summaryrefslogtreecommitdiffstats
path: root/ld/NEWS
diff options
context:
space:
mode:
authorJoern Rennecke <joern.rennecke@embecosm.com>2004-05-19 14:15:55 +0000
committerJoern Rennecke <joern.rennecke@embecosm.com>2004-05-19 14:15:55 +0000
commit35835446e4115755fc48a367d58ea57f1b88dd1e (patch)
treefcd06a51d247a81c7afc8060349263ca3ea57815 /ld/NEWS
parent0841712ea90a7633d54bbc1f00fa82e068815b66 (diff)
downloadppe42-binutils-35835446e4115755fc48a367d58ea57f1b88dd1e.tar.gz
ppe42-binutils-35835446e4115755fc48a367d58ea57f1b88dd1e.zip
* NEWS: Mention new linker map file generation and the
--reduce-memory-overheads option. * ld.texinfo: Document --reduce-memory-overheads option. * ld.h (map_symbol_def): New struct. (struct user_section_struct, section_userdata_type): Rename to: (struct lean_user_section_struct, lean_section_userdata_type). (struct fat_user_section_struct, fat_section_userdata_type): New. (SECTION_USERDATA_SIZE): Define. (args_type): New member reduce_memory_overheads. * ldlang.c (map_obstack): New static variable. (init_map_userdata, print_all_symbols, sort_def_symbol): New functions. (lang_map): Unless command_line.reduce_memory_overheads is set, initialize lists of defined symbols for each section. (print_input_section): Unless command_line.reduce_memory_overheads is set, use print_all_symbols. (init_os): Use lean_section_userdata_type / SECTION_USERDATA_SIZE. * ldmain.c (main): Initialize command_line.reduce_memory_overheads. * lexsup.c (enum option_values): Add OPTION_REDUCE_MEMORY_OVERHEADS. (ld_options): Add entry for --reduce-memory-overheads. (parse_args): Handle OPTION_REDUCE_MEMORY_OVERHEADS.
Diffstat (limited to 'ld/NEWS')
-rw-r--r--ld/NEWS6
1 files changed, 6 insertions, 0 deletions
diff --git a/ld/NEWS b/ld/NEWS
index cec229be64..a33c6dc70c 100644
--- a/ld/NEWS
+++ b/ld/NEWS
@@ -1,5 +1,11 @@
-*- text -*-
+* Linker map files are now generated with an O(N) algorithm for finding symbols
+ that are defined in each section. This uses about 40% more memory for
+ symbols than the old O(N^2) algorithm. You can use the new
+ --reduce-memory-overheads option to select the old algorithm; this option
+ might also be used in the future to select similar tradeoffs.
+
* New PE --large-address-aware option to indicate executables support virtual
addresses greater than 2 gigabytes.
OpenPOWER on IntegriCloud