Pablo Labs

[log.txt](https://github.com/jart/blink/files/13626353/log.txt)

Some syscalls are missing from what I saw

on my local tests, without dynarec, test27 (sse4.2 test) is working proper, and with dynarec (i tried with safeflags=2 and strongmem=3) not working proper [box64_android_test27_output_with_dynarec.log](https://github.com/ptitSeb/box64/files/15270017/box64_android_test27_output_with_dynarec.log) [box64_android_test27_output_without_dynarec.log](https://github.com/ptitSeb/box64/files/15270018/box64_android_test27_output_without_dynarec.log) Using diff get this...

> Try to use `BOX64_DYNAREC_TEST=1 box64 ./test27` on your local machine to see if there is any `DIFF xxx` logs. I will test when I get home

> Try to use `BOX64_DYNAREC_TEST=1 box64 ./test27` on your local machine to see if there is any `DIFF xxx` logs. [box64-android-dynarec-test.log](https://github.com/ptitSeb/box64/files/15277220/box64-android-dynarec-test.log)

Don't know, I will search about

> Looks like there is an issue with VAARG flot/double/long double parameter somehow. Do you know if long double (128bits) exist on android arm? I have read https://developer.android.com/ndk/guides/abis and 32bit...

-RPI5ARM64PS16K -DARM_DYNAREC=1

I checked here again today about the test30 failure, comparing the logs running natively on android-x86-64 and box64, there was a difference in psqrtpd ``` [pablo@archlinux tests]$ diff box64-android-output-test30.txt ref30.txt...

Running with FASTNAN=0 gives: ``` [pablo@archlinux tests]$ BOX64_DYNAREC_FASTNAN=0 BOX64_DYNAREC_FASTROUND=0 BOX64_AVX=2 ../build/box64 test30_android > box64-android-output-test30.txt [pablo@archlinux tests]$ diff box64-android-output-test30.txt ref30.txt 198,199c198,199 < psqrtpd(0 -2 ) = 0 0x7ff8000000000000 < psqrtpd(inf -inf...