libstdc++: Adjust probabilities of hashmap loop conditions

We are currently generating a loop which has more comparisons than you'd
typically need as the probablities on the small size loop are such that it
assumes the likely case is that an element is not found.

This again generates a pattern that's harder for branch predictors to follow,
but also just generates more instructions for the what one could say is the
typical case: That your hashtable contains the entry you are looking for.

This patch adds a __builtin_expect in _M_find_before_node where at the moment
the loop is optimized for the case where we don't do any iterations.

A simple testcase is (compiled with -fno-split-path to simulate the loop
in libstdc++):

#include <stdbool.h>

bool foo (int **a, int n, int val, int *tkn)
{
    for (int i = 0; i < n; i++)
    {
        if (!a[i] || a[i]==tkn)
          return false;

        if (*a[i] == val)
          return true;
    }
}

which generataes:

foo:
        cmp     w1, 0
        ble     .L1
        add     x1, x0, w1, uxtw 3
        b       .L4
.L9:
        ldr     w4, [x4]
        cmp     w4, w2
        beq     .L6
        cmp     x0, x1
        beq     .L1
.L4:
        ldr     x4, [x0]
        add     x0, x0, 8
        cmp     x4, 0
        ccmp    x4, x3, 4, ne
        bne     .L9
        mov     w0, 0
.L1:
        ret
.L6:
        mov     w0, 1
        ret

i.e. BB rotation makes is generate an unconditional branch to a conditional
branch. However this method is only called when the size is above a certain
threshold, and so it's likely that we have to do that first iteration.

Adding:

#include <stdbool.h>

bool foo (int **a, int n, int val, int *tkn)
{
    for (int i = 0; i < n; i++)
    {
        if (__builtin_expect(!a[i] || a[i]==tkn, 0))
          return false;

        if (*a[i] == val)
          return true;
    }
}

to indicate that we will likely do an iteration more generates:

foo:
        cmp     w1, 0
        ble     .L1
        add     x1, x0, w1, uxtw 3
.L4:
        ldr     x4, [x0]
        add     x0, x0, 8
        cmp     x4, 0
        ccmp    x4, x3, 4, ne
        beq     .L5
        ldr     w4, [x4]
        cmp     w4, w2
        beq     .L6
        cmp     x0, x1
        bne     .L4
.L1:
        ret
.L5:
        mov     w0, 0
        ret
.L6:
        mov     w0, 1
        ret

which results in ~0-10% extra on top of the previous patch.

In table form:

+-------------+---------------+-------+--------------------+-------------------+-----------------+
| benchmark   | Type          | Size  | Inline vs baseline | final vs baseline | final vs inline |
+-------------+---------------+-------+--------------------+-------------------+-----------------+
| find many   | uint64_t      | 11253 | -15.67%            | -22.96%           | -8.65%          |
| find many   | uint64_t      | 11253 | -16.74%            | -23.37%           | -7.96%          |
| find single | uint64_t      | 345   | -5.88%             | -11.54%           | -6.02%          |
| find many   | string        | 11253 | -4.50%             | -9.56%            | -5.29%          |
| find single | uint64_t      | 345   | -4.38%             | -9.41%            | -5.26%          |
| find single | shared string | 11253 | -6.67%             | -11.00%           | -4.64%          |
| find single | shared string | 11253 | -4.63%             | -9.03%            | -4.61%          |
| find single | shared string | 345   | -10.41%            | -14.44%           | -4.50%          |
| find many   | string        | 11253 | -3.41%             | -7.51%            | -4.24%          |
| find many   | shared string | 11253 | -2.30%             | -5.72%            | -3.50%          |
| find many   | string        | 13    | 2.86%              | -0.30%            | -3.07%          |
| find single | string        | 11253 | 4.47%              | 1.34%             | -3.00%          |
| find many   | custom string | 11253 | 0.25%              | -2.75%            | -2.99%          |
| find single | uint64_t      | 345   | 2.99%              | 0.01%             | -2.90%          |
| find single | shared string | 345   | -11.53%            | -13.67%           | -2.41%          |
| find single | uint64_t      | 11253 | 0.49%              | -1.59%            | -2.07%          |
+-------------+---------------+-------+--------------------+-------------------+-----------------+

libstdc++-v3/ChangeLog:

	* include/bits/hashtable.h
	(_M_find_before_node): Make it likely that the map has at least one
	entry and so we do at least one iteration.

1 files changed, 2 insertions, 2 deletions

diff --git a/libstdc++-v3/include/bits/hashtable.h b/libstdc++-v3/include/bits/hashtable.h
index cd60fb5..5792cc5 100644
--- a/libstdc++-v3/include/bits/hashtable.h
+++ b/libstdc++-v3/include/bits/hashtable.h
@@ -2177,7 +2177,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
 	  if (this->_M_equals(__k, __code, *__p))
 	    return __prev_p;
 
-	  if (!__p->_M_nxt || _M_bucket_index(*__p->_M_next()) != __bkt)
+	  if (__builtin_expect (!__p->_M_nxt || _M_bucket_index(*__p->_M_next()) != __bkt, 0))
 	    break;
 	  __prev_p = __p;
 	}
@@ -2207,7 +2207,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
 	    if (this->_M_equals_tr(__k, __code, *__p))
 	      return __prev_p;
 
-	    if (!__p->_M_nxt || _M_bucket_index(*__p->_M_next()) != __bkt)
+	    if (__builtin_expect (!__p->_M_nxt || _M_bucket_index(*__p->_M_next()) != __bkt, 0))
 	      break;
 	    __prev_p = __p;
 	  }