Some aspect of GCC memory consumption increased by recent patch

[Jan Hubicka]

Hi,
the increase in:

> 
> comparing Gerald's testcase PR8361 compilation at -O1 level:
> Peak amount of GGC memory allocated before garbage collecting run decreased from 97174k to 88921k, overall -9.28%
> Amount of produced GGC garbage increased from 545290k to 671734k, overall 23.19%
> Amount of memory still referenced at the end of compilation decreased from 63318k to 60153k, overall -5.26%
> Overall memory needed: 0k -> 2666k
> Peak memory use before GGC: 97174k -> 88921k
> Peak memory use after GGC: 90561k -> 88039k
> Maximum of released memory in single GGC run: 19488k -> 19486k
> Garbage: 545290k -> 671734k
> Leak: 63318k -> 60153k
> Overhead: 75450k -> 144644k
> GGC runs: 763 -> 841
> 
> comparing Gerald's testcase PR8361 compilation at -O2 level:
> Peak amount of GGC memory allocated before garbage collecting run decreased from 97175k to 88922k, overall -9.28%
> Amount of produced GGC garbage increased from 600955k to 726419k, overall 20.88%
> Amount of memory still referenced at the end of compilation decreased from 63940k to 60762k, overall -5.23%
> Overall memory needed: 0k -> 2666k
> Peak memory use before GGC: 97175k -> 88922k
> Peak memory use after GGC: 90562k -> 88039k
> Maximum of released memory in single GGC run: 19489k -> 19487k
> Garbage: 600955k -> 726419k
> Leak: 63940k -> 60762k
> Overhead: 93911k -> 166904k
> GGC runs: 821 -> 876
> 
> comparing Gerald's testcase PR8361 compilation at -O3 level:
> Amount of produced GGC garbage increased from 616982k to 770533k, overall 24.89%
> Amount of memory still referenced at the end of compilation decreased from 64593k to 61369k, overall -5.25%
> Overall memory needed: 0k -> 2666k
> Peak memory use before GGC: 92864k -> 90120k
> Peak memory use after GGC: 91396k -> 89108k
> Maximum of released memory in single GGC run: 20194k -> 19982k
> Garbage: 616982k -> 770533k
> Leak: 64593k -> 61369k
> Overhead: 95868k -> 181353k
> GGC runs: 803 -> 866

Seems to come mostly from increased memory usage of VARRAYS.
block_defs varray now consume 19MB instead of 7MB.  Overall there is
118MB instead of 105MB.  We now produce 547817 instead of 233466 SSA
names.  Can this be caused by ...
> +
> +2004-09-10  Diego Novillo  <dnovillo@redhat.com>
> +
> +	Revert
> +
> +	2004-09-08  Jie Zhang  <zhangjie@magima.com.cn>
> +
> +	    * tree-ssa-alias.c (compute_flow_insensitive_aliasing): If type
> +	    memory tag is call clobbered, so are its aliases.
> +	    (group_aliases): When two memory tags being grouped, if one is
> +	    call clobbered, so are the other and its aliases.
> +	    (add_may_alias): Remove call-clobbering stuff.
> +	    (replace_may_alias): Likewise.
> +	    (merge_pointed_to_info): Merge pt_global_mem


Honza