Instead of creating a new kernel for each block we pre-build 25 kernel within a chunk.
This reduces the calls to `clCreateProgramWithSource` which creates the leaks and will crash the miner after a few hours.

Workaround was first introduced by @SChernykh in xmrig-amd

increase version to 2.10.2

OpenCL: fix memory leak

add cli option `amdCacheDir`

add option to control where the amd cache data should be stored

Repair OSX CUDA searched-path for NVRTC

Fix a memory leak that the OpenCL programs was not freed correctly.
First solved by @xmrig

increase version to 2.10.1

cryptonight_r without mediaops

add cryptonight_v8_double

Add support for non media ops for cryptonight_r

fix avx detection

fix cn_gpu rounding mode selection

If cryptonight_gpu is used in the dev pool and a cryptonight_v8 derived
algorithm than cryptonight_gpu will create wrong results if v8 was
running as non ASM version.

Add new algorithm alias for cryptonight_v8_double. The difference to v8
is that the number of iterations is increased.

Add also the currency X-CASH which use cryptonight_double.

CPU: fix cryptonight_r slowdown

If the miner changed the pool (user pool or dev pool) and the network
block of monero has not changed the miner switch to the non asm version.

Add check if the hash function has changed. If the hash function not
point to the asm version the ASM function is recreated.

update currency pow table

remove wrong documentation

The OS is able to disable avx. In that case our avx detection is wrong.
See xmrig issue: https://github.com/xmrig/xmrig/issues/951

Import fix: https://github.com/xmrig/xmrig/commit/6052da3c4399514c95245b2d342f04baa20958db#diff-92fa29935c136762d35981769ac79f52

thx @xmrig for the notification.

fix opencl and cuda cryptonight_r caching

use double hash for cryptonight_r

OpenCL: fix windows vega and vii

fix masari

Remove fork description for monero.

Remove a warning about unsupported CUDA from an older release.

The original implementation of the cache release and create always new
kernel, this can lead into performance issues and my crashes.

add support for cnR double thread assembler

Since masari increased the block size the miner crashed each time it
gets connected with a masari pool.

This PR extent the possible size of a block to 128 byte and updated the
kernel.

Fix again issues with vega and vii with the latest windows driver.

increase version to 2.10.0

dev pool change

Edited comments in configs, set default verbose to 4

activate nicehash

Support of CryptoNight v8 ReverseWaltz