Olaf Bernstein

Amazing work! I ran a quick benchmark on the kendryte k230 (thead C908) with [this neon mandelbrot code](https://github.com/skeeto/mandel-simd) and my handwritten [rvv mandelbrot code](https://github.com/camel-cdr/rvv-bench/blob/main/bench/mandelbrot.S#L3) (slightly adjusted to fit the neon,...

> Thanks @camel-cdr ! Can you run all the SIMDe tests from this PR on the k230? I'm currently working on that, however I run into problems with the glibc...

I couldn't figure out how to get the glibc versions to align.

@eric900115 For neon RVV codegen can be 100% portable. If we require the standard V extension (VLEN>=128 and ELEN=64), then we can use LMUL=1 on all implementations, because even for...

> Here's my `meson setup --cross ...` config for using https://packages.debian.org/unstable/clang-18 and running on the [official Debian image](https://kendryte-download.canaan-creative.com/developer/k230/canmv_debian_sdcard_sdk_1.3.img.gz): Thanks it worked. I didn't know about the debian image, and was...

Sounds interesting, Berlin is only 2-3 hours away from me. But I'm not really involved with Debian (except for running it). Btw, do you know how Debian deals with compiler...

@mr-c Yes, I get similar errors when testing `master`: ``` ../test/arm/neon/abs.c:711: assertion failed: r[0] == simde_vld1q_s32(test_vec[i].r)[0] (0 == -2147483648) test/arm/neon/abs.cpp:711: assertion failed: r[0] == simde_vld1q_s32(test_vec[i].r)[0] (0 == -2147483648) timeout qabs-native-c...

@mr-c Here we go, looks like the abs errors are gone, great work. ``` ../test/arm/neon/qrdmlah_lane.c:475: assertion failed: r[0] == simde_vld1_s16(test_vec[i].r)[0] (-18972 == -13752) ../test/arm/neon/qrdmlah_lane.c:678: assertion failed: r[0] == simde_vld1_s16(test_vec[i].r)[0] (26528...

@mr-c the errors are still there, but the values are different now: ``` ../test/arm/neon/qrdmlah.c:193: assertion failed: r[0] == simde_vld1_s16(test_vec[i].r)[0] (17480 == 3378) test/arm/neon/qrdmlah.cpp:193: assertion failed: r[0] == simde_vld1_s16(test_vec[i].r)[0] (17480 ==...

@mr-c it was the native tests, I ran it via: `for i in *native*; do ./$i; done > /dev/null`