Debian/Ubuntu Intel OpenCL installation

oneapi.sh

# To make a long story short

# Install apt source
curl -fsSL https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | sudo gpg --dearmor --output /usr/share/keyrings/intel-sw-products.gpg
echo "deb [signed-by=/usr/share/keyrings/intel-sw-products.gpg] https://apt.repos.intel.com/oneapi all main" | sudo tee /etc/apt/sources.list.d/intel-oneapi.list
apt update

# Install oneapi
apt install intel-oneapi-runtime-opencl intel-basekit clinfo opencl-headers

# source environment variables, put this in .bashrc
source /opt/intel/oneapi/setvars.sh

# Compile john
$ sudo apt install opencl-headers
$ cd JohnTheRipper/src
$ ./configure LDFLAGS=-L/opt/intel/opencl
...
OpenCL support ..................................... yes
$ ldconfig -v
...
libOpenCL.so.1 -> libOpenCL.so.1.2
make clean && make -sj8
$ cd ../run
$ ./john -list:opencl-devices
Platform #0 name: Intel(R) FPGA Emulation Platform for OpenCL(TM), version: OpenCL 1.2 Intel(R) FPGA SDK for OpenCL(TM), Version 20.3
...
Platform #1 name: Intel(R) OpenCL, version: OpenCL 3.0 LINUX
...
Platform #2 name: Portable Computing Language, version: OpenCL 2.0 pocl 1.8  Linux, None+Asserts, RELOC, LLVM 11.1.0, SLEEF, DISTRO, POCL_DEBUG

# The second device is fastest in my case and can be specified with -device=2
# Benchmark for 60 seconds
$ ./john -form=raw-MD5-opencl --device=2 -test=60
...
Device 2: 13th Gen Intel(R) Core(TM) i7-1355U
Benchmarking: raw-MD5-opencl [MD5 OpenCL]... Binary Build log: Device build started
Options used by backend compiler:
Device build done
Reload Program Binary Object.
Binary Build log: Device build started
Options used by backend compiler:
Device build done
Reload Program Binary Object.
LWS=512 GWS=4194304 DONE
Raw:    32806K c/s real, 9181K c/s virtual

# This is not very fast. The GPU is from a Lenovo Thinkpad T16 Gen 2 and the GPU is called 
# Intel® Iris® Xe Graphics
# Architecture Xe
# Codename Raptor Lake-P

# Benchmark hashcat
$ cd hashcat
$ make
$ ./hashcat -D 1 -b -m 0
hashcat (v6.2.6-827-g46ce637d3) starting in benchmark mode
...
OpenCL API (OpenCL 1.2 Intel(R) FPGA SDK for OpenCL(TM), Version 20.3) - Platform #1 [Intel(R) Corporation]
===========================================================================================================
* Device #1: Intel(R) FPGA Emulation Device, skipped

OpenCL API (OpenCL 3.0 LINUX) - Platform #2 [Intel(R) Corporation]
==================================================================
* Device #2: 13th Gen Intel(R) Core(TM) i7-1355U, 15843/31751 MB (7937 MB allocatable), 12MCU

OpenCL API (OpenCL 2.0 pocl 1.8  Linux, None+Asserts, RELOC, LLVM 11.1.0, SLEEF, DISTRO, POCL_DEBUG) - Platform #3 [The pocl project]
=====================================================================================================================================
* Device #3: pthread-13th Gen Intel(R) Core(TM) i7-1355U, skipped

Benchmark relevant options:
===========================
* --backend-devices-virtual=1
* --opencl-device-types=1
* --optimized-kernel-enable

-------------------
* Hash-Mode 0 (MD5)
-------------------

Speed.#2.........:   369.7 MH/s (33.67ms) @ Accel:1024 Loops:1024 Thr:1 Vec:8

# If I interpret the result correctly hashcat is faster but it's still slow.
# I have seen benchmarks on this same card in Windows that is 10 times faster
# with hashcat. Something is wrong. It should be possible to get it to run faster.