Add A32 support in CI #620
Conversation
README.md
Outdated
| @@ -126,6 +126,10 @@ or | |||
| ```shell | |||
| $ make CROSS_COMPILE=arm-linux-gnueabihf- check # ARMv7-A | |||
| ``` | |||
| or | |||
| ``` shell | |||
| $ make CROSS_COMPILE=arm-linux-gnueabihf- A32=1 check # ARMv8-A running in 32-bit mode, i.e., A32 instruction set | |||
There was a problem hiding this comment.
In this case, A32 is always 1, right?
There was a problem hiding this comment.
Correct. A32 always sets to 1 if we want to run test program compiled under A32 instruction set.
There was a problem hiding this comment.
If toolchain is set to arm-linux-gnueabihf-, A32 should not appear since the former implies the latter.
There was a problem hiding this comment.
The goal of using A32 for indicating A32 instruction set build is that we can not only test our code on AArch32 but also check whether the NEON optimization on A32 will create any impacts or not.
Nevertheless, we may set the -mcpu flag (link to readme of tinymembenchmark) in other Makefile input parameters such as CFLAGS, and I consider I can start to implement from this.
Cuda-Chen
force-pushed
the
add-armv8a-32bit-ci
branch
2 times, most recently
from
16ad16a
to
8bbd6d0
Compare
Cuda-Chen
force-pushed
the
add-armv8a-32bit-ci
branch
from
20a7216
to
ef2f8f0
Compare
Add A32 support in CI. Choose Cortex-A32 as test target because it only supports AArch32. Moreover, Cortex-A32 has Armv8-A Cryptographic Extension, which is an ideal target for testing Cryptographic Extension conversions.
Cuda-Chen
force-pushed
the
add-armv8a-32bit-ci
branch
from
ef2f8f0
to
4134dec
Compare
|
Thank @Cuda-Chen for contributing! |
Add A32 support in CI.
For @jonathanhue, if possible, kindly leave some suggestions as I have little experience on A32 binary running on Armv8-A.