diff options
author | hubicka <hubicka@138bc75d-0d04-0410-961f-82ee72b054a4> | 2009-11-23 20:01:29 +0000 |
---|---|---|
committer | hubicka <hubicka@138bc75d-0d04-0410-961f-82ee72b054a4> | 2009-11-23 20:01:29 +0000 |
commit | 5fce5299af4810e08744160f9387d11ef1623a94 (patch) | |
tree | c9663c118c473b5fc50a9f3781ed3452e9f94056 /gcc/ipa-cp.c | |
parent | e74b076da6338645b4e2010f97ef50577b314d5f (diff) | |
download | gcc-5fce5299af4810e08744160f9387d11ef1623a94.tar.gz |
* ipa-cp.c (ipcp_compute_node_scale): Work around completely
wrong profile updates.
* predict.c (counts_to_freqs): Be expected for ENTRY/EXIT block
having largest frequency.
* ira-live.c (ira_implicitly_set_insn_hard_regs): Silecne
used uninitalized warning.
* tree-optimize.c (execute_fixup_cfg): Rescale entry and exit block
frequencies.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@154462 138bc75d-0d04-0410-961f-82ee72b054a4
Diffstat (limited to 'gcc/ipa-cp.c')
-rw-r--r-- | gcc/ipa-cp.c | 14 |
1 files changed, 13 insertions, 1 deletions
diff --git a/gcc/ipa-cp.c b/gcc/ipa-cp.c index 4b632c0d9fb..2af2c7f2d6f 100644 --- a/gcc/ipa-cp.c +++ b/gcc/ipa-cp.c @@ -578,7 +578,13 @@ build_const_val (struct ipcp_lattice *lat, tree tree_type) /* Compute the proper scale for NODE. It is the ratio between the number of direct calls (represented on the incoming cgraph_edges) and sum of all - invocations of NODE (represented as count in cgraph_node). */ + invocations of NODE (represented as count in cgraph_node). + + FIXME: This code is wrong. Since the callers can be also clones and + the clones are not scaled yet, the sums gets unrealistically high. + To properly compute the counts, we would need to do propagation across + callgraph (as external call to A might imply call to non-clonned B + if A's clone calls clonned B). */ static void ipcp_compute_node_scale (struct cgraph_node *node) { @@ -589,6 +595,12 @@ ipcp_compute_node_scale (struct cgraph_node *node) /* Compute sum of all counts of callers. */ for (cs = node->callers; cs != NULL; cs = cs->next_caller) sum += cs->count; + /* Work around the unrealistically high sum problem. We just don't want + the non-cloned body to have negative or very low frequency. Since + majority of execution time will be spent in clones anyway, this should + give good enough profile. */ + if (sum > node->count * 9 / 10) + sum = node->count * 9 / 10; if (node->count == 0) ipcp_set_node_scale (node, 0); else |