Re: [PATCH 06/13] range: Introduce range_inverse_array()

[Eric Auger <eric.auger@redhat.com>]

Hi Philippe,
On 9/4/23 10:18, Philippe Mathieu-Daudé wrote:
> On 4/9/23 10:03, Eric Auger wrote:
>> This helper reverses an array of regions, turning original
>> regions into holes and original holes into actual regions,
>> covering the whole UINT64_MAX span.
>>
>> Signed-off-by: Eric Auger <eric.auger@redhat.com>
>> ---
>>   include/qemu/range.h |  3 +++
>>   util/range.c         | 35 +++++++++++++++++++++++++++++++++++
>>   2 files changed, 38 insertions(+)
>>
>> diff --git a/include/qemu/range.h b/include/qemu/range.h
>> index 7e2b1cc447..fc1d3dabe6 100644
>> --- a/include/qemu/range.h
>> +++ b/include/qemu/range.h
>> @@ -219,4 +219,7 @@ static inline int ranges_overlap(uint64_t first1,
>> uint64_t len1,
>>     GList *range_list_insert(GList *list, Range *data);
>>   +void range_inverse_array(uint32_t nr_ranges, Range *ranges,
>> +                         uint32_t *nr_inv_ranges, Range **inv_ranges);
>> +
>>   #endif
>> diff --git a/util/range.c b/util/range.c
>> index 098d9d2dc0..11c4ff0b78 100644
>> --- a/util/range.c
>> +++ b/util/range.c
>> @@ -70,3 +70,38 @@ GList *range_list_insert(GList *list, Range *data)
>>         return list;
>>   }
>> +
>> +/*
>> + * Inverse an array of sorted ranges over the UINT64_MAX span, ie.
>> + * original ranges becomes holes in the newly allocated inv_ranges
>> + */
>
> Most of the functions are described in the header; could you move this
> description with the declaration?
this is the case for all static inline primitives but not for
range_list_insert(), hence that choice. Now I don't have a strong opinion.
>
>> +void range_inverse_array(uint32_t nr_ranges, Range *ranges,
>> +                         uint32_t *nr_inv_ranges, Range **inv_ranges)
>> +{
>> +    Range *resv;
>> +    int i = 0, j = 0;
>> +
>> +    resv = g_malloc0_n(nr_ranges + 1, sizeof(Range));
>> +
>> +    /* first range lob is greater than 0, insert a first range */
>> +    if (range_lob(&ranges[0]) > 0) {
>> +        range_set_bounds(&resv[i++], 0,
>> +                         range_lob(&ranges[0]) - 1);
>> +    }
>> +
>> +    /* insert a range inbetween each original range */
>> +    for (; j < nr_ranges - 1; j++) {
>> +        if (range_compare(&ranges[j], &ranges[j + 1])) {
>> +            range_set_bounds(&resv[i++], range_upb(&ranges[j]) + 1,
>> +                             range_lob(&ranges[j + 1]) - 1);
>> +        }
>> +    }
>> +    /* last range upb is less than UINT64_MAX, insert a last range */
>
> In order to use this new function with variable range sizes,
> can we pass UINT64_MAX as an 'inv_range_upb' argument?
Indeed I hesitated to bring enhanced comodity by letting the caller pass
the upper bound and allow values less than UINT64_MAX. But I was afraid
this would complexify the implementation, hence the current choice. I
will have a look & see.

Thanks

Eric
>
>> +    if (range_upb(&ranges[nr_ranges - 1]) <  UINT64_MAX) {
>> +        range_set_bounds(&resv[i++],
>> +                          range_upb(&ranges[nr_ranges - 1]) + 1,
>> UINT64_MAX);
>> +    }
>> +    *nr_inv_ranges = i;
>> +    resv = g_realloc(resv, i * sizeof(Range));
>> +    *inv_ranges = resv;
>> +}
>