diff --git a/src/libsodium/crypto_generichash/blake2b/ref/blake2b-compress-avx2.h b/src/libsodium/crypto_generichash/blake2b/ref/blake2b-compress-avx2.h index d08603a8..7c11321b 100644 --- a/src/libsodium/crypto_generichash/blake2b/ref/blake2b-compress-avx2.h +++ b/src/libsodium/crypto_generichash/blake2b/ref/blake2b-compress-avx2.h @@ -68,17 +68,17 @@ LOADU64(const void *p) #define BLAKE2B_DIAG_V1(a, b, c, d) \ do { \ - d = _mm256_permute4x64_epi64(d, _MM_SHUFFLE(2, 1, 0, 3)); \ - c = _mm256_permute4x64_epi64(c, _MM_SHUFFLE(1, 0, 3, 2)); \ - b = _mm256_permute4x64_epi64(b, _MM_SHUFFLE(0, 3, 2, 1)); \ - } while (0) + a = _mm256_permute4x64_epi64(a, _MM_SHUFFLE(2, 1, 0, 3)); \ + d = _mm256_permute4x64_epi64(d, _MM_SHUFFLE(1, 0, 3, 2)); \ + c = _mm256_permute4x64_epi64(c, _MM_SHUFFLE(0, 3, 2, 1)); \ + } while(0) #define BLAKE2B_UNDIAG_V1(a, b, c, d) \ do { \ - d = _mm256_permute4x64_epi64(d, _MM_SHUFFLE(0, 3, 2, 1)); \ - c = _mm256_permute4x64_epi64(c, _MM_SHUFFLE(1, 0, 3, 2)); \ - b = _mm256_permute4x64_epi64(b, _MM_SHUFFLE(2, 1, 0, 3)); \ - } while (0) + a = _mm256_permute4x64_epi64(a, _MM_SHUFFLE(0, 3, 2, 1)); \ + d = _mm256_permute4x64_epi64(d, _MM_SHUFFLE(1, 0, 3, 2)); \ + c = _mm256_permute4x64_epi64(c, _MM_SHUFFLE(2, 1, 0, 3)); \ + } while(0) #include "blake2b-load-avx2.h" diff --git a/src/libsodium/crypto_generichash/blake2b/ref/blake2b-load-avx2.h b/src/libsodium/crypto_generichash/blake2b/ref/blake2b-load-avx2.h index 8c15f177..12a5d189 100644 --- a/src/libsodium/crypto_generichash/blake2b/ref/blake2b-load-avx2.h +++ b/src/libsodium/crypto_generichash/blake2b/ref/blake2b-load-avx2.h @@ -17,15 +17,15 @@ #define BLAKE2B_LOAD_MSG_0_3(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m4, m5); \ - t1 = _mm256_unpacklo_epi64(m6, m7); \ + t0 = _mm256_unpacklo_epi64(m7, m4); \ + t1 = _mm256_unpacklo_epi64(m5, m6); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_0_4(b0) \ do { \ - t0 = _mm256_unpackhi_epi64(m4, m5); \ - t1 = _mm256_unpackhi_epi64(m6, m7); \ + t0 = _mm256_unpackhi_epi64(m7, m4); \ + t1 = _mm256_unpackhi_epi64(m5, m6); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -43,17 +43,17 @@ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) -#define BLAKE2B_LOAD_MSG_1_3(b0) \ - do { \ - t0 = _mm256_shuffle_epi32(m0, _MM_SHUFFLE(1, 0, 3, 2)); \ - t1 = _mm256_unpackhi_epi64(m5, m2); \ - b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ +#define BLAKE2B_LOAD_MSG_1_3(b0) \ + do { \ + t0 = _mm256_unpackhi_epi64(m2, m0); \ + t1 = _mm256_blend_epi32(m5, m0, 0x33); \ + b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_1_4(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m6, m1); \ - t1 = _mm256_unpackhi_epi64(m3, m1); \ + t0 = _mm256_alignr_epi8(m6, m1, 8); \ + t1 = _mm256_blend_epi32(m3, m1, 0x33); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -73,15 +73,15 @@ #define BLAKE2B_LOAD_MSG_2_3(b0) \ do { \ - t0 = _mm256_blend_epi32(m1, m5, 0x33); \ - t1 = _mm256_unpackhi_epi64(m3, m4); \ + t0 = _mm256_alignr_epi8(m5, m4, 8); \ + t1 = _mm256_unpackhi_epi64(m1, m3); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_2_4(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m7, m3); \ - t1 = _mm256_alignr_epi8(m2, m0, 8); \ + t0 = _mm256_unpacklo_epi64(m2, m7); \ + t1 = _mm256_blend_epi32(m0, m3, 0x33); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -99,17 +99,17 @@ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) -#define BLAKE2B_LOAD_MSG_3_3(b0) \ - do { \ - t0 = _mm256_blend_epi32(m2, m1, 0x33); \ - t1 = _mm256_blend_epi32(m7, m2, 0x33); \ - b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ +#define BLAKE2B_LOAD_MSG_3_3(b0) \ + do { \ + t0 = _mm256_alignr_epi8(m1, m7, 8); \ + t1 = _mm256_shuffle_epi32(m2, _MM_SHUFFLE(1, 0, 3, 2)); \ + b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_3_4(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m3, m5); \ - t1 = _mm256_unpacklo_epi64(m0, m4); \ + t0 = _mm256_unpacklo_epi64(m4, m3); \ + t1 = _mm256_unpacklo_epi64(m5, m0); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -129,15 +129,15 @@ #define BLAKE2B_LOAD_MSG_4_3(b0) \ do { \ - t0 = _mm256_blend_epi32(m5, m7, 0x33); \ - t1 = _mm256_blend_epi32(m1, m3, 0x33); \ + t0 = _mm256_alignr_epi8(m7, m1, 8); \ + t1 = _mm256_alignr_epi8(m3, m5, 8); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_4_4(b0) \ do { \ - t0 = _mm256_alignr_epi8(m6, m0, 8); \ - t1 = _mm256_blend_epi32(m6, m4, 0x33); \ + t0 = _mm256_unpackhi_epi64(m6, m0); \ + t1 = _mm256_unpacklo_epi64(m6, m4); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -157,15 +157,15 @@ #define BLAKE2B_LOAD_MSG_5_3(b0) \ do { \ - t0 = _mm256_blend_epi32(m3, m2, 0x33); \ - t1 = _mm256_unpackhi_epi64(m7, m0); \ + t0 = _mm256_alignr_epi8(m2, m0, 8); \ + t1 = _mm256_unpackhi_epi64(m3, m7); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_5_4(b0) \ do { \ - t0 = _mm256_unpackhi_epi64(m6, m2); \ - t1 = _mm256_blend_epi32(m4, m7, 0x33); \ + t0 = _mm256_unpackhi_epi64(m4, m6); \ + t1 = _mm256_alignr_epi8(m7, m2, 8); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -183,18 +183,18 @@ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) -#define BLAKE2B_LOAD_MSG_6_3(b0) \ - do { \ - t0 = _mm256_unpacklo_epi64(m0, m3); \ - t1 = _mm256_shuffle_epi32(m4, _MM_SHUFFLE(1, 0, 3, 2)); \ - b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ +#define BLAKE2B_LOAD_MSG_6_3(b0) \ + do { \ + t0 = _mm256_unpacklo_epi64(m4, m0); \ + t1 = _mm256_blend_epi32(m4, m3, 0x33); \ + b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) -#define BLAKE2B_LOAD_MSG_6_4(b0) \ - do { \ - t0 = _mm256_unpackhi_epi64(m3, m1); \ - t1 = _mm256_blend_epi32(m5, m1, 0x33); \ - b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ +#define BLAKE2B_LOAD_MSG_6_4(b0) \ + do { \ + t0 = _mm256_unpackhi_epi64(m5, m3); \ + t1 = _mm256_shuffle_epi32(m1, _MM_SHUFFLE(1, 0, 3, 2)); \ + b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_7_1(b0) \ @@ -213,15 +213,15 @@ #define BLAKE2B_LOAD_MSG_7_3(b0) \ do { \ - t0 = _mm256_unpackhi_epi64(m2, m7); \ - t1 = _mm256_unpacklo_epi64(m4, m1); \ + t0 = _mm256_blend_epi32(m2, m1, 0x33); \ + t1 = _mm256_alignr_epi8(m4, m7, 8); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_7_4(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m0, m2); \ - t1 = _mm256_unpacklo_epi64(m3, m5); \ + t0 = _mm256_unpacklo_epi64(m5, m0); \ + t1 = _mm256_unpacklo_epi64(m2, m3); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -241,15 +241,15 @@ #define BLAKE2B_LOAD_MSG_8_3(b0) \ do { \ - t0 = m6; \ - t1 = _mm256_alignr_epi8(m5, m0, 8); \ + t0 = _mm256_unpacklo_epi64(m5, m6); \ + t1 = _mm256_unpackhi_epi64(m6, m0); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_8_4(b0) \ do { \ - t0 = _mm256_blend_epi32(m3, m1, 0x33); \ - t1 = m2; \ + t0 = _mm256_alignr_epi8(m1, m2, 8); \ + t1 = _mm256_alignr_epi8(m2, m3, 8); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -269,15 +269,15 @@ #define BLAKE2B_LOAD_MSG_9_3(b0) \ do { \ - t0 = _mm256_unpackhi_epi64(m7, m4); \ - t1 = _mm256_unpackhi_epi64(m1, m6); \ + t0 = _mm256_unpackhi_epi64(m6, m7); \ + t1 = _mm256_unpackhi_epi64(m4, m1); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_9_4(b0) \ do { \ - t0 = _mm256_alignr_epi8(m7, m5, 8); \ - t1 = _mm256_unpacklo_epi64(m6, m0); \ + t0 = _mm256_blend_epi32(m5, m0, 0x33); \ + t1 = _mm256_unpacklo_epi64(m7, m6); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -297,15 +297,15 @@ #define BLAKE2B_LOAD_MSG_10_3(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m4, m5); \ - t1 = _mm256_unpacklo_epi64(m6, m7); \ + t0 = _mm256_unpacklo_epi64(m7, m4); \ + t1 = _mm256_unpacklo_epi64(m5, m6); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_10_4(b0) \ do { \ - t0 = _mm256_unpackhi_epi64(m4, m5); \ - t1 = _mm256_unpackhi_epi64(m6, m7); \ + t0 = _mm256_unpackhi_epi64(m7, m4); \ + t1 = _mm256_unpackhi_epi64(m5, m6); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) @@ -323,17 +323,17 @@ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) -#define BLAKE2B_LOAD_MSG_11_3(b0) \ - do { \ - t0 = _mm256_shuffle_epi32(m0, _MM_SHUFFLE(1, 0, 3, 2)); \ - t1 = _mm256_unpackhi_epi64(m5, m2); \ - b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ +#define BLAKE2B_LOAD_MSG_11_3(b0) \ + do { \ + t0 = _mm256_unpackhi_epi64(m2, m0); \ + t1 = _mm256_blend_epi32(m5, m0, 0x33); \ + b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0) #define BLAKE2B_LOAD_MSG_11_4(b0) \ do { \ - t0 = _mm256_unpacklo_epi64(m6, m1); \ - t1 = _mm256_unpackhi_epi64(m3, m1); \ + t0 = _mm256_alignr_epi8(m6, m1, 8); \ + t1 = _mm256_blend_epi32(m3, m1, 0x33); \ b0 = _mm256_blend_epi32(t0, t1, 0xF0); \ } while (0)