Re: RFA: Small PATCH to add pow2p_hwi to hwint.h

[Jeff Law]

On 09/08/2016 02:59 PM, Jason Merrill wrote:

On Thu, Sep 8, 2016 at 11:55 AM, Joseph Myers  wrote:

> On Thu, 8 Sep 2016, Jason Merrill wrote:
>

>> Various places in GCC use negate, bit-and and compare to test whether
>> an integer is a power of 2, but I think it would be clearer for this
>> test to be wrapped in a function.

>
> (x & -x) == x is also true for 0.  Whatever the correct function semantics
> for 0, the comment needs to reflect them.

Yep, I was just realizing that, too.  This much larger patch introduces:

 least_bit_hwi: (x & -x)
 pow2_or_zerop: (x & -x) == x
 pow2p_hwi: x && x == (x & -x)
 ctz_or_zero: floor_log2 (x & -x)

and replaces these patterns accordingly.  I'm not at all attached to the names.

Tested x86_64-pc-linux-gnu.


hwint.diff


commit 4b76501b72f3953ac57cb077b4e25a90afb6d9a9
Author: Jason Merrill 
Date:   Thu Sep 8 13:10:12 2016 -0400

Add inline functions for various bitwise operations.

* hwint.h (least_bit_hwi, pow2_or_zerop, pow2p_hwi, ctz_or_zero):
New.
* hwint.c (exact_log2): Use pow2p_hwi.
(ctz_hwi, ffs_hwi): Use least_bit_hwi.
* alias.c (memrefs_conflict_p): Use pow2_or_zerop.
* builtins.c (get_object_alignment_2, get_object_alignment)
(get_pointer_alignment, fold_builtin_atomic_always_lock_free): Use
least_bit_hwi.
* calls.c (compute_argument_addresses, store_one_arg): Use
least_bit_hwi.
* cfgexpand.c (expand_one_stack_var_at): Use least_bit_hwi.
* combine.c (force_to_mode): Use least_bit_hwi.
* emit-rtl.c (set_mem_attributes_minus_bitpos, adjust_address_1):
Use least_bit_hwi.
* expmed.c (synth_mult, expand_divmod): Use ctz_or_zero, ctz_hwi.
(init_expmed_one_conv): Use pow2p_hwi.
* fold-const.c (round_up_loc, round_down_loc): Use pow2_or_zerop.
(fold_binary_loc): Use pow2p_hwi.
* function.c (assign_parm_find_stack_rtl): Use least_bit_hwi.
* gimple-fold.c (gimple_fold_builtin_memory_op): Use pow2p_hwi.
* gimple-ssa-strength-reduction.c (replace_ref): Use least_bit_hwi.
* hsa-gen.c (gen_hsa_addr_with_align, hsa_bitmemref_alignment):
Use least_bit_hwi.
* ipa-cp.c (ipcp_alignment_lattice::meet_with_1): Use least_bit_hwi.
* ipa-prop.c (ipa_modify_call_arguments): Use least_bit_hwi.
* omp-low.c (oacc_loop_fixed_partitions)
(oacc_loop_auto_partitions): Use least_bit_hwi.
* rtlanal.c (nonzero_bits1): Use ctz_or_zero.
* stor-layout.c (place_field): Use least_bit_hwi.
* tree-pretty-print.c (dump_generic_node): Use pow2p_hwi.
* tree-sra.c (build_ref_for_offset): Use least_bit_hwi.
* tree-ssa-ccp.c (ccp_finalize): Use least_bit_hwi.
* tree-ssa-math-opts.c (bswap_replace): Use least_bit_hwi.
* tree-ssa-strlen.c (handle_builtin_memcmp): Use pow2p_hwi.
* tree-vect-data-refs.c (vect_analyze_group_access_1)
(vect_grouped_store_supported, vect_grouped_load_supported)
(vect_permute_load_chain, vect_shift_permute_load_chain)
(vect_transform_grouped_load): Use pow2p_hwi.
* tree-vect-generic.c (expand_vector_divmod): Use ctz_or_zero.
* tree-vect-patterns.c (vect_recog_divmod_pattern): Use ctz_or_zero.
* tree-vect-stmts.c (vectorizable_mask_load_store): Use
least_bit_hwi.
* tsan.c (instrument_expr): Use least_bit_hwi.
* var-tracking.c (negative_power_of_two_p): Use pow2_or_zerop.
I was briefly worried about some of the expmed changes, but managed to 
convince myself they were correct.


Ok for the trunk.

Thanks,
jeff

Re: RFA: Small PATCH to add pow2p_hwi to hwint.h

[Jason Merrill]

On Thu, Sep 8, 2016 at 11:55 AM, Joseph Myers  wrote:
> On Thu, 8 Sep 2016, Jason Merrill wrote:
>
>> Various places in GCC use negate, bit-and and compare to test whether
>> an integer is a power of 2, but I think it would be clearer for this
>> test to be wrapped in a function.
>
> (x & -x) == x is also true for 0.  Whatever the correct function semantics
> for 0, the comment needs to reflect them.

Yep, I was just realizing that, too.  This much larger patch introduces:

 least_bit_hwi: (x & -x)
 pow2_or_zerop: (x & -x) == x
 pow2p_hwi: x && x == (x & -x)
 ctz_or_zero: floor_log2 (x & -x)

and replaces these patterns accordingly.  I'm not at all attached to the names.

Tested x86_64-pc-linux-gnu.
commit 4b76501b72f3953ac57cb077b4e25a90afb6d9a9
Author: Jason Merrill 
Date:   Thu Sep 8 13:10:12 2016 -0400

Add inline functions for various bitwise operations.

* hwint.h (least_bit_hwi, pow2_or_zerop, pow2p_hwi, ctz_or_zero):
New.
* hwint.c (exact_log2): Use pow2p_hwi.
(ctz_hwi, ffs_hwi): Use least_bit_hwi.
* alias.c (memrefs_conflict_p): Use pow2_or_zerop.
* builtins.c (get_object_alignment_2, get_object_alignment)
(get_pointer_alignment, fold_builtin_atomic_always_lock_free): Use
least_bit_hwi.
* calls.c (compute_argument_addresses, store_one_arg): Use
least_bit_hwi.
* cfgexpand.c (expand_one_stack_var_at): Use least_bit_hwi.
* combine.c (force_to_mode): Use least_bit_hwi.
* emit-rtl.c (set_mem_attributes_minus_bitpos, adjust_address_1):
Use least_bit_hwi.
* expmed.c (synth_mult, expand_divmod): Use ctz_or_zero, ctz_hwi.
(init_expmed_one_conv): Use pow2p_hwi.
* fold-const.c (round_up_loc, round_down_loc): Use pow2_or_zerop.
(fold_binary_loc): Use pow2p_hwi.
* function.c (assign_parm_find_stack_rtl): Use least_bit_hwi.
* gimple-fold.c (gimple_fold_builtin_memory_op): Use pow2p_hwi.
* gimple-ssa-strength-reduction.c (replace_ref): Use least_bit_hwi.
* hsa-gen.c (gen_hsa_addr_with_align, hsa_bitmemref_alignment):
Use least_bit_hwi.
* ipa-cp.c (ipcp_alignment_lattice::meet_with_1): Use least_bit_hwi.
* ipa-prop.c (ipa_modify_call_arguments): Use least_bit_hwi.
* omp-low.c (oacc_loop_fixed_partitions)
(oacc_loop_auto_partitions): Use least_bit_hwi.
* rtlanal.c (nonzero_bits1): Use ctz_or_zero.
* stor-layout.c (place_field): Use least_bit_hwi.
* tree-pretty-print.c (dump_generic_node): Use pow2p_hwi.
* tree-sra.c (build_ref_for_offset): Use least_bit_hwi.
* tree-ssa-ccp.c (ccp_finalize): Use least_bit_hwi.
* tree-ssa-math-opts.c (bswap_replace): Use least_bit_hwi.
* tree-ssa-strlen.c (handle_builtin_memcmp): Use pow2p_hwi.
* tree-vect-data-refs.c (vect_analyze_group_access_1)
(vect_grouped_store_supported, vect_grouped_load_supported)
(vect_permute_load_chain, vect_shift_permute_load_chain)
(vect_transform_grouped_load): Use pow2p_hwi.
* tree-vect-generic.c (expand_vector_divmod): Use ctz_or_zero.
* tree-vect-patterns.c (vect_recog_divmod_pattern): Use ctz_or_zero.
* tree-vect-stmts.c (vectorizable_mask_load_store): Use
least_bit_hwi.
* tsan.c (instrument_expr): Use least_bit_hwi.
* var-tracking.c (negative_power_of_two_p): Use pow2_or_zerop.

diff --git a/gcc/alias.c b/gcc/alias.c
index f4b5a92..277125e 100644
--- a/gcc/alias.c
+++ b/gcc/alias.c
@@ -2534,7 +2534,7 @@ memrefs_conflict_p (int xsize, rtx x, int ysize, rtx y, 
HOST_WIDE_INT c)
 {
   HOST_WIDE_INT sc = INTVAL (XEXP (x, 1));
   unsigned HOST_WIDE_INT uc = sc;
-  if (sc < 0 && -uc == (uc & -uc))
+  if (sc < 0 && pow2_or_zerop (-uc))
{
  if (xsize > 0)
xsize = -xsize;
@@ -2549,7 +2549,7 @@ memrefs_conflict_p (int xsize, rtx x, int ysize, rtx y, 
HOST_WIDE_INT c)
 {
   HOST_WIDE_INT sc = INTVAL (XEXP (y, 1));
   unsigned HOST_WIDE_INT uc = sc;
-  if (sc < 0 && -uc == (uc & -uc))
+  if (sc < 0 && pow2_or_zerop (-uc))
{
  if (ysize > 0)
ysize = -ysize;
diff --git a/gcc/builtins.c b/gcc/builtins.c
index 1073e35..0ccba15 100644
--- a/gcc/builtins.c
+++ b/gcc/builtins.c
@@ -305,7 +305,7 @@ get_object_alignment_2 (tree exp, unsigned int *alignp,
{
  ptr_bitmask = TREE_INT_CST_LOW (TREE_OPERAND (addr, 1));
  ptr_bitmask *= BITS_PER_UNIT;
- align = ptr_bitmask & -ptr_bitmask;
+ align = least_bit_hwi (ptr_bitmask);
  addr = TREE_OPERAND (addr, 0);
}
 
@@ -325,7 +325,7 @@ get_object_alignment_2 (tree exp, unsigned int *alignp,
  unsigned HOST_WIDE_INT step = 1;
  

Re: RFA: Small PATCH to add pow2p_hwi to hwint.h

[Joseph Myers]

On Thu, 8 Sep 2016, Jason Merrill wrote:

> Various places in GCC use negate, bit-and and compare to test whether
> an integer is a power of 2, but I think it would be clearer for this
> test to be wrapped in a function.

(x & -x) == x is also true for 0.  Whatever the correct function semantics 
for 0, the comment needs to reflect them.

-- 
Joseph S. Myers
jos...@codesourcery.com

Re: RFA: Small PATCH to add pow2p_hwi to hwint.h

[Jakub Jelinek]

On Thu, Sep 08, 2016 at 08:53:26AM -0600, Jeff Law wrote:
> On 09/07/2016 11:56 PM, Jason Merrill wrote:
> > Various places in GCC use negate, bit-and and compare to test whether
> > an integer is a power of 2, but I think it would be clearer for this
> > test to be wrapped in a function.
> > 
> > OK for trunk?
> > 
> I think the canonical way we've written that is
> 
> exact_log2 (x) != -1

We have integer_pow2p which works on trees (and is actually implemented by
testing if popcnt is 1 on the wide-int).
Calling exact_log2 for such a test if you don't need the value is a waste of
compile time, ctz_hwi doesn't need to be computed.  And, exact_log2 is -1
even on 0, which is not pow2p_hwi.

I think having pow2p_hwi would be nice, if we really start using it where
appropriate.

Jakub

Re: RFA: Small PATCH to add pow2p_hwi to hwint.h

[Jeff Law]

On 09/07/2016 11:56 PM, Jason Merrill wrote:

Various places in GCC use negate, bit-and and compare to test whether
an integer is a power of 2, but I think it would be clearer for this
test to be wrapped in a function.

OK for trunk?


I think the canonical way we've written that is

exact_log2 (x) != -1


Though you could argue that's far from obvious.

Presumably you have a follow-up to start using pow2p_hwi?

jeff