pq_crypto 0.6.1 → 0.6.2

data/ext/pqcrypto/vendor/mldsa-native/mldsa/src/native/x86_64/src/pointwise_acc_l7.S

@@ -1,187 +0,0 @@
-/*
- * Copyright (c) The mldsa-native project authors
- * SPDX-License-Identifier: Apache-2.0 OR ISC OR MIT
- */
-
-/* References
- * ==========
- *
- * - [REF_AVX2]
- *   CRYSTALS-Dilithium optimized AVX2 implementation
- *   Bai, Ducas, Kiltz, Lepoint, Lyubashevsky, Schwabe, Seiler, Stehlé
- *   https://github.com/pq-crystals/dilithium/tree/master/avx2
- */
-
-/*
- * This file is derived from the public domain
- * AVX2 Dilithium implementation @[REF_AVX2].
- */
-
-#include "../../../common.h"
-#if defined(MLD_ARITH_BACKEND_X86_64_DEFAULT) && \
-    !defined(MLD_CONFIG_MULTILEVEL_NO_SHARED) && \
-    (defined(MLD_CONFIG_MULTILEVEL_WITH_SHARED) || MLDSA_L == 7)
-
-/*
- * WARNING: This file is auto-derived from the mldsa-native source file
- *   dev/x86_64/src/pointwise_acc_l7.S using scripts/simpasm. Do not modify it directly.
- */
-
-#if defined(__ELF__)
-.section .note.GNU-stack,"",@progbits
-#endif
-
-.text
-.balign 4
-.global MLD_ASM_NAMESPACE(pointwise_acc_l7_avx2)
-MLD_ASM_FN_SYMBOL(pointwise_acc_l7_avx2)
-
-        .cfi_startproc
-        vmovdqa	0x20(%rcx), %ymm0
-        vmovdqa	(%rcx), %ymm1
-        xorl	%eax, %eax
-
-Lpointwise_acc_l7_avx2_looptop2:
-        vmovdqa	(%rsi), %ymm6
-        vmovdqa	0x20(%rsi), %ymm8
-        vmovdqa	(%rdx), %ymm10
-        vmovdqa	0x20(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vmovdqa	%ymm6, %ymm2
-        vmovdqa	%ymm7, %ymm3
-        vmovdqa	%ymm8, %ymm4
-        vmovdqa	%ymm9, %ymm5
-        vmovdqa	0x400(%rsi), %ymm6
-        vmovdqa	0x420(%rsi), %ymm8
-        vmovdqa	0x400(%rdx), %ymm10
-        vmovdqa	0x420(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vpaddq	%ymm2, %ymm6, %ymm2
-        vpaddq	%ymm3, %ymm7, %ymm3
-        vpaddq	%ymm4, %ymm8, %ymm4
-        vpaddq	%ymm5, %ymm9, %ymm5
-        vmovdqa	0x800(%rsi), %ymm6
-        vmovdqa	0x820(%rsi), %ymm8
-        vmovdqa	0x800(%rdx), %ymm10
-        vmovdqa	0x820(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vpaddq	%ymm2, %ymm6, %ymm2
-        vpaddq	%ymm3, %ymm7, %ymm3
-        vpaddq	%ymm4, %ymm8, %ymm4
-        vpaddq	%ymm5, %ymm9, %ymm5
-        vmovdqa	0xc00(%rsi), %ymm6
-        vmovdqa	0xc20(%rsi), %ymm8
-        vmovdqa	0xc00(%rdx), %ymm10
-        vmovdqa	0xc20(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vpaddq	%ymm2, %ymm6, %ymm2
-        vpaddq	%ymm3, %ymm7, %ymm3
-        vpaddq	%ymm4, %ymm8, %ymm4
-        vpaddq	%ymm5, %ymm9, %ymm5
-        vmovdqa	0x1000(%rsi), %ymm6
-        vmovdqa	0x1020(%rsi), %ymm8
-        vmovdqa	0x1000(%rdx), %ymm10
-        vmovdqa	0x1020(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vpaddq	%ymm2, %ymm6, %ymm2
-        vpaddq	%ymm3, %ymm7, %ymm3
-        vpaddq	%ymm4, %ymm8, %ymm4
-        vpaddq	%ymm5, %ymm9, %ymm5
-        vmovdqa	0x1400(%rsi), %ymm6
-        vmovdqa	0x1420(%rsi), %ymm8
-        vmovdqa	0x1400(%rdx), %ymm10
-        vmovdqa	0x1420(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vpaddq	%ymm2, %ymm6, %ymm2
-        vpaddq	%ymm3, %ymm7, %ymm3
-        vpaddq	%ymm4, %ymm8, %ymm4
-        vpaddq	%ymm5, %ymm9, %ymm5
-        vmovdqa	0x1800(%rsi), %ymm6
-        vmovdqa	0x1820(%rsi), %ymm8
-        vmovdqa	0x1800(%rdx), %ymm10
-        vmovdqa	0x1820(%rdx), %ymm12
-        vpsrlq	$0x20, %ymm6, %ymm7
-        vpsrlq	$0x20, %ymm8, %ymm9
-        vmovshdup	%ymm10, %ymm11  # ymm11 = ymm10[1,1,3,3,5,5,7,7]
-        vmovshdup	%ymm12, %ymm13  # ymm13 = ymm12[1,1,3,3,5,5,7,7]
-        vpmuldq	%ymm10, %ymm6, %ymm6
-        vpmuldq	%ymm11, %ymm7, %ymm7
-        vpmuldq	%ymm12, %ymm8, %ymm8
-        vpmuldq	%ymm13, %ymm9, %ymm9
-        vpaddq	%ymm2, %ymm6, %ymm2
-        vpaddq	%ymm3, %ymm7, %ymm3
-        vpaddq	%ymm4, %ymm8, %ymm4
-        vpaddq	%ymm5, %ymm9, %ymm5
-        vpmuldq	%ymm2, %ymm0, %ymm6
-        vpmuldq	%ymm3, %ymm0, %ymm7
-        vpmuldq	%ymm4, %ymm0, %ymm8
-        vpmuldq	%ymm5, %ymm0, %ymm9
-        vpmuldq	%ymm6, %ymm1, %ymm6
-        vpmuldq	%ymm7, %ymm1, %ymm7
-        vpmuldq	%ymm8, %ymm1, %ymm8
-        vpmuldq	%ymm9, %ymm1, %ymm9
-        vpsubq	%ymm6, %ymm2, %ymm2
-        vpsubq	%ymm7, %ymm3, %ymm3
-        vpsubq	%ymm8, %ymm4, %ymm4
-        vpsubq	%ymm9, %ymm5, %ymm5
-        vpsrlq	$0x20, %ymm2, %ymm2
-        vmovshdup	%ymm4, %ymm4    # ymm4 = ymm4[1,1,3,3,5,5,7,7]
-        vpblendd	$0xaa, %ymm3, %ymm2, %ymm2 # ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[3],ymm2[4],ymm3[5],ymm2[6],ymm3[7]
-        vpblendd	$0xaa, %ymm5, %ymm4, %ymm4 # ymm4 = ymm4[0],ymm5[1],ymm4[2],ymm5[3],ymm4[4],ymm5[5],ymm4[6],ymm5[7]
-        vmovdqa	%ymm2, (%rdi)
-        vmovdqa	%ymm4, 0x20(%rdi)
-        addq	$0x40, %rsi
-        addq	$0x40, %rdx
-        addq	$0x40, %rdi
-        addl	$0x1, %eax
-        cmpl	$0x10, %eax
-        jb	Lpointwise_acc_l7_avx2_looptop2
-        retq
-        .cfi_endproc
-
-MLD_ASM_FN_SIZE(pointwise_acc_l7_avx2)
-
-#endif /* MLD_ARITH_BACKEND_X86_64_DEFAULT && !MLD_CONFIG_MULTILEVEL_NO_SHARED \
-          && (MLD_CONFIG_MULTILEVEL_WITH_SHARED || MLDSA_L == 7) */

data/ext/pqcrypto/vendor/mldsa-native/mldsa/src/native/x86_64/src/poly_caddq_avx2.c

@@ -1,61 +0,0 @@
-/*
- * Copyright (c) The mldsa-native project authors
- * SPDX-License-Identifier: Apache-2.0 OR ISC OR MIT
- */
-
-/* References
- * ==========
- *
- * - [REF_AVX2]
- *   CRYSTALS-Dilithium optimized AVX2 implementation
- *   Bai, Ducas, Kiltz, Lepoint, Lyubashevsky, Schwabe, Seiler, Stehlé
- *   https://github.com/pq-crystals/dilithium/tree/master/avx2
- */
-
-/*
- * This file is derived from the public domain
- * AVX2 Dilithium implementation @[REF_AVX2].
- */
-
-#include "../../../common.h"
-
-#if defined(MLD_ARITH_BACKEND_X86_64_DEFAULT) && \
-    !defined(MLD_CONFIG_MULTILEVEL_NO_SHARED)
-
-#include <immintrin.h>
-#include "arith_native_x86_64.h"
-#include "consts.h"
-
-/*************************************************
- * Name:        mld_poly_caddq_avx2
- *
- * Description: For all coefficients of in/out polynomial add Q if
- *              coefficient is negative.
- *
- * Arguments:   - int32_t *r: pointer to input/output polynomial
- **************************************************/
-void mld_poly_caddq_avx2(int32_t *r)
-{
-  unsigned int i;
-  __m256i f, g;
-  const __m256i q = _mm256_set1_epi32(MLDSA_Q);
-  const __m256i zero = _mm256_setzero_si256();
-  __m256i *rr = (__m256i *)r;
-
-  for (i = 0; i < MLDSA_N / 8; i++)
-  {
-    f = _mm256_load_si256(&rr[i]);
-    g = _mm256_cmpgt_epi32(zero, f);
-    g = _mm256_and_si256(g, q);
-    f = _mm256_add_epi32(f, g);
-    _mm256_store_si256(&rr[i], f);
-  }
-}
-
-#else /* MLD_ARITH_BACKEND_X86_64_DEFAULT && !MLD_CONFIG_MULTILEVEL_NO_SHARED \
-       */
-
-MLD_EMPTY_CU(avx2_reduce)
-
-#endif /* !(MLD_ARITH_BACKEND_X86_64_DEFAULT && \
-          !MLD_CONFIG_MULTILEVEL_NO_SHARED) */