On 11/19/21 13:13, Richard Biener wrote:
On November 19, 2021 4:00:01 PM GMT+01:00, Andrew MacLeod
wrote:
On 11/19/21 04:21, Richard Biener wrote:
On Thu, Nov 18, 2021 at 8:30 PM Andrew MacLeod via Gcc-patches
wrote:
At issue here is the dynamic approach we currently use for outgoing edge
c
On November 19, 2021 4:00:01 PM GMT+01:00, Andrew MacLeod
wrote:
>On 11/19/21 04:21, Richard Biener wrote:
>> On Thu, Nov 18, 2021 at 8:30 PM Andrew MacLeod via Gcc-patches
>> wrote:
>>> At issue here is the dynamic approach we currently use for outgoing edge
>>> calculations. It isn't normally
On 11/19/21 04:21, Richard Biener wrote:
On Thu, Nov 18, 2021 at 8:30 PM Andrew MacLeod via Gcc-patches
wrote:
At issue here is the dynamic approach we currently use for outgoing edge
calculations. It isn't normally a problem, but once you get a very
large number of possible outgoing values (i
On 11/19/21 04:21, Richard Biener wrote:
On Thu, Nov 18, 2021 at 8:30 PM Andrew MacLeod via Gcc-patches
wrote:
At issue here is the dynamic approach we currently use for outgoing edge
calculations. It isn't normally a problem, but once you get a very
large number of possible outgoing values (i
On 11/19/21 05:15, Aldy Hernandez wrote:
On 11/18/21 8:28 PM, Andrew MacLeod wrote:
@@ -376,6 +366,14 @@ range_def_chain::get_def_chain (tree name)
return NULL;
}
+ // Terminate the def chains if we see too many cascading stmts.
+ if (m_logical_depth == param_ranger_logical_dep
On 11/18/21 8:28 PM, Andrew MacLeod wrote:
@@ -376,6 +366,14 @@ range_def_chain::get_def_chain (tree name)
return NULL;
}
+ // Terminate the def chains if we see too many cascading stmts.
+ if (m_logical_depth == param_ranger_logical_depth)
+return NULL;
+
+ // Increase
On Thu, Nov 18, 2021 at 8:30 PM Andrew MacLeod via Gcc-patches
wrote:
>
> At issue here is the dynamic approach we currently use for outgoing edge
> calculations. It isn't normally a problem, but once you get a very
> large number of possible outgoing values (ie very long unrolled blocks)
> with
At issue here is the dynamic approach we currently use for outgoing edge
calculations. It isn't normally a problem, but once you get a very
large number of possible outgoing values (ie very long unrolled blocks)
with pairs of values on a statement, and individual queries for each
one, it becom
