From 16219eef8ab0fc7e6378113b3be0783ec3519893 Mon Sep 17 00:00:00 2001
From: Sun Yimin <emmansun@users.noreply.github.com>
Date: Thu, 7 Nov 2024 17:19:53 +0800
Subject: [PATCH] ppc64x: sm4/zuc reduce VAND

---
 sm4/aesni_macros_ppc64x.s | 6 ++----
 sm4/asm_ppc64x.s          | 6 +++---
 sm4/cbc_ppc64x.s          | 2 +-
 sm4/ecb_ppc64x.s          | 2 +-
 zuc/asm_ppc64x.s          | 5 ++---
 5 files changed, 9 insertions(+), 12 deletions(-)

diff --git a/sm4/aesni_macros_ppc64x.s b/sm4/aesni_macros_ppc64x.s
index c2a37f8..3d7ab9b 100644
--- a/sm4/aesni_macros_ppc64x.s
+++ b/sm4/aesni_macros_ppc64x.s
@@ -86,8 +86,7 @@
 #define AFFINE_TRANSFORM(L, H, V_FOUR, x, y, z)  \
 	VAND NIBBLE_MASK, x, z;              \
 	VPERM L, L, z, y;                    \
-	VSRD x, V_FOUR, x;                   \
-	VAND NIBBLE_MASK, x, z;              \
+	VSRB x, V_FOUR, z;                   \
 	VPERM H, H, z, x;                    \
 	VXOR y, x, x
 
@@ -102,8 +101,7 @@
 	VNOR  x, x, z;                       \ // z = NOT(x)
 	VAND  NIBBLE_MASK, z, z;             \	
 	VPERM L, L, z, y;                    \
-	VSRD x, V_FOUR, x;                   \
-	VAND NIBBLE_MASK, x, z;              \
+	VSRB x, V_FOUR, z;                   \
 	VPERM H, H, z, x;                    \
 	VXOR y, x, x
 
diff --git a/sm4/asm_ppc64x.s b/sm4/asm_ppc64x.s
index 54f092b..85dc631 100644
--- a/sm4/asm_ppc64x.s
+++ b/sm4/asm_ppc64x.s
@@ -60,7 +60,7 @@ GLOBL ·rcon(SB), RODATA, $112
 // func expandKeyAsm(key *byte, ck, enc, dec *uint32, inst int)
 TEXT ·expandKeyAsm(SB),NOSPLIT,$0
 	// prepare/load constants
-	VSPLTISW $4, V_FOUR;
+	VSPLTISB $4, V_FOUR;
 #ifdef NEEDS_PERMW
 	MOVD	$·rcon(SB), R4
 	LVX	(R4), ESPERMW
@@ -115,7 +115,7 @@ ksLoop:
 // func encryptBlockAsm(xk *uint32, dst, src *byte, inst int)
 TEXT ·encryptBlockAsm(SB),NOSPLIT,$0
 	// prepare/load constants
-	VSPLTISW $4, V_FOUR;
+	VSPLTISB $4, V_FOUR;
 #ifdef NEEDS_PERMW
 	MOVD	$·rcon(SB), R4
 	LVX	(R4), ESPERMW
@@ -156,7 +156,7 @@ encryptBlockLoop:
 // func encryptBlocksAsm(xk *uint32, dst, src []byte, inst int)
 TEXT ·encryptBlocksAsm(SB),NOSPLIT,$0
 	// prepare/load constants
-	VSPLTISW $4, V_FOUR;
+	VSPLTISB $4, V_FOUR;
 #ifdef NEEDS_PERMW
 	MOVD	$·rcon(SB), R4
 	LVX	(R4), ESPERMW
diff --git a/sm4/cbc_ppc64x.s b/sm4/cbc_ppc64x.s
index 8371224..e1f9b0e 100644
--- a/sm4/cbc_ppc64x.s
+++ b/sm4/cbc_ppc64x.s
@@ -45,7 +45,7 @@ TEXT ·decryptBlocksChain(SB),NOSPLIT,$0
 #define rk R5
 #define srcLen R6
 	// prepare/load constants
-	VSPLTISW $4, V_FOUR;
+	VSPLTISB $4, V_FOUR;
 #ifdef NEEDS_PERMW
 	MOVD	$·rcon(SB), R4
 	LVX	(R4), ESPERMW
diff --git a/sm4/ecb_ppc64x.s b/sm4/ecb_ppc64x.s
index 14b3316..8d61b07 100644
--- a/sm4/ecb_ppc64x.s
+++ b/sm4/ecb_ppc64x.s
@@ -30,7 +30,7 @@ TEXT ·encryptSm4Ecb(SB),NOSPLIT,$0
 #define rk R5
 #define srcLen R6
 	// prepare/load constants
-	VSPLTISW $4, V_FOUR;
+	VSPLTISB $4, V_FOUR;
 #ifdef NEEDS_PERMW
 	MOVD	$·rcon(SB), R4
 	LVX	(R4), ESPERMW
diff --git a/zuc/asm_ppc64x.s b/zuc/asm_ppc64x.s
index b32b64d..d8e6999 100644
--- a/zuc/asm_ppc64x.s
+++ b/zuc/asm_ppc64x.s
@@ -42,7 +42,7 @@ GLOBL rcon<>(SB), RODATA, $160
 #define P3 V30
 
 #define LOAD_CONSTS \
-	VSPLTISW $4, V_FOUR \
+	VSPLTISB $4, V_FOUR \
 	MOVD $rcon<>+0x00(SB), R4 \
 	LXVD2X (R4)(R0), NIBBLE_MASK \
 	MOVD $0x10, R5 \
@@ -89,8 +89,7 @@ GLOBL rcon<>(SB), RODATA, $160
 #define AFFINE_TRANSFORM(L, H, V_FOUR, x, y, z)  \
 	VAND NIBBLE_MASK, x, z;              \
 	VPERM L, L, z, y;                    \
-	VSRD x, V_FOUR, x;                   \
-	VAND NIBBLE_MASK, x, z;              \
+	VSRB x, V_FOUR, z;                   \
 	VPERM H, H, z, x;                    \
 	VXOR y, x, x