[committed] Fix bitvector_clear_bits

[Tom de Vries <tdevries@suse.de>]

Hi,

The commit 94e7891 "Reimplement seen flag of struct import_cu as bitvector"
introduces the function bitvector_clear_bits.

This function calculates the start element to be cleared, and the number of
elements, but then proceeds to use the number of elements as number of bytes
argument to the memset call.

Fix this by converting the memset argument from number of elements into number
of bytes.

This brings the improvement by the mentioned commit of 15% on vmlinux down
to 4%:
...
real:  mean:  16684.40 100.00%  stddev:  219.82
       mean:  15962.80  95.68%  stddev:  161.35
user:  mean:  16239.40 100.00%  stddev:   93.90
       mean:  15556.00  95.79%  stddev:  155.83
sys:   mean:    326.40 100.00%  stddev:   21.09
       mean:    313.60  96.08%  stddev:   33.66
...

Committed to trunk.

Thanks,
- Tom

Fix bitvector_clear_bits

2020-01-06  Tom de Vries  <tdevries@suse.de>

	* dwz.c (bitvector_clear_bits): Fix cleared range.

---
 dwz.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/dwz.c b/dwz.c
index 2f1f59f..33aea2a 100644
--- a/dwz.c
+++ b/dwz.c
@@ -6890,7 +6890,8 @@ bitvector_clear_bits (BITVECTOR_TYPE *vector, unsigned int a, unsigned int b)
 {
   unsigned int range_min = a / (sizeof (BITVECTOR_TYPE) * 8);
   unsigned int range_max = b / (sizeof (BITVECTOR_TYPE) * 8);
-  memset (&vector[range_min], 0, range_max - range_min + 1);
+  memset (&vector[range_min], 0,
+	  (range_max - range_min + 1) * sizeof (BITVECTOR_TYPE));
 }
 
 /* Function to optimize the size of DW_TAG_imported_unit DIEs by