Re: Increase inlining limits for inline functions with builtin_constant_p on parameter

[Martin Liška]

On 10/21/20 5:11 PM, Jan Hubicka wrote:

return ( __builtin_constant_p((size) - 1) ? ( __builtin_constant_p((size) - 1)
? ( ((size) - 1) < 2 ? 0 : ((size) - 1) & (1ULL << 63) ? 63 : ((size) - 1) &
(1ULL << 62) ? 62 : ((size) - 1) & (1ULL << 61) ? 61 : ((size) - 1) & (1ULL <<
60) ? 60 : ((size) - 1) & (1ULL << 59) ? 59 : ((size) - 1) & (1ULL << 58) ? 58
: ((size) - 1) & (1ULL << 57) ? 57 : ((size) - 1) & (1ULL << 56) ? 56 : ((size)
- 1) & (1ULL << 55) ? 55 : ((size) - 1) & (1ULL << 54) ? 54 : ((size) - 1) &
(1ULL << 53) ? 53 : ((size) - 1) & (1ULL << 52) ? 52 : ((size) - 1) & (1ULL <<
51) ? 51 : ((size) - 1) & (1ULL << 50) ? 50 : ((size) - 1) & (1ULL << 49) ? 49
: ((size) - 1) & (1ULL << 48) ? 48 : ((size) - 1) & (1ULL << 47) ? 47 : ((size)
- 1) & (1ULL << 46) ? 46 : ((size) - 1) & (1ULL << 45) ? 45 : ((size) - 1) &
(1ULL << 44) ? 44 : ((size) - 1) & (1ULL << 43) ? 43 : ((size) - 1) & (1ULL <<
42) ? 42 : ((size) - 1) & (1ULL << 41) ? 41 : ((size) - 1) & (1ULL << 40) ? 40
: ((size) - 1) & (1ULL << 39) ? 39 : ((size) - 1) & (1ULL << 38) ? 38 : ((size)
- 1) & (1ULL << 37) ? 37 : ((size) - 1) & (1ULL << 36) ? 36 : ((size) - 1) &
(1ULL << 35) ? 35 : ((size) - 1) & (1ULL << 34) ? 34 : ((size) - 1) & (1ULL <<
33) ? 33 : ((size) - 1) & (1ULL << 32) ? 32 : ((size) - 1) & (1ULL << 31) ? 31
: ((size) - 1) & (1ULL << 30) ? 30 : ((size) - 1) & (1ULL << 29) ? 29 : ((size)
- 1) & (1ULL << 28) ? 28 : ((size) - 1) & (1ULL << 27) ? 27 : ((size) - 1) &
(1ULL << 26) ? 26 : ((size) - 1) & (1ULL << 25) ? 25 : ((size) - 1) & (1ULL <<
24) ? 24 : ((size) - 1) & (1ULL << 23) ? 23 : ((size) - 1) & (1ULL << 22) ? 22
: ((size) - 1) & (1ULL << 21) ? 21 : ((size) - 1) & (1ULL << 20) ? 20 : ((size)
- 1) & (1ULL << 19) ? 19 : ((size) - 1) & (1ULL << 18) ? 18 : ((size) - 1) &
(1ULL << 17) ? 17 : ((size) - 1) & (1ULL << 16) ? 16 : ((size) - 1) & (1ULL <<
15) ? 15 : ((size) - 1) & (1ULL << 14) ? 14 : ((size) - 1) & (1ULL << 13) ? 13
: ((size) - 1) & (1ULL << 12) ? 12 : ((size) - 1) & (1ULL << 11) ? 11 : ((size)
- 1) & (1ULL << 10) ? 10 : ((size) - 1) & (1ULL << 9) ? 9 : ((size) - 1) &
(1ULL << 8) ? 8 : ((size) - 1) & (1ULL << 7) ? 7 : ((size) - 1) & (1ULL << 6) ?
6 : ((size) - 1) & (1ULL << 5) ? 5 : ((size) - 1) & (1ULL << 4) ? 4 : ((size) -
1) & (1ULL << 3) ? 3 : ((size) - 1) & (1ULL << 2) ? 2 : 1) : -1) :
(sizeof((size) - 1) <= 4) ? __ilog2_u32((size) - 1) : __ilog2_u64((size) - 1) )
- 12 + 1;

Hello.

Maybe a silly question, but isn't the code only an obfuscation?
What about doing:

if (__builtin_constant_p(x))
  return x < 2 ? 0 : __builtin_clz(x);

which is fully eliminated in a c.031t.ccp1 pass right after einline happens
for a call context where 'x' is a constant.

Isn't the right place for fix in Linux kernel?

Martin