block-sha1: remove use of obsolete x86 assembly

In the block SHA-1 code, we have special assembly code for i386 and
amd64 to perform rotations with assembly.  This is supposed to help pick
the correct rotation operation depending on which rotation is smaller,
which can help some systems perform slightly better, since any circular
rotation can be specified as either a rotate left or a rotate right.
However, this isn't needed, so we should remove it.

First, SHA-1, like SHA-2, uses fixed constant rotates.  Thus, all
rotation amounts are known at compile time and are in fact baked into
the code.  Fortunately, peephole optimizers recognize rotations
specified in the normal way and automatically emit the correct code,
including a preference for choosing a rotate left versus a rotate right.
This has been the case for well over a decade, and is a standard example
of the utility of a peephole optimizer.

Moreover, all modern CPUs, with the exception of extremely limited
embedded CPUs such as some Cortex-M processors, provide a barrel
shifter, which lets the CPU perform rotates of any bit amount in
constant time.  This is valuable for many cryptographic algorithms to
improve performance, and is required to prevent timing attacks in
algorithms which use data-dependent rotations (which don't include the
hash algorithms we use).  As a result, even though the compiler does the
correct optimization, it isn't even needed here and either a left or a
right rotate is equally acceptable.

In fact, the SHA-256 code already takes this into account and just
writes the simple code using an inline function to let the compiler
optimize it for us.

The downside of using this code, however, is that it uses a GCC
extension, which makes the compiler complain when using -pedantic unless
it's prefixed with __extension__.  We could fix that, but since it's
not needed, let's just remove it.  We haven't noticed this because
almost everyone uses the SHA1DC code instead, but it still shows up for
some people.

Signed-off-by: brian m. carlson <>
Signed-off-by: Junio C Hamano <>
This commit is contained in:
brian m. carlson 2022-03-10 17:47:50 +00:00 committed by Junio C Hamano
parent 4c53a8c20f
commit 544d93bc3b
1 changed files with 0 additions and 17 deletions

View File

@ -11,27 +11,10 @@
#include "sha1.h"
#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
* Force usage of rol or ror by selecting the one with the smaller constant.
* It _can_ generate slightly smaller code (a constant of 1 is special), but
* perhaps more importantly it's possibly faster on any uarch that does a
* rotate with a loop.
#define SHA_ASM(op, x, n) ({ unsigned int __res; __asm__(op " %1,%0":"=r" (__res):"i" (n), "0" (x)); __res; })
#define SHA_ROL(x,n) SHA_ASM("rol", x, n)
#define SHA_ROR(x,n) SHA_ASM("ror", x, n)
#define SHA_ROT(X,l,r) (((X) << (l)) | ((X) >> (r)))
#define SHA_ROL(X,n) SHA_ROT(X,n,32-(n))
#define SHA_ROR(X,n) SHA_ROT(X,32-(n),n)
* If you have 32 registers or more, the compiler can (and should)
* try to change the array[] accesses into registers. However, on