chacha/asm/chacha-x86.pl: improve [backward] portability.

In order to minimize dependency on assembler version a number of
post-SSE2 instructions are encoded manually. But in order to simplify
the procedure only register operands are considered. Non-register
operands are passed down to assembler. Module in question uses pshufb
with memory operands, and old [GNU] assembler can't handle it.
Fortunately in this case it's possible skip just the problematic
segment without skipping SSSE3 support altogether.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
This commit is contained in:
Andy Polyakov 2016-11-08 11:11:58 +01:00
parent cebb186989
commit d89773d659

View File

@ -50,7 +50,7 @@ for (@ARGV) { $xmm=1 if (/-DOPENSSL_IA32_SSE2/); }
$ymm=1 if ($xmm &&
`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
=~ /GNU assembler version ([2-9]\.[0-9]+)/ &&
$1>=2.19); # first version supporting AVX
($gasver=$1)>=2.19); # first version supporting AVX
$ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32n" &&
`nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/ &&
@ -465,6 +465,12 @@ if ($ymm) {
&label("pic_point"),"eax"));
&movdqu ("xmm3",&QWP(0,"ebx")); # counter and nonce
if (defined($gasver) && $gasver>=2.17) { # even though we encode
# pshufb manually, we
# handle only register
# operands, while this
# segment uses memory
# operand...
&cmp ($len,64*4);
&jb (&label("1x"));
@ -646,6 +652,7 @@ if ($ymm) {
&paddd ("xmm2",&QWP(16*6,"eax")); # +four
&pand ("xmm3",&QWP(16*7,"eax"));
&por ("xmm3","xmm2"); # counter value
}
{
my ($a,$b,$c,$d,$t,$t1,$rot16,$rot24)=map("xmm$_",(0..7));