02:42:11 lol... no worries. 18:48:03 Is it possible to launch 64-bit termux within 32-bit Android? 19:13:40 Is it possible to launch any 64-bit app on 32-bit Android? 19:16:41 cpu should be 64-bit here 19:17:18 I suppose that user space binary can use 64-bit asm 19:23:30 There's AArch64 execution state and AArch32 execution state. 64-bit state can run 32-bit code, but not the other way. 19:24:42 https://community.arm.com/cfs-file/__key/communityserver-discussions-components-files/10/mixing_5F00_AArch32_5F00_and_5F00_AArch64.png 19:26:49 I've tried to find simple asm source to compile and test on the phone. 19:27:07 It would be some kind of runtime check 19:27:14 aarch64 specific asm * 19:27:20 But didn't find anything 19:51:39 aarch32 OS cannot host an aarch64 userspace application 19:54:52 The phone isn't rooted therefore I don't know for sure what OS is being running. I'm trying to test it in runtime 19:55:32 If you're right then I'll see any error 19:59:37 install CPU-Z from the playstore and go to System Tab 20:00:12 you'll see either aarch64 or armv7l 20:01:35 armv8l 20:01:53 that's 64-bit 20:02:15 Therefore I'm continue my road to compile runtime test 20:02:28 it's likely not reliable test 20:04:01 I take it back, it looks like the 32-bit mode is called armv8l sometimes 20:04:39 should say aarch64 in 64-bit mode 20:13:03 Can you modify this source to add dummy aes instruction ? https://raw.githubusercontent.com/matja/asm-examples/master/aarch64/hello.aarch64.linux.syscall.gas.asm 20:13:17 the same way as you did in c++ to test hardware aes support in runtime? 20:13:29 because hello wrold without aes is running fine here 20:26:37 or even better just static randomx-benchmark build for aarch64 20:28:49 I've added 3 instructions from randomx-benchmark and it works fine: http://paste.debian.net/hidden/c0786122/ 20:29:22 Unfortunately, there is no static randomx-benchmark for aarch64 on github 20:29:38 all runtime tests have passed 20:30:34 good, so you have a 64-bit ARM chip with crypto extensions 20:31:00 but that android app says armv8l 20:31:09 and termux is a shitty enviroment 20:31:14 the same level as cygwin 20:31:35 can't you just try to compile xmrig in termux? 20:31:39 I'm going to cross-compile for aarch64 20:31:46 xmrig didn't even compiled due to lack of some headers 20:32:00 works on my android phone, but it's 64 bit 20:32:05 xmrig hasn't been compiled due to errors * 20:38:11 randomx-benchmark won't compile? 20:38:51 I've compiled randomx-benchmark successfuly in termux but it fails to work without --softAes 20:39:09 I need correct static build for aarch64 to test with hardware aes 20:39:20 it crashes without softAes? 20:39:25 263ms per hash without largepages, without hardware aes 20:39:41 No crash, your c++ code detects exception and prints 'try ... --softAes' 20:39:53 without dataset * 20:40:20 what cmake options did you use? 20:40:52 it looks like AES code was not even compiled 20:41:36 ( No need to investigate it since termux environment is a shit ) 20:42:26 you have to use just "cmake .." without ARCH to get the full build 20:42:41 native can exclude AES if it's wrongly detected 20:43:07 another option is that the compiler doesn't support it 20:44:35 here: https://github.com/tevador/RandomX/blob/master/src/intrin_portable.h#L451 20:45:30 you can try to comment out othe #ifdef and see what happens 20:57:54 cross-compiled binary said 'Bad system call' via stderr within termux 20:58:14 Can't find any rwx place to upload it via 'adb push' 20:58:24 and test outside of termux 21:12:01 It's likely that cross-compile environment with 5.x kernel isn't compatible with 4.x android kernel