Update assembly code for lsh256 and lsh512 implementations

RyuaNerin · Jun 19, 2024 · 5adc6be · 5adc6be
1 parent 0297c3c
commit 5adc6be
Show file tree

Hide file tree

Showing 6 changed files with 23 additions and 61 deletions.
diff --git a/lsh256/asm_amd64_avx2.s b/lsh256/asm_amd64_avx2.s
@@ -9,7 +9,7 @@
 
 #include "textflag.h"
 
-TEXT ·__lsh256_avx2_init(SB), NOSPLIT, $16
+TEXT ·__lsh256_avx2_init(SB), NOSPLIT, $0-16
 	MOVQ	ctx+0(FP), DI
 	MOVQ	algtype+8(FP), SI
 
@@ -54,7 +54,7 @@ DATA LCDATA2<>+0x070(SB)/8, $0x0000000400000007
 DATA LCDATA2<>+0x078(SB)/8, $0x0000000600000005
 GLOBL LCDATA2<>(SB), RODATA|NOPTR, $128
 
-TEXT ·__lsh256_avx2_update(SB), NOSPLIT, $32
+TEXT ·__lsh256_avx2_update(SB), NOSPLIT, $0-32
 	MOVQ ctx+0(FP), DI
 	MOVQ data_base+8(FP), SI
 	MOVQ data_len+16(FP), DX
@@ -543,7 +543,7 @@ LBB1_34:
 	JE   LBB1_41
 	JMP  LBB1_36
 
-TEXT ·__lsh256_avx2_final(SB), NOSPLIT, $16
+TEXT ·__lsh256_avx2_final(SB), NOSPLIT, $0-16
 	MOVQ ctx+0(FP), DI
 	MOVQ hashval+8(FP), SI
 

diff --git a/lsh256/asm_amd64_sse2.s b/lsh256/asm_amd64_sse2.s
@@ -9,7 +9,7 @@
 
 #include "textflag.h"
 
-TEXT ·__lsh256_sse2_init(SB), NOSPLIT, $16
+TEXT ·__lsh256_sse2_init(SB), NOSPLIT, $0-16
 	MOVQ	ctx+0(FP), DI
 	MOVQ	algtype+8(FP), SI
 
@@ -56,17 +56,13 @@ DATA LCDATA2<>+0x070(SB)/8, $0x0000000000000000
 DATA LCDATA2<>+0x078(SB)/8, $0x0000000000000000
 GLOBL LCDATA2<>(SB), RODATA|NOPTR, $128
 
-TEXT ·__lsh256_sse2_update(SB), NOSPLIT, $128-32
+// stacksize 96
+TEXT ·__lsh256_sse2_update(SB), NOSPLIT, $96-32
 	MOVQ ctx+0(FP), DI
 	MOVQ data_base+8(FP), SI
 	MOVQ data_len+16(FP), DX
 	//   data_cap+24
 
-	// stack 96
-	MOVQ SP, BP
-	ADDQ $16, SP
-	ANDQ $-16, SP
-	MOVQ BP, 96(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	WORD $0x4f8b; BYTE $0x10 // mov    ecx, dword [rdi + 16]
@@ -1040,7 +1036,6 @@ LBB1_68:
 	WORD $0x4f89; BYTE $0x10 // mov    dword [rdi + 16], ecx
 
 LBB1_69:
-	MOVQ 96(SP), SP
 	RET
 
 LBB1_12:
@@ -1169,15 +1164,11 @@ LBB1_34:
 	JMP  LBB1_36
 
 
-TEXT ·__lsh256_sse2_final(SB), NOSPLIT, $80-16
+// stacksize 64
+TEXT ·__lsh256_sse2_final(SB), NOSPLIT, $64-16
 	MOVQ ctx+0(FP), DI
 	MOVQ hashval+8(FP), SI
 
-	// stack size: 64
-	MOVQ SP, BP
-	ADDQ $16, SP
-	ANDQ $-16, SP
-	MOVQ BP, 64(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	LONG $0x10478b44               // mov    r8d, dword [rdi + 16]
@@ -1633,6 +1624,4 @@ LBB2_14:
 	LONG $0xef0f4166; BYTE $0xc5               // pxor    xmm0, xmm13
 	LONG $0x0e7f0ff3                           // movdqu    oword [rsi], xmm1
 	LONG $0x467f0ff3; BYTE $0x10               // movdqu    oword [rsi + 16], xmm0
-
-	MOVQ 64(SP), SP
 	RET
diff --git a/lsh256/asm_amd64_ssse3.s b/lsh256/asm_amd64_ssse3.s
@@ -23,7 +23,7 @@ DATA LCDATA2<>+0x050(SB)/8, $0x1359621b185fe69e
 DATA LCDATA2<>+0x058(SB)/8, $0x1a116870263fccb2
 GLOBL LCDATA2<>(SB), RODATA|NOPTR, $96
 
-TEXT ·__lsh256_ssse3_update(SB), NOSPLIT, $32
+TEXT ·__lsh256_ssse3_update(SB), NOSPLIT, $0-32
 	MOVQ ctx+0(FP), DI
 	MOVQ data_base+8(FP), SI
 	MOVQ data_len+16(FP), DX
@@ -708,7 +708,7 @@ LBB1_34:
 	JMP  LBB1_36
 
 
-TEXT ·__lsh256_ssse3_final(SB), NOSPLIT, $16
+TEXT ·__lsh256_ssse3_final(SB), NOSPLIT, $0-16
 	MOVQ ctx+0(FP), DI
 	MOVQ hashval_base+8(FP), SI
 

diff --git a/lsh512/asm_amd64_avx2.s b/lsh512/asm_amd64_avx2.s
@@ -9,7 +9,7 @@
 
 #include "textflag.h"
 
-TEXT ·__lsh512_avx2_init(SB), NOSPLIT, $16
+TEXT ·__lsh512_avx2_init(SB), NOSPLIT, $0-16
 	MOVQ	ctx+0(FP), DI
 	MOVQ	algtype+8(FP), SI
 
@@ -84,17 +84,13 @@ DATA LCDATA2<>+0x0e0(SB)/8, $0x0605040302010007
 DATA LCDATA2<>+0x0e8(SB)/8, $0x080f0e0d0c0b0a09
 GLOBL LCDATA2<>(SB), RODATA|NOPTR, $240
 
-TEXT ·__lsh512_avx2_update(SB), NOSPLIT, $224-32
+	// stack size: 192
+TEXT ·__lsh512_avx2_update(SB), NOSPLIT, $192-32
 	MOVQ ctx+0(FP), DI
 	MOVQ data_base+8(FP), SI
 	MOVQ data_len+16(FP), DX
 	//   data_cap+24
 
-	// stack size: 192
-	MOVQ SP, BP
-	ADDQ $32, SP
-	ANDQ $-32, SP
-	MOVQ BP, 192(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	WORD $0x4f8b; BYTE $0x10                   // mov    ecx, dword [rdi + 16]
@@ -648,7 +644,6 @@ LBB1_68:
 	WORD $0x4f89; BYTE $0x10 // mov    dword [rdi + 16], ecx
 
 LBB1_69:
-	MOVQ 192(SP), SP
 	VZEROUPPER
 	RET
 
@@ -789,7 +784,7 @@ LBB1_34:
 	JE   LBB1_41
 	JMP  LBB1_36
 
-TEXT ·__lsh512_avx2_final(SB), NOSPLIT, $16
+TEXT ·__lsh512_avx2_final(SB), NOSPLIT, $0-16
 	MOVQ ctx+0(FP), DI
 	MOVQ hashval+8(FP), SI
 

diff --git a/lsh512/asm_amd64_sse2.s b/lsh512/asm_amd64_sse2.s
@@ -9,7 +9,7 @@
 
 #include "textflag.h"
 
-TEXT ·__lsh512_sse2_init(SB), NOSPLIT, $16
+TEXT ·__lsh512_sse2_init(SB), NOSPLIT, $0-16
 	MOVQ	ctx+0(FP), DI
 	MOVQ	algtype+8(FP), SI
 
@@ -77,17 +77,13 @@ DATA LCDATA2<>+0x070(SB)/8, $0x6cc37895f4ad9e70
 DATA LCDATA2<>+0x078(SB)/8, $0x448304c8d7f3f4d5
 GLOBL LCDATA2<>(SB), RODATA|NOPTR, $128
 
-TEXT ·__lsh512_sse2_update(SB), NOSPLIT, $320-32
+// stack size: 288
+TEXT ·__lsh512_sse2_update(SB), NOSPLIT, $288-32
 	MOVQ ctx+0(FP), DI
 	MOVQ data_base+8(FP), SI
 	MOVQ data_len+16(FP), DX
 	//   data_cap+24
 
-	// stack size: 288
-	MOVQ SP, BP
-	ADDQ $16, SP
-	ANDQ $-16, SP
-	MOVQ BP, 288(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	WORD $0x4f8b; BYTE $0x10                   // mov    ecx, dword [rdi + 16]
@@ -1467,7 +1463,6 @@ LBB1_69:
 	WORD $0x4f89; BYTE $0x10 // mov    dword [rdi + 16], ecx
 
 LBB1_70:
-	MOVQ 288(SP), SP
 	RET
 
 LBB1_12:
@@ -1595,15 +1590,11 @@ LBB1_34:
 	JE   LBB1_41
 	JMP  LBB1_36
 
-TEXT ·__lsh512_sse2_final(SB), NOSPLIT, $288-16
+// Stack size: 272
+TEXT ·__lsh512_sse2_final(SB), NOSPLIT, $272-16
 	MOVQ ctx+0(FP), DI
 	MOVQ hashval+8(FP), SI
 
-	// Stack size: 272
-	MOVQ SP, BP
-	ADDQ $16, SP
-	ANDQ $-16, SP
-	MOVQ BP, 272(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	LONG $0x10478b44                     // mov    r8d, dword [rdi + 16]
@@ -2265,6 +2256,4 @@ LBB2_14:
 	LONG $0x5e7f0ff3; BYTE $0x10                       // movdqu    oword [rsi + 16], xmm3
 	LONG $0x567f0ff3; BYTE $0x20                       // movdqu    oword [rsi + 32], xmm2
 	LONG $0x7f0f44f3; WORD $0x3046                     // movdqu    oword [rsi + 48], xmm8
-
-	MOVQ 272(SP), SP
 	RET
diff --git a/lsh512/asm_amd64_ssse3.s b/lsh512/asm_amd64_ssse3.s
@@ -31,17 +31,13 @@ DATA LCDATA2<>+0x090(SB)/8, $0x6cc37895f4ad9e70
 DATA LCDATA2<>+0x098(SB)/8, $0x448304c8d7f3f4d5
 GLOBL LCDATA2<>(SB), RODATA|NOPTR, $160
 
-TEXT ·__lsh512_ssse3_update(SB), NOSPLIT, $320-32
+// stack size: 288
+TEXT ·__lsh512_ssse3_update(SB), NOSPLIT, $288-32
 	MOVQ ctx+0(FP), DI
 	MOVQ data_base+8(FP), SI
 	MOVQ data_len+16(FP), DX
 	//   data_cap+24
 
-	// stack size: 288
-	MOVQ SP, BP
-	ADDQ $16, SP
-	ANDQ $-16, SP
-	MOVQ BP, 288(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	WORD $0x4f8b; BYTE $0x10                   // mov    ecx, dword [rdi + 16]
@@ -1138,7 +1134,6 @@ LBB1_69:
 	WORD $0x4f89; BYTE $0x10 // mov    dword [rdi + 16], ecx
 
 LBB1_70:
-	MOVQ 288(SP), SP
 	RET
 
 LBB1_14:
@@ -1266,15 +1261,11 @@ LBB1_36:
 	JE   LBB1_43
 	JMP  LBB1_38
 
-TEXT ·__lsh512_ssse3_final(SB), NOSPLIT, $288-16
+// Stack size: 272
+TEXT ·__lsh512_ssse3_final(SB), NOSPLIT, $272-16
 	MOVQ ctx+0(FP), DI
 	MOVQ hashval+8(FP), SI
 
-	// Stack size: 272
-	MOVQ SP, BP
-	ADDQ $16, SP
-	ANDQ $-16, SP
-	MOVQ BP, 272(SP)
 	LEAQ LCDATA2<>(SB), BP
 
 	LONG $0x10478b44                     // mov    r8d, dword [rdi + 16]
@@ -1785,6 +1776,4 @@ LBB2_14:
 	LONG $0x5e7f0ff3; BYTE $0x10               // movdqu    oword [rsi + 16], xmm3
 	LONG $0x6e7f0ff3; BYTE $0x20               // movdqu    oword [rsi + 32], xmm5
 	LONG $0x7f0f44f3; WORD $0x3066             // movdqu    oword [rsi + 48], xmm12
-
-	MOVQ 272(SP), SP
 	RET