| ccf126a5 | 30-Mar-2020 |
Jens Wiklander <jens.wiklander@linaro.org> |
core: ltc: remove Arm AES CE routines
Removes the Arm AES CE routines from the library.
This will later be added in common code to be shared with other crypto
libraries etc.
Acked-by: Etienne Carr
core: ltc: remove Arm AES CE routines
Removes the Arm AES CE routines from the library.
This will later be added in common code to be shared with other crypto
libraries etc.
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
show more ...
|
| bb1d5c32 | 10-Sep-2019 |
Jerome Forissier <jerome@forissier.org> |
core: ltc: aes_modes_armv8a_ce_a64.S: get rid of literal load of addend vector
Cherry pick of Linux kernel commit ed6ed11830a9 ("crypto:
arm64/aes-modes - get rid of literal load of addend vector").
core: ltc: aes_modes_armv8a_ce_a64.S: get rid of literal load of addend vector
Cherry pick of Linux kernel commit ed6ed11830a9 ("crypto:
arm64/aes-modes - get rid of literal load of addend vector"). Original
commit message:
"
Replace the literal load of the addend vector with a sequence that
performs each add individually. This sequence is only 2 instructions
longer than the original, and 2% faster on Cortex-A53.
This is an improvement by itself, but also works around a Clang issue,
whose integrated assembler does not implement the GNU ARM asm syntax
completely, and does not support the =literal notation for FP registers
(more info at https://bugs.llvm.org/show_bug.cgi?id=38642)
"
Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
show more ...
|
| 9ff4f2cc | 05-Aug-2016 |
Jerome Forissier <jerome.forissier@linaro.org> |
arm32: AES using ARMv8-A cryptographic extensions
Add AES acceleration to LibTomCrypt in 32-bit mode using ARMv8
Cryptographic Extensions. Enabled when CFG_CRYPTO_AES_ARM32_CE=y.
The ARMv8 CE assem
arm32: AES using ARMv8-A cryptographic extensions
Add AES acceleration to LibTomCrypt in 32-bit mode using ARMv8
Cryptographic Extensions. Enabled when CFG_CRYPTO_AES_ARM32_CE=y.
The ARMv8 CE assembler code comes from the Linux kernel's
arch/arm64/crypto/aes-ce-core.S (contributed by Ard Biesheuvel,
Linaro) with minor updates:
- Removed the alignment hints in vld1/vst1 operations, for instance
'vld1.8 {q1}, [r1, :64]!' becomes 'vld1.8 {q1}, [r1]!'. This is
required because the input and output buffers may not be 8-byte
aligned.
- ce_aes_ctr_encrypt(): do not increment the counter on the first block
because the C wrapper already did. Do not increment the counter after
the last block has been processed either, because the C wrapper will do
it before calling ce_aes_ctr_encrypt() again.
- aes_xts_decrypt(): drop the 'first' parameter which is not used (it
would always be set to 1 by the caller).
Tested on HiKey with xtest [1] for functional tests and aes-perf [2] for
speed. In the table below, a mode name followed by a '+' means
accelerated with crypto extensions, while no '+' is with acceleration
disabled (i.e., CFG_CRYPTO_AES_ARM32_CE=n).
Average encryption speed (MiB/s):
Size | Mode
(KiB) | ECB CBC CTR XTS ECB+ CBC+ CTR+ XTS+
------+--------------------------------------------------
1 | 14.9 12.7 12.3 10.9 33.8 32.7 33.3 29.1
2 | 18.2 15.4 14.9 13.3 65.0 60.9 62.6 55.4
4 | 20.8 17.3 16.6 15.0 119.3 106.0 112.6 101.0
8 | 22.4 18.2 17.6 16.0 201.6 166.5 182.3 166.7
16 | 23.3 19.0 18.1 16.6 308.1 232.9 264.3 248.3
32 | 23.8 19.3 18.4 16.9 427.7 295.1 346.4 335.1
64 | 24.0 19.5 18.6 17.1 532.8 341.3 411.0 407.6
128 | 24.2 19.6 18.7 17.2 608.1 371.3 454.9 457.5
256 | 24.2 19.6 18.7 17.1 624.4 382.0 463.5 481.6
512 | 24.0 19.4 18.6 17.1 658.8 389.6 477.3 498.3
[1] https://github.com/OP-TEE/optee_test
[2] https://github.com/linaro-swg/aes-perf.git
Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: David Brown <david.brown@linaro.org>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>
show more ...
|
| 34777d46 | 04-Aug-2016 |
Jerome Forissier <jerome.forissier@linaro.org> |
arm64: libtomcrypt: rename AES CE files
aes_arm64_ce.c contains wrapper code for the assembly routines that
implement AES using the ARMv8-A cryptographic extensions. It is
currently used only in 64-
arm64: libtomcrypt: rename AES CE files
aes_arm64_ce.c contains wrapper code for the assembly routines that
implement AES using the ARMv8-A cryptographic extensions. It is
currently used only in 64-bit builds, but could be used for 32-bits
as well. So, rename it to aes_armv8a_ce.c to better reflect its content
and prepare for 32-bit support. Also rename the 64-bit assembly file
(aes_modes_arm64_ce_a64.S -> aes_modes_armv8a_ce_a64.S) for
consistency.
Note: this naming scheme (xyz_armv8a_ce.c / xyz_armv8a_ce_a{32,64}.S)
is already used for SHA1 and SHA2.
Signed-off-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: David Brown <david.brown@linaro.org>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>
show more ...
|