Optimization Pragmas

This section describes the optimization pragmas.

• aggressive_hoisting
  Improves the number of variables that get hoisted out of a loop.
• c9x_alias_by_type
  Allows back-end optimizations to use alias type information.
• epilogue_helper_functions
  Controls size optimization for function termination instructions.
• fp_contract
  Controls the SPE additional fused multiply-add instructions codegen, when -spe2_vector is selected.
• fp_contract_aggressive
  Enables the peephole pattern to detect and convert the seperate instructions.
• ipa_rescopes_globals
  Rescopes the application global variables, that are only used in one function, to local static.
• peephole
  Controls the use peephole optimization.
• peephole_enable_16bit_load_store_inst
  Enables use of 16-bit load/store instructions instead of 32-bit load/store instructions.
• ppc_opt_defuse_mem_limit
  Controls memory consumed by compiler optimizations on the host computer.
• ppc_unroll_instructions_limit
  Limits number of instructions in an unrolled loop to value.
• ppc_unroll_speculative
  Controls speculative unrolling of counting loops which do not have fixed counts.
• processor
  Specifies the scheduling model used for instruction scheduling optimization.
• prologue_helper_functions
  Controls size optimization for function initialization instructions.
• remove_frsp_aggressive
  Improves code optimization by transforming Power Architecture LFS and FRSP instructions into the equivalent FMR instructions.
• schedule
  Specifies the use of instruction scheduling optimization.
• schedule z760
  Controls the rearrangement of instructions to reduce the effects of instruction latency.
• scheduling
  Specifies the scheduling model used for instruction scheduling optimization.
• switch_op
  Minimizes comparisions for "if" and "switch" statements.
• switch_cases_array
  
• volatileasm
  Controls the optimization of inline assembly statements.