15:34 < szt_> https://github.com/radare/radare2/blob/master/libr/anal/esil.c#L186 what reason to do this?
15:35 < szt_> esil->cur & (ut64) ( 1 << 1 ); esil->cur & (ut64) ( 1 << 2 ); etc
15:36 < szt_> maybe just use loop
15:38 < szt_> or lookup table
15:39 < szt_> https://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetNaive
15:39 < szt_> pancake
15:59 < jvoisin> pancake: I already have some customers for the calendar.
16:08 < jvoisin> szt_: send a patch ♥
16:11 < szt_> jvoisin, https://graphics.stanford.edu/~seander/bithacks.html#ParityNaive what parity computing method you prefer?
16:12 < szt_> Can use 64-bit integer?
16:12 < szt_> unsigned char b;  // byte value to compute the parity of
16:12 < szt_> bool parity =
16:12 < szt_>   (((b * 0x0101010101010101ULL) & 0x8040201008040201ULL) % 0x1FF) & 1;
16:13 < jvoisin> just use a loop
16:14 < jvoisin> plz no magic numbers :/
16:14 < szt_> loop - slow
16:14 < szt_> bithacks - fun
16:15 < jvoisin> maintainance--
16:15 < szt_> / comments what this code do
16:15 < szt_> and maybe some loop code for reference
16:16 < szt_> commented
16:17 < jvoisin> the compiler will optimize it anyway
16:17 < jvoisin> write readable code :/
16:21 < szt_> no, compiler is unable to optimize this things
16:22 < szt_> http://gcc.godbolt.org/ you can play with this
16:31 < szt_> http://goo.gl/TYsMrA assembly comparsion
16:39 < szt_> btw look at clang clang http://goo.gl/1VnP23
16:39 < szt_> redare2_parity_test(unsigned char):               # @redare2_parity_test(unsigned char)
16:39 < szt_>     xor    edi, 1
16:39 < szt_>     mov    eax, edi
16:39 < szt_>     ret
16:43 < szt_> xvilka__, look at http://goo.gl/1VnP23 and https://github.com/radare/radare2/blob/master/libr/anal/esil.c#L186
16:44 < szt_> why code in clang compiles into
16:44 < szt_>     xor    edi, 1
16:44 < szt_>     mov    eax, edi
16:44 < szt_>     ret
16:46 < szt_> bug?
16:53 < xvilka__> szt_: yep, bug
16:57 < szt_> consider to use faster parity check algorithm.
16:57 < szt_> and bug-free
16:57 < szt_> I don't know what algo is better for this case
16:59 < xvilka___> szt_: send a bugreport to the clang/llvm pls
17:07 < szt_> xvilka___, hmm... I think this is not clang bug
17:07 < szt_> something wrong with this code
17:08 < szt_> It does not check parity
17:11 < szt_> To check parity you need something like
17:11 < szt_>     bits ^= !!(esil->cur & (ut64)   1);
17:11 < szt_>     bits ^= !!(esil->cur & (ut64) ( 1 << 1 ));
17:11 < szt_>     bits ^= !!(esil->cur & (ut64) ( 1 << 2 ));
17:11 < szt_>     etc
17:12 < xvilka___> szt_: despite the real purpose of that code, it is still a clang bug. because that code _can't_ be compressed in just xor 1
17:14 < szt_> xvilka___, no, it can. You xor-ing every bit of some variable with variable which equal 1 at start
17:15 < szt_> so xor edi, 1 do exactly the same
17:15 < szt_> not a clang bug
17:16 < dx> i'm skimming that conversation, but are you saying that something behaves correctly with gcc, misbehaves with clang, and isn't clang's fault?
17:18 < szt_> I think all OK with clang and gcc. Just bug in radare2
17:19 < szt_> btw sometimes i finding some bugs in GCC
17:19 < szt_> sorry my english
17:25 < xvilka___> hm
17:25 < xvilka___> szt_: missed the bits = 1 part
17:25 < xvilka___> szt_: yes, you right
17:25 < xvilka___> szt_: pls send PR
17:26 < szt_> What means "send PR"?
17:27 < dx> yeah i just finished doing my own testing
17:28 < szt_> pull request
17:29 < szt_> xvilka___, but what bit parity check is better for this case?
17:29 < szt_> I am not sure
17:29 < xvilka___> szt_: the thing you've suggested looks o
17:29 < xvilka___> k
17:52 < szt_> xvilka___, https://github.com/radare/radare2/pull/3139
17:54 < dx> fancy
17:55 < dx> "The method above takes around 4 operations, but only works on bytes"
17:57 < dx> esil->cur is ut64
18:09 -!- condret [~condret@unaffiliated/condret] has joined #radare
18:10 < szt_> (esil->cur & 0xff)
18:10 < szt_> 8 bit
18:10 < condret> waaat
18:10 < condret> where does dat happen
18:10 < condret> pancake: ping
18:13 < dx> szt_: oh i see it says least significant byte. nevermind
18:13 < crowell> morn
18:13 < dx> condret: context is https://github.com/radare/radare2/pull/3139
18:17 < condret> black magic
18:19 < condret> i did a comment on it
18:19 < dx> condret: context is: it didn't