[Patch GCC 5/Vect] Partial backport of r228751 (pr68238)

[Patch GCC 5/Vect] Partial backport of r228751 (pr68238)

[James Greenhalgh]

[-- Attachment #1: Type: text/plain, Size: 785 bytes --]


Hi,

As requested in the PR, this patch is a partial backport of r228751.

I can't claim any responsibility for it, but I did take it through the
paces on an aarch64-none-linux-gnu and x86_64-none-linux-gnu bootstrap/
test run and found no issues.

Applied as r230092 on gcc-5-branch (pre-approved in the PR) after checking
that it gives the right results for the code I derived the PR from.

I'll start a test cycle for a 4.9 backport.

Thanks,
James

---
2015-11-09  James Greenhalgh  <james.greenhalgh@arm.com>

	Partial backport from trunk r228751.
	PR tree-optimization/68238
	2015-10-13  Richard Biener  <rguenther@suse.de>

	* tree-vect-loop.c (vect_estimate_min_profitable_iters): Use
	LOOP_VINFO_COMP_ALIAS_DDRS to estimate alias versioning cost.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Patch-GCC-5-Vect-Partial-backport-of-r228751-pr68238.patch --]
[-- Type: text/x-patch;  name=0001-Patch-GCC-5-Vect-Partial-backport-of-r228751-pr68238.patch, Size: 648 bytes --]

diff --git a/gcc/tree-vect-loop.c b/gcc/tree-vect-loop.c
index 88ef251..05515b5 100644
--- a/gcc/tree-vect-loop.c
+++ b/gcc/tree-vect-loop.c
@@ -2825,7 +2825,7 @@ vect_estimate_min_profitable_iters (loop_vec_info loop_vinfo,
   if (LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
     {
       /*  FIXME: Make cost depend on complexity of individual check.  */
-      unsigned len = LOOP_VINFO_MAY_ALIAS_DDRS (loop_vinfo).length ();
+      unsigned len = LOOP_VINFO_COMP_ALIAS_DDRS (loop_vinfo).length ();
       (void) add_stmt_cost (target_cost_data, len, vector_stmt, NULL, 0,
 			    vect_prologue);
       dump_printf (MSG_NOTE,

Re: [Patch GCC 5/Vect] Partial backport of r228751 (pr68238)

[James Greenhalgh]

On Tue, Nov 10, 2015 at 10:11:30AM +0000, James Greenhalgh wrote:
> 
> Hi,
> 
> As requested in the PR, this patch is a partial backport of r228751.
> 
> I can't claim any responsibility for it, but I did take it through the
> paces on an aarch64-none-linux-gnu and x86_64-none-linux-gnu bootstrap/
> test run and found no issues.
> 
> Applied as r230092 on gcc-5-branch (pre-approved in the PR) after checking
> that it gives the right results for the code I derived the PR from.
> 
> I'll start a test cycle for a 4.9 backport.

Now done and committed as r230110, after a clean run on
aarch64-none-linux-gnu and x86_64-none-linux-gnu.

Thanks,
James

> 2015-11-10  James Greenhalgh  <james.greenhalgh@arm.com>
> 
> 	Partial backport from trunk r228751.
> 	PR tree-optimization/68238
> 	2015-10-13  Richard Biener  <rguenther@suse.de>
> 
> 	* tree-vect-loop.c (vect_estimate_min_profitable_iters): Use
> 	LOOP_VINFO_COMP_ALIAS_DDRS to estimate alias versioning cost.
> 

> diff --git a/gcc/tree-vect-loop.c b/gcc/tree-vect-loop.c
> index 88ef251..05515b5 100644
> --- a/gcc/tree-vect-loop.c
> +++ b/gcc/tree-vect-loop.c
> @@ -2825,7 +2825,7 @@ vect_estimate_min_profitable_iters (loop_vec_info loop_vinfo,
>    if (LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
>      {
>        /*  FIXME: Make cost depend on complexity of individual check.  */
> -      unsigned len = LOOP_VINFO_MAY_ALIAS_DDRS (loop_vinfo).length ();
> +      unsigned len = LOOP_VINFO_COMP_ALIAS_DDRS (loop_vinfo).length ();
>        (void) add_stmt_cost (target_cost_data, len, vector_stmt, NULL, 0,
>  			    vect_prologue);
>        dump_printf (MSG_NOTE,