This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
On 03/02/2016 02:53 PM, Vladimir Makarov wrote:
I suspect you're right -- when I was looking at handling reload via splitting and calling back into IRA, cost updates were key to getting good code for the newly created pseudos. It wasn't a matter of getting them perfect, but a matter of getting them "reasonable".The overall effect of this patch seems to be fairly minimal in practice, which is slightly disappointing. On the whole, it seems to be a very slight net positive, with few instances where we generate additional instructions.I am just speculating may be the most important thing for performance is updating costs not their exact values (although the values can be important in some cases too).
jeff
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |