aboutsummaryrefslogtreecommitdiff
path: root/gcc/flow.c
diff options
context:
space:
mode:
authorRichard Earnshaw <erich@gnu.org>1997-07-16 14:54:40 +0000
committerRichard Earnshaw <erich@gnu.org>1997-07-16 14:54:40 +0000
commitb111229af91d7ab08f334dfb16c55b4266e14a28 (patch)
tree7f6d9336ac40efa03f7d2ccada1f3684896e60d0 /gcc/flow.c
parentd23f41585d5cc3aae5a97ad2f5933d0ef653c312 (diff)
downloadgcc-b111229af91d7ab08f334dfb16c55b4266e14a28.zip
gcc-b111229af91d7ab08f334dfb16c55b4266e14a28.tar.gz
gcc-b111229af91d7ab08f334dfb16c55b4266e14a28.tar.bz2
(tune_flags): New variable.
(target_{cpu,fpe}_name): Delete. (arm_fpu_arch): New variable. (arm_select): Also allow -march=... to configure just the architecture. (all_procs): Allow armv{2,2a,3,3m,4,4t} for use with -march=. (arm_override_options): Handle -march=, but don't let -mcpu= and -mtune= match the architecture names, since we can only tune for an implementation. Rework selection of tuning options for floating point. (use_return_insn): Support interworking with Thumb code. (arm_rtx_costs): Rework multiply costs so that cost is based on the tune, not the architecture. (f_register_operand): New function. (output_return_instruction): Support interworking with Thumb code. (output_func_epilogue): Support interworking with Thumb code. Remove redundant calculation of code_size. Use floating-point load-multiples if permitted. (emit_sfm): New function. (arm_expand_prologue): Use floating-point store-multiples if permitted. From-SVN: r14457
Diffstat (limited to 'gcc/flow.c')
0 files changed, 0 insertions, 0 deletions