Eric Richter authored Sep 11, 2024 and Niels Möller committed Sep 13, 2024

Originally, the 16 input words were loaded with 16 individual vector load
instructions. This has a side effect where the last three loads would
overread 1/2/3 extra words.

Fix the overread by replacing unnecessary overlapped reads with shifts.
As a consequence, the constant registers for 4,8,12 can be removed, and
also gain about 1~2% in performance.

Signed-off-by: Eric Richter <erichte@linux.ibm.com>

Brad Smith authored Sep 01, 2024 and Niels Möller committed Sep 07, 2024

Signed-off-by: Brad Smith <brad@comstyle.com>

Eric Richter authored Aug 29, 2024 and Niels Möller committed Aug 30, 2024

According to the ABI, the stack pointer is quadword aligned, so starting
the stack storage at offset -8, may cause the return address to be
stepped on. Adjusting to use -16 as the starting point, which also
matches other POWER assembly code.

Signed-off-by: Eric Richter <erichte@linux.ibm.com>

Eric Richter authored Jun 20, 2024 and Niels Möller committed Aug 04, 2024

By passing in the constant offset value into the LOAD macro, the use of
m4_unquote to calculate the correct constant GPR can be avoided,
improving readability.

Signed-off-by: Eric Richter <erichte@linux.ibm.com>

Foolbar authored Jun 10, 2024 and Niels Möller committed Jun 11, 2024

getauxval() is available on Android since API 18.
https://developer.android.com/ndk/guides/cpu-features#features_using_libcs_getauxval3

Eric Richter authored Jun 04, 2024 and Niels Möller committed Jun 09, 2024

This patch introduces an optimized powerpc64 assembly implementation for
sha256-compress-n. This takes advantage of the vshasigma instruction, as
well as unrolling loops to best take advantage of running instructions
in parallel.

The following data was captured on a POWER 10 LPAR @ ~3.896GHz

Current C implementation:
         Algorithm         mode Mbyte/s
            sha256       update  280.97
       hmac-sha256     64 bytes   80.81
       hmac-sha256    256 bytes  170.50
       hmac-sha256   1024 bytes  241.92
       hmac-sha256   4096 bytes  268.54
       hmac-sha256   single msg  276.16

With optimized assembly:
         Algorithm         mode Mbyte/s
            sha256       update  461.45
       hmac-sha256     64 bytes  123.88
       hmac-sha256    256 bytes  268.81
       hmac-sha256   1024 bytes  390.91
       hmac-sha256   4096 bytes  438.02
       hmac-sha256   single msg  453.83

Signed-off-by: Eric Richter <erichte@linux.ibm.com>

See
https://docs.gitlab.com/ee/update/deprecations.html#removal-of-tags-from-small-saas-runners-on-linux,
and corresponding gnutls update
gnutls@642c39ba.

Daiki Ueno authored Apr 13, 2024 and Niels Möller committed Apr 14, 2024

Signed-off-by: Daiki Ueno <dueno@redhat.com>

For fat builds, move definition to fat-ppc.c. For builds where the
functions are unavailable, define as macros returning zero, and rely
on the compiler to eliminate the code that uses the return value.

Change type of the rounds argument from size_t to unsigned.