← Commit list

4923f74e5

Denys Vlasenko

2022-02-08

SUCCESS Duration: 155s

Subject

libbb/sha1: shrink unrolled x86-64 code

Binary

No binary (build failed)

Architecture & Compiler Flags

Architecture: x86_64-linux-gnu Compiler: gcc (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0

—

Build Log

(log yok)

Commit Diff GitHub →

libbb/hash_md5_sha_x86-64.S +10 -23 libbb/hash_md5_sha_x86-64.S.sh +11 -23

libbb/hash_md5_sha_x86-64.S+10 -23

@@ -24,6 +24,7 @@ sha1_process_block64:
	# xmm0..xmm3: W[]
	# xmm4,xmm5: temps
	# xmm6: current round constant
+	# xmm7: all round constants
	# -64(%rsp): area for passing RCONST + W[] from vector to integer units

	movl 80(%rdi), %eax # a = ctx->hash[0]
@@ -32,16 +33,17 @@ sha1_process_block64:
	movl 92(%rdi), %edx # d = ctx->hash[3]
	movl 96(%rdi), %ebp # e = ctx->hash[4]

-	movaps rconst0x5A827999(%rip), %xmm6
+	movaps sha1const(%rip), %xmm7
+	pshufd $0x00, %xmm7, %xmm6

	# Load W[] to xmm registers, byteswapping on the fly.
	#
	# For iterations 0..15, we pass W[] in rsi,r8..r14
-	# for use in RD1A's instead of spilling them to stack.
+	# for use in RD1As instead of spilling them to stack.
	# We lose parallelized addition of RCONST, but LEA
-	# can do two additions at once, so it's probably a wash.
+	# can do two additions at once, so it is probably a wash.
	# (We use rsi instead of rN because this makes two
-	# LEAs in two first RD1A's shorter by one byte).
+	# LEAs in two first RD1As shorter by one byte).
	movq 4*0(%rdi), %rsi
	movq 4*2(%rdi), %r8
	bswapq %rsi
@@ -253,7 +255,7 @@ sha1_process_block64:
	roll $5, %edi # rotl32(a,5)
	addl %edi, %edx # e += rotl32(a,5)
	rorl $2, %eax # b = rotl32(b,30)
-	movaps rconst0x6ED9EBA1(%rip), %xmm6
+	pshufd $0x55, %xmm7, %xmm6
	# PREP %xmm1 %xmm2 %xmm3 %xmm0 -64+16*1(%rsp)
	movaps %xmm0, %xmm4
	psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0)
@@ -614,7 +616,7 @@ sha1_process_block64:
	roll $5, %esi # rotl32(a,5)
	addl %esi, %edx # e += rotl32(a,5)
	rorl $2, %eax # b = rotl32(b,30)
-	movaps rconst0x8F1BBCDC(%rip), %xmm6
+	pshufd $0xaa, %xmm7, %xmm6
	# PREP %xmm2 %xmm3 %xmm0 %xmm1 -64+16*2(%rsp)
	movaps %xmm1, %xmm4
	psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0)
@@ -1001,7 +1003,7 @@ sha1_process_block64:
	roll $5, %esi # rotl32(a,5)
	addl %esi, %edx # e += rotl32(a,5)
	rorl $2, %eax # b = rotl32(b,30)
-	movaps rconst0xCA62C1D6(%rip), %xmm6
+	pshufd $0xff, %xmm7, %xmm6
	# PREP %xmm3 %xmm0 %xmm1 %xmm2 -64+16*3(%rsp)
	movaps %xmm2, %xmm4
	psrldq $4, %xmm4 # rshift by 4 bytes: T1 = ([13],[14],[15],0)
@@ -1475,25 +1477,10 @@ sha1_process_block64:

	.section .rodata.cst16.sha1const, "aM", @progbits, 16
	.balign 16
-	rconst0x5A827999:
+	sha1const:
	.long 0x5A827999
-	.long 0x5A827999
-	.long 0x5A827999
-	.long 0x5A827999
-	rconst0x6ED9EBA1:
-	.long 0x6ED9EBA1
-	.long 0x6ED9EBA1
-	.long 0x6ED9EBA1
	.long 0x6ED9EBA1
-	rconst0x8F1BBCDC:
	.long 0x8F1BBCDC
-	.long 0x8F1BBCDC
-	.long 0x8F1BBCDC
-	.long 0x8F1BBCDC
-	rconst0xCA62C1D6:
-	.long 0xCA62C1D6
-	.long 0xCA62C1D6
-	.long 0xCA62C1D6
	.long 0xCA62C1D6

	#endif

libbb/hash_md5_sha_x86-64.S.sh+11 -23

@@ -34,6 +34,7 @@ exec >hash_md5_sha_x86-64.S
	xmmT1="%xmm4"
	xmmT2="%xmm5"
	xmmRCONST="%xmm6"
+	xmmALLRCONST="%xmm7"
	T=`printf '\t'`

	# SSE instructions are longer than 4 bytes on average.
@@ -125,6 +126,7 @@ sha1_process_block64:
	# xmm0..xmm3: W[]
	# xmm4,xmm5: temps
	# xmm6: current round constant
+	# xmm7: all round constants
	# -64(%rsp): area for passing RCONST + W[] from vector to integer units

	movl 80(%rdi), %eax # a = ctx->hash[0]
@@ -133,16 +135,17 @@ sha1_process_block64:
	movl 92(%rdi), %edx # d = ctx->hash[3]
	movl 96(%rdi), %ebp # e = ctx->hash[4]

-	movaps rconst0x5A827999(%rip), $xmmRCONST
+	movaps sha1const(%rip), $xmmALLRCONST
+	pshufd \$0x00, $xmmALLRCONST, $xmmRCONST

	# Load W[] to xmm registers, byteswapping on the fly.
	#
	# For iterations 0..15, we pass W[] in rsi,r8..r14
-	# for use in RD1A's instead of spilling them to stack.
+	# for use in RD1As instead of spilling them to stack.
	# We lose parallelized addition of RCONST, but LEA
-	# can do two additions at once, so it's probably a wash.
+	# can do two additions at once, so it is probably a wash.
	# (We use rsi instead of rN because this makes two
-	# LEAs in two first RD1A's shorter by one byte).
+	# LEAs in two first RD1As shorter by one byte).
	movq 4*0(%rdi), %rsi
	movq 4*2(%rdi), %r8
	bswapq %rsi
@@ -359,7 +362,7 @@ RD1A bx cx dx bp ax 4; RD1A ax bx cx dx bp 5; RD1A bp ax bx cx dx 6; RD1A dx
	a=`PREP %xmm0 %xmm1 %xmm2 %xmm3 "-64+16*0(%rsp)"`
	b=`RD1A cx dx bp ax bx 8; RD1A bx cx dx bp ax 9; RD1A ax bx cx dx bp 10; RD1A bp ax bx cx dx 11;`
	INTERLEAVE "$a" "$b"
-	a=`echo " movaps rconst0x6ED9EBA1(%rip), $xmmRCONST"
+	a=`echo " pshufd \\$0x55, $xmmALLRCONST, $xmmRCONST"
	PREP %xmm1 %xmm2 %xmm3 %xmm0 "-64+16*1(%rsp)"`
	b=`RD1A dx bp ax bx cx 12; RD1A cx dx bp ax bx 13; RD1A bx cx dx bp ax 14; RD1A ax bx cx dx bp 15;`
	INTERLEAVE "$a" "$b"
@@ -378,7 +381,7 @@ INTERLEAVE "$a" "$b"
	a=`PREP %xmm1 %xmm2 %xmm3 %xmm0 "-64+16*1(%rsp)"`
	b=`RD2 cx dx bp ax bx 28; RD2 bx cx dx bp ax 29; RD2 ax bx cx dx bp 30; RD2 bp ax bx cx dx 31;`
	INTERLEAVE "$a" "$b"
-	a=`echo " movaps rconst0x8F1BBCDC(%rip), $xmmRCONST"
+	a=`echo " pshufd \\$0xaa, $xmmALLRCONST, $xmmRCONST"
	PREP %xmm2 %xmm3 %xmm0 %xmm1 "-64+16*2(%rsp)"`
	b=`RD2 dx bp ax bx cx 32; RD2 cx dx bp ax bx 33; RD2 bx cx dx bp ax 34; RD2 ax bx cx dx bp 35;`
	INTERLEAVE "$a" "$b"
@@ -397,7 +400,7 @@ INTERLEAVE "$a" "$b"
	a=`PREP %xmm2 %xmm3 %xmm0 %xmm1 "-64+16*2(%rsp)"`
	b=`RD3 cx dx bp ax bx 48; RD3 bx cx dx bp ax 49; RD3 ax bx cx dx bp 50; RD3 bp ax bx cx dx 51;`
	INTERLEAVE "$a" "$b"
-	a=`echo " movaps rconst0xCA62C1D6(%rip), $xmmRCONST"
+	a=`echo " pshufd \\$0xff, $xmmALLRCONST, $xmmRCONST"
	PREP %xmm3 %xmm0 %xmm1 %xmm2 "-64+16*3(%rsp)"`
	b=`RD3 dx bp ax bx cx 52; RD3 cx dx bp ax bx 53; RD3 bx cx dx bp ax 54; RD3 ax bx cx dx bp 55;`
	INTERLEAVE "$a" "$b"
@@ -439,25 +442,10 @@ echo "

	.section .rodata.cst16.sha1const, \"aM\", @progbits, 16
	.balign 16
-	rconst0x5A827999:
+	sha1const:
	.long 0x5A827999
-	.long 0x5A827999
-	.long 0x5A827999
-	.long 0x5A827999
-	rconst0x6ED9EBA1:
-	.long 0x6ED9EBA1
-	.long 0x6ED9EBA1
-	.long 0x6ED9EBA1
	.long 0x6ED9EBA1
-	rconst0x8F1BBCDC:
	.long 0x8F1BBCDC
-	.long 0x8F1BBCDC
-	.long 0x8F1BBCDC
-	.long 0x8F1BBCDC
-	rconst0xCA62C1D6:
-	.long 0xCA62C1D6
-	.long 0xCA62C1D6
-	.long 0xCA62C1D6
	.long 0xCA62C1D6

	#endif"