[git commit] libbb/sha1: shrink x86_64 version - use r8..15 for W[8..15]
Denys Vlasenko
vda.linux at googlemail.com
Sat Jan 1 14:01:53 UTC 2022
commit: https://git.busybox.net/busybox/commit/?id=d643010feeef312c77d7f51c3dd476d4e605c982
branch: https://git.busybox.net/busybox/commit/?id=refs/heads/master
function old new delta
sha1_process_block64 3683 3562 -121
Signed-off-by: Denys Vlasenko <vda.linux at googlemail.com>
---
libbb/Config.src | 2 +-
libbb/hash_md5_sha.c | 299 ++++++++++++++++++++++++++++++++++++++++-----------
2 files changed, 240 insertions(+), 61 deletions(-)
diff --git a/libbb/Config.src b/libbb/Config.src
index e027c14a8..f66f65f81 100644
--- a/libbb/Config.src
+++ b/libbb/Config.src
@@ -59,7 +59,7 @@ config SHA1_SMALL
Trade binary size versus speed for the sha1 algorithm.
throughput MB/s size of sha1_process_block64
value 486 x86-64 486 x86-64
- 0 367 367 3657 3683
+ 0 367 367 3657 3562
1 224 229 654 732
2,3 200 195 358 380
diff --git a/libbb/hash_md5_sha.c b/libbb/hash_md5_sha.c
index 9de30dfe6..a4e36066a 100644
--- a/libbb/hash_md5_sha.c
+++ b/libbb/hash_md5_sha.c
@@ -700,22 +700,194 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
{
BUILD_BUG_ON(offsetof(sha1_ctx_t, hash) != 80);
asm(
-// TODO: store W[] in r8..r15? (r8..r11 are callee-clobbered, no need to save)
"\n\
- ##pushq %r15 # \n\
- ##pushq %r14 # \n\
- ##pushq %r13 # \n\
- ##pushq %r12 # \n\
- ##pushq %rbp # \n\
- ##pushq %rbx # \n\
- movq %rbp, %r8 # callee-saved \n\
- movq %rbx, %r9 # callee-saved \n\
- movq %rdi, %r10 # we need ctx at the end \n\
- movl $15, %eax \n\
+ pushq %r15 # \n\
+ pushq %r14 # \n\
+ pushq %r13 # \n\
+ pushq %r12 # \n\
+ pushq %rbp # \n\
+ pushq %rbx # \n\
+ pushq %rdi # we need ctx at the end \n\
+ \n\
+#Register and stack use: \n\
+# eax..edx: a..d \n\
+# ebp: e \n\
+# esi,edi: temps \n\
+# -32+4*n(%rsp),r8...r15: W[0..7,8..15] \n\
+ .macro loadW n,r \n\
+ .if \\n == 0 \n\
+ movl -32+4*0(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 1 \n\
+ movl -32+4*1(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 2 \n\
+ movl -32+4*2(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 3 \n\
+ movl -32+4*3(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 4 \n\
+ movl -32+4*4(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 5 \n\
+ movl -32+4*5(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 6 \n\
+ movl -32+4*6(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 7 \n\
+ movl -32+4*7(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 8 \n\
+ movl %r8d,\\r \n\
+ .endif \n\
+ .if \\n == 9 \n\
+ movl %r9d,\\r \n\
+ .endif \n\
+ .if \\n == 10 \n\
+ movl %r10d,\\r \n\
+ .endif \n\
+ .if \\n == 11 \n\
+ movl %r11d,\\r \n\
+ .endif \n\
+ .if \\n == 12 \n\
+ movl %r12d,\\r \n\
+ .endif \n\
+ .if \\n == 13 \n\
+ movl %r13d,\\r \n\
+ .endif \n\
+ .if \\n == 14 \n\
+ movl %r14d,\\r \n\
+ .endif \n\
+ .if \\n == 15 \n\
+ movl %r15d,\\r \n\
+ .endif \n\
+ .endm \n\
+ \n\
+ .macro storeW r,n \n\
+ .if \\n == 0 \n\
+ movl \\r,-32+4*0(%rsp) \n\
+ .endif \n\
+ .if \\n == 1 \n\
+ movl \\r,-32+4*1(%rsp) \n\
+ .endif \n\
+ .if \\n == 2 \n\
+ movl \\r,-32+4*2(%rsp) \n\
+ .endif \n\
+ .if \\n == 3 \n\
+ movl \\r,-32+4*3(%rsp) \n\
+ .endif \n\
+ .if \\n == 4 \n\
+ movl \\r,-32+4*4(%rsp) \n\
+ .endif \n\
+ .if \\n == 5 \n\
+ movl \\r,-32+4*5(%rsp) \n\
+ .endif \n\
+ .if \\n == 6 \n\
+ movl \\r,-32+4*6(%rsp) \n\
+ .endif \n\
+ .if \\n == 7 \n\
+ movl \\r,-32+4*7(%rsp) \n\
+ .endif \n\
+ .if \\n == 8 \n\
+ movl \\r,%r8d \n\
+ .endif \n\
+ .if \\n == 9 \n\
+ movl \\r,%r9d \n\
+ .endif \n\
+ .if \\n == 10 \n\
+ movl \\r,%r10d \n\
+ .endif \n\
+ .if \\n == 11 \n\
+ movl \\r,%r11d \n\
+ .endif \n\
+ .if \\n == 12 \n\
+ movl \\r,%r12d \n\
+ .endif \n\
+ .if \\n == 13 \n\
+ movl \\r,%r13d \n\
+ .endif \n\
+ .if \\n == 14 \n\
+ movl \\r,%r14d \n\
+ .endif \n\
+ .if \\n == 15 \n\
+ movl \\r,%r15d \n\
+ .endif \n\
+ .endm \n\
+ \n\
+ .macro xorW n,r \n\
+ .if \\n == 0 \n\
+ xorl -32+4*0(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 1 \n\
+ xorl -32+4*1(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 2 \n\
+ xorl -32+4*2(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 3 \n\
+ xorl -32+4*3(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 4 \n\
+ xorl -32+4*4(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 5 \n\
+ xorl -32+4*5(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 6 \n\
+ xorl -32+4*6(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 7 \n\
+ xorl -32+4*7(%rsp),\\r \n\
+ .endif \n\
+ .if \\n == 8 \n\
+ xorl %r8d,\\r \n\
+ .endif \n\
+ .if \\n == 9 \n\
+ xorl %r9d,\\r \n\
+ .endif \n\
+ .if \\n == 10 \n\
+ xorl %r10d,\\r \n\
+ .endif \n\
+ .if \\n == 11 \n\
+ xorl %r11d,\\r \n\
+ .endif \n\
+ .if \\n == 12 \n\
+ xorl %r12d,\\r \n\
+ .endif \n\
+ .if \\n == 13 \n\
+ xorl %r13d,\\r \n\
+ .endif \n\
+ .if \\n == 14 \n\
+ xorl %r14d,\\r \n\
+ .endif \n\
+ .if \\n == 15 \n\
+ xorl %r15d,\\r \n\
+ .endif \n\
+ .endm \n\
+ \n\
+ movl 4*8(%rdi), %r8d \n\
+ bswap %r8d \n\
+ movl 4*9(%rdi), %r9d \n\
+ bswap %r9d \n\
+ movl 4*10(%rdi), %r10d \n\
+ bswap %r10d \n\
+ movl 4*11(%rdi), %r11d \n\
+ bswap %r11d \n\
+ movl 4*12(%rdi), %r12d \n\
+ bswap %r12d \n\
+ movl 4*13(%rdi), %r13d \n\
+ bswap %r13d \n\
+ movl 4*14(%rdi), %r14d \n\
+ bswap %r14d \n\
+ movl 4*15(%rdi), %r15d \n\
+ bswap %r15d \n\
+ movl $7, %eax \n\
1: \n\
movl (%rdi,%rax,4), %esi \n\
bswap %esi \n\
- movl %esi, -64(%rsp,%rax,4) \n\
+ movl %esi, -32(%rsp,%rax,4) \n\
decl %eax \n\
jns 1b \n\
movl 80(%rdi), %eax # a = ctx->hash[0] \n\
@@ -723,15 +895,10 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
movl 88(%rdi), %ecx # c = ctx->hash[2] \n\
movl 92(%rdi), %edx # d = ctx->hash[3] \n\
movl 96(%rdi), %ebp # e = ctx->hash[4] \n\
-#Register and stack use: \n\
-# eax..edx: a..d \n\
-# ebp: e \n\
-# esi,edi: temps \n\
-# -64+4*n(%rsp): W[n] \n\
"
#define RD1As(a,b,c,d,e, n, RCONST) \
"\n\
- ##movl -64+4*"n"(%rsp), %esi # n=0, W[0] already in %esi \n\
+ ##loadW "n", %esi # n=0, W[0] already in %esi \n\
movl %e"c", %edi # c \n\
xorl %e"d", %edi # ^d \n\
andl %e"b", %edi # &b \n\
@@ -745,7 +912,7 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
"
#define RD1Bs(a,b,c,d,e, n, RCONST) \
"\n\
- movl -64+4*"n"(%rsp), %esi # W[n] \n\
+ loadW "n", %esi # W[n] \n\
movl %e"c", %edi # c \n\
xorl %e"d", %edi # ^d \n\
andl %e"b", %edi # &b \n\
@@ -757,14 +924,27 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
addl %esi, %e"e" # e += rotl32(a,5) \n\
rorl $2, %e"b" # b = rotl32(b,30) \n\
"
-#define RD1Cs(a,b,c,d,e, n13,n8,n2,n, RCONST) \
+#define RD1Cs(a,b,c,d,e, n, RCONST) \
"\n\
- movl -64+4*"n13"(%rsp), %esi # W[(n+13) & 15] \n\
- xorl -64+4*"n8"(%rsp), %esi # ^W[(n+8) & 15] \n\
- xorl -64+4*"n2"(%rsp), %esi # ^W[(n+2) & 15] \n\
- xorl -64+4*"n"(%rsp), %esi # ^W[n & 15] \n\
+ movl %e"c", %edi # c \n\
+ xorl %e"d", %edi # ^d \n\
+ andl %e"b", %edi # &b \n\
+ xorl %e"d", %edi # (((c ^ d) & b) ^ d) \n\
+ leal "RCONST"(%r"e",%r"n"), %e"e" # e += RCONST + W[n] \n\
+ addl %edi, %e"e" # e += (((c ^ d) & b) ^ d) \n\
+ movl %e"a", %esi # \n\
+ roll $5, %esi # rotl32(a,5) \n\
+ addl %esi, %e"e" # e += rotl32(a,5) \n\
+ rorl $2, %e"b" # b = rotl32(b,30) \n\
+"
+#define RD1Ds(a,b,c,d,e, n13,n8,n2,n, RCONST) \
+"\n\
+ loadW "n13", %esi # W[(n+13) & 15] \n\
+ xorW "n8", %esi # ^W[(n+8) & 15] \n\
+ xorW "n2", %esi # ^W[(n+2) & 15] \n\
+ xorW "n", %esi # ^W[n & 15] \n\
roll %esi # \n\
- movl %esi, -64+4*"n"(%rsp) # store to W[n & 15] \n\
+ storeW %esi, "n" # store to W[n & 15] \n\
movl %e"c", %edi # c \n\
xorl %e"d", %edi # ^d \n\
andl %e"b", %edi # &b \n\
@@ -776,23 +956,24 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
addl %esi, %e"e" # e += rotl32(a,5) \n\
rorl $2, %e"b" # b = rotl32(b,30) \n\
"
-#define RD1A(a,b,c,d,e, n) RD1As(STR(a),STR(b),STR(c),STR(d),STR(e), STR((n)), STR(RCONST))
-#define RD1B(a,b,c,d,e, n) RD1Bs(STR(a),STR(b),STR(c),STR(d),STR(e), STR((n)), STR(RCONST))
-#define RD1C(a,b,c,d,e, n) RD1Cs(STR(a),STR(b),STR(c),STR(d),STR(e), STR(((n+13)&15)), STR(((n+8)&15)), STR(((n+2)&15)), STR(((n)&15)), STR(RCONST))
+#define RD1A(a,b,c,d,e, n) RD1As(STR(a),STR(b),STR(c),STR(d),STR(e), STR(n), STR(RCONST))
+#define RD1B(a,b,c,d,e, n) RD1Bs(STR(a),STR(b),STR(c),STR(d),STR(e), STR(n), STR(RCONST))
+#define RD1C(a,b,c,d,e, n) RD1Cs(STR(a),STR(b),STR(c),STR(d),STR(e), STR(n), STR(RCONST))
+#define RD1D(a,b,c,d,e, n) RD1Ds(STR(a),STR(b),STR(c),STR(d),STR(e), STR(((n+13)&15)), STR(((n+8)&15)), STR(((n+2)&15)), STR(((n)&15)), STR(RCONST))
#undef RCONST
#define RCONST 0x5A827999
RD1A(ax,bx,cx,dx,bp, 0) RD1B(bp,ax,bx,cx,dx, 1) RD1B(dx,bp,ax,bx,cx, 2) RD1B(cx,dx,bp,ax,bx, 3) RD1B(bx,cx,dx,bp,ax, 4)
- RD1B(ax,bx,cx,dx,bp, 5) RD1B(bp,ax,bx,cx,dx, 6) RD1B(dx,bp,ax,bx,cx, 7) RD1B(cx,dx,bp,ax,bx, 8) RD1B(bx,cx,dx,bp,ax, 9)
- RD1B(ax,bx,cx,dx,bp,10) RD1B(bp,ax,bx,cx,dx,11) RD1B(dx,bp,ax,bx,cx,12) RD1B(cx,dx,bp,ax,bx,13) RD1B(bx,cx,dx,bp,ax,14)
- RD1B(ax,bx,cx,dx,bp,15) RD1C(bp,ax,bx,cx,dx,16) RD1C(dx,bp,ax,bx,cx,17) RD1C(cx,dx,bp,ax,bx,18) RD1C(bx,cx,dx,bp,ax,19)
+ RD1B(ax,bx,cx,dx,bp, 5) RD1B(bp,ax,bx,cx,dx, 6) RD1B(dx,bp,ax,bx,cx, 7) RD1C(cx,dx,bp,ax,bx, 8) RD1C(bx,cx,dx,bp,ax, 9)
+ RD1C(ax,bx,cx,dx,bp,10) RD1C(bp,ax,bx,cx,dx,11) RD1C(dx,bp,ax,bx,cx,12) RD1C(cx,dx,bp,ax,bx,13) RD1C(bx,cx,dx,bp,ax,14)
+ RD1C(ax,bx,cx,dx,bp,15) RD1D(bp,ax,bx,cx,dx,16) RD1D(dx,bp,ax,bx,cx,17) RD1D(cx,dx,bp,ax,bx,18) RD1D(bx,cx,dx,bp,ax,19)
#define RD2s(a,b,c,d,e, n13,n8,n2,n, RCONST) \
"\n\
- movl -64+4*"n13"(%rsp), %esi # W[(n+13) & 15] \n\
- xorl -64+4*"n8"(%rsp), %esi # ^W[(n+8) & 15] \n\
- xorl -64+4*"n2"(%rsp), %esi # ^W[(n+2) & 15] \n\
- xorl -64+4*"n"(%rsp), %esi # ^W[n & 15] \n\
+ loadW "n13", %esi # W[(n+13) & 15] \n\
+ xorW "n8", %esi # ^W[(n+8) & 15] \n\
+ xorW "n2", %esi # ^W[(n+2) & 15] \n\
+ xorW "n", %esi # ^W[n & 15] \n\
roll %esi # \n\
- movl %esi, -64+4*"n"(%rsp) # store to W[n & 15] \n\
+ storeW %esi, "n" # store to W[n & 15] \n\
movl %e"c", %edi # c \n\
xorl %e"d", %edi # ^d \n\
xorl %e"b", %edi # ^b \n\
@@ -819,12 +1000,12 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
andl %e"c", %esi # si: b & c \n\
andl %e"d", %edi # di: (b | c) & d \n\
orl %esi, %edi # ((b | c) & d) | (b & c) \n\
- movl -64+4*"n13"(%rsp), %esi # W[(n+13) & 15] \n\
- xorl -64+4*"n8"(%rsp), %esi # ^W[(n+8) & 15] \n\
- xorl -64+4*"n2"(%rsp), %esi # ^W[(n+2) & 15] \n\
- xorl -64+4*"n"(%rsp), %esi # ^W[n & 15] \n\
+ loadW "n13", %esi # W[(n+13) & 15] \n\
+ xorW "n8", %esi # ^W[(n+8) & 15] \n\
+ xorW "n2", %esi # ^W[(n+2) & 15] \n\
+ xorW "n", %esi # ^W[n & 15] \n\
roll %esi # \n\
- movl %esi, -64+4*"n"(%rsp) # store to W[n & 15] \n\
+ storeW %esi, "n" # store to W[n & 15] \n\
addl %edi, %e"e" # += ((b | c) & d) | (b & c)\n\
leal "RCONST"(%r"e",%rsi), %e"e" # e += RCONST + mixed_W \n\
movl %e"a", %esi # \n\
@@ -843,12 +1024,12 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
#define RD4As(a,b,c,d,e, n13,n8,n2,n, RCONST) \
"\n\
- movl -64+4*"n13"(%rsp), %esi # W[(n+13) & 15] \n\
- xorl -64+4*"n8"(%rsp), %esi # ^W[(n+8) & 15] \n\
- xorl -64+4*"n2"(%rsp), %esi # ^W[(n+2) & 15] \n\
- xorl -64+4*"n"(%rsp), %esi # ^W[n & 15] \n\
+ loadW "n13", %esi # W[(n+13) & 15] \n\
+ xorW "n8", %esi # ^W[(n+8) & 15] \n\
+ xorW "n2", %esi # ^W[(n+2) & 15] \n\
+ xorW "n", %esi # ^W[n & 15] \n\
roll %esi # \n\
- movl %esi, -64+4*"n"(%rsp) # store to W[n & 15] \n\
+ storeW %esi, "n" # store to W[n & 15] \n\
movl %e"c", %edi # c \n\
xorl %e"d", %edi # ^d \n\
xorl %e"b", %edi # ^b \n\
@@ -861,12 +1042,12 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
"
#define RD4Bs(a,b,c,d,e, n13,n8,n2,n, RCONST) \
"\n\
- movl -64+4*"n13"(%rsp), %esi # W[(n+13) & 15] \n\
- xorl -64+4*"n8"(%rsp), %esi # ^W[(n+8) & 15] \n\
- xorl -64+4*"n2"(%rsp), %esi # ^W[(n+2) & 15] \n\
- xorl -64+4*"n"(%rsp), %esi # ^W[n & 15] \n\
+ loadW "n13", %esi # W[(n+13) & 15] \n\
+ xorW "n8", %esi # ^W[(n+8) & 15] \n\
+ xorW "n2", %esi # ^W[(n+2) & 15] \n\
+ xorW "n", %esi # ^W[n & 15] \n\
roll %esi # \n\
- ##movl %esi, -64+4*"n"(%rsp) # store to W[n & 15] elided \n\
+ #storeW %esi, "n" # store to W[n & 15] elided \n\
movl %e"c", %edi # c \n\
xorl %e"d", %edi # ^d \n\
xorl %e"b", %edi # ^b \n\
@@ -888,20 +1069,18 @@ static void FAST_FUNC sha1_process_block64(sha1_ctx_t *ctx UNUSED_PARAM)
RD4A(ax,bx,cx,dx,bp,15) RD4A(bp,ax,bx,cx,dx,16) RD4B(dx,bp,ax,bx,cx,17) RD4B(cx,dx,bp,ax,bx,18) RD4B(bx,cx,dx,bp,ax,19)
"\n\
- movq %r10, %rdi # \n\
+ popq %rdi # \n\
addl %eax, 80(%rdi) # ctx->hash[0] += a \n\
addl %ebx, 84(%rdi) # ctx->hash[1] += b \n\
addl %ecx, 88(%rdi) # ctx->hash[2] += c \n\
addl %edx, 92(%rdi) # ctx->hash[3] += d \n\
addl %ebp, 96(%rdi) # ctx->hash[4] += e \n\
- movq %r9, %rbx # callee-saved \n\
- movq %r8, %rbp # callee-saved \n\
- ##popq %rbx # \n\
- ##popq %rbp # \n\
- ##popq %r12 # \n\
- ##popq %r13 # \n\
- ##popq %r14 # \n\
- ##popq %r15 # \n\
+ popq %rbx # \n\
+ popq %rbp # \n\
+ popq %r12 # \n\
+ popq %r13 # \n\
+ popq %r14 # \n\
+ popq %r15 # \n\
"
); /* asm */
#undef RCONST
More information about the busybox-cvs
mailing list