GCC is smart enough to substitute the final result for FLS calculations
as implemented in the fallback C code we have in `__fls' and `fls'
applied to constant values. The presence of inline asm defeats the
compiler though, forcing it to emit extraneous CLZ/DCLZ calculation for
processors that support these instructions.
Use `__builtin_constant_p' then to avoid inline asm altogether for
constants.
Signed-off-by: Maciej W. Rozycki <macro@xxxxxxxxxxxxxx>
---
Ralf,
Even my good old trusty 4.1.2 can do it...
Maciej
linux-mips-const-fls.diff
Index: linux/arch/mips/include/asm/bitops.h
===================================================================
--- linux.orig/arch/mips/include/asm/bitops.h 2015-04-02 20:18:51.384515000 +0100
+++ linux/arch/mips/include/asm/bitops.h 2015-04-02 20:27:54.273191000 +0100
@@ -481,7 +481,7 @@ static inline unsigned long __fls(unsign
{
int num;
- if (BITS_PER_LONG == 32 &&
+ if (BITS_PER_LONG == 32 && !__builtin_constant_p(word) &&
__builtin_constant_p(cpu_has_clo_clz) && cpu_has_clo_clz) {
__asm__(
" .set push \n"
@@ -494,7 +494,7 @@ static inline unsigned long __fls(unsign
return 31 - num;
}
- if (BITS_PER_LONG == 64 &&
+ if (BITS_PER_LONG == 64 && !__builtin_constant_p(word) &&
__builtin_constant_p(cpu_has_mips64) && cpu_has_mips64) {
__asm__(
" .set push \n"
@@ -559,7 +559,8 @@ static inline int fls(int x)
{
int r;
- if (__builtin_constant_p(cpu_has_clo_clz) && cpu_has_clo_clz) {
+ if (!__builtin_constant_p(x) &&
+ __builtin_constant_p(cpu_has_clo_clz) && cpu_has_clo_clz) {
__asm__(
" .set push \n"
" .set "MIPS_ISA_LEVEL" \n"
