QPQ-AG/libsodium
12
0
Fork 0
mirror of https://github.com/jedisct1/libsodium.git synced 2026-08-28 20:57:13 +09:00
Code Issues Packages Projects Releases Wiki Activity

Constify

Browse Source
This commit is contained in:
Frank Denis
2018-12-26 18:25:16 +01:00
parent 7bc5a3da66
commit 0a6e10f75f
7 changed files with 118 additions and 118 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/libsodium/crypto_generichash/blake2b/ref/blake2b-compress-avx2.h
+3 -3
View File
@@ -2,13 +2,13 @@
#ifndef blake2b_compress_avx2_H
#define blake2b_compress_avx2_H
#define LOADU128(p) _mm_loadu_si128((__m128i *) (p))
#define LOADU128(p) _mm_loadu_si128((const __m128i *) (p))
#define STOREU128(p, r) _mm_storeu_si128((__m128i *) (p), r)
#define LOAD(p) _mm256_load_si256((__m256i *) (p))
#define LOAD(p) _mm256_load_si256((const __m256i *) (p))
#define STORE(p, r) _mm256_store_si256((__m256i *) (p), r)
#define LOADU(p) _mm256_loadu_si256((__m256i *) (p))
#define LOADU(p) _mm256_loadu_si256((const __m256i *) (p))
#define STOREU(p, r) _mm256_storeu_si256((__m256i *) (p), r)
static inline uint64_t
src/libsodium/crypto_stream/chacha20/dolbeau/u0.h
+4 -4
View File
@@ -69,10 +69,10 @@ if (bytes > 0) {
t_1 = _mm_srli_epi32(t_1, 25);
x_1 = _mm_xor_si128(x_1, t_1);
}
x_0 = _mm_add_epi32(x_0, _mm_loadu_si128((__m128i*) (x + 0)));
x_1 = _mm_add_epi32(x_1, _mm_loadu_si128((__m128i*) (x + 4)));
x_2 = _mm_add_epi32(x_2, _mm_loadu_si128((__m128i*) (x + 8)));
x_3 = _mm_add_epi32(x_3, _mm_loadu_si128((__m128i*) (x + 12)));
x_0 = _mm_add_epi32(x_0, _mm_loadu_si128((const __m128i*) (x + 0)));
x_1 = _mm_add_epi32(x_1, _mm_loadu_si128((const __m128i*) (x + 4)));
x_2 = _mm_add_epi32(x_2, _mm_loadu_si128((const __m128i*) (x + 8)));
x_3 = _mm_add_epi32(x_3, _mm_loadu_si128((const __m128i*) (x + 12)));
_mm_storeu_si128((__m128i*) (partialblock + 0), x_0);
_mm_storeu_si128((__m128i*) (partialblock + 16), x_1);
_mm_storeu_si128((__m128i*) (partialblock + 32), x_2);
src/libsodium/crypto_stream/salsa20/xmm6int/u0.h
+22 -22
View File
@@ -1,8 +1,8 @@
if (bytes > 0) {
__m128i diag0 = _mm_loadu_si128((__m128i *) (x + 0));
__m128i diag1 = _mm_loadu_si128((__m128i *) (x + 4));
__m128i diag2 = _mm_loadu_si128((__m128i *) (x + 8));
__m128i diag3 = _mm_loadu_si128((__m128i *) (x + 12));
__m128i diag0 = _mm_loadu_si128((const __m128i *) (x + 0));
__m128i diag1 = _mm_loadu_si128((const __m128i *) (x + 4));
__m128i diag2 = _mm_loadu_si128((const __m128i *) (x + 8));
__m128i diag3 = _mm_loadu_si128((const __m128i *) (x + 12));
__m128i a0, a1, a2, a3, a4, a5, a6, a7;
__m128i b0, b1, b2, b3, b4, b5, b6, b7;
uint8_t partialblock[64];
@@ -156,25 +156,25 @@ if (bytes > 0) {
diag0 = _mm_xor_si128(diag0, b7);
}
diag0 = _mm_add_epi32(diag0, _mm_loadu_si128((__m128i *) (x + 0)));
diag1 = _mm_add_epi32(diag1, _mm_loadu_si128((__m128i *) (x + 4)));
diag2 = _mm_add_epi32(diag2, _mm_loadu_si128((__m128i *) (x + 8)));
diag3 = _mm_add_epi32(diag3, _mm_loadu_si128((__m128i *) (x + 12)));
diag0 = _mm_add_epi32(diag0, _mm_loadu_si128((const __m128i *) (x + 0)));
diag1 = _mm_add_epi32(diag1, _mm_loadu_si128((const __m128i *) (x + 4)));
diag2 = _mm_add_epi32(diag2, _mm_loadu_si128((const __m128i *) (x + 8)));
diag3 = _mm_add_epi32(diag3, _mm_loadu_si128((const __m128i *) (x + 12)));
#define ONEQUAD_SHUFFLE(A, B, C, D) \
do { \
uint32_t in##A = _mm_cvtsi128_si32(diag0); \
uint32_t in##B = _mm_cvtsi128_si32(diag1); \
uint32_t in##C = _mm_cvtsi128_si32(diag2); \
uint32_t in##D = _mm_cvtsi128_si32(diag3); \
diag0 = _mm_shuffle_epi32(diag0, 0x39); \
diag1 = _mm_shuffle_epi32(diag1, 0x39); \
diag2 = _mm_shuffle_epi32(diag2, 0x39); \
diag3 = _mm_shuffle_epi32(diag3, 0x39); \
*(uint32_t *) (partialblock + (A * 4)) = in##A; \
*(uint32_t *) (partialblock + (B * 4)) = in##B; \
*(uint32_t *) (partialblock + (C * 4)) = in##C; \
*(uint32_t *) (partialblock + (D * 4)) = in##D; \
#define ONEQUAD_SHUFFLE(A, B, C, D) \
do { \
uint32_t in##A = _mm_cvtsi128_si32(diag0); \
uint32_t in##B = _mm_cvtsi128_si32(diag1); \
uint32_t in##C = _mm_cvtsi128_si32(diag2); \
uint32_t in##D = _mm_cvtsi128_si32(diag3); \
diag0 = _mm_shuffle_epi32(diag0, 0x39); \
diag1 = _mm_shuffle_epi32(diag1, 0x39); \
diag2 = _mm_shuffle_epi32(diag2, 0x39); \
diag3 = _mm_shuffle_epi32(diag3, 0x39); \
*(uint32_t *) (partialblock + (A * 4)) = in##A; \
*(uint32_t *) (partialblock + (B * 4)) = in##B; \
*(uint32_t *) (partialblock + (C * 4)) = in##C; \
*(uint32_t *) (partialblock + (D * 4)) = in##D; \
} while (0)
#define ONEQUAD(A, B, C, D) ONEQUAD_SHUFFLE(A, B, C, D)
src/libsodium/crypto_stream/salsa20/xmm6int/u1.h
+8 -8
View File
@@ -1,8 +1,8 @@
while (bytes >= 64) {
__m128i diag0 = _mm_loadu_si128((__m128i *) (x + 0));
__m128i diag1 = _mm_loadu_si128((__m128i *) (x + 4));
__m128i diag2 = _mm_loadu_si128((__m128i *) (x + 8));
__m128i diag3 = _mm_loadu_si128((__m128i *) (x + 12));
__m128i diag0 = _mm_loadu_si128((const __m128i *) (x + 0));
__m128i diag1 = _mm_loadu_si128((const __m128i *) (x + 4));
__m128i diag2 = _mm_loadu_si128((const __m128i *) (x + 8));
__m128i diag3 = _mm_loadu_si128((const __m128i *) (x + 12));
__m128i a0, a1, a2, a3, a4, a5, a6, a7;
__m128i b0, b1, b2, b3, b4, b5, b6, b7;
@@ -157,10 +157,10 @@ while (bytes >= 64) {
diag0 = _mm_xor_si128(diag0, b7);
}
diag0 = _mm_add_epi32(diag0, _mm_loadu_si128((__m128i *) (x + 0)));
diag1 = _mm_add_epi32(diag1, _mm_loadu_si128((__m128i *) (x + 4)));
diag2 = _mm_add_epi32(diag2, _mm_loadu_si128((__m128i *) (x + 8)));
diag3 = _mm_add_epi32(diag3, _mm_loadu_si128((__m128i *) (x + 12)));
diag0 = _mm_add_epi32(diag0, _mm_loadu_si128((const __m128i *) (x + 0)));
diag1 = _mm_add_epi32(diag1, _mm_loadu_si128((const __m128i *) (x + 4)));
diag2 = _mm_add_epi32(diag2, _mm_loadu_si128((const __m128i *) (x + 8)));
diag3 = _mm_add_epi32(diag3, _mm_loadu_si128((const __m128i *) (x + 12)));
#define ONEQUAD_SHUFFLE(A, B, C, D) \
do { \
src/libsodium/crypto_stream/salsa20/xmm6int/u4.h
+24 -24
View File
@@ -12,22 +12,22 @@ if (bytes >= 256) {
/* element broadcast immediate for _mm_shuffle_epi32 are in order:
0x00, 0x55, 0xaa, 0xff */
z0 = _mm_loadu_si128((__m128i *) (x + 0));
z0 = _mm_loadu_si128((const __m128i *) (x + 0));
z5 = _mm_shuffle_epi32(z0, 0x55);
z10 = _mm_shuffle_epi32(z0, 0xaa);
z15 = _mm_shuffle_epi32(z0, 0xff);
z0 = _mm_shuffle_epi32(z0, 0x00);
z1 = _mm_loadu_si128((__m128i *) (x + 4));
z1 = _mm_loadu_si128((const __m128i *) (x + 4));
z6 = _mm_shuffle_epi32(z1, 0xaa);
z11 = _mm_shuffle_epi32(z1, 0xff);
z12 = _mm_shuffle_epi32(z1, 0x00);
z1 = _mm_shuffle_epi32(z1, 0x55);
z2 = _mm_loadu_si128((__m128i *) (x + 8));
z2 = _mm_loadu_si128((const __m128i *) (x + 8));
z7 = _mm_shuffle_epi32(z2, 0xff);
z13 = _mm_shuffle_epi32(z2, 0x55);
z2 = _mm_shuffle_epi32(z2, 0xaa);
/* no z8 -> first half of the nonce, will fill later */
z3 = _mm_loadu_si128((__m128i *) (x + 12));
z3 = _mm_loadu_si128((const __m128i *) (x + 12));
z4 = _mm_shuffle_epi32(z3, 0x00);
z14 = _mm_shuffle_epi32(z3, 0xaa);
z3 = _mm_shuffle_epi32(z3, 0xff);
@@ -498,26 +498,26 @@ if (bytes >= 256) {
/* store data ; this macro first transpose data in-registers, and then store
* them in memory. much faster with icc. */
#define ONEQUAD_TRANSPOSE(A, B, C, D) \
z##A = _mm_add_epi32(z##A, orig##A); \
z##B = _mm_add_epi32(z##B, orig##B); \
z##C = _mm_add_epi32(z##C, orig##C); \
z##D = _mm_add_epi32(z##D, orig##D); \
y##A = _mm_unpacklo_epi32(z##A, z##B); \
y##B = _mm_unpacklo_epi32(z##C, z##D); \
y##C = _mm_unpackhi_epi32(z##A, z##B); \
y##D = _mm_unpackhi_epi32(z##C, z##D); \
z##A = _mm_unpacklo_epi64(y##A, y##B); \
z##B = _mm_unpackhi_epi64(y##A, y##B); \
z##C = _mm_unpacklo_epi64(y##C, y##D); \
z##D = _mm_unpackhi_epi64(y##C, y##D); \
y##A = _mm_xor_si128(z##A, _mm_loadu_si128((__m128i *) (m + 0))); \
_mm_storeu_si128((__m128i *) (c + 0), y##A); \
y##B = _mm_xor_si128(z##B, _mm_loadu_si128((__m128i *) (m + 64))); \
_mm_storeu_si128((__m128i *) (c + 64), y##B); \
y##C = _mm_xor_si128(z##C, _mm_loadu_si128((__m128i *) (m + 128))); \
_mm_storeu_si128((__m128i *) (c + 128), y##C); \
y##D = _mm_xor_si128(z##D, _mm_loadu_si128((__m128i *) (m + 192))); \
#define ONEQUAD_TRANSPOSE(A, B, C, D) \
z##A = _mm_add_epi32(z##A, orig##A); \
z##B = _mm_add_epi32(z##B, orig##B); \
z##C = _mm_add_epi32(z##C, orig##C); \
z##D = _mm_add_epi32(z##D, orig##D); \
y##A = _mm_unpacklo_epi32(z##A, z##B); \
y##B = _mm_unpacklo_epi32(z##C, z##D); \
y##C = _mm_unpackhi_epi32(z##A, z##B); \
y##D = _mm_unpackhi_epi32(z##C, z##D); \
z##A = _mm_unpacklo_epi64(y##A, y##B); \
z##B = _mm_unpackhi_epi64(y##A, y##B); \
z##C = _mm_unpacklo_epi64(y##C, y##D); \
z##D = _mm_unpackhi_epi64(y##C, y##D); \
y##A = _mm_xor_si128(z##A, _mm_loadu_si128((const __m128i *) (m + 0))); \
_mm_storeu_si128((__m128i *) (c + 0), y##A); \
y##B = _mm_xor_si128(z##B, _mm_loadu_si128((const __m128i *) (m + 64))); \
_mm_storeu_si128((__m128i *) (c + 64), y##B); \
y##C = _mm_xor_si128(z##C, _mm_loadu_si128((const __m128i *) (m + 128))); \
_mm_storeu_si128((__m128i *) (c + 128), y##C); \
y##D = _mm_xor_si128(z##D, _mm_loadu_si128((const __m128i *) (m + 192))); \
_mm_storeu_si128((__m128i *) (c + 192), y##D)
#define ONEQUAD(A, B, C, D) ONEQUAD_TRANSPOSE(A, B, C, D)
src/libsodium/crypto_stream/salsa20/xmm6int/u8.h
+53 -53
View File
@@ -361,45 +361,45 @@ if (bytes >= 512) {
/* store data ; this macro first transpose data in-registers, and then store
* them in memory. much faster with icc. */
#define ONEQUAD_TRANSPOSE(A, B, C, D) \
{ \
__m128i t0, t1, t2, t3; \
z##A = _mm256_add_epi32(z##A, orig##A); \
z##B = _mm256_add_epi32(z##B, orig##B); \
z##C = _mm256_add_epi32(z##C, orig##C); \
z##D = _mm256_add_epi32(z##D, orig##D); \
y##A = _mm256_unpacklo_epi32(z##A, z##B); \
y##B = _mm256_unpacklo_epi32(z##C, z##D); \
y##C = _mm256_unpackhi_epi32(z##A, z##B); \
y##D = _mm256_unpackhi_epi32(z##C, z##D); \
z##A = _mm256_unpacklo_epi64(y##A, y##B); \
z##B = _mm256_unpackhi_epi64(y##A, y##B); \
z##C = _mm256_unpacklo_epi64(y##C, y##D); \
z##D = _mm256_unpackhi_epi64(y##C, y##D); \
t0 = _mm_xor_si128(_mm256_extracti128_si256(z##A, 0), \
_mm_loadu_si128((__m128i*) (m + 0))); \
_mm_storeu_si128((__m128i*) (c + 0), t0); \
t1 = _mm_xor_si128(_mm256_extracti128_si256(z##B, 0), \
_mm_loadu_si128((__m128i*) (m + 64))); \
_mm_storeu_si128((__m128i*) (c + 64), t1); \
t2 = _mm_xor_si128(_mm256_extracti128_si256(z##C, 0), \
_mm_loadu_si128((__m128i*) (m + 128))); \
_mm_storeu_si128((__m128i*) (c + 128), t2); \
t3 = _mm_xor_si128(_mm256_extracti128_si256(z##D, 0), \
_mm_loadu_si128((__m128i*) (m + 192))); \
_mm_storeu_si128((__m128i*) (c + 192), t3); \
t0 = _mm_xor_si128(_mm256_extracti128_si256(z##A, 1), \
_mm_loadu_si128((__m128i*) (m + 256))); \
_mm_storeu_si128((__m128i*) (c + 256), t0); \
t1 = _mm_xor_si128(_mm256_extracti128_si256(z##B, 1), \
_mm_loadu_si128((__m128i*) (m + 320))); \
_mm_storeu_si128((__m128i*) (c + 320), t1); \
t2 = _mm_xor_si128(_mm256_extracti128_si256(z##C, 1), \
_mm_loadu_si128((__m128i*) (m + 384))); \
_mm_storeu_si128((__m128i*) (c + 384), t2); \
t3 = _mm_xor_si128(_mm256_extracti128_si256(z##D, 1), \
_mm_loadu_si128((__m128i*) (m + 448))); \
_mm_storeu_si128((__m128i*) (c + 448), t3); \
#define ONEQUAD_TRANSPOSE(A, B, C, D) \
{ \
__m128i t0, t1, t2, t3; \
z##A = _mm256_add_epi32(z##A, orig##A); \
z##B = _mm256_add_epi32(z##B, orig##B); \
z##C = _mm256_add_epi32(z##C, orig##C); \
z##D = _mm256_add_epi32(z##D, orig##D); \
y##A = _mm256_unpacklo_epi32(z##A, z##B); \
y##B = _mm256_unpacklo_epi32(z##C, z##D); \
y##C = _mm256_unpackhi_epi32(z##A, z##B); \
y##D = _mm256_unpackhi_epi32(z##C, z##D); \
z##A = _mm256_unpacklo_epi64(y##A, y##B); \
z##B = _mm256_unpackhi_epi64(y##A, y##B); \
z##C = _mm256_unpacklo_epi64(y##C, y##D); \
z##D = _mm256_unpackhi_epi64(y##C, y##D); \
t0 = _mm_xor_si128(_mm256_extracti128_si256(z##A, 0), \
_mm_loadu_si128((const __m128i*) (m + 0))); \
_mm_storeu_si128((__m128i*) (c + 0), t0); \
t1 = _mm_xor_si128(_mm256_extracti128_si256(z##B, 0), \
_mm_loadu_si128((const __m128i*) (m + 64))); \
_mm_storeu_si128((__m128i*) (c + 64), t1); \
t2 = _mm_xor_si128(_mm256_extracti128_si256(z##C, 0), \
_mm_loadu_si128((const __m128i*) (m + 128))); \
_mm_storeu_si128((__m128i*) (c + 128), t2); \
t3 = _mm_xor_si128(_mm256_extracti128_si256(z##D, 0), \
_mm_loadu_si128((const __m128i*) (m + 192))); \
_mm_storeu_si128((__m128i*) (c + 192), t3); \
t0 = _mm_xor_si128(_mm256_extracti128_si256(z##A, 1), \
_mm_loadu_si128((const __m128i*) (m + 256))); \
_mm_storeu_si128((__m128i*) (c + 256), t0); \
t1 = _mm_xor_si128(_mm256_extracti128_si256(z##B, 1), \
_mm_loadu_si128((const __m128i*) (m + 320))); \
_mm_storeu_si128((__m128i*) (c + 320), t1); \
t2 = _mm_xor_si128(_mm256_extracti128_si256(z##C, 1), \
_mm_loadu_si128((const __m128i*) (m + 384))); \
_mm_storeu_si128((__m128i*) (c + 384), t2); \
t3 = _mm_xor_si128(_mm256_extracti128_si256(z##D, 1), \
_mm_loadu_si128((const __m128i*) (m + 448))); \
_mm_storeu_si128((__m128i*) (c + 448), t3); \
}
#define ONEQUAD(A, B, C, D) ONEQUAD_TRANSPOSE(A, B, C, D)
@@ -433,20 +433,20 @@ if (bytes >= 512) {
y##D = _mm256_permute2x128_si256(z##D, z##D2, 0x20); \
y##D2 = _mm256_permute2x128_si256(z##D, z##D2, 0x31); \
y##A = _mm256_xor_si256(y##A, _mm256_loadu_si256((__m256i*) (m + 0))); \
y##B = \
_mm256_xor_si256(y##B, _mm256_loadu_si256((__m256i*) (m + 64))); \
y##C = \
_mm256_xor_si256(y##C, _mm256_loadu_si256((__m256i*) (m + 128))); \
y##D = \
_mm256_xor_si256(y##D, _mm256_loadu_si256((__m256i*) (m + 192))); \
y##A2 = \
_mm256_xor_si256(y##A2, _mm256_loadu_si256((__m256i*) (m + 256))); \
y##B2 = \
_mm256_xor_si256(y##B2, _mm256_loadu_si256((__m256i*) (m + 320))); \
y##C2 = \
_mm256_xor_si256(y##C2, _mm256_loadu_si256((__m256i*) (m + 384))); \
y##D2 = \
_mm256_xor_si256(y##D2, _mm256_loadu_si256((__m256i*) (m + 448))); \
y##B = _mm256_xor_si256( \
y##B, _mm256_loadu_si256((const __m256i*) (m + 64))); \
y##C = _mm256_xor_si256( \
y##C, _mm256_loadu_si256((const __m256i*) (m + 128))); \
y##D = _mm256_xor_si256( \
y##D, _mm256_loadu_si256((const __m256i*) (m + 192))); \
y##A2 = _mm256_xor_si256( \
y##A2, _mm256_loadu_si256((const __m256i*) (m + 256))); \
y##B2 = _mm256_xor_si256( \
y##B2, _mm256_loadu_si256((const __m256i*) (m + 320))); \
y##C2 = _mm256_xor_si256( \
y##C2, _mm256_loadu_si256((const __m256i*) (m + 384))); \
y##D2 = _mm256_xor_si256( \
y##D2, _mm256_loadu_si256((const __m256i*) (m + 448))); \
_mm256_storeu_si256((__m256i*) (c + 0), y##A); \
_mm256_storeu_si256((__m256i*) (c + 64), y##B); \
_mm256_storeu_si256((__m256i*) (c + 128), y##C); \
src/libsodium/crypto_verify/sodium/verify.c
+4 -4
View File
@@ -44,12 +44,12 @@ crypto_verify_n(const unsigned char *x_, const unsigned char *y_,
(const volatile __m128i *volatile) (const void *) x_;
const volatile __m128i *volatile y =
(const volatile __m128i *volatile) (const void *) y_;
v1 = _mm_loadu_si128((const volatile __m128i *) &x[0]);
v2 = _mm_loadu_si128((const volatile __m128i *) &y[0]);
v1 = _mm_loadu_si128((const __m128i *) &x[0]);
v2 = _mm_loadu_si128((const __m128i *) &y[0]);
z = _mm_xor_si128(v1, v2);
for (i = 1; i < n / 16; i++) {
v1 = _mm_loadu_si128((const volatile __m128i *) &x[i]);
v2 = _mm_loadu_si128((const volatile __m128i *) &y[i]);
v1 = _mm_loadu_si128((const __m128i *) &x[i]);
v2 = _mm_loadu_si128((const __m128i *) &y[i]);
z = _mm_or_si128(z, _mm_xor_si128(v1, v2));
}
m = _mm_movemask_epi8(_mm_cmpeq_epi32(z, zero));