New config.h option to make SHA-512 smaller #178
Conversation
Saves 356 bytes on sha512.o compiling for Cortex-M0+ with ARM-GCC Size measured with: arm-none-eabi-gcc -Wall -Wextra -Iinclude -Os -mcpu=cortex-m0plus -mthumb -c library/sha512.c arm-none-eabi-size sha512.o GCC version: arm-none-eabi-gcc (GNU Tools for Arm Embedded Processors 7-2018-q2-update) 7.3.1 20180622 (release) [ARM/embedded-7-branch revision 261907]
Saves 108 bytes (measured as in previous commit).
Saves 1924 bytes (same measurement as before).
mpg
added
enhancement
Patater
left a comment
Patater
left a comment
There was a problem hiding this comment.
I'm happy with this. Will MBEDTLS_SHA512_SMALLER be tested in the full config tests in all.sh?
|
@Patater Thanks for your review! Yes, since the option is not excluded from |
gilles-peskine-arm
left a comment
gilles-peskine-arm
left a comment
There was a problem hiding this comment.
Looks good to me, but I left suggestions for improvements.
| PUT_UINT64_BE(n, b, i); | ||
| } | ||
| #else | ||
| #define sha512_put_uint64_be PUT_UINT64_BE |
There was a problem hiding this comment.
Can't we make it a function all the time and count on compilers to inline when compiling for performance?
|
|
||
| P( A[0], A[1], A[2], A[3], A[4], A[5], A[6], A[7], W[i], K[i] ); | ||
|
|
||
| temp1 = A[7]; A[7] = A[6]; A[6] = A[5]; A[5] = A[4]; A[4] = A[3]; |
There was a problem hiding this comment.
You can save a few more bytes with
temp1 = A[7];
for( k = 1; k < 8; k++ )
A[k] = A[k - 1];
A[0] = temp1;
On the other hand, I tried removing the rotation of A and running
P( A[(80-i+0)%8], A[(80-i+1)%8], A[(80-i+2)%8], A[(80-i+3)%8], A[(80-i+4)%8], A[(80-i+5)%8], A[(80-i+6)%8], A[(80-i+7)%8],
W[i], K[i] );
but that has almost the same code size as your version. I wonder if there's more to gain there by tweaking P and its auxiliary macros to use temporaries for certain calculations.
Patater
removed
needs: review
|
CI on merge result is all passing. Jenkins failed to notify GitHub with the result. |
3 participants
Add
MBEDTLS_SHA512_SMALLER, parallel toMBEDTLS_SHA256_SMALLER.Sizes, as measured with
arm-none-eabi-gcc -Wall -Wextra -Iinclude -Os -mcpu=cortex-m0plus -mthumb -c library/sha512.c && arm-none-eabi-size sha512.o:MBEDTLS_SHA512_SMALLER: 5691MBEDTLS_SHA512_SMALLER: 3411Speed impact on my laptop's i7, building with GCC -O2 and running
programs/test/benchmark sha512:Todo: