From 20b88bb160865b0c2998056fe841b1cbacd6d4c7 Mon Sep 17 00:00:00 2001
From: Andy Polyakov <appro@openssl.org>
Date: Thu, 31 Mar 2016 18:47:17 +0200
Subject: PPC assembly pack: remove branch hints.

As it turns out branch hints grew as kind of a misconception. In
addition their interpretation by GNU assembler is affected by
assembler flags and can end up with opposite meaning on different
processors. As we have to loose quite a lot on misinterprerations,
especially on newer processors, we just omit them altogether.

Reviewed-by: Tim Hudson <tjh@openssl.org>
---
 crypto/bn/asm/ppc-mont.pl   | 10 +++++-----
 crypto/bn/asm/ppc.pl        | 10 +++++-----
 crypto/bn/asm/ppc64-mont.pl | 12 ++++++------
 3 files changed, 16 insertions(+), 16 deletions(-)

(limited to 'crypto/bn')

diff --git a/crypto/bn/asm/ppc-mont.pl b/crypto/bn/asm/ppc-mont.pl
index da69c6aaaf..6930a3aceb 100644
--- a/crypto/bn/asm/ppc-mont.pl
+++ b/crypto/bn/asm/ppc-mont.pl
@@ -191,7 +191,7 @@ L1st:
 
 	addi	$j,$j,$BNSZ	; j++
 	addi	$tp,$tp,$BNSZ	; tp++
-	bdnz-	L1st
+	bdnz	L1st
 ;L1st
 	addc	$lo0,$alo,$hi0
 	addze	$hi0,$ahi
@@ -253,7 +253,7 @@ Linner:
 	addze	$hi1,$hi1
 	$ST	$lo1,0($tp)	; tp[j-1]
 	addi	$tp,$tp,$BNSZ	; tp++
-	bdnz-	Linner
+	bdnz	Linner
 ;Linner
 	$LD	$tj,$BNSZ($tp)	; tp[j]
 	addc	$lo0,$alo,$hi0
@@ -276,7 +276,7 @@ Linner:
 	slwi	$tj,$num,`log($BNSZ)/log(2)`
 	$UCMP	$i,$tj
 	addi	$i,$i,$BNSZ
-	ble-	Louter
+	ble	Louter
 
 	addi	$num,$num,2	; restore $num
 	subfc	$j,$j,$j	; j=0 and "clear" XER[CA]
@@ -289,7 +289,7 @@ Lsub:	$LDX	$tj,$tp,$j
 	subfe	$aj,$nj,$tj	; tp[j]-np[j]
 	$STX	$aj,$rp,$j
 	addi	$j,$j,$BNSZ
-	bdnz-	Lsub
+	bdnz	Lsub
 
 	li	$j,0
 	mtctr	$num
@@ -304,7 +304,7 @@ Lcopy:				; copy or in-place refresh
 	$STX	$tj,$rp,$j
 	$STX	$j,$tp,$j	; zap at once
 	addi	$j,$j,$BNSZ
-	bdnz-	Lcopy
+	bdnz	Lcopy
 
 	$POP	$tj,0($sp)
 	li	r3,1
diff --git a/crypto/bn/asm/ppc.pl b/crypto/bn/asm/ppc.pl
index 04df1fe5cc..446d8ba949 100644
--- a/crypto/bn/asm/ppc.pl
+++ b/crypto/bn/asm/ppc.pl
@@ -1556,7 +1556,7 @@ Lppcasm_sub_mainloop:
 				# if carry = 1 this is r7-r8. Else it
 				# is r7-r8 -1 as we need.
 	$STU	r6,$BNSZ(r3)
-	bdnz-	Lppcasm_sub_mainloop
+	bdnz	Lppcasm_sub_mainloop
 Lppcasm_sub_adios:	
 	subfze	r3,r0		# if carry bit is set then r3 = 0 else -1
 	andi.	r3,r3,1         # keep only last bit.
@@ -1603,7 +1603,7 @@ Lppcasm_add_mainloop:
 	$LDU	r8,$BNSZ(r5)
 	adde	r8,r7,r8
 	$STU	r8,$BNSZ(r3)
-	bdnz-	Lppcasm_add_mainloop
+	bdnz	Lppcasm_add_mainloop
 Lppcasm_add_adios:	
 	addze	r3,r0			#return carry bit.
 	blr
@@ -1762,7 +1762,7 @@ Lppcasm_sqr_mainloop:
 	$UMULH  r8,r6,r6
 	$STU	r7,$BNSZ(r3)
 	$STU	r8,$BNSZ(r3)
-	bdnz-	Lppcasm_sqr_mainloop
+	bdnz	Lppcasm_sqr_mainloop
 Lppcasm_sqr_adios:	
 	blr
 	.long	0
@@ -1827,7 +1827,7 @@ Lppcasm_mw_LOOP:
 	
 	addi	r3,r3,`4*$BNSZ`
 	addi	r4,r4,`4*$BNSZ`
-	bdnz-	Lppcasm_mw_LOOP
+	bdnz	Lppcasm_mw_LOOP
 
 Lppcasm_mw_REM:
 	andi.	r5,r5,0x3
@@ -1951,7 +1951,7 @@ Lppcasm_maw_mainloop:
 	$ST	r11,`3*$BNSZ`(r3)
 	addi	r3,r3,`4*$BNSZ`
 	addi	r4,r4,`4*$BNSZ`
-	bdnz-	Lppcasm_maw_mainloop
+	bdnz	Lppcasm_maw_mainloop
 	
 Lppcasm_maw_leftover:
 	andi.	r5,r5,0x3
diff --git a/crypto/bn/asm/ppc64-mont.pl b/crypto/bn/asm/ppc64-mont.pl
index 9e3c12d788..595fc6d31f 100644
--- a/crypto/bn/asm/ppc64-mont.pl
+++ b/crypto/bn/asm/ppc64-mont.pl
@@ -734,7 +734,7 @@ $code.=<<___;
 ___
 }
 $code.=<<___;
-	bdnz-	L1st
+	bdnz	L1st
 
 	fctid	$dota,$dota
 	fctid	$dotb,$dotb
@@ -1280,7 +1280,7 @@ $code.=<<___;
 ___
 }
 $code.=<<___;
-	bdnz-	Linner
+	bdnz	Linner
 
 	fctid	$dota,$dota
 	fctid	$dotb,$dotb
@@ -1490,7 +1490,7 @@ Lsub:	ldx	$t0,$tp,$i
 	stdx	$t0,$rp,$i
 	stdx	$t2,$t6,$i
 	addi	$i,$i,16
-	bdnz-	Lsub
+	bdnz	Lsub
 
 	li	$i,0
 	subfe	$ovf,$i,$ovf	; handle upmost overflow bit
@@ -1517,7 +1517,7 @@ Lcopy:				; copy or in-place refresh
 	stdx	$i,$tp,$i	; zap tp at once
 	stdx	$i,$t4,$i
 	addi	$i,$i,16
-	bdnz-	Lcopy
+	bdnz	Lcopy
 ___
 $code.=<<___ if ($SIZE_T==4);
 	subf	$np,$num,$np	; rewind np
@@ -1550,7 +1550,7 @@ Lsub:	lwz	$t0,12($tp)	; load tp[j..j+3] in 64-bit word order
 	stw	$t5,8($rp)
 	stw	$t6,12($rp)
 	stwu	$t7,16($rp)
-	bdnz-	Lsub
+	bdnz	Lsub
 
 	li	$i,0
 	subfe	$ovf,$i,$ovf	; handle upmost overflow bit
@@ -1582,7 +1582,7 @@ Lcopy:				; copy or in-place refresh
 	stwu	$t3,16($rp)
 	std	$i,8($tp)	; zap tp at once
 	stdu	$i,16($tp)
-	bdnz-	Lcopy
+	bdnz	Lcopy
 ___
 
 $code.=<<___;
-- 
cgit v1.2.3