3.17.4 AVR32 Options
These options are defined for AVR32 implementations:
-muse-rodata-section
- Use section .rodata for read-only data instead of .text.
-mhard-float
- Use floating point coprocessor instructions.
-msoft-float
- Use software floating-point library for floating-point operations.
-mforce-double-align
- Force double-word alignment for double-word memory accesses.
-masm-addr-pseudos
- Use assembler pseudo-instructions lda.w and call for handling direct
addresses. (Enabled by default)
-mno-init-got
- Do not initialize the GOT register before using it when compiling PIC
code.
-mrelax
- Let invoked assembler and linker do relaxing
(Enabled by default when optimization level is >1).
This means that when the address of symbols are known at link time,
the linker can optimize icall and mcall
instructions into a rcall instruction if possible.
Loading the address of a symbol can also be optimized.
-mmd-reorg-opt
- Perform machine dependent optimizations in reorg stage.
-mpart=
part- Generate code for the specified part. Permissible parts are:
ap7000,
ap7001,
ap7002,
ap7200,
uc3a0128,
uc3a0256,
uc3a0512,
uc3a0512es,
uc3a1128,
uc3a1256,
uc3a1512,
uc3a1512es,
uc3a3revd,
uc3a364,
uc3a364s,
uc3a3128,
uc3a3128s,
uc3a3256,
uc3a3256s,
uc3b064,
uc3b0128,
uc3b0256,
uc3b0256es,
uc3b0512,
uc3b0512revc,
uc3b164,
uc3b1128,
uc3b1256,
uc3b1256es,
uc3b1512
uc3b1512revc
uc3c0512crevc,
uc3c0256c,
uc3c0128c,
uc3c064c,
uc3c1512crevc,
uc3c1256c,
uc3c1128c,
uc3c164c,
uc3c2512crevc,
uc3c2256c,
uc3c2128c,
uc3c264c,
uc3l064,
uc3l032,
uc3l016,
uc3l064revb.
-mcpu=
cpu-type- Same as -mpart. Obsolete.
-march=
arch- Generate code for the specified architecture. Permissible architectures are:
ap, uc and ucr2.
-mfast-float
- Enable fast floating-point library that does not conform to IEEE-754 but is still good enough
for most applications. The fast floating-point library does not round to the nearest even
but away from zero. Enabled by default if the -funsafe-math-optimizations switch is specified.
-mimm-in-const-pool
- Put large immediates in constant pool. This is enabled by default for archs with insn-cache.