From 7abb2bb96e0cd584f44dd8b219ad16d0232a6485 Mon Sep 17 00:00:00 2001
From: Denys Vlasenko <vda.linux@googlemail.com>
Date: Mon, 3 Jan 2022 17:02:48 +0100
Subject: [PATCH] libbb/sha1: x86_64 version: tidying up, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
---
 libbb/hash_md5_sha_x86-64.S    | 32 ++++++++++++++++----------------
 libbb/hash_md5_sha_x86-64.S.sh | 33 ++++++++++++++++++++-------------
 2 files changed, 36 insertions(+), 29 deletions(-)

diff --git a/libbb/hash_md5_sha_x86-64.S b/libbb/hash_md5_sha_x86-64.S
index ec4e63765..95b85d80a 100644
--- a/libbb/hash_md5_sha_x86-64.S
+++ b/libbb/hash_md5_sha_x86-64.S
@@ -60,7 +60,7 @@ sha1_process_block64:
 	xorl	%edx, %edi		# ^d
 	andl	%ebx, %edi		# &b
 	xorl	%edx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbp,%rsi),%ebp # e += RCONST + W[n]
+	leal	0x5A827999(%rbp,%rsi), %ebp # e += RCONST + W[n]
 	addl	%edi, %ebp		# e += (((c ^ d) & b) ^ d)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -72,7 +72,7 @@ sha1_process_block64:
 	xorl	%ecx, %edi		# ^d
 	andl	%eax, %edi		# &b
 	xorl	%ecx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rdx,%rsi),%edx # e += RCONST + W[n]
+	leal	0x5A827999(%rdx,%rsi), %edx # e += RCONST + W[n]
 	addl	%edi, %edx		# e += (((c ^ d) & b) ^ d)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -84,7 +84,7 @@ sha1_process_block64:
 	xorl	%ebx, %edi		# ^d
 	andl	%ebp, %edi		# &b
 	xorl	%ebx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rcx,%rsi),%ecx # e += RCONST + W[n]
+	leal	0x5A827999(%rcx,%rsi), %ecx # e += RCONST + W[n]
 	addl	%edi, %ecx		# e += (((c ^ d) & b) ^ d)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -96,7 +96,7 @@ sha1_process_block64:
 	xorl	%eax, %edi		# ^d
 	andl	%edx, %edi		# &b
 	xorl	%eax, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbx,%rsi),%ebx # e += RCONST + W[n]
+	leal	0x5A827999(%rbx,%rsi), %ebx # e += RCONST + W[n]
 	addl	%edi, %ebx		# e += (((c ^ d) & b) ^ d)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -108,7 +108,7 @@ sha1_process_block64:
 	xorl	%ebp, %edi		# ^d
 	andl	%ecx, %edi		# &b
 	xorl	%ebp, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rax,%rsi),%eax # e += RCONST + W[n]
+	leal	0x5A827999(%rax,%rsi), %eax # e += RCONST + W[n]
 	addl	%edi, %eax		# e += (((c ^ d) & b) ^ d)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -120,7 +120,7 @@ sha1_process_block64:
 	xorl	%edx, %edi		# ^d
 	andl	%ebx, %edi		# &b
 	xorl	%edx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbp,%rsi),%ebp # e += RCONST + W[n]
+	leal	0x5A827999(%rbp,%rsi), %ebp # e += RCONST + W[n]
 	addl	%edi, %ebp		# e += (((c ^ d) & b) ^ d)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -132,7 +132,7 @@ sha1_process_block64:
 	xorl	%ecx, %edi		# ^d
 	andl	%eax, %edi		# &b
 	xorl	%ecx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rdx,%rsi),%edx # e += RCONST + W[n]
+	leal	0x5A827999(%rdx,%rsi), %edx # e += RCONST + W[n]
 	addl	%edi, %edx		# e += (((c ^ d) & b) ^ d)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -144,7 +144,7 @@ sha1_process_block64:
 	xorl	%ebx, %edi		# ^d
 	andl	%ebp, %edi		# &b
 	xorl	%ebx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rcx,%rsi),%ecx # e += RCONST + W[n]
+	leal	0x5A827999(%rcx,%rsi), %ecx # e += RCONST + W[n]
 	addl	%edi, %ecx		# e += (((c ^ d) & b) ^ d)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -156,7 +156,7 @@ sha1_process_block64:
 	xorl	%eax, %edi		# ^d
 	andl	%edx, %edi		# &b
 	xorl	%eax, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbx,%r8),%ebx # e += RCONST + W[n]
+	leal	0x5A827999(%rbx,%r8), %ebx # e += RCONST + W[n]
 	addl	%edi, %ebx		# e += (((c ^ d) & b) ^ d)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -168,7 +168,7 @@ sha1_process_block64:
 	xorl	%ebp, %edi		# ^d
 	andl	%ecx, %edi		# &b
 	xorl	%ebp, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rax,%r9),%eax # e += RCONST + W[n]
+	leal	0x5A827999(%rax,%r9), %eax # e += RCONST + W[n]
 	addl	%edi, %eax		# e += (((c ^ d) & b) ^ d)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -180,7 +180,7 @@ sha1_process_block64:
 	xorl	%edx, %edi		# ^d
 	andl	%ebx, %edi		# &b
 	xorl	%edx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbp,%r10),%ebp # e += RCONST + W[n]
+	leal	0x5A827999(%rbp,%r10), %ebp # e += RCONST + W[n]
 	addl	%edi, %ebp		# e += (((c ^ d) & b) ^ d)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -192,7 +192,7 @@ sha1_process_block64:
 	xorl	%ecx, %edi		# ^d
 	andl	%eax, %edi		# &b
 	xorl	%ecx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rdx,%r11),%edx # e += RCONST + W[n]
+	leal	0x5A827999(%rdx,%r11), %edx # e += RCONST + W[n]
 	addl	%edi, %edx		# e += (((c ^ d) & b) ^ d)
 	movl	%ebp, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -204,7 +204,7 @@ sha1_process_block64:
 	xorl	%ebx, %edi		# ^d
 	andl	%ebp, %edi		# &b
 	xorl	%ebx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rcx,%r12),%ecx # e += RCONST + W[n]
+	leal	0x5A827999(%rcx,%r12), %ecx # e += RCONST + W[n]
 	addl	%edi, %ecx		# e += (((c ^ d) & b) ^ d)
 	movl	%edx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -216,7 +216,7 @@ sha1_process_block64:
 	xorl	%eax, %edi		# ^d
 	andl	%edx, %edi		# &b
 	xorl	%eax, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbx,%r13),%ebx # e += RCONST + W[n]
+	leal	0x5A827999(%rbx,%r13), %ebx # e += RCONST + W[n]
 	addl	%edi, %ebx		# e += (((c ^ d) & b) ^ d)
 	movl	%ecx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -228,7 +228,7 @@ sha1_process_block64:
 	xorl	%ebp, %edi		# ^d
 	andl	%ecx, %edi		# &b
 	xorl	%ebp, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rax,%r14),%eax # e += RCONST + W[n]
+	leal	0x5A827999(%rax,%r14), %eax # e += RCONST + W[n]
 	addl	%edi, %eax		# e += (((c ^ d) & b) ^ d)
 	movl	%ebx, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
@@ -240,7 +240,7 @@ sha1_process_block64:
 	xorl	%edx, %edi		# ^d
 	andl	%ebx, %edi		# &b
 	xorl	%edx, %edi		# (((c ^ d) & b) ^ d)
-	leal	0x5A827999(%rbp,%r15),%ebp # e += RCONST + W[n]
+	leal	0x5A827999(%rbp,%r15), %ebp # e += RCONST + W[n]
 	addl	%edi, %ebp		# e += (((c ^ d) & b) ^ d)
 	movl	%eax, %esi		#
 	roll	$5, %esi		# rotl32(a,5)
diff --git a/libbb/hash_md5_sha_x86-64.S.sh b/libbb/hash_md5_sha_x86-64.S.sh
index 5f09546b2..c5f0ef504 100755
--- a/libbb/hash_md5_sha_x86-64.S.sh
+++ b/libbb/hash_md5_sha_x86-64.S.sh
@@ -74,22 +74,24 @@ test "$1" -ge 8 && echo "%r${1}d"
 RD1A() {
 local a=$1;local b=$2;local c=$3;local d=$4;local e=$5
 local n=$(($6))
-echo "# $n"
-test $n = 0 && echo "
+local n0=$(((n+0) & 15))
+echo "
+# $n
+";test $n0 = 0 && echo "
 	# W[0], already in %esi
-";test $n != 0 && test $n -lt 8 && echo "
-	movl	`W32 $n`, %esi		# W[n]
-";test $n -ge 8 && echo "
-	# W[n], in %r$n
+";test $n0 != 0 && test $n0 -lt 8 && echo "
+	movl	`W32 $n0`, %esi		# W[n]
+";test $n0 -ge 8 && echo "
+	# W[n], in %r$n0
 ";echo "
 	movl	%e$c, %edi		# c
 	xorl	%e$d, %edi		# ^d
 	andl	%e$b, %edi		# &b
 	xorl	%e$d, %edi		# (((c ^ d) & b) ^ d)
-";test $n -lt 8 && echo "
-	leal	$RCONST(%r$e,%rsi),%e$e # e += RCONST + W[n]
-";test $n -ge 8 && echo "
-	leal	$RCONST(%r$e,%r$n),%e$e # e += RCONST + W[n]
+";test $n0 -lt 8 && echo "
+	leal	$RCONST(%r$e,%rsi), %e$e # e += RCONST + W[n]
+";test $n0 -ge 8 && echo "
+	leal	$RCONST(%r$e,%r$n0), %e$e # e += RCONST + W[n]
 ";echo "
 	addl	%edi, %e$e		# e += (((c ^ d) & b) ^ d)
 	movl	%e$a, %esi		#
@@ -119,7 +121,7 @@ echo "
 	xorl	`W32 $n8`, `W32 $n0`	# ^W[(n+8) & 15]
 	xorl	`W32 $n2`, `W32 $n0`	# ^W[(n+2) & 15]
 	roll	`W32 $n0`		#
-"; echo "
+";echo "
 	movl	%e$c, %edi		# c
 	xorl	%e$d, %edi		# ^d
 	andl	%e$b, %edi		# &b
@@ -165,7 +167,7 @@ echo "
 	xorl	`W32 $n8`, `W32 $n0`	# ^W[(n+8) & 15]
 	xorl	`W32 $n2`, `W32 $n0`	# ^W[(n+2) & 15]
 	roll	`W32 $n0`		#
-"; echo "
+";echo "
 	movl	%e$c, %edi		# c
 	xorl	%e$d, %edi		# ^d
 	xorl	%e$b, %edi		# ^b
@@ -216,7 +218,7 @@ echo "
 	xorl	`W32 $n8`, `W32 $n0`	# ^W[(n+8) & 15]
 	xorl	`W32 $n2`, `W32 $n0`	# ^W[(n+2) & 15]
 	roll	`W32 $n0`		#
-"; echo "
+";echo "
 	addl	%edi, %e$e		# += ((b | c) & d) | (b & c)
 ";test $n0 -lt 8 && echo "
 	leal	$RCONST(%r$e,%rsi), %e$e # e += RCONST + W[n & 15]
@@ -246,6 +248,11 @@ RD2 ax bx cx dx bp 60; RD2 bp ax bx cx dx 61; RD2 dx bp ax bx cx 62; RD2 cx dx b
 RD2 ax bx cx dx bp 65; RD2 bp ax bx cx dx 66; RD2 dx bp ax bx cx 67; RD2 cx dx bp ax bx 68; RD2 bx cx dx bp ax 69
 RD2 ax bx cx dx bp 70; RD2 bp ax bx cx dx 71; RD2 dx bp ax bx cx 72; RD2 cx dx bp ax bx 73; RD2 bx cx dx bp ax 74
 RD2 ax bx cx dx bp 75; RD2 bp ax bx cx dx 76; RD2 dx bp ax bx cx 77; RD2 cx dx bp ax bx 78; RD2 bx cx dx bp ax 79
+# Note: new W[n&15] values generated in last 3 iterations
+# (W[13,14,15]) are unused after each of these iterations.
+# Since we use r8..r15 for W[8..15], this does not matter.
+# If we switch to e.g. using r8..r15 for W[0..7], then saving of W[13,14,15]
+# (the "movl %esi, `W32 $n0`" insn) is a dead store and can be removed.
 } | grep -v '^$'
 
 echo "