi,
-		 insert_after, &index_before_incr, &index_after_incr);
+      create_iv (nitems_total, MINUS_EXPR, nitems_step, NULL_TREE, loop,
+		 &incr_gsi, insert_after, &index_before_incr,
+		 &index_after_incr);
       gimple_seq_add_stmt (header_seq, gimple_build_assign (step, MIN_EXPR,
 							    index_before_incr,
 							    nitems_step));
       *iv_step = step;
-      return index_after_incr;
+      *compare_step = nitems_step;
+      return index_before_incr;
     }
 
   /* Create increment IV.  */
@@ -825,6 +827,7 @@  vect_set_loop_condition_partial_vectors (class loop *loop,
      arbitrarily pick the last.  */
   tree test_ctrl = NULL_TREE;
   tree iv_step = NULL_TREE;
+  tree compare_step = NULL_TREE;
   rgroup_controls *rgc;
   rgroup_controls *iv_rgc = nullptr;
   unsigned int i;
@@ -861,7 +864,7 @@  vect_set_loop_condition_partial_vectors (class loop *loop,
 						 &preheader_seq, &header_seq,
 						 loop_cond_gsi, rgc, niters,
 						 niters_skip, might_wrap_p,
-						 &iv_step);
+						 &iv_step, &compare_step);
 
 	    iv_rgc = rgc;
 	  }
@@ -884,10 +887,21 @@  vect_set_loop_condition_partial_vectors (class loop *loop,
 
   /* Get a boolean result that tells us whether to iterate.  */
   edge exit_edge = single_exit (loop);
-  tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
-  tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
-  gcond *cond_stmt = gimple_build_cond (code, test_ctrl, zero_ctrl,
-					NULL_TREE, NULL_TREE);
+  gcond *cond_stmt;
+  if (LOOP_VINFO_USING_DECREMENTING_IV_P (loop_vinfo))
+    {
+      gcc_assert (compare_step);
+      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? LE_EXPR : GT_EXPR;
+      cond_stmt = gimple_build_cond (code, test_ctrl, compare_step, NULL_TREE,
+				     NULL_TREE);
+    }
+  else
+    {
+      tree_code code = (exit_edge->flags & EDGE_TRUE_VALUE) ? EQ_EXPR : NE_EXPR;
+      tree zero_ctrl = build_zero_cst (TREE_TYPE (test_ctrl));
+      cond_stmt
+	= gimple_build_cond (code, test_ctrl, zero_ctrl, NULL_TREE, NULL_TREE);
+    }
   gsi_insert_before (&loop_cond_gsi, cond_stmt, GSI_SAME_STMT);
 
   /* The loop iterates (NITERS - 1) / VF + 1 times.