This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Switch statement optimization
- From: Christian Hildner <christian dot hildner at hob dot de>
- To: gcc at gnu dot org
- Date: Mon, 10 Jul 2006 11:38:53 +0200
- Subject: Switch statement optimization
Hi,
while dealing with autogenerated code I found that GCC often outputs compare trees instead of branch tables for switch statements. I found that GCC uses the density of case labels within their spanned range as the key criterion. If the density of labels is smaller than 10% (33% for size optimization) then GCC uses compare instruction rather than a branch table. What I am wondering is why no real cost estimation is done here. In my case there are around 200 to 1500 labels within the 16K range, so they are completely resolved to compares. In contrast the Intel compiler outputs branch tables more likely.
The question now is whether the factors 10 (3 for size opt) have been found using any standard benchmark. And shouldn't at least the factor be optimized for a standard benchmark (like the upcoming SPEC CPU 2006 INT) if this simple linear method is used for modeling a logarithmic behavior? Or should a cost estimation be introduced?
For size optimization there could be a quite accurate architecture dependent size estimation.
Opinions, Comments?
Christian
PS: Please CC me.