RPM Package Manager, CVS Repository
http://rpm5.org/cvs/
____________________________________________________________________________
Server: rpm5.org Name: Jeff Johnson
Root: /v/rpm/cvs Email: j...@rpm5.org
Module: rpm Date: 09-Oct-2014 04:41:04
Branch: rpm-5_4 Handle: 20141009023641240
Modified files: (Branch: rpm-5_4)
rpm CHANGES
rpm/rpmio Makefile.am blake2.h blake2b.c blake2bp.c
blake2s.c blake2sp.c tgit.c
Log:
- blake2: merge SSE2 (and later) optimizations.
- blake2: fix: x86_64 alignment issues for blake2s #pragma pack.
Summary:
Revision Changes Path
1.3501.2.453+2 -0 rpm/CHANGES
1.293.2.43 +2 -0 rpm/rpmio/Makefile.am
1.1.2.3 +5 -6 rpm/rpmio/blake2.h
1.1.2.3 +839 -43 rpm/rpmio/blake2b.c
1.1.2.5 +4 -4 rpm/rpmio/blake2bp.c
1.1.2.3 +735 -31 rpm/rpmio/blake2s.c
1.1.2.5 +4 -4 rpm/rpmio/blake2sp.c
1.1.2.45 +8 -7 rpm/rpmio/tgit.c
____________________________________________________________________________
patch -p0 <<'@@ .'
Index: rpm/CHANGES
============================================================================
$ cvs diff -u -r1.3501.2.452 -r1.3501.2.453 CHANGES
--- rpm/CHANGES 3 Oct 2014 19:20:10 -0000 1.3501.2.452
+++ rpm/CHANGES 9 Oct 2014 02:36:41 -0000 1.3501.2.453
@@ -1,4 +1,6 @@
5.4.15 -> 5.4.16:
+ - jbj: blake2: merge SSE2 (and later) optimizations.
+ - jbj: blake2: fix: x86_64 alignment issues for blake2s #pragma pack.
- jbj: rpmgfs: stub in macros and put/list/get/dump methods.
- jbj: mongo: upgrade to mongo-c-driver 1.0.1
- jbj: mongo: resurrect unit tests.
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/Makefile.am
============================================================================
$ cvs diff -u -r1.293.2.42 -r1.293.2.43 Makefile.am
--- rpm/rpmio/Makefile.am 3 Oct 2014 19:20:14 -0000 1.293.2.42
+++ rpm/rpmio/Makefile.am 9 Oct 2014 02:37:38 -0000 1.293.2.43
@@ -7,6 +7,8 @@
SUBDIRS = auto # tests
+CLEANFILES = *.gcov .libs/*.gcda .libs/*.gcno *.gcno *.gcda
+
EXTRA_DIST = librpmio.vers \
fnmatch_loop.c getdate.y html-parse.c html-parse.h libsqlio.c \
rpmcpio.c rpmcpio.h rpmgenbasedir.c rpmgenpkglist.c rpmgensrclist.c \
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/blake2.h
============================================================================
$ cvs diff -u -r1.1.2.2 -r1.1.2.3 blake2.h
--- rpm/rpmio/blake2.h 31 Oct 2013 20:45:44 -0000 1.1.2.2
+++ rpm/rpmio/blake2.h 9 Oct 2014 02:38:24 -0000 1.1.2.3
@@ -20,7 +20,7 @@
#if defined(_MSC_VER)
#define ALIGN(x) __declspec(align(x))
#else
-#define ALIGN(x) __attribute__((aligned(x)))
+#define ALIGN(x) __attribute__ ((__aligned__(x)))
#endif
#if defined(__cplusplus)
@@ -43,7 +43,6 @@
BLAKE2B_PERSONALBYTES = 16
};
-#pragma pack(push, 1)
typedef struct __blake2s_param {
uint8_t digest_length; // 1
uint8_t key_length; // 2
@@ -53,7 +52,7 @@
uint8_t node_offset[6]; // 14
uint8_t node_depth; // 15
uint8_t inner_length; // 16
- // uint8_t reserved[0];
+ uint8_t reserved[0];
uint8_t salt[BLAKE2S_SALTBYTES]; // 24
uint8_t personal[BLAKE2S_PERSONALBYTES]; // 32
} blake2s_param;
@@ -65,6 +64,7 @@
uint8_t buf[2 * BLAKE2S_BLOCKBYTES];
size_t buflen;
uint8_t last_node;
+ uint8_t pad[3];
} blake2s_state;
typedef struct __blake2b_param {
@@ -90,20 +90,19 @@
uint8_t last_node;
} blake2b_state;
- typedef struct __blake2sp_state {
+ ALIGN(64) typedef struct __blake2sp_state {
blake2s_state S[8][1];
blake2s_state R[1];
uint8_t buf[8 * BLAKE2S_BLOCKBYTES];
size_t buflen;
} blake2sp_state;
- typedef struct __blake2bp_state {
+ ALIGN(64) typedef struct __blake2bp_state {
blake2b_state S[4][1];
blake2b_state R[1];
uint8_t buf[4 * BLAKE2B_BLOCKBYTES];
size_t buflen;
} blake2bp_state;
-#pragma pack(pop)
// Streaming API
int blake2s_init(blake2s_state * S, const uint8_t outlen);
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/blake2b.c
============================================================================
$ cvs diff -u -r1.1.2.2 -r1.1.2.3 blake2b.c
--- rpm/rpmio/blake2b.c 31 Oct 2013 20:45:44 -0000 1.1.2.2
+++ rpm/rpmio/blake2b.c 9 Oct 2014 02:39:16 -0000 1.1.2.3
@@ -10,21 +10,668 @@
You should have received a copy of the CC0 Public Domain Dedication along
with
this software. If not, see
<http://creativecommons.org/publicdomain/zero/1.0/>.
*/
+#define XXXSSE
#include "system.h"
#include "blake2.h"
#include "blake2-impl.h"
+#ifdef XXXSSE /* compiler intrinsic defines may not be portable */
+#if defined(__SSE2__)
+#define HAVE_SSE2
+#endif
+
+#if defined(__SSSE3__)
+#define HAVE_SSSE3
+#endif
+
+#if defined(__SSE4_1__)
+#define HAVE_SSE41
+#endif
+
+#if defined(__AVX__)
+#define HAVE_AVX
+#endif
+
+#if defined(__XOP__)
+#define HAVE_XOP
+#endif
+
+
+#ifdef HAVE_AVX2
+#ifndef HAVE_AVX
+#define HAVE_AVX
+#endif
+#endif
+
+#ifdef HAVE_XOP
+#ifndef HAVE_AVX
+#define HAVE_AVX
+#endif
+#endif
+
+#ifdef HAVE_AVX
+#ifndef HAVE_SSE41
+#define HAVE_SSE41
+#endif
+#endif
+
+#ifdef HAVE_SSE41
+#ifndef HAVE_SSSE3
+#define HAVE_SSSE3
+#endif
+#endif
+
+#ifdef HAVE_SSSE3
+#define HAVE_SSE2
+#endif
+
+#if !defined(HAVE_SSE2)
+#error "This code requires at least SSE2."
+#endif
+
+#include <emmintrin.h>
+#if defined(HAVE_SSSE3)
+#include <tmmintrin.h>
+#endif
+#if defined(HAVE_SSE41)
+#include <smmintrin.h>
+#endif
+#if defined(HAVE_AVX)
+#include <immintrin.h>
+#endif
+#if defined(HAVE_XOP)
+#include <x86intrin.h>
+#endif
+#endif /* XXXSSE */
+
#include "debug.h"
-static const uint64_t blake2b_IV[8] = {
- 0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
- 0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
- 0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
- 0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
-};
+#ifdef XXXSSE
+/*==============================================================*/
+/* --- blake2b-round.h */
+
+#define LOAD(p) _mm_load_si128( (__m128i *)(p) )
+#define STORE(p,r) _mm_store_si128((__m128i *)(p), r)
+
+#define LOADU(p) _mm_loadu_si128( (__m128i *)(p) )
+#define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r)
+
+#define TOF(reg) _mm_castsi128_ps((reg))
+#define TOI(reg) _mm_castps_si128((reg))
+
+#define LIKELY(x) __builtin_expect((x),1)
+
+
+/* Microarchitecture-specific macros */
+#ifndef HAVE_XOP
+#ifdef HAVE_SSSE3
+#define _mm_roti_epi64(x, c) \
+ (-(c) == 32) ? _mm_shuffle_epi32((x), _MM_SHUFFLE(2,3,0,1)) \
+ : (-(c) == 24) ? _mm_shuffle_epi8((x), r24) \
+ : (-(c) == 16) ? _mm_shuffle_epi8((x), r16) \
+ : (-(c) == 63) ? _mm_xor_si128(_mm_srli_epi64((x), -(c)),
_mm_add_epi64((x), (x))) \
+ : _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_slli_epi64((x),
64-(-(c))))
+#else
+#define _mm_roti_epi64(r, c) _mm_xor_si128(_mm_srli_epi64( (r), -(c)
),_mm_slli_epi64( (r), 64-(-c) ))
+#endif
+#else
+/* ... */
+#endif
+
+
+
+#define G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
+ row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
+ row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
+ \
+ row4l = _mm_xor_si128(row4l, row1l); \
+ row4h = _mm_xor_si128(row4h, row1h); \
+ \
+ row4l = _mm_roti_epi64(row4l, -32); \
+ row4h = _mm_roti_epi64(row4h, -32); \
+ \
+ row3l = _mm_add_epi64(row3l, row4l); \
+ row3h = _mm_add_epi64(row3h, row4h); \
+ \
+ row2l = _mm_xor_si128(row2l, row3l); \
+ row2h = _mm_xor_si128(row2h, row3h); \
+ \
+ row2l = _mm_roti_epi64(row2l, -24); \
+ row2h = _mm_roti_epi64(row2h, -24); \
+
+#define G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
+ row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
+ row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
+ \
+ row4l = _mm_xor_si128(row4l, row1l); \
+ row4h = _mm_xor_si128(row4h, row1h); \
+ \
+ row4l = _mm_roti_epi64(row4l, -16); \
+ row4h = _mm_roti_epi64(row4h, -16); \
+ \
+ row3l = _mm_add_epi64(row3l, row4l); \
+ row3h = _mm_add_epi64(row3h, row4h); \
+ \
+ row2l = _mm_xor_si128(row2l, row3l); \
+ row2h = _mm_xor_si128(row2h, row3h); \
+ \
+ row2l = _mm_roti_epi64(row2l, -63); \
+ row2h = _mm_roti_epi64(row2h, -63); \
+
+#if defined(HAVE_SSSE3)
+#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+ t0 = _mm_alignr_epi8(row2h, row2l, 8); \
+ t1 = _mm_alignr_epi8(row2l, row2h, 8); \
+ row2l = t0; \
+ row2h = t1; \
+ \
+ t0 = row3l; \
+ row3l = row3h; \
+ row3h = t0; \
+ \
+ t0 = _mm_alignr_epi8(row4h, row4l, 8); \
+ t1 = _mm_alignr_epi8(row4l, row4h, 8); \
+ row4l = t1; \
+ row4h = t0;
+
+#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+ t0 = _mm_alignr_epi8(row2l, row2h, 8); \
+ t1 = _mm_alignr_epi8(row2h, row2l, 8); \
+ row2l = t0; \
+ row2h = t1; \
+ \
+ t0 = row3l; \
+ row3l = row3h; \
+ row3h = t0; \
+ \
+ t0 = _mm_alignr_epi8(row4l, row4h, 8); \
+ t1 = _mm_alignr_epi8(row4h, row4l, 8); \
+ row4l = t1; \
+ row4h = t0;
+#else
+
+#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+ t0 = row4l;\
+ t1 = row2l;\
+ row4l = row3l;\
+ row3l = row3h;\
+ row3h = row4l;\
+ row4l = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t0, t0)); \
+ row4h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row4h, row4h)); \
+ row2l = _mm_unpackhi_epi64(row2l, _mm_unpacklo_epi64(row2h, row2h)); \
+ row2h = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(t1, t1))
+
+#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
+ t0 = row3l;\
+ row3l = row3h;\
+ row3h = t0;\
+ t0 = row2l;\
+ t1 = row4l;\
+ row2l = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(row2l, row2l)); \
+ row2h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row2h, row2h)); \
+ row4l = _mm_unpackhi_epi64(row4l, _mm_unpacklo_epi64(row4h, row4h)); \
+ row4h = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t1, t1))
+
+#endif
+#if defined(HAVE_SSE41)
+/*==============================================================*/
+/* --- blake2b-load-sse41.h */
+
+#define LOAD_MSG_0_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m1); \
+b1 = _mm_unpacklo_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_0_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m0, m1); \
+b1 = _mm_unpackhi_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_0_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m4, m5); \
+b1 = _mm_unpacklo_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_0_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m5); \
+b1 = _mm_unpackhi_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_1_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m7, m2); \
+b1 = _mm_unpackhi_epi64(m4, m6); \
+} while(0)
+
+
+#define LOAD_MSG_1_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m5, m4); \
+b1 = _mm_alignr_epi8(m3, m7, 8); \
+} while(0)
+
+
+#define LOAD_MSG_1_3(b0, b1) \
+do \
+{ \
+b0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
+b1 = _mm_unpackhi_epi64(m5, m2); \
+} while(0)
+
+
+#define LOAD_MSG_1_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m6, m1); \
+b1 = _mm_unpackhi_epi64(m3, m1); \
+} while(0)
+
+
+#define LOAD_MSG_2_1(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m6, m5, 8); \
+b1 = _mm_unpackhi_epi64(m2, m7); \
+} while(0)
+
+
+#define LOAD_MSG_2_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m4, m0); \
+b1 = _mm_blend_epi16(m1, m6, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_2_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m5, m1, 0xF0); \
+b1 = _mm_unpackhi_epi64(m3, m4); \
+} while(0)
+
+
+#define LOAD_MSG_2_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m7, m3); \
+b1 = _mm_alignr_epi8(m2, m0, 8); \
+} while(0)
+
+
+#define LOAD_MSG_3_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m3, m1); \
+b1 = _mm_unpackhi_epi64(m6, m5); \
+} while(0)
+
+
+#define LOAD_MSG_3_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m0); \
+b1 = _mm_unpacklo_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_3_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m1, m2, 0xF0); \
+b1 = _mm_blend_epi16(m2, m7, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_3_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m3, m5); \
+b1 = _mm_unpacklo_epi64(m0, m4); \
+} while(0)
+
+
+#define LOAD_MSG_4_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m2); \
+b1 = _mm_unpacklo_epi64(m1, m5); \
+} while(0)
+
+
+#define LOAD_MSG_4_2(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m0, m3, 0xF0); \
+b1 = _mm_blend_epi16(m2, m7, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_4_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m7, m5, 0xF0); \
+b1 = _mm_blend_epi16(m3, m1, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_4_4(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m6, m0, 8); \
+b1 = _mm_blend_epi16(m4, m6, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_5_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m1, m3); \
+b1 = _mm_unpacklo_epi64(m0, m4); \
+} while(0)
+
+
+#define LOAD_MSG_5_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m6, m5); \
+b1 = _mm_unpackhi_epi64(m5, m1); \
+} while(0)
+
+
+#define LOAD_MSG_5_3(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m2, m3, 0xF0); \
+b1 = _mm_unpackhi_epi64(m7, m0); \
+} while(0)
+
+
+#define LOAD_MSG_5_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m6, m2); \
+b1 = _mm_blend_epi16(m7, m4, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_6_1(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m6, m0, 0xF0); \
+b1 = _mm_unpacklo_epi64(m7, m2); \
+} while(0)
+
+
+#define LOAD_MSG_6_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m2, m7); \
+b1 = _mm_alignr_epi8(m5, m6, 8); \
+} while(0)
+
+
+#define LOAD_MSG_6_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m3); \
+b1 = _mm_shuffle_epi32(m4, _MM_SHUFFLE(1,0,3,2)); \
+} while(0)
+
+
+#define LOAD_MSG_6_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m3, m1); \
+b1 = _mm_blend_epi16(m1, m5, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_7_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m6, m3); \
+b1 = _mm_blend_epi16(m6, m1, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_7_2(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m7, m5, 8); \
+b1 = _mm_unpackhi_epi64(m0, m4); \
+} while(0)
+
+
+#define LOAD_MSG_7_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m2, m7); \
+b1 = _mm_unpacklo_epi64(m4, m1); \
+} while(0)
+
+
+#define LOAD_MSG_7_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m2); \
+b1 = _mm_unpacklo_epi64(m3, m5); \
+} while(0)
+
+
+#define LOAD_MSG_8_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m3, m7); \
+b1 = _mm_alignr_epi8(m0, m5, 8); \
+} while(0)
+
+
+#define LOAD_MSG_8_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m7, m4); \
+b1 = _mm_alignr_epi8(m4, m1, 8); \
+} while(0)
+
+
+#define LOAD_MSG_8_3(b0, b1) \
+do \
+{ \
+b0 = m6; \
+b1 = _mm_alignr_epi8(m5, m0, 8); \
+} while(0)
+
+
+#define LOAD_MSG_8_4(b0, b1) \
+do \
+{ \
+b0 = _mm_blend_epi16(m1, m3, 0xF0); \
+b1 = m2; \
+} while(0)
+
+
+#define LOAD_MSG_9_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m5, m4); \
+b1 = _mm_unpackhi_epi64(m3, m0); \
+} while(0)
+
+
+#define LOAD_MSG_9_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m1, m2); \
+b1 = _mm_blend_epi16(m3, m2, 0xF0); \
+} while(0)
+
+
+#define LOAD_MSG_9_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m7, m4); \
+b1 = _mm_unpackhi_epi64(m1, m6); \
+} while(0)
+
+
+#define LOAD_MSG_9_4(b0, b1) \
+do \
+{ \
+b0 = _mm_alignr_epi8(m7, m5, 8); \
+b1 = _mm_unpacklo_epi64(m6, m0); \
+} while(0)
+
+
+#define LOAD_MSG_10_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m0, m1); \
+b1 = _mm_unpacklo_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_10_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m0, m1); \
+b1 = _mm_unpackhi_epi64(m2, m3); \
+} while(0)
+
+
+#define LOAD_MSG_10_3(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m4, m5); \
+b1 = _mm_unpacklo_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_10_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpackhi_epi64(m4, m5); \
+b1 = _mm_unpackhi_epi64(m6, m7); \
+} while(0)
+
+
+#define LOAD_MSG_11_1(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m7, m2); \
+b1 = _mm_unpackhi_epi64(m4, m6); \
+} while(0)
+
+
+#define LOAD_MSG_11_2(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m5, m4); \
+b1 = _mm_alignr_epi8(m3, m7, 8); \
+} while(0)
+
+
+#define LOAD_MSG_11_3(b0, b1) \
+do \
+{ \
+b0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
+b1 = _mm_unpackhi_epi64(m5, m2); \
+} while(0)
+
+
+#define LOAD_MSG_11_4(b0, b1) \
+do \
+{ \
+b0 = _mm_unpacklo_epi64(m6, m1); \
+b1 = _mm_unpackhi_epi64(m3, m1); \
+} while(0)
+/*==============================================================*/
+#else /* defined(HAVE_SSE41) */
+/*==============================================================*/
+/* --- blake2b-load-sse2.h */
+
+#define LOAD_MSG_0_1(b0, b1) b0 = _mm_set_epi64x(m2, m0); b1 =
_mm_set_epi64x(m6, m4)
+#define LOAD_MSG_0_2(b0, b1) b0 = _mm_set_epi64x(m3, m1); b1 =
_mm_set_epi64x(m7, m5)
+#define LOAD_MSG_0_3(b0, b1) b0 = _mm_set_epi64x(m10, m8); b1 =
_mm_set_epi64x(m14, m12)
+#define LOAD_MSG_0_4(b0, b1) b0 = _mm_set_epi64x(m11, m9); b1 =
_mm_set_epi64x(m15, m13)
+#define LOAD_MSG_1_1(b0, b1) b0 = _mm_set_epi64x(m4, m14); b1 =
_mm_set_epi64x(m13, m9)
+#define LOAD_MSG_1_2(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 =
_mm_set_epi64x(m6, m15)
+#define LOAD_MSG_1_3(b0, b1) b0 = _mm_set_epi64x(m0, m1); b1 =
_mm_set_epi64x(m5, m11)
+#define LOAD_MSG_1_4(b0, b1) b0 = _mm_set_epi64x(m2, m12); b1 =
_mm_set_epi64x(m3, m7)
+#define LOAD_MSG_2_1(b0, b1) b0 = _mm_set_epi64x(m12, m11); b1 =
_mm_set_epi64x(m15, m5)
+#define LOAD_MSG_2_2(b0, b1) b0 = _mm_set_epi64x(m0, m8); b1 =
_mm_set_epi64x(m13, m2)
+#define LOAD_MSG_2_3(b0, b1) b0 = _mm_set_epi64x(m3, m10); b1 =
_mm_set_epi64x(m9, m7)
+#define LOAD_MSG_2_4(b0, b1) b0 = _mm_set_epi64x(m6, m14); b1 =
_mm_set_epi64x(m4, m1)
+#define LOAD_MSG_3_1(b0, b1) b0 = _mm_set_epi64x(m3, m7); b1 =
_mm_set_epi64x(m11, m13)
+#define LOAD_MSG_3_2(b0, b1) b0 = _mm_set_epi64x(m1, m9); b1 =
_mm_set_epi64x(m14, m12)
+#define LOAD_MSG_3_3(b0, b1) b0 = _mm_set_epi64x(m5, m2); b1 =
_mm_set_epi64x(m15, m4)
+#define LOAD_MSG_3_4(b0, b1) b0 = _mm_set_epi64x(m10, m6); b1 =
_mm_set_epi64x(m8, m0)
+#define LOAD_MSG_4_1(b0, b1) b0 = _mm_set_epi64x(m5, m9); b1 =
_mm_set_epi64x(m10, m2)
+#define LOAD_MSG_4_2(b0, b1) b0 = _mm_set_epi64x(m7, m0); b1 =
_mm_set_epi64x(m15, m4)
+#define LOAD_MSG_4_3(b0, b1) b0 = _mm_set_epi64x(m11, m14); b1 =
_mm_set_epi64x(m3, m6)
+#define LOAD_MSG_4_4(b0, b1) b0 = _mm_set_epi64x(m12, m1); b1 =
_mm_set_epi64x(m13, m8)
+#define LOAD_MSG_5_1(b0, b1) b0 = _mm_set_epi64x(m6, m2); b1 =
_mm_set_epi64x(m8, m0)
+#define LOAD_MSG_5_2(b0, b1) b0 = _mm_set_epi64x(m10, m12); b1 =
_mm_set_epi64x(m3, m11)
+#define LOAD_MSG_5_3(b0, b1) b0 = _mm_set_epi64x(m7, m4); b1 =
_mm_set_epi64x(m1, m15)
+#define LOAD_MSG_5_4(b0, b1) b0 = _mm_set_epi64x(m5, m13); b1 =
_mm_set_epi64x(m9, m14)
+#define LOAD_MSG_6_1(b0, b1) b0 = _mm_set_epi64x(m1, m12); b1 =
_mm_set_epi64x(m4, m14)
+#define LOAD_MSG_6_2(b0, b1) b0 = _mm_set_epi64x(m15, m5); b1 =
_mm_set_epi64x(m10, m13)
+#define LOAD_MSG_6_3(b0, b1) b0 = _mm_set_epi64x(m6, m0); b1 =
_mm_set_epi64x(m8, m9)
+#define LOAD_MSG_6_4(b0, b1) b0 = _mm_set_epi64x(m3, m7); b1 =
_mm_set_epi64x(m11, m2)
+#define LOAD_MSG_7_1(b0, b1) b0 = _mm_set_epi64x(m7, m13); b1 =
_mm_set_epi64x(m3, m12)
+#define LOAD_MSG_7_2(b0, b1) b0 = _mm_set_epi64x(m14, m11); b1 =
_mm_set_epi64x(m9, m1)
+#define LOAD_MSG_7_3(b0, b1) b0 = _mm_set_epi64x(m15, m5); b1 =
_mm_set_epi64x(m2, m8)
+#define LOAD_MSG_7_4(b0, b1) b0 = _mm_set_epi64x(m4, m0); b1 =
_mm_set_epi64x(m10, m6)
+#define LOAD_MSG_8_1(b0, b1) b0 = _mm_set_epi64x(m14, m6); b1 =
_mm_set_epi64x(m0, m11)
+#define LOAD_MSG_8_2(b0, b1) b0 = _mm_set_epi64x(m9, m15); b1 =
_mm_set_epi64x(m8, m3)
+#define LOAD_MSG_8_3(b0, b1) b0 = _mm_set_epi64x(m13, m12); b1 =
_mm_set_epi64x(m10, m1)
+#define LOAD_MSG_8_4(b0, b1) b0 = _mm_set_epi64x(m7, m2); b1 =
_mm_set_epi64x(m5, m4)
+#define LOAD_MSG_9_1(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 =
_mm_set_epi64x(m1, m7)
+#define LOAD_MSG_9_2(b0, b1) b0 = _mm_set_epi64x(m4, m2); b1 =
_mm_set_epi64x(m5, m6)
+#define LOAD_MSG_9_3(b0, b1) b0 = _mm_set_epi64x(m9, m15); b1 =
_mm_set_epi64x(m13, m3)
+#define LOAD_MSG_9_4(b0, b1) b0 = _mm_set_epi64x(m14, m11); b1 =
_mm_set_epi64x(m0, m12)
+#define LOAD_MSG_10_1(b0, b1) b0 = _mm_set_epi64x(m2, m0); b1 =
_mm_set_epi64x(m6, m4)
+#define LOAD_MSG_10_2(b0, b1) b0 = _mm_set_epi64x(m3, m1); b1 =
_mm_set_epi64x(m7, m5)
+#define LOAD_MSG_10_3(b0, b1) b0 = _mm_set_epi64x(m10, m8); b1 =
_mm_set_epi64x(m14, m12)
+#define LOAD_MSG_10_4(b0, b1) b0 = _mm_set_epi64x(m11, m9); b1 =
_mm_set_epi64x(m15, m13)
+#define LOAD_MSG_11_1(b0, b1) b0 = _mm_set_epi64x(m4, m14); b1 =
_mm_set_epi64x(m13, m9)
+#define LOAD_MSG_11_2(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 =
_mm_set_epi64x(m6, m15)
+#define LOAD_MSG_11_3(b0, b1) b0 = _mm_set_epi64x(m0, m1); b1 =
_mm_set_epi64x(m5, m11)
+#define LOAD_MSG_11_4(b0, b1) b0 = _mm_set_epi64x(m2, m12); b1 =
_mm_set_epi64x(m3, m7)
+/*==============================================================*/
+#endif /* defined(HAVE_SSE41) */
+
+#define ROUND(r) \
+ LOAD_MSG_ ##r ##_1(b0, b1); \
+ G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+ LOAD_MSG_ ##r ##_2(b0, b1); \
+ G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+ DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
+ LOAD_MSG_ ##r ##_3(b0, b1); \
+ G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+ LOAD_MSG_ ##r ##_4(b0, b1); \
+ G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
+ UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
+/*==============================================================*/
+#else /* XXXSSE */
+/*==============================================================*/
static const uint8_t blake2b_sigma[12][16] = {
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
@@ -40,6 +687,39 @@
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3}
};
+#define G(r,i,a,b,c,d) \
+ do { \
+ a = a + b + m[blake2b_sigma[r][2*i+0]]; \
+ d = rotr64(d ^ a, 32); \
+ c = c + d; \
+ b = rotr64(b ^ c, 24); \
+ a = a + b + m[blake2b_sigma[r][2*i+1]]; \
+ d = rotr64(d ^ a, 16); \
+ c = c + d; \
+ b = rotr64(b ^ c, 63); \
+ } while(0)
+#define ROUND(r) \
+ do { \
+ G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
+ G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
+ G(r,2,v[ 2],v[ 6],v[10],v[14]); \
+ G(r,3,v[ 3],v[ 7],v[11],v[15]); \
+ G(r,4,v[ 0],v[ 5],v[10],v[15]); \
+ G(r,5,v[ 1],v[ 6],v[11],v[12]); \
+ G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
+ G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
+ } while(0)
+/*==============================================================*/
+#endif /* XXXSSE */
+
+ALIGN(64)
+static const uint64_t blake2b_IV[8] = {
+ 0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
+ 0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
+ 0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
+ 0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
+};
+
static inline int blake2b_set_lastnode(blake2b_state * S)
{
@@ -47,13 +727,13 @@
return 0;
}
+/* Some helper functions, not necessarily useful */
static inline int blake2b_clear_lastnode(blake2b_state * S)
{
S->f[1] = 0ULL;
return 0;
}
-/* Some helper functions, not necessarily useful */
static inline int blake2b_set_lastblock(blake2b_state * S)
{
if (S->last_node)
@@ -75,8 +755,21 @@
static inline int blake2b_increment_counter(blake2b_state * S,
const uint64_t inc)
{
+#ifdef XXXSSE
+#if __x86_64__
+ // ADD/ADC chain
+ __uint128_t t = ((__uint128_t) S->t[1] << 64) | S->t[0];
+ t += inc;
+ S->t[0] = (uint64_t) (t >> 0);
+ S->t[1] = (uint64_t) (t >> 64);
+#else
+ S->t[0] += inc;
+ S->t[1] += (S->t[0] < inc);
+#endif
+#else /* XXXSSE */
S->t[0] += inc;
S->t[1] += (S->t[0] < inc);
+#endif /* XXXSSE */
return 0;
}
@@ -165,6 +858,18 @@
/* init xors IV with input parameter block */
int blake2b_init_param(blake2b_state * S, const blake2b_param * P)
{
+#ifdef XXXSSE
+ uint8_t *p, *h, *v;
+ //blake2b_init0( S );
+ v = (uint8_t *) (blake2b_IV);
+ h = (uint8_t *) (S->h);
+ p = (uint8_t *) (P);
+ /* IV XOR ParamBlock */
+ memset(S, 0, sizeof(blake2b_state));
+
+ for (int i = 0; i < BLAKE2B_OUTBYTES; ++i)
+ h[i] = v[i] ^ p[i];
+#else /* XXXSSE */
blake2b_init0(S);
uint8_t *p = (uint8_t *) (P);
size_t i;
@@ -172,6 +877,7 @@
/* IV XOR ParamBlock */
for (i = 0; i < 8; ++i)
S->h[i] ^= load64(p + sizeof(S->h[i]) * i);
+#endif /* XXXSSE */
return 0;
}
@@ -185,6 +891,23 @@
if (outlen == 0 || outlen > BLAKE2B_OUTBYTES)
return -1;
+#ifdef XXXSSE
+ { const blake2b_param _P = {
+ .digest_length = outlen,
+ .key_length = 0,
+ .fanout = 1,
+ .depth = 1,
+ .leaf_length = 0,
+ .node_offset = 0,
+ .node_depth = 0,
+ .inner_length = 0,
+ .reserved = {0},
+ .salt = {0},
+ .personal = {0}
+ };
+ P[0] = *(blake2b_param *)&_P; /* structure assignment */
+ }
+#else /* XXXSSE */
P->digest_length = outlen;
P->key_length = 0;
P->fanout = 1;
@@ -193,9 +916,11 @@
store64(&P->node_offset, 0);
P->node_depth = 0;
P->inner_length = 0;
- memset(P->reserved, 0, sizeof(P->reserved));
+ memset(P->reserved, 0, sizeof(P->reserved) );
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
+#endif /* XXXSSE */
+
return blake2b_init_param(S, P);
}
@@ -211,6 +936,23 @@
if (!key || !keylen || keylen > BLAKE2B_KEYBYTES)
return -1;
+#ifdef XXXSSE
+ { const blake2b_param _P = {
+ .digest_length = outlen,
+ .key_length = keylen,
+ .fanout = 1,
+ .depth = 1,
+ .leaf_length = 0,
+ .node_offset = 0,
+ .node_depth = 0,
+ .inner_length = 0,
+ .reserved = {0},
+ .salt = {0},
+ .personal = {0}
+ };
+ P[0] = *(blake2b_param *)&_P; /* structure assignment */
+ }
+#else /* XXXSSE */
P->digest_length = outlen;
P->key_length = keylen;
P->fanout = 1;
@@ -219,9 +961,10 @@
store64(&P->node_offset, 0);
P->node_depth = 0;
P->inner_length = 0;
- memset(P->reserved, 0, sizeof(P->reserved));
+ memset(P->reserved, 0, sizeof(P->reserved) );
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
+#endif /* XXXSSE */
if (blake2b_init_param(S, P) < 0)
return -1;
@@ -236,9 +979,60 @@
return 0;
}
-static int blake2b_compress(blake2b_state * S,
+static inline int blake2b_compress(blake2b_state * S,
const uint8_t block[BLAKE2B_BLOCKBYTES])
{
+#ifdef XXXSSE
+ __m128i row1l, row1h;
+ __m128i row2l, row2h;
+ __m128i row3l, row3h;
+ __m128i row4l, row4h;
+ __m128i b0, b1;
+ __m128i t0, t1;
+#if defined(HAVE_SSSE3) && !defined(HAVE_XOP)
+ const __m128i r16 =
+ _mm_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8,
+ 9);
+ const __m128i r24 =
+ _mm_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9,
+ 10);
+#endif
+#if defined(HAVE_SSE41)
+ const __m128i m0 = LOADU(block + 00);
+ const __m128i m1 = LOADU(block + 16);
+ const __m128i m2 = LOADU(block + 32);
+ const __m128i m3 = LOADU(block + 48);
+ const __m128i m4 = LOADU(block + 64);
+ const __m128i m5 = LOADU(block + 80);
+ const __m128i m6 = LOADU(block + 96);
+ const __m128i m7 = LOADU(block + 112);
+#else
+ const uint64_t m0 = ((uint64_t *) block)[0];
+ const uint64_t m1 = ((uint64_t *) block)[1];
+ const uint64_t m2 = ((uint64_t *) block)[2];
+ const uint64_t m3 = ((uint64_t *) block)[3];
+ const uint64_t m4 = ((uint64_t *) block)[4];
+ const uint64_t m5 = ((uint64_t *) block)[5];
+ const uint64_t m6 = ((uint64_t *) block)[6];
+ const uint64_t m7 = ((uint64_t *) block)[7];
+ const uint64_t m8 = ((uint64_t *) block)[8];
+ const uint64_t m9 = ((uint64_t *) block)[9];
+ const uint64_t m10 = ((uint64_t *) block)[10];
+ const uint64_t m11 = ((uint64_t *) block)[11];
+ const uint64_t m12 = ((uint64_t *) block)[12];
+ const uint64_t m13 = ((uint64_t *) block)[13];
+ const uint64_t m14 = ((uint64_t *) block)[14];
+ const uint64_t m15 = ((uint64_t *) block)[15];
+#endif
+ row1l = LOADU(&S->h[0]);
+ row1h = LOADU(&S->h[2]);
+ row2l = LOADU(&S->h[4]);
+ row2h = LOADU(&S->h[6]);
+ row3l = LOADU(&blake2b_IV[0]);
+ row3h = LOADU(&blake2b_IV[2]);
+ row4l = _mm_xor_si128(LOADU(&blake2b_IV[4]), LOADU(&S->t[0]));
+ row4h = _mm_xor_si128(LOADU(&blake2b_IV[6]), LOADU(&S->f[0]));
+#else /* XXXSSE */
uint64_t m[16];
uint64_t v[16];
int i;
@@ -257,28 +1051,8 @@
v[13] = S->t[1] ^ blake2b_IV[5];
v[14] = S->f[0] ^ blake2b_IV[6];
v[15] = S->f[1] ^ blake2b_IV[7];
-#define G(r,i,a,b,c,d) \
- do { \
- a = a + b + m[blake2b_sigma[r][2*i+0]]; \
- d = rotr64(d ^ a, 32); \
- c = c + d; \
- b = rotr64(b ^ c, 24); \
- a = a + b + m[blake2b_sigma[r][2*i+1]]; \
- d = rotr64(d ^ a, 16); \
- c = c + d; \
- b = rotr64(b ^ c, 63); \
- } while(0)
-#define ROUND(r) \
- do { \
- G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
- G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
- G(r,2,v[ 2],v[ 6],v[10],v[14]); \
- G(r,3,v[ 3],v[ 7],v[11],v[15]); \
- G(r,4,v[ 0],v[ 5],v[10],v[15]); \
- G(r,5,v[ 1],v[ 6],v[11],v[12]); \
- G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
- G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
- } while(0)
+#endif /* XXXSSE */
+
ROUND(0);
ROUND(1);
ROUND(2);
@@ -292,15 +1066,23 @@
ROUND(10);
ROUND(11);
+#ifdef XXXSSE
+ row1l = _mm_xor_si128(row3l, row1l);
+ row1h = _mm_xor_si128(row3h, row1h);
+ STOREU(&S->h[0], _mm_xor_si128(LOADU(&S->h[0]), row1l));
+ STOREU(&S->h[2], _mm_xor_si128(LOADU(&S->h[2]), row1h));
+ row2l = _mm_xor_si128(row4l, row2l);
+ row2h = _mm_xor_si128(row4h, row2h);
+ STOREU(&S->h[4], _mm_xor_si128(LOADU(&S->h[4]), row2l));
+ STOREU(&S->h[6], _mm_xor_si128(LOADU(&S->h[6]), row2h));
+#else /* XXXSSE */
for (i = 0; i < 8; ++i)
S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
+#endif /* XXXSSE */
-#undef G
-#undef ROUND
return 0;
}
-/* inlen now in bytes */
int blake2b_update(blake2b_state * S, const uint8_t * in, uint64_t inlen)
{
while (inlen > 0) {
@@ -328,11 +1110,8 @@
return 0;
}
-/* Is this correct? */
int blake2b_final(blake2b_state * S, uint8_t * out, uint8_t outlen)
{
- uint8_t buffer[BLAKE2B_OUTBYTES];
- int i;
if (S->buflen > BLAKE2B_BLOCKBYTES) {
blake2b_increment_counter(S, BLAKE2B_BLOCKBYTES);
@@ -346,10 +1125,17 @@
memset(S->buf + S->buflen, 0, 2 * BLAKE2B_BLOCKBYTES - S->buflen);
/* Padding */
blake2b_compress(S, S->buf);
- for (i = 0; i < 8; ++i) /* Output full hash to temp buffer */
- store64(buffer + sizeof(S->h[i]) * i, S->h[i]);
+#ifdef XXXSSE
+ memcpy( out, &S->h[0], outlen );
+#else /* XXXSSE */
+ { uint8_t buffer[BLAKE2B_OUTBYTES];
+ int i;
+ for (i = 0; i < 8; ++i) /* Output full hash to temp buffer */
+ store64(buffer + sizeof(S->h[i]) * i, S->h[i]);
- memcpy(out, buffer, outlen);
+ memcpy(out, buffer, outlen);
+ }
+#endif /* XXXSSE */
return 0;
}
@@ -382,6 +1168,16 @@
return 0;
}
+
+#if defined(SUPERCOP)
+int crypto_hash(unsigned char *out, unsigned char *in,
+ unsigned long long inlen)
+{
+ return blake2b(out, in, NULL, BLAKE2B_OUTBYTES, inlen, 0);
+}
+#endif
+
+
#if defined(BLAKE2B_SELFTEST)
#include "blake2-kat.h"
int main(int argc, char **argv)
@@ -398,9 +1194,9 @@
for (i = 0; i < KAT_LENGTH; ++i) {
uint8_t hash[BLAKE2B_OUTBYTES];
- blake2b(hash, buf, key, BLAKE2B_OUTBYTES, i, BLAKE2B_KEYBYTES);
- if (memcmp(hash, blake2b_keyed_kat[i], BLAKE2B_OUTBYTES)) {
+ if (blake2b(hash, buf, key, BLAKE2B_OUTBYTES, i, BLAKE2B_KEYBYTES) < 0
+ || memcmp(hash, blake2b_keyed_kat[i], BLAKE2B_OUTBYTES)) {
puts("error");
return -1;
}
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/blake2bp.c
============================================================================
$ cvs diff -u -r1.1.2.4 -r1.1.2.5 blake2bp.c
--- rpm/rpmio/blake2bp.c 27 Mar 2014 19:14:04 -0000 1.1.2.4
+++ rpm/rpmio/blake2bp.c 9 Oct 2014 02:39:34 -0000 1.1.2.5
@@ -35,7 +35,7 @@
store32(&P->leaf_length, 0);
store64(&P->node_offset, offset);
P->node_depth = 0;
- P->inner_length = outlen;
+ P->inner_length = BLAKE2B_OUTBYTES;
memset(P->reserved, 0, sizeof(P->reserved));
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
@@ -53,7 +53,7 @@
store32(&P->leaf_length, 0);
store64(&P->node_offset, 0);
P->node_depth = 1;
- P->inner_length = outlen;
+ P->inner_length = BLAKE2B_OUTBYTES;
memset(P->reserved, 0, sizeof(P->reserved));
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
@@ -289,9 +289,9 @@
for (i = 0; i < KAT_LENGTH; ++i) {
uint8_t hash[BLAKE2B_OUTBYTES];
- blake2bp(hash, buf, key, BLAKE2B_OUTBYTES, i, BLAKE2B_KEYBYTES);
- if (memcmp(hash, blake2bp_keyed_kat[i], BLAKE2B_OUTBYTES)) {
+ if (blake2bp(hash, buf, key, BLAKE2B_OUTBYTES, i, BLAKE2B_KEYBYTES) < 0
+ || memcmp(hash, blake2bp_keyed_kat[i], BLAKE2B_OUTBYTES)) {
puts("error");
return -1;
}
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/blake2s.c
============================================================================
$ cvs diff -u -r1.1.2.2 -r1.1.2.3 blake2s.c
--- rpm/rpmio/blake2s.c 31 Oct 2013 20:45:44 -0000 1.1.2.2
+++ rpm/rpmio/blake2s.c 9 Oct 2014 02:39:54 -0000 1.1.2.3
@@ -10,19 +10,596 @@
You should have received a copy of the CC0 Public Domain Dedication along
with
this software. If not, see
<http://creativecommons.org/publicdomain/zero/1.0/>.
*/
+#define XXXSSE
#include "system.h"
#include "blake2.h"
#include "blake2-impl.h"
+#ifdef XXXSSE /* compiler intrinsic defines may not be portable */
+#if defined(__SSE2__)
+#define HAVE_SSE2
+#endif
+
+#if defined(__SSSE3__)
+#define HAVE_SSSE3
+#endif
+
+#if defined(__SSE4_1__)
+#define HAVE_SSE41
+#endif
+
+#if defined(__AVX__)
+#define HAVE_AVX
+#endif
+
+#if defined(__XOP__)
+#define HAVE_XOP
+#endif
+
+
+#ifdef HAVE_AVX2
+#ifndef HAVE_AVX
+#define HAVE_AVX
+#endif
+#endif
+
+#ifdef HAVE_XOP
+#ifndef HAVE_AVX
+#define HAVE_AVX
+#endif
+#endif
+
+#ifdef HAVE_AVX
+#ifndef HAVE_SSE41
+#define HAVE_SSE41
+#endif
+#endif
+
+#ifdef HAVE_SSE41
+#ifndef HAVE_SSSE3
+#define HAVE_SSSE3
+#endif
+#endif
+
+#ifdef HAVE_SSSE3
+#define HAVE_SSE2
+#endif
+
+#if !defined(HAVE_SSE2)
+#error "This code requires at least SSE2."
+#endif
+#include <emmintrin.h>
+#if defined(HAVE_SSSE3)
+#include <tmmintrin.h>
+#endif
+#if defined(HAVE_SSE41)
+#include <smmintrin.h>
+#endif
+#if defined(HAVE_AVX)
+#include <immintrin.h>
+#endif
+#if defined(HAVE_XOP)
+#include <x86intrin.h>
+#endif /* XXXSSE */
+
+#endif
+
#include "debug.h"
-static const uint32_t blake2s_IV[8] = {
- 0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
- 0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
-};
+#ifdef XXXSSE
+/*==============================================================*/
+/* --- blake2s-round.h */
+
+#define LOAD(p) _mm_load_si128( (__m128i *)(p) )
+#define STORE(p,r) _mm_store_si128((__m128i *)(p), r)
+
+#define LOADU(p) _mm_loadu_si128( (__m128i *)(p) )
+#define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r)
+
+#define TOF(reg) _mm_castsi128_ps((reg))
+#define TOI(reg) _mm_castps_si128((reg))
+
+#define LIKELY(x) __builtin_expect((x),1)
+
+
+/* Microarchitecture-specific macros */
+#ifndef HAVE_XOP
+#ifdef HAVE_SSSE3
+#define _mm_roti_epi32(r, c) ( \
+ (8==-(c)) ? _mm_shuffle_epi8(r,r8) \
+ : (16==-(c)) ? _mm_shuffle_epi8(r,r16) \
+ : _mm_xor_si128(_mm_srli_epi32( (r), -(c) ),_mm_slli_epi32(
(r), 32-(-(c)) )) )
+#else
+#define _mm_roti_epi32(r, c) _mm_xor_si128(_mm_srli_epi32( (r), -(c)
),_mm_slli_epi32( (r), 32-(-c) ))
+#endif
+#else
+/* ... */
+#endif
+
+#define G1(row1,row2,row3,row4,buf) \
+ row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \
+ row4 = _mm_xor_si128( row4, row1 ); \
+ row4 = _mm_roti_epi32(row4, -16); \
+ row3 = _mm_add_epi32( row3, row4 ); \
+ row2 = _mm_xor_si128( row2, row3 ); \
+ row2 = _mm_roti_epi32(row2, -12);
+
+#define G2(row1,row2,row3,row4,buf) \
+ row1 = _mm_add_epi32( _mm_add_epi32( row1, buf), row2 ); \
+ row4 = _mm_xor_si128( row4, row1 ); \
+ row4 = _mm_roti_epi32(row4, -8); \
+ row3 = _mm_add_epi32( row3, row4 ); \
+ row2 = _mm_xor_si128( row2, row3 ); \
+ row2 = _mm_roti_epi32(row2, -7);
+
+#define DIAGONALIZE(row1,row2,row3,row4) \
+ row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(2,1,0,3) ); \
+ row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \
+ row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(0,3,2,1) );
+
+#define UNDIAGONALIZE(row1,row2,row3,row4) \
+ row4 = _mm_shuffle_epi32( row4, _MM_SHUFFLE(0,3,2,1) ); \
+ row3 = _mm_shuffle_epi32( row3, _MM_SHUFFLE(1,0,3,2) ); \
+ row2 = _mm_shuffle_epi32( row2, _MM_SHUFFLE(2,1,0,3) );
+
+#if defined(HAVE_XOP)
+/*==============================================================*/
+/* --- blake2s-load-xop.h */
+
+#define TOB(x) ((x)*4*0x01010101 + 0x03020100) // ..or not TOB
+
+/* Basic VPPERM emulation, for testing purposes */
+/*static __m128i _mm_perm_epi8(const __m128i src1, const __m128i src2, const
__m128i sel)
+{
+ const __m128i sixteen = _mm_set1_epi8(16);
+ const __m128i t0 = _mm_shuffle_epi8(src1, sel);
+ const __m128i s1 = _mm_shuffle_epi8(src2, _mm_sub_epi8(sel, sixteen));
+ const __m128i mask = _mm_or_si128(_mm_cmpeq_epi8(sel, sixteen),
+ _mm_cmpgt_epi8(sel, sixteen)); //
(>=16) = 0xff : 00
+ return _mm_blendv_epi8(t0, s1, mask);
+}*/
+
+#define LOAD_MSG_0_1(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(6),TOB(4),TOB(2),TOB(0)) );
+
+#define LOAD_MSG_0_2(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(7),TOB(5),TOB(3),TOB(1)) );
+
+#define LOAD_MSG_0_3(buf) \
+buf = _mm_perm_epi8(m2, m3, _mm_set_epi32(TOB(6),TOB(4),TOB(2),TOB(0)) );
+
+#define LOAD_MSG_0_4(buf) \
+buf = _mm_perm_epi8(m2, m3, _mm_set_epi32(TOB(7),TOB(5),TOB(3),TOB(1)) );
+
+#define LOAD_MSG_1_1(buf) \
+t0 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(0),TOB(5),TOB(0),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_1_2(buf) \
+t1 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(2),TOB(0),TOB(4),TOB(6)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_1_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(5),TOB(0),TOB(0),TOB(1)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_1_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(3),TOB(7),TOB(2),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(4)) );
+
+#define LOAD_MSG_2_1(buf) \
+t0 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(0),TOB(1),TOB(0),TOB(7)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(7),TOB(2),TOB(4),TOB(0)) );
+
+#define LOAD_MSG_2_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(2),TOB(0),TOB(4)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_2_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(7),TOB(3),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_2_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(4),TOB(1),TOB(6),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_3_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(3),TOB(7)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(5),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_3_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(0),TOB(1),TOB(5)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(6),TOB(4),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_3_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(4),TOB(5),TOB(2)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_3_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(0),TOB(6)) ); \
+buf = _mm_perm_epi8(t1, m2, _mm_set_epi32(TOB(4),TOB(2),TOB(6),TOB(0)) );
+
+#define LOAD_MSG_4_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(2),TOB(5),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(6),TOB(2),TOB(1),TOB(5)) );
+
+#define LOAD_MSG_4_2(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(4),TOB(7),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_4_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(3),TOB(6),TOB(0),TOB(0)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(2),TOB(7),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(6)) );
+
+#define LOAD_MSG_4_4(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(4),TOB(0),TOB(1)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(4),TOB(0)) );
+
+#define LOAD_MSG_5_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(6),TOB(2)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(4),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_5_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(3),TOB(7),TOB(6),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(4)) );
+
+#define LOAD_MSG_5_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(1),TOB(0),TOB(7),TOB(4)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_5_4(buf) \
+t1 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(5),TOB(0),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(6),TOB(1),TOB(5)) );
+
+#define LOAD_MSG_6_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(4),TOB(0),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(6),TOB(1),TOB(4)) );
+
+#define LOAD_MSG_6_2(buf) \
+t1 = _mm_perm_epi8(m1, m2, _mm_set_epi32(TOB(6),TOB(0),TOB(0),TOB(1)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(5),TOB(7),TOB(0)) );
+
+#define LOAD_MSG_6_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(6),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(4),TOB(5),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_6_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(2),TOB(3),TOB(7)) ); \
+buf = _mm_perm_epi8(t1, m2, _mm_set_epi32(TOB(7),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_7_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(3),TOB(0),TOB(7),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(4),TOB(1),TOB(5)) );
+
+#define LOAD_MSG_7_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(5),TOB(1),TOB(0),TOB(7)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(6),TOB(0)) );
+
+#define LOAD_MSG_7_3(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(2),TOB(0),TOB(0),TOB(5)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(4),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(7),TOB(0)) );
+
+#define LOAD_MSG_7_4(buf) \
+t1 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(6),TOB(4),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m2, _mm_set_epi32(TOB(6),TOB(2),TOB(1),TOB(0)) );
+
+#define LOAD_MSG_8_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(0),TOB(0),TOB(0),TOB(6)) ); \
+t0 = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(7),TOB(1),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(6),TOB(0)) );
+
+#define LOAD_MSG_8_2(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(4),TOB(3),TOB(5),TOB(0)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(1),TOB(7)) );
+
+#define LOAD_MSG_8_3(buf) \
+t0 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(6),TOB(1),TOB(0),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(3),TOB(2),TOB(5),TOB(4)) ); \
+
+#define LOAD_MSG_8_4(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(5),TOB(4),TOB(7),TOB(2)) );
+
+#define LOAD_MSG_9_1(buf) \
+t0 = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(1),TOB(7),TOB(0),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m2, _mm_set_epi32(TOB(3),TOB(2),TOB(4),TOB(6)) );
+
+#define LOAD_MSG_9_2(buf) \
+buf = _mm_perm_epi8(m0, m1, _mm_set_epi32(TOB(5),TOB(6),TOB(4),TOB(2)) );
+
+#define LOAD_MSG_9_3(buf) \
+t0 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(3),TOB(5),TOB(0)) ); \
+buf = _mm_perm_epi8(t0, m3, _mm_set_epi32(TOB(5),TOB(2),TOB(1),TOB(7)) );
+
+#define LOAD_MSG_9_4(buf) \
+t1 = _mm_perm_epi8(m0, m2, _mm_set_epi32(TOB(0),TOB(0),TOB(0),TOB(7)) ); \
+buf = _mm_perm_epi8(t1, m3, _mm_set_epi32(TOB(3),TOB(4),TOB(6),TOB(0)) );
+
+/*==============================================================*/
+#elif defined(HAVE_SSE41)
+/*==============================================================*/
+/* --- blake2s-load-sse41.h */
+
+#define LOAD_MSG_0_1(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m0), TOF(m1), _MM_SHUFFLE(2,0,2,0)));
+
+#define LOAD_MSG_0_2(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m0), TOF(m1), _MM_SHUFFLE(3,1,3,1)));
+
+#define LOAD_MSG_0_3(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m2), TOF(m3), _MM_SHUFFLE(2,0,2,0)));
+
+#define LOAD_MSG_0_4(buf) \
+buf = TOI(_mm_shuffle_ps(TOF(m2), TOF(m3), _MM_SHUFFLE(3,1,3,1)));
+
+#define LOAD_MSG_1_1(buf) \
+t0 = _mm_blend_epi16(m1, m2, 0x0C); \
+t1 = _mm_slli_si128(m3, 4); \
+t2 = _mm_blend_epi16(t0, t1, 0xF0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,1,0,3));
+
+#define LOAD_MSG_1_2(buf) \
+t0 = _mm_shuffle_epi32(m2,_MM_SHUFFLE(0,0,2,0)); \
+t1 = _mm_blend_epi16(m1,m3,0xC0); \
+t2 = _mm_blend_epi16(t0, t1, 0xF0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
+
+#define LOAD_MSG_1_3(buf) \
+t0 = _mm_slli_si128(m1, 4); \
+t1 = _mm_blend_epi16(m2, t0, 0x30); \
+t2 = _mm_blend_epi16(m0, t1, 0xF0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
+
+#define LOAD_MSG_1_4(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_slli_si128(m3, 4); \
+t2 = _mm_blend_epi16(t0, t1, 0x0C); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,3,0,1));
+
+#define LOAD_MSG_2_1(buf) \
+t0 = _mm_unpackhi_epi32(m2,m3); \
+t1 = _mm_blend_epi16(m3,m1,0x0C); \
+t2 = _mm_blend_epi16(t0, t1, 0x0F); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(3,1,0,2));
+
+#define LOAD_MSG_2_2(buf) \
+t0 = _mm_unpacklo_epi32(m2,m0); \
+t1 = _mm_blend_epi16(t0, m0, 0xF0); \
+t2 = _mm_slli_si128(m3, 8); \
+buf = _mm_blend_epi16(t1, t2, 0xC0);
+
+#define LOAD_MSG_2_3(buf) \
+t0 = _mm_blend_epi16(m0, m2, 0x3C); \
+t1 = _mm_srli_si128(m1, 12); \
+t2 = _mm_blend_epi16(t0,t1,0x03); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,0,3,2));
+
+#define LOAD_MSG_2_4(buf) \
+t0 = _mm_slli_si128(m3, 4); \
+t1 = _mm_blend_epi16(m0, m1, 0x33); \
+t2 = _mm_blend_epi16(t1, t0, 0xC0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(0,1,2,3));
+
+#define LOAD_MSG_3_1(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_unpackhi_epi32(t0, m2); \
+t2 = _mm_blend_epi16(t1, m3, 0x0C); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(3,1,0,2));
+
+#define LOAD_MSG_3_2(buf) \
+t0 = _mm_slli_si128(m2, 8); \
+t1 = _mm_blend_epi16(m3,m0,0x0C); \
+t2 = _mm_blend_epi16(t1, t0, 0xC0); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,0,1,3));
+
+#define LOAD_MSG_3_3(buf) \
+t0 = _mm_blend_epi16(m0,m1,0x0F); \
+t1 = _mm_blend_epi16(t0, m3, 0xC0); \
+buf = _mm_shuffle_epi32(t1, _MM_SHUFFLE(3,0,1,2));
+
+#define LOAD_MSG_3_4(buf) \
+t0 = _mm_unpacklo_epi32(m0,m2); \
+t1 = _mm_unpackhi_epi32(m1,m2); \
+buf = _mm_unpacklo_epi64(t1,t0);
+
+#define LOAD_MSG_4_1(buf) \
+t0 = _mm_unpacklo_epi64(m1,m2); \
+t1 = _mm_unpackhi_epi64(m0,m2); \
+t2 = _mm_blend_epi16(t0,t1,0x33); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,0,1,3));
+
+#define LOAD_MSG_4_2(buf) \
+t0 = _mm_unpackhi_epi64(m1,m3); \
+t1 = _mm_unpacklo_epi64(m0,m1); \
+buf = _mm_blend_epi16(t0,t1,0x33);
+
+#define LOAD_MSG_4_3(buf) \
+t0 = _mm_unpackhi_epi64(m3,m1); \
+t1 = _mm_unpackhi_epi64(m2,m0); \
+buf = _mm_blend_epi16(t1,t0,0x33);
+
+#define LOAD_MSG_4_4(buf) \
+t0 = _mm_blend_epi16(m0,m2,0x03); \
+t1 = _mm_slli_si128(t0, 8); \
+t2 = _mm_blend_epi16(t1,m3,0x0F); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,2,0,3));
+
+#define LOAD_MSG_5_1(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_unpacklo_epi32(m0,m2); \
+buf = _mm_unpacklo_epi64(t0,t1);
+
+#define LOAD_MSG_5_2(buf) \
+t0 = _mm_srli_si128(m2, 4); \
+t1 = _mm_blend_epi16(m0,m3,0x03); \
+buf = _mm_blend_epi16(t1,t0,0x3C);
+
+#define LOAD_MSG_5_3(buf) \
+t0 = _mm_blend_epi16(m1,m0,0x0C); \
+t1 = _mm_srli_si128(m3, 4); \
+t2 = _mm_blend_epi16(t0,t1,0x30); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,2,3,0));
+
+#define LOAD_MSG_5_4(buf) \
+t0 = _mm_unpacklo_epi64(m1,m2); \
+t1= _mm_shuffle_epi32(m3, _MM_SHUFFLE(0,2,0,1)); \
+buf = _mm_blend_epi16(t0,t1,0x33);
+
+#define LOAD_MSG_6_1(buf) \
+t0 = _mm_slli_si128(m1, 12); \
+t1 = _mm_blend_epi16(m0,m3,0x33); \
+buf = _mm_blend_epi16(t1,t0,0xC0);
+
+#define LOAD_MSG_6_2(buf) \
+t0 = _mm_blend_epi16(m3,m2,0x30); \
+t1 = _mm_srli_si128(m1, 4); \
+t2 = _mm_blend_epi16(t0,t1,0x03); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(2,1,3,0));
+
+#define LOAD_MSG_6_3(buf) \
+t0 = _mm_unpacklo_epi64(m0,m2); \
+t1 = _mm_srli_si128(m1, 4); \
+buf = _mm_shuffle_epi32(_mm_blend_epi16(t0,t1,0x0C), _MM_SHUFFLE(2,3,1,0));
+
+#define LOAD_MSG_6_4(buf) \
+t0 = _mm_unpackhi_epi32(m1,m2); \
+t1 = _mm_unpackhi_epi64(m0,t0); \
+buf = _mm_shuffle_epi32(t1, _MM_SHUFFLE(3,0,1,2));
+
+#define LOAD_MSG_7_1(buf) \
+t0 = _mm_unpackhi_epi32(m0,m1); \
+t1 = _mm_blend_epi16(t0,m3,0x0F); \
+buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(2,0,3,1));
+
+#define LOAD_MSG_7_2(buf) \
+t0 = _mm_blend_epi16(m2,m3,0x30); \
+t1 = _mm_srli_si128(m0,4); \
+t2 = _mm_blend_epi16(t0,t1,0x03); \
+buf = _mm_shuffle_epi32(t2, _MM_SHUFFLE(1,0,2,3));
+
+#define LOAD_MSG_7_3(buf) \
+t0 = _mm_unpackhi_epi64(m0,m3); \
+t1 = _mm_unpacklo_epi64(m1,m2); \
+t2 = _mm_blend_epi16(t0,t1,0x3C); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(0,2,3,1));
+
+#define LOAD_MSG_7_4(buf) \
+t0 = _mm_unpacklo_epi32(m0,m1); \
+t1 = _mm_unpackhi_epi32(m1,m2); \
+buf = _mm_unpacklo_epi64(t0,t1);
+
+#define LOAD_MSG_8_1(buf) \
+t0 = _mm_unpackhi_epi32(m1,m3); \
+t1 = _mm_unpacklo_epi64(t0,m0); \
+t2 = _mm_blend_epi16(t1,m2,0xC0); \
+buf = _mm_shufflehi_epi16(t2,_MM_SHUFFLE(1,0,3,2));
+
+#define LOAD_MSG_8_2(buf) \
+t0 = _mm_unpackhi_epi32(m0,m3); \
+t1 = _mm_blend_epi16(m2,t0,0xF0); \
+buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(0,2,1,3));
+
+#define LOAD_MSG_8_3(buf) \
+t0 = _mm_blend_epi16(m2,m0,0x0C); \
+t1 = _mm_slli_si128(t0,4); \
+buf = _mm_blend_epi16(t1,m3,0x0F);
+
+#define LOAD_MSG_8_4(buf) \
+t0 = _mm_blend_epi16(m1,m0,0x30); \
+buf = _mm_shuffle_epi32(t0,_MM_SHUFFLE(1,0,3,2));
+
+#define LOAD_MSG_9_1(buf) \
+t0 = _mm_blend_epi16(m0,m2,0x03); \
+t1 = _mm_blend_epi16(m1,m2,0x30); \
+t2 = _mm_blend_epi16(t1,t0,0x0F); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(1,3,0,2));
+
+#define LOAD_MSG_9_2(buf) \
+t0 = _mm_slli_si128(m0,4); \
+t1 = _mm_blend_epi16(m1,t0,0xC0); \
+buf = _mm_shuffle_epi32(t1,_MM_SHUFFLE(1,2,0,3));
+
+#define LOAD_MSG_9_3(buf) \
+t0 = _mm_unpackhi_epi32(m0,m3); \
+t1 = _mm_unpacklo_epi32(m2,m3); \
+t2 = _mm_unpackhi_epi64(t0,t1); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(3,0,2,1));
+
+#define LOAD_MSG_9_4(buf) \
+t0 = _mm_blend_epi16(m3,m2,0xC0); \
+t1 = _mm_unpacklo_epi32(m0,m3); \
+t2 = _mm_blend_epi16(t0,t1,0x0F); \
+buf = _mm_shuffle_epi32(t2,_MM_SHUFFLE(0,1,2,3));
+
+/*==============================================================*/
+#else /* defined(HAVE_SSE41) */
+/*==============================================================*/
+/* --- blake2s-load-sse2.h */
+
+#define LOAD_MSG_0_1(buf) buf = _mm_set_epi32(m6,m4,m2,m0)
+#define LOAD_MSG_0_2(buf) buf = _mm_set_epi32(m7,m5,m3,m1)
+#define LOAD_MSG_0_3(buf) buf = _mm_set_epi32(m14,m12,m10,m8)
+#define LOAD_MSG_0_4(buf) buf = _mm_set_epi32(m15,m13,m11,m9)
+#define LOAD_MSG_1_1(buf) buf = _mm_set_epi32(m13,m9,m4,m14)
+#define LOAD_MSG_1_2(buf) buf = _mm_set_epi32(m6,m15,m8,m10)
+#define LOAD_MSG_1_3(buf) buf = _mm_set_epi32(m5,m11,m0,m1)
+#define LOAD_MSG_1_4(buf) buf = _mm_set_epi32(m3,m7,m2,m12)
+#define LOAD_MSG_2_1(buf) buf = _mm_set_epi32(m15,m5,m12,m11)
+#define LOAD_MSG_2_2(buf) buf = _mm_set_epi32(m13,m2,m0,m8)
+#define LOAD_MSG_2_3(buf) buf = _mm_set_epi32(m9,m7,m3,m10)
+#define LOAD_MSG_2_4(buf) buf = _mm_set_epi32(m4,m1,m6,m14)
+#define LOAD_MSG_3_1(buf) buf = _mm_set_epi32(m11,m13,m3,m7)
+#define LOAD_MSG_3_2(buf) buf = _mm_set_epi32(m14,m12,m1,m9)
+#define LOAD_MSG_3_3(buf) buf = _mm_set_epi32(m15,m4,m5,m2)
+#define LOAD_MSG_3_4(buf) buf = _mm_set_epi32(m8,m0,m10,m6)
+#define LOAD_MSG_4_1(buf) buf = _mm_set_epi32(m10,m2,m5,m9)
+#define LOAD_MSG_4_2(buf) buf = _mm_set_epi32(m15,m4,m7,m0)
+#define LOAD_MSG_4_3(buf) buf = _mm_set_epi32(m3,m6,m11,m14)
+#define LOAD_MSG_4_4(buf) buf = _mm_set_epi32(m13,m8,m12,m1)
+#define LOAD_MSG_5_1(buf) buf = _mm_set_epi32(m8,m0,m6,m2)
+#define LOAD_MSG_5_2(buf) buf = _mm_set_epi32(m3,m11,m10,m12)
+#define LOAD_MSG_5_3(buf) buf = _mm_set_epi32(m1,m15,m7,m4)
+#define LOAD_MSG_5_4(buf) buf = _mm_set_epi32(m9,m14,m5,m13)
+#define LOAD_MSG_6_1(buf) buf = _mm_set_epi32(m4,m14,m1,m12)
+#define LOAD_MSG_6_2(buf) buf = _mm_set_epi32(m10,m13,m15,m5)
+#define LOAD_MSG_6_3(buf) buf = _mm_set_epi32(m8,m9,m6,m0)
+#define LOAD_MSG_6_4(buf) buf = _mm_set_epi32(m11,m2,m3,m7)
+#define LOAD_MSG_7_1(buf) buf = _mm_set_epi32(m3,m12,m7,m13)
+#define LOAD_MSG_7_2(buf) buf = _mm_set_epi32(m9,m1,m14,m11)
+#define LOAD_MSG_7_3(buf) buf = _mm_set_epi32(m2,m8,m15,m5)
+#define LOAD_MSG_7_4(buf) buf = _mm_set_epi32(m10,m6,m4,m0)
+#define LOAD_MSG_8_1(buf) buf = _mm_set_epi32(m0,m11,m14,m6)
+#define LOAD_MSG_8_2(buf) buf = _mm_set_epi32(m8,m3,m9,m15)
+#define LOAD_MSG_8_3(buf) buf = _mm_set_epi32(m10,m1,m13,m12)
+#define LOAD_MSG_8_4(buf) buf = _mm_set_epi32(m5,m4,m7,m2)
+#define LOAD_MSG_9_1(buf) buf = _mm_set_epi32(m1,m7,m8,m10)
+#define LOAD_MSG_9_2(buf) buf = _mm_set_epi32(m5,m6,m4,m2)
+#define LOAD_MSG_9_3(buf) buf = _mm_set_epi32(m13,m3,m9,m15)
+#define LOAD_MSG_9_4(buf) buf = _mm_set_epi32(m0,m12,m14,m11)
+
+/*==============================================================*/
+#endif /* defined(HAVE_SSE41) */
+
+#define ROUND(r) \
+ LOAD_MSG_ ##r ##_1(buf1); \
+ G1(row1,row2,row3,row4,buf1); \
+ LOAD_MSG_ ##r ##_2(buf2); \
+ G2(row1,row2,row3,row4,buf2); \
+ DIAGONALIZE(row1,row2,row3,row4); \
+ LOAD_MSG_ ##r ##_3(buf3); \
+ G1(row1,row2,row3,row4,buf3); \
+ LOAD_MSG_ ##r ##_4(buf4); \
+ G2(row1,row2,row3,row4,buf4); \
+ UNDIAGONALIZE(row1,row2,row3,row4); \
+
+/*==============================================================*/
+#else /* XXXSSE */
+/*==============================================================*/
static const uint8_t blake2s_sigma[10][16] = {
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
@@ -36,6 +613,36 @@
{10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0},
};
+#define G(r,i,a,b,c,d) \
+ do { \
+ a = a + b + m[blake2s_sigma[r][2*i+0]]; \
+ d = rotr32(d ^ a, 16); \
+ c = c + d; \
+ b = rotr32(b ^ c, 12); \
+ a = a + b + m[blake2s_sigma[r][2*i+1]]; \
+ d = rotr32(d ^ a, 8); \
+ c = c + d; \
+ b = rotr32(b ^ c, 7); \
+ } while(0)
+#define ROUND(r) \
+ do { \
+ G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
+ G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
+ G(r,2,v[ 2],v[ 6],v[10],v[14]); \
+ G(r,3,v[ 3],v[ 7],v[11],v[15]); \
+ G(r,4,v[ 0],v[ 5],v[10],v[15]); \
+ G(r,5,v[ 1],v[ 6],v[11],v[12]); \
+ G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
+ G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
+ } while(0)
+/*==============================================================*/
+#endif /* XXXSSE */
+
+static const uint32_t blake2s_IV[8] = {
+ 0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
+ 0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
+};
+
static inline int blake2s_set_lastnode(blake2s_state * S)
{
S->f[1] = ~0U;
@@ -70,8 +677,15 @@
static inline int blake2s_increment_counter(blake2s_state * S,
const uint32_t inc)
{
+#ifdef XXXSSE
+ uint64_t t = ((uint64_t) S->t[1] << 32) | S->t[0];
+ t += inc;
+ S->t[0] = (uint32_t) (t >> 0);
+ S->t[1] = (uint32_t) (t >> 32);
+#else /* XXXSSE */
S->t[0] += inc;
S->t[1] += (S->t[0] < inc);
+#endif /* XXXSSE */
return 0;
}
@@ -159,6 +773,18 @@
/* init2 xors IV with input parameter block */
int blake2s_init_param(blake2s_state * S, const blake2s_param * P)
{
+#ifdef XXXSSE
+ uint8_t *p, *h, *v;
+ //blake2s_init0( S );
+ v = (uint8_t *) (blake2s_IV);
+ h = (uint8_t *) (S->h);
+ p = (uint8_t *) (P);
+ /* IV XOR ParamBlock */
+ memset(S, 0, sizeof(blake2s_state));
+
+ for (int i = 0; i < BLAKE2S_OUTBYTES; ++i)
+ h[i] = v[i] ^ p[i];
+#else /* XXXSSE */
blake2s_init0(S);
uint32_t *p = (uint32_t *) (P);
size_t i;
@@ -166,6 +792,7 @@
/* IV XOR ParamBlock */
for (i = 0; i < 8; ++i)
S->h[i] ^= load32(&p[i]);
+#endif /* XXXSSE */
return 0;
}
@@ -180,6 +807,22 @@
if (outlen == 0 || outlen > BLAKE2S_OUTBYTES)
return -1;
+#ifdef XXXSSE
+ { const blake2s_param _P = {
+ .digest_length = outlen,
+ .key_length = 0,
+ .fanout = 1,
+ .depth = 1,
+ .leaf_length = 0,
+ .node_offset = {0},
+ .node_depth = 0,
+ .inner_length = 0,
+ .salt = {0},
+ .personal = {0}
+ };
+ P[0] = *(blake2s_param *)&_P; /* structure assignment */
+ }
+#else /* XXXSSE */
P->digest_length = outlen;
P->key_length = 0;
P->fanout = 1;
@@ -191,12 +834,15 @@
// memset(P->reserved, 0, sizeof(P->reserved) );
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
+#endif /* XXXSSE */
+
return blake2s_init_param(S, P);
}
int blake2s_init_key(blake2s_state * S, const uint8_t outlen,
const void *key, const uint8_t keylen)
{
+ /* Move interval verification here? */
blake2s_param P[1];
if (outlen == 0 || outlen > BLAKE2S_OUTBYTES)
@@ -205,6 +851,22 @@
if (!key || !keylen || keylen > BLAKE2S_KEYBYTES)
return -1;
+#ifdef XXXSSE
+ { const blake2s_param _P = {
+ .digest_length = outlen,
+ .key_length = keylen,
+ .fanout = 1,
+ .depth = 1,
+ .leaf_length = 0,
+ .node_offset = {0},
+ .node_depth = 0,
+ .inner_length = 0,
+ .salt = {0},
+ .personal = {0}
+ };
+ P[0] = *(blake2s_param *)&_P; /* structure assignment */
+ }
+#else /* XXXSSE */
P->digest_length = outlen;
P->key_length = keylen;
P->fanout = 1;
@@ -216,6 +878,7 @@
// memset(P->reserved, 0, sizeof(P->reserved) );
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
+#endif /* XXXSSE */
if (blake2s_init_param(S, P) < 0)
return -1;
@@ -230,9 +893,58 @@
return 0;
}
-static int blake2s_compress(blake2s_state * S,
+static inline int blake2s_compress(blake2s_state * S,
const uint8_t block[BLAKE2S_BLOCKBYTES])
{
+#ifdef XXXSSE
+ __m128i row1, row2, row3, row4;
+ __m128i buf1, buf2, buf3, buf4;
+#if defined(HAVE_SSE41)
+ __m128i t0, t1;
+#if !defined(HAVE_XOP)
+ __m128i t2;
+#endif
+#endif
+ __m128i ff0, ff1;
+#if defined(HAVE_SSSE3) && !defined(HAVE_XOP)
+ const __m128i r8 =
+ _mm_set_epi8(12, 15, 14, 13, 8, 11, 10, 9, 4, 7, 6, 5, 0, 3, 2, 1);
+ const __m128i r16 =
+ _mm_set_epi8(13, 12, 15, 14, 9, 8, 11, 10, 5, 4, 7, 6, 1, 0, 3, 2);
+#endif
+#if defined(HAVE_SSE41)
+ const __m128i m0 = LOADU(block + 00);
+ const __m128i m1 = LOADU(block + 16);
+ const __m128i m2 = LOADU(block + 32);
+ const __m128i m3 = LOADU(block + 48);
+#else
+ const uint32_t m0 = ((uint32_t *) block)[0];
+ const uint32_t m1 = ((uint32_t *) block)[1];
+ const uint32_t m2 = ((uint32_t *) block)[2];
+ const uint32_t m3 = ((uint32_t *) block)[3];
+ const uint32_t m4 = ((uint32_t *) block)[4];
+ const uint32_t m5 = ((uint32_t *) block)[5];
+ const uint32_t m6 = ((uint32_t *) block)[6];
+ const uint32_t m7 = ((uint32_t *) block)[7];
+ const uint32_t m8 = ((uint32_t *) block)[8];
+ const uint32_t m9 = ((uint32_t *) block)[9];
+ const uint32_t m10 = ((uint32_t *) block)[10];
+ const uint32_t m11 = ((uint32_t *) block)[11];
+ const uint32_t m12 = ((uint32_t *) block)[12];
+ const uint32_t m13 = ((uint32_t *) block)[13];
+ const uint32_t m14 = ((uint32_t *) block)[14];
+ const uint32_t m15 = ((uint32_t *) block)[15];
+#endif
+ row1 = ff0 = LOADU(&S->h[0]);
+ row2 = ff1 = LOADU(&S->h[4]);
+ row3 = _mm_setr_epi32(0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A);
+ row4 =
+ _mm_xor_si128(_mm_setr_epi32
+ (0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19),
+ LOADU(&S->t[0]));
+
+#else /* XXXSSE */
+
uint32_t m[16];
uint32_t v[16];
size_t i;
@@ -251,28 +963,8 @@
v[13] = S->t[1] ^ blake2s_IV[5];
v[14] = S->f[0] ^ blake2s_IV[6];
v[15] = S->f[1] ^ blake2s_IV[7];
-#define G(r,i,a,b,c,d) \
- do { \
- a = a + b + m[blake2s_sigma[r][2*i+0]]; \
- d = rotr32(d ^ a, 16); \
- c = c + d; \
- b = rotr32(b ^ c, 12); \
- a = a + b + m[blake2s_sigma[r][2*i+1]]; \
- d = rotr32(d ^ a, 8); \
- c = c + d; \
- b = rotr32(b ^ c, 7); \
- } while(0)
-#define ROUND(r) \
- do { \
- G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
- G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
- G(r,2,v[ 2],v[ 6],v[10],v[14]); \
- G(r,3,v[ 3],v[ 7],v[11],v[15]); \
- G(r,4,v[ 0],v[ 5],v[10],v[15]); \
- G(r,5,v[ 1],v[ 6],v[11],v[12]); \
- G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
- G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
- } while(0)
+#endif /* XXXSSE */
+
ROUND(0);
ROUND(1);
ROUND(2);
@@ -284,11 +976,14 @@
ROUND(8);
ROUND(9);
+#ifdef XXXSSE
+ STOREU( &S->h[0], _mm_xor_si128( ff0, _mm_xor_si128( row1, row3 ) ) );
+ STOREU( &S->h[4], _mm_xor_si128( ff1, _mm_xor_si128( row2, row4 ) ) );
+#else
for (i = 0; i < 8; ++i)
S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
+#endif
-#undef G
-#undef ROUND
return 0;
}
@@ -344,6 +1039,7 @@
return 0;
}
+/* inlen, at least, should be uint64_t. Others can be size_t. */
int blake2s(uint8_t * out, const void *in, const void *key,
const uint8_t outlen, const uint64_t inlen, uint8_t keylen)
{
@@ -372,6 +1068,14 @@
return 0;
}
+#if defined(SUPERCOP)
+int crypto_hash(unsigned char *out, unsigned char *in,
+ unsigned long long inlen)
+{
+ return blake2s(out, in, NULL, BLAKE2S_OUTBYTES, inlen, 0);
+}
+#endif
+
#if defined(BLAKE2S_SELFTEST)
#include "blake2-kat.h"
int main(int argc, char **argv)
@@ -388,9 +1092,9 @@
for (i = 0; i < KAT_LENGTH; ++i) {
uint8_t hash[BLAKE2S_OUTBYTES];
- blake2s(hash, buf, key, BLAKE2S_OUTBYTES, i, BLAKE2S_KEYBYTES);
- if (memcmp(hash, blake2s_keyed_kat[i], BLAKE2S_OUTBYTES)) {
+ if (blake2s(hash, buf, key, BLAKE2S_OUTBYTES, i, BLAKE2S_KEYBYTES) < 0
+ || memcmp(hash, blake2s_keyed_kat[i], BLAKE2S_OUTBYTES)) {
puts("error");
return -1;
}
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/blake2sp.c
============================================================================
$ cvs diff -u -r1.1.2.4 -r1.1.2.5 blake2sp.c
--- rpm/rpmio/blake2sp.c 27 Mar 2014 19:14:04 -0000 1.1.2.4
+++ rpm/rpmio/blake2sp.c 9 Oct 2014 02:40:41 -0000 1.1.2.5
@@ -35,7 +35,7 @@
store32(&P->leaf_length, 0);
store48(P->node_offset, offset);
P->node_depth = 0;
- P->inner_length = outlen;
+ P->inner_length = BLAKE2S_OUTBYTES;
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
return blake2s_init_param(S, P);
@@ -52,7 +52,7 @@
store32(&P->leaf_length, 0);
store48(P->node_offset, 0ULL);
P->node_depth = 1;
- P->inner_length = outlen;
+ P->inner_length = BLAKE2S_OUTBYTES;
memset(P->salt, 0, sizeof(P->salt));
memset(P->personal, 0, sizeof(P->personal));
return blake2s_init_param(S, P);
@@ -289,9 +289,9 @@
for (i = 0; i < KAT_LENGTH; ++i) {
uint8_t hash[BLAKE2S_OUTBYTES];
- blake2sp(hash, buf, key, BLAKE2S_OUTBYTES, i, BLAKE2S_KEYBYTES);
- if (memcmp(hash, blake2sp_keyed_kat[i], BLAKE2S_OUTBYTES)) {
+ if (blake2sp(hash, buf, key, BLAKE2S_OUTBYTES, i, BLAKE2S_KEYBYTES) < 0
+ || memcmp(hash, blake2sp_keyed_kat[i], BLAKE2S_OUTBYTES)) {
puts("error");
return -1;
}
@@ .
patch -p0 <<'@@ .'
Index: rpm/rpmio/tgit.c
============================================================================
$ cvs diff -u -r1.1.2.44 -r1.1.2.45 tgit.c
--- rpm/rpmio/tgit.c 7 Oct 2014 18:16:56 -0000 1.1.2.44
+++ rpm/rpmio/tgit.c 9 Oct 2014 02:40:41 -0000 1.1.2.45
@@ -59,13 +59,13 @@
FILE * fp = stdout;
if (pd->total_steps > 0) {
- fprintf(fp, "\rCheckouts: %d/%d", pd->completed_steps, pd->total_steps);
+ fprintf(fp, "\rCheckouts: %u/%u", (unsigned)pd->completed_steps,
(unsigned)pd->total_steps);
} else
if (pd->fetch_progress.total_deltas > 0
&& pd->fetch_progress.received_objects ==
pd->fetch_progress.total_objects) {
- fprintf(fp, "\rResolving deltas %d/%d",
- pd->fetch_progress.indexed_deltas,
- pd->fetch_progress.total_deltas);
+ fprintf(fp, "\rResolving deltas %u/%u",
+ (unsigned)pd->fetch_progress.indexed_deltas,
+ (unsigned)pd->fetch_progress.total_deltas);
if (pd->fetch_progress.indexed_deltas ==
pd->fetch_progress.total_deltas)
fprintf(fp, ", done.\n");
} else {
@@ -79,10 +79,11 @@
Receiving objects: 100% (53569/53569), 20.52 MiB | 5.24 MiB/s, done.
Resolving deltas: 100% (36892/36892), done.
#endif
- fprintf(fp, "\rReceiving objects: %3d%% (%d/%d) indexed (%d), %6.2f
MiB",
+ fprintf(fp, "\rReceiving objects: %3d%% (%u/%u) indexed (%u), %6.2f
MiB",
network_percent,
- pd->fetch_progress.received_objects,
pd->fetch_progress.total_objects,
- pd->fetch_progress.indexed_objects,
+ (unsigned)pd->fetch_progress.received_objects,
+ (unsigned)pd->fetch_progress.total_objects,
+ (unsigned)pd->fetch_progress.indexed_objects,
(pd->fetch_progress.received_bytes / (1024. * 1024.)) );
if (pd->fetch_progress.received_objects ==
pd->fetch_progress.total_objects)
fprintf(fp, ", done.\n");
@@ .
______________________________________________________________________
RPM Package Manager http://rpm5.org
CVS Sources Repository rpm-cvs@rpm5.org
