WW07'25 source code update

Author: ipl_ci

CHANGELOG.md

@@ -11,6 +11,8 @@ Cmake build options `-DMERGED_BLD:BOOL=off -DMBX_PLATFORM_LIST="k1;l9"` may be u
 [BUILD.md](./BUILD.md) for the details.
 - Fixed AVX512 IFMA implementation (k1 branch) of SM2 signature and verification single-buffer algorithm. The optimized path is re-enabled.
 - Added `ippsHashMethod_SM3_NI` and `ippsHashMethod_SM3_TT` methods for SM3 hash algorithm optimization with the new SM3 instructions for Lunar Lake and Arrow Lake S CPUs. The runtime dispatch introduced in Intel(R) Cryptography Primitives Library 1.0.0 release `ippsHashMethod_SM3` is moved to `ippsHashMethod_SM3_TT` and the behavior of `ippsHashMethod` API is aligned with SHA hash family.
+- Deprecated `fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys` and `fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size`.
+Please see [DEPRECATION_NOTES](DEPRECATION_NOTES.md) for more details.
 
 ## Intel(R) Cryptography Primitives Library 1.0.1
 - Fixed an issue with invalid memory access for AES-GCM algorithm with Intel® Advanced Vector Extensions 2 (Intel® AVX2) vector extensions of Intel® AES New Instructions (Intel® AES-NI) in case of corner sizes.

DEPRECATION_NOTES.md

@@ -4,6 +4,20 @@ This document describes deprecated API in different Intel® Cryptography Primiti
 
 The deprecated API means it is obsolete and will be removed in one of future Intel® Cryptography Primitives Library releases. If you have any concerns, please use the following link for opening a ticket and providing feedback:  <https://supporttickets.intel.com.>
 
+## Intel® Cryptography Primitives Library v1.1.0
+
+### FIPS self-tests
+
+The common `fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size*` functions have been specialized for signing and signature verification.
+The memory footprint of the corresponding FIPS self-tests was reduced.
+
+| Deprecated                                                 | Recommended replacement                                | Context                                                                |
+| :--------------------------------------------------------- | :----------------------------------------------------: | :--------------------------------------------------------------------: |
+| fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys | fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size_keys   | Understand memory requirements of RSA PKCS1 v1.5 signature generation  |
+| fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size      | fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size        | Understand memory requirements of RSA PKCS1 v1.5 signature generation  |
+| fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys | fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size_keys | Understand memory requirements of RSA PKCS1 v1.5 signature verfication |
+| fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size      | fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size      | Understand memory requirements of RSA PKCS1 v1.5 signature verfication |
+
 ## Intel® Cryptography Primitives Library v1.0.0
 
 ### Service Functions

README_FIPS.md

@@ -276,15 +276,31 @@ fips_test_status fips_selftest_ippsHMACMessage_rmf (void);
 ##### RSA sign/verify
 
 ```cpp
-fips_test_status fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys (int *pKeysBufferSize);
-fips_test_status fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size (int *pBufferSize Ipp8u *pKeysBuffer);
+fips_test_status fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size_keys (int *pKeysBufferSize);
+fips_test_status fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size (int *pBufferSize Ipp8u *pKeysBuffer);
 fips_test_status fips_selftest_ippsRSASign_PKCS1v15_rmf (Ipp8u *pBuffer Ipp8u *pKeysBuffer);
+```
+
+, where `pBuffer` is the valid buffer for selftest of size indicated by
+`fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size` and `pKeysBuffer` is the
+valid buffer for selftest of size indicated by `fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size_keys`.
+
+```cpp
+fips_test_status fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size_keys (int *pKeysBufferSize);
+fips_test_status fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size (int *pBufferSize Ipp8u *pKeysBuffer);
 fips_test_status fips_selftest_ippsRSAVerify_PKCS1v15_rmf (Ipp8u *pBuffer Ipp8u *pKeysBuffer);
 ```
 
 , where `pBuffer` is the valid buffer for selftest of size indicated by
-`fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size` and `pKeysBuffer` is the
-valid buffer for selftest of size indicated by `fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys`.
+`fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size` and `pKeysBuffer` is the
+valid buffer for selftest of size indicated by `fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size_keys`.
+
+Following APIs have been deprecated:
+```cpp
+fips_test_status fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys (int *pKeysBufferSize);
+fips_test_status fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size (int *pBufferSize Ipp8u *pKeysBuffer);
+```
+Their transition plan can be found in [DEPRECATION_NOTES](./DEPRECATION_NOTES.md).
 
 ```cpp
 fips_test_status fips_selftest_ippsRSASignVerify_PSS_rmf_get_size_keys (int *pKeysBufferSize);

include/ippcp/fips_cert.h

@@ -84,8 +84,20 @@ IPPAPI(fips_test_status, fips_selftest_ippsHMACUpdate_rmf, (Ipp8u *pBuffer))
 IPPAPI(fips_test_status, fips_selftest_ippsHMACMessage_rmf, (void))
 
 /* RSA sign/verify */
+
+#define LEAN_GET_SIZE "Function reports more memory than required. A leaner alternative is available: \
+fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size_keys, fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size, \
+fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size_keys, fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size."
+
+IPP_DEPRECATED(LEAN_GET_SIZE) \
 IPPAPI(fips_test_status, fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size_keys, (int *pKeysBufferSize))
+IPP_DEPRECATED(LEAN_GET_SIZE) \
 IPPAPI(fips_test_status, fips_selftest_ippsRSASignVerify_PKCS1v15_rmf_get_size, (int *pBufferSize, Ipp8u *pKeysBuffer))
+
+IPPAPI(fips_test_status, fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size_keys, (int *pKeysBufferSize))
+IPPAPI(fips_test_status, fips_selftest_ippsRSASign_PKCS1v15_rmf_get_size, (int *pBufferSize, Ipp8u *pKeysBuffer))
+IPPAPI(fips_test_status, fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size_keys, (int *pKeysBufferSize))
+IPPAPI(fips_test_status, fips_selftest_ippsRSAVerify_PKCS1v15_rmf_get_size, (int *pBufferSize, Ipp8u *pKeysBuffer))
 IPPAPI(fips_test_status, fips_selftest_ippsRSASign_PKCS1v15_rmf, (Ipp8u *pBuffer, Ipp8u *pKeysBuffer))
 IPPAPI(fips_test_status, fips_selftest_ippsRSAVerify_PKCS1v15_rmf, (Ipp8u *pBuffer, Ipp8u *pKeysBuffer))
 

sources/ippcp/asm_ia32/pcpsha512l9ni.asm

@@ -76,7 +76,6 @@ IPPASM UpdateSHA512ni,PUBLIC
         ;; hash infrastructure (caller) sends the block size in bytes
         ;; the algorithm requires the number of 2^7=128 byte blocks
         shr             arg_num_blks, 7
-        or              arg_num_blks, arg_num_blks
         je              .done_hash
 
         mov      arg_hash, pDigest

sources/ippcp/asm_intel64/pcpsha1l9as.asm

@@ -35,7 +35,7 @@
 ;;
 ;; assignments
 ;;
-%xdefine hA  eax        ;; hash values into GPR registers
+%xdefine hA  eax        ;; hash values into general purpose registers (GPRs)
 %xdefine F  ebp
 %xdefine hB  ebx
 %xdefine hC  ecx

sources/ippcp/asm_intel64/pcpsha256l9as.asm

@@ -34,7 +34,7 @@
 ;;
 ;; assignments
 ;;
-%xdefine hA  eax  ;; hash values into GPR registers
+%xdefine hA  eax  ;; hash values into general purpose registers (GPRs)
 %xdefine hB  ebx
 %xdefine hC  ecx
 %xdefine hD  edx

sources/ippcp/asm_intel64/pcpsha512l9ni.asm

@@ -78,7 +78,6 @@ IPPASM UpdateSHA512ni,PUBLIC
         ;; hash infrastructure (caller) sends the block size in bytes
         ;; the algorithm requires the number of 2^7=128 byte blocks
         shr             arg_num_blks, 7
-        or              arg_num_blks, arg_num_blks
         je              .done_hash
 
 ;; ===========================================================

sources/ippcp/asm_intel64/pcpsm4l9_ni_as.asm

@@ -90,6 +90,9 @@ IPPASM cpSMS4_SetRoundKeys_ni,PUBLIC
     %assign i (i + 1)
     %endrep
 
+    ; zeroize
+    vpxor xmm0, xmm0
+
     REST_XMM_AVX
     REST_GPR
     ret
@@ -127,6 +130,9 @@ IPPASM cpSMS4_ECB_ni,PUBLIC
     vpshufb xmm0, [rel out_shufb]
     vmovdqu [pOut], xmm0
 
+    ; zeroize
+    vpxor xmm0, xmm0
+
     REST_XMM_AVX
     REST_GPR
     ret
@@ -165,6 +171,10 @@ IPPASM cpSMS4_ECB_ni_256,PUBLIC
     vpshufb ymm0, [rel out_shufb]
     vmovdqu [pOut], ymm0
 
+    ; zeroize
+    vpxor ymm0, ymm0
+    vpxor ymm1, ymm1
+
     REST_XMM_AVX
     REST_GPR
     ret

sources/ippcp/crypto_mb/src/rsa/internal_avx2/ifma_ams52x10_diagonal_mb4.c

@@ -64,108 +64,104 @@ static void ams52x10_square_diagonal_mb4(__m256i* res, const int64u* inpA_mb)
     // 1st triangle - sum the products, double and square
     r0 = zero_simd;
 
-    res[0]  = _mm256_madd52lo_epu64(r0, inpA[0], inpA[0]);
-    r1      = zero_simd;
-    r2      = zero_simd;
-    r3      = zero_simd;
-    r4      = zero_simd;
-    r5      = zero_simd;
-    r6      = zero_simd;
-    r7      = zero_simd;
-    r8      = zero_simd;
-    AL      = inpA[0];
-    r0      = _mm256_madd52lo_epu64(r0, AL, inpA[1]);      // Sum(1)
-    r1      = _mm256_madd52lo_epu64(r1, AL, inpA[2]);      // Sum(2)
-    r2      = _mm256_madd52lo_epu64(r2, AL, inpA[3]);      // Sum(3)
-    r3      = _mm256_madd52lo_epu64(r3, AL, inpA[4]);      // Sum(4)
-    r4      = _mm256_madd52lo_epu64(r4, AL, inpA[5]);      // Sum(5)
-    r5      = _mm256_madd52lo_epu64(r5, AL, inpA[6]);      // Sum(6)
-    r6      = _mm256_madd52lo_epu64(r6, AL, inpA[7]);      // Sum(7)
-    r7      = _mm256_madd52lo_epu64(r7, AL, inpA[8]);      // Sum(8)
-    r1      = _mm256_madd52hi_epu64(r1, AL, inpA[1]);      // Sum(1)
-    r2      = _mm256_madd52hi_epu64(r2, AL, inpA[2]);      // Sum(2)
-    r3      = _mm256_madd52hi_epu64(r3, AL, inpA[3]);      // Sum(3)
-    r4      = _mm256_madd52hi_epu64(r4, AL, inpA[4]);      // Sum(4)
-    r5      = _mm256_madd52hi_epu64(r5, AL, inpA[5]);      // Sum(5)
-    r6      = _mm256_madd52hi_epu64(r6, AL, inpA[6]);      // Sum(6)
-    r7      = _mm256_madd52hi_epu64(r7, AL, inpA[7]);      // Sum(7)
-    r8      = _mm256_madd52hi_epu64(r8, AL, inpA[8]);      // Sum(8)
-    AL      = inpA[1];
-    r2      = _mm256_madd52lo_epu64(r2, AL, inpA[2]);      // Sum(3)
-    r3      = _mm256_madd52lo_epu64(r3, AL, inpA[3]);      // Sum(4)
-    r4      = _mm256_madd52lo_epu64(r4, AL, inpA[4]);      // Sum(5)
-    r5      = _mm256_madd52lo_epu64(r5, AL, inpA[5]);      // Sum(6)
-    r6      = _mm256_madd52lo_epu64(r6, AL, inpA[6]);      // Sum(7)
-    r7      = _mm256_madd52lo_epu64(r7, AL, inpA[7]);      // Sum(8)
-    r3      = _mm256_madd52hi_epu64(r3, AL, inpA[2]);      // Sum(3)
-    r4      = _mm256_madd52hi_epu64(r4, AL, inpA[3]);      // Sum(4)
-    r5      = _mm256_madd52hi_epu64(r5, AL, inpA[4]);      // Sum(5)
-    r6      = _mm256_madd52hi_epu64(r6, AL, inpA[5]);      // Sum(6)
-    r7      = _mm256_madd52hi_epu64(r7, AL, inpA[6]);      // Sum(7)
-    r8      = _mm256_madd52hi_epu64(r8, AL, inpA[7]);      // Sum(8)
-    AL      = inpA[2];
-    r4      = _mm256_madd52lo_epu64(r4, AL, inpA[3]);      // Sum(5)
-    r5      = _mm256_madd52lo_epu64(r5, AL, inpA[4]);      // Sum(6)
-    r6      = _mm256_madd52lo_epu64(r6, AL, inpA[5]);      // Sum(7)
-    r7      = _mm256_madd52lo_epu64(r7, AL, inpA[6]);      // Sum(8)
-    r5      = _mm256_madd52hi_epu64(r5, AL, inpA[3]);      // Sum(5)
-    r6      = _mm256_madd52hi_epu64(r6, AL, inpA[4]);      // Sum(6)
-    r7      = _mm256_madd52hi_epu64(r7, AL, inpA[5]);      // Sum(7)
-    r8      = _mm256_madd52hi_epu64(r8, AL, inpA[6]);      // Sum(8)
-    AL      = inpA[3];
-    r6      = _mm256_madd52lo_epu64(r6, AL, inpA[4]);      // Sum(7)
-    r7      = _mm256_madd52lo_epu64(r7, AL, inpA[5]);      // Sum(8)
-    r7      = _mm256_madd52hi_epu64(r7, AL, inpA[4]);      // Sum(7)
-    r8      = _mm256_madd52hi_epu64(r8, AL, inpA[5]);      // Sum(8)
-    r0      = _mm256_add_epi64(r0, r0);                    // Double(1)
-    r0      = _mm256_madd52hi_epu64(r0, inpA[0], inpA[0]); // Add square(1)
-    res[1]  = r0;
-    r1      = _mm256_add_epi64(r1, r1);                    // Double(2)
-    r1      = _mm256_madd52lo_epu64(r1, inpA[1], inpA[1]); // Add square(2)
-    res[2]  = r1;
-    r2      = _mm256_add_epi64(r2, r2);                    // Double(3)
-    r2      = _mm256_madd52hi_epu64(r2, inpA[1], inpA[1]); // Add square(3)
-    res[3]  = r2;
-    r3      = _mm256_add_epi64(r3, r3);                    // Double(4)
-    r3      = _mm256_madd52lo_epu64(r3, inpA[2], inpA[2]); // Add square(4)
-    res[4]  = r3;
-    r4      = _mm256_add_epi64(r4, r4);                    // Double(5)
-    r4      = _mm256_madd52hi_epu64(r4, inpA[2], inpA[2]); // Add square(5)
-    res[5]  = r4;
-    r5      = _mm256_add_epi64(r5, r5);                    // Double(6)
-    r5      = _mm256_madd52lo_epu64(r5, inpA[3], inpA[3]); // Add square(6)
-    res[6]  = r5;
-    r6      = _mm256_add_epi64(r6, r6);                    // Double(7)
-    r6      = _mm256_madd52hi_epu64(r6, inpA[3], inpA[3]); // Add square(7)
-    res[7]  = r6;
-    r7      = _mm256_add_epi64(r7, r7);                    // Double(8)
-    r7      = _mm256_madd52lo_epu64(r7, inpA[4], inpA[4]); // Add square(8)
-    res[8]  = r7;
-    r0      = r8;
-    r1      = zero_simd;
-    AL      = inpA[0];
-    r0      = _mm256_madd52lo_epu64(r0, AL, inpA[9]); // Sum(9)
-    r1      = _mm256_madd52hi_epu64(r1, AL, inpA[9]); // Sum(9)
-    AL      = inpA[1];
-    r0      = _mm256_madd52lo_epu64(r0, AL, inpA[8]); // Sum(9)
-    r1      = _mm256_madd52hi_epu64(r1, AL, inpA[8]); // Sum(9)
-    AL      = inpA[2];
-    r0      = _mm256_madd52lo_epu64(r0, AL, inpA[7]); // Sum(9)
-    r1      = _mm256_madd52hi_epu64(r1, AL, inpA[7]); // Sum(9)
-    AL      = inpA[3];
-    r0      = _mm256_madd52lo_epu64(r0, AL, inpA[6]); // Sum(9)
-    r1      = _mm256_madd52hi_epu64(r1, AL, inpA[6]); // Sum(9)
-    AL      = inpA[4];
-    r0      = _mm256_madd52lo_epu64(r0, AL, inpA[5]); // Sum(9)
-    r1      = _mm256_madd52hi_epu64(r1, AL, inpA[5]); // Sum(9)
-    AL      = inpA[5];
-    AL      = inpA[6];
-    AL      = inpA[7];
-    r0      = _mm256_add_epi64(r0, r0);                    // Double(9)
-    r0      = _mm256_madd52hi_epu64(r0, inpA[4], inpA[4]); // Add square(9)
-    res[9]  = r0;
-    r0      = r1;
-    res[10] = r0;                                          // finish up 1st triangle
+    res[0] = _mm256_madd52lo_epu64(r0, inpA[0], inpA[0]);
+    r1     = zero_simd;
+    r2     = zero_simd;
+    r3     = zero_simd;
+    r4     = zero_simd;
+    r5     = zero_simd;
+    r6     = zero_simd;
+    r7     = zero_simd;
+    r8     = zero_simd;
+    AL     = inpA[0];
+    r0     = _mm256_madd52lo_epu64(r0, AL, inpA[1]);      // Sum(1)
+    r1     = _mm256_madd52lo_epu64(r1, AL, inpA[2]);      // Sum(2)
+    r2     = _mm256_madd52lo_epu64(r2, AL, inpA[3]);      // Sum(3)
+    r3     = _mm256_madd52lo_epu64(r3, AL, inpA[4]);      // Sum(4)
+    r4     = _mm256_madd52lo_epu64(r4, AL, inpA[5]);      // Sum(5)
+    r5     = _mm256_madd52lo_epu64(r5, AL, inpA[6]);      // Sum(6)
+    r6     = _mm256_madd52lo_epu64(r6, AL, inpA[7]);      // Sum(7)
+    r7     = _mm256_madd52lo_epu64(r7, AL, inpA[8]);      // Sum(8)
+    r1     = _mm256_madd52hi_epu64(r1, AL, inpA[1]);      // Sum(1)
+    r2     = _mm256_madd52hi_epu64(r2, AL, inpA[2]);      // Sum(2)
+    r3     = _mm256_madd52hi_epu64(r3, AL, inpA[3]);      // Sum(3)
+    r4     = _mm256_madd52hi_epu64(r4, AL, inpA[4]);      // Sum(4)
+    r5     = _mm256_madd52hi_epu64(r5, AL, inpA[5]);      // Sum(5)
+    r6     = _mm256_madd52hi_epu64(r6, AL, inpA[6]);      // Sum(6)
+    r7     = _mm256_madd52hi_epu64(r7, AL, inpA[7]);      // Sum(7)
+    r8     = _mm256_madd52hi_epu64(r8, AL, inpA[8]);      // Sum(8)
+    AL     = inpA[1];
+    r2     = _mm256_madd52lo_epu64(r2, AL, inpA[2]);      // Sum(3)
+    r3     = _mm256_madd52lo_epu64(r3, AL, inpA[3]);      // Sum(4)
+    r4     = _mm256_madd52lo_epu64(r4, AL, inpA[4]);      // Sum(5)
+    r5     = _mm256_madd52lo_epu64(r5, AL, inpA[5]);      // Sum(6)
+    r6     = _mm256_madd52lo_epu64(r6, AL, inpA[6]);      // Sum(7)
+    r7     = _mm256_madd52lo_epu64(r7, AL, inpA[7]);      // Sum(8)
+    r3     = _mm256_madd52hi_epu64(r3, AL, inpA[2]);      // Sum(3)
+    r4     = _mm256_madd52hi_epu64(r4, AL, inpA[3]);      // Sum(4)
+    r5     = _mm256_madd52hi_epu64(r5, AL, inpA[4]);      // Sum(5)
+    r6     = _mm256_madd52hi_epu64(r6, AL, inpA[5]);      // Sum(6)
+    r7     = _mm256_madd52hi_epu64(r7, AL, inpA[6]);      // Sum(7)
+    r8     = _mm256_madd52hi_epu64(r8, AL, inpA[7]);      // Sum(8)
+    AL     = inpA[2];
+    r4     = _mm256_madd52lo_epu64(r4, AL, inpA[3]);      // Sum(5)
+    r5     = _mm256_madd52lo_epu64(r5, AL, inpA[4]);      // Sum(6)
+    r6     = _mm256_madd52lo_epu64(r6, AL, inpA[5]);      // Sum(7)
+    r7     = _mm256_madd52lo_epu64(r7, AL, inpA[6]);      // Sum(8)
+    r5     = _mm256_madd52hi_epu64(r5, AL, inpA[3]);      // Sum(5)
+    r6     = _mm256_madd52hi_epu64(r6, AL, inpA[4]);      // Sum(6)
+    r7     = _mm256_madd52hi_epu64(r7, AL, inpA[5]);      // Sum(7)
+    r8     = _mm256_madd52hi_epu64(r8, AL, inpA[6]);      // Sum(8)
+    AL     = inpA[3];
+    r6     = _mm256_madd52lo_epu64(r6, AL, inpA[4]);      // Sum(7)
+    r7     = _mm256_madd52lo_epu64(r7, AL, inpA[5]);      // Sum(8)
+    r7     = _mm256_madd52hi_epu64(r7, AL, inpA[4]);      // Sum(7)
+    r8     = _mm256_madd52hi_epu64(r8, AL, inpA[5]);      // Sum(8)
+    r0     = _mm256_add_epi64(r0, r0);                    // Double(1)
+    r0     = _mm256_madd52hi_epu64(r0, inpA[0], inpA[0]); // Add square(1)
+    res[1] = r0;
+    r1     = _mm256_add_epi64(r1, r1);                    // Double(2)
+    r1     = _mm256_madd52lo_epu64(r1, inpA[1], inpA[1]); // Add square(2)
+    res[2] = r1;
+    r2     = _mm256_add_epi64(r2, r2);                    // Double(3)
+    r2     = _mm256_madd52hi_epu64(r2, inpA[1], inpA[1]); // Add square(3)
+    res[3] = r2;
+    r3     = _mm256_add_epi64(r3, r3);                    // Double(4)
+    r3     = _mm256_madd52lo_epu64(r3, inpA[2], inpA[2]); // Add square(4)
+    res[4] = r3;
+    r4     = _mm256_add_epi64(r4, r4);                    // Double(5)
+    r4     = _mm256_madd52hi_epu64(r4, inpA[2], inpA[2]); // Add square(5)
+    res[5] = r4;
+    r5     = _mm256_add_epi64(r5, r5);                    // Double(6)
+    r5     = _mm256_madd52lo_epu64(r5, inpA[3], inpA[3]); // Add square(6)
+    res[6] = r5;
+    r6     = _mm256_add_epi64(r6, r6);                    // Double(7)
+    r6     = _mm256_madd52hi_epu64(r6, inpA[3], inpA[3]); // Add square(7)
+    res[7] = r6;
+    r7     = _mm256_add_epi64(r7, r7);                    // Double(8)
+    r7     = _mm256_madd52lo_epu64(r7, inpA[4], inpA[4]); // Add square(8)
+    res[8] = r7;
+    r0     = r8;                                     // First, lo_ of 9th chunk is hi_ of 8th chunk
+    r1     = zero_simd;
+    AL     = inpA[0];
+    r0     = _mm256_madd52lo_epu64(r0, AL, inpA[9]); // Sum(9)
+    r1     = _mm256_madd52hi_epu64(r1, AL, inpA[9]); // Sum(9)
+    AL     = inpA[1];
+    r0     = _mm256_madd52lo_epu64(r0, AL, inpA[8]); // Sum(9)
+    r1     = _mm256_madd52hi_epu64(r1, AL, inpA[8]); // Sum(9)
+    AL     = inpA[2];
+    r0     = _mm256_madd52lo_epu64(r0, AL, inpA[7]); // Sum(9)
+    r1     = _mm256_madd52hi_epu64(r1, AL, inpA[7]); // Sum(9)
+    AL     = inpA[3];
+    r0     = _mm256_madd52lo_epu64(r0, AL, inpA[6]); // Sum(9)
+    r1     = _mm256_madd52hi_epu64(r1, AL, inpA[6]); // Sum(9)
+    AL     = inpA[4];
+    r0     = _mm256_madd52lo_epu64(r0, AL, inpA[5]); // Sum(9)
+    r1     = _mm256_madd52hi_epu64(r1, AL, inpA[5]); // Sum(9)
+    r0     = _mm256_add_epi64(r0, r0);               // Double(9)
+    r0     = _mm256_madd52hi_epu64(r0, inpA[4], inpA[4]); // Add square(9)
+    res[9] = r0;
+    r0     = r1;                                          // finish up 1st triangle
 
     ASM("jmp l0\nl0:\n");