Signed-off-by: hugo-syn <hugo.vincent@synacktiv.com>

Signed-off-by: hugo-syn <hugo.vincent@synacktiv.com>

plase -> please

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>

Vulkan deps are optional. vulkan-utils is replaced with vulkan-tools
since Debian 10.

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>

Clang prior to 13 no longer fails to build ncnn since #4845.

Clang 18 fixes SSE4.1 translation, which yields a major speedup.

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>

* vulkan-driver-loader.md
* static vulkan on apple

* pnnx support grid_sample op

* complete the permute and gridsample operator fusion

* spilt calculation into two stages and support permute fusion

* Add POWER8 VSX toolchains

POWER8, though slower than POWER9, is still used in the wild; these
toolchains should still be much faster on POWER8 than POWER8 without VSX
optimizations.

* VSX toolchains: set -cpu arg in QEMU CI tests

* Add POWER9 VSX toolchains

Translating x86_64 SSE to ppc64le VSX intrinsics yields a quite large
speedup on POWER9. See this article for background:

https://www.talospace.com/2019/07/easier-power-vectorizing-for-fun-and.html

* Add power9le docs

* power9le clang toolchain: Document Clang 13+ requirement

---------

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>

* simplify binaryop

* less gpu test

* update binaryop broadcast doc

* do not test atan2 zero

* stb_image_write 1.16

* stb_image v2.28

* update emsdk 3.1.28

* enable stb arm neon

* update doc

Co-authored-by: ncnnnnn <67086033+ncnnnnn@users.noreply.github.com>

* mha is now permute and reshape free
* gemm user defined tile mnk param

* fix doc, sync x86 gemm fix

Co-authored-by: EdVince <EdVince@users.noreply.github.com>
Co-authored-by: nihuini <nihuini@tencent.com>