From e6918187568dbd01842d8d1d2c808ce16a894239 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 21 Apr 2024 13:54:28 +0200 Subject: Adding upstream version 18.2.2. Signed-off-by: Daniel Baumann --- .../intel-ipsec-mb/avx512/sha256_x16_avx512.asm | 758 +++++++++++++++++++++ 1 file changed, 758 insertions(+) create mode 100644 src/spdk/intel-ipsec-mb/avx512/sha256_x16_avx512.asm (limited to 'src/spdk/intel-ipsec-mb/avx512/sha256_x16_avx512.asm') diff --git a/src/spdk/intel-ipsec-mb/avx512/sha256_x16_avx512.asm b/src/spdk/intel-ipsec-mb/avx512/sha256_x16_avx512.asm new file mode 100644 index 000000000..cdbb61ea3 --- /dev/null +++ b/src/spdk/intel-ipsec-mb/avx512/sha256_x16_avx512.asm @@ -0,0 +1,758 @@ +;; +;; Copyright (c) 2017-2018, Intel Corporation +;; +;; Redistribution and use in source and binary forms, with or without +;; modification, are permitted provided that the following conditions are met: +;; +;; * Redistributions of source code must retain the above copyright notice, +;; this list of conditions and the following disclaimer. +;; * Redistributions in binary form must reproduce the above copyright +;; notice, this list of conditions and the following disclaimer in the +;; documentation and/or other materials provided with the distribution. +;; * Neither the name of Intel Corporation nor the names of its contributors +;; may be used to endorse or promote products derived from this software +;; without specific prior written permission. +;; +;; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +;; AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +;; IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +;; DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE +;; FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +;; DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +;; SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +;; CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +;; OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +;; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +;; + +;; Stack must be aligned to 32 bytes before call +;; +;; Registers: RAX RBX RCX RDX RBP RSI RDI R8 R9 R10 R11 R12 R13 R14 R15 +;; ----------------------------------------------------------- +;; Windows clobbers: RAX RDX RSI RDI R9 R10 R11 R12 R13 R14 R15 +;; Windows preserves: RCX +;; ----------------------------------------------------------- +;; Linux clobbers: RAX RCX RDX RSI R9 R10 R11 R12 R13 R14 R15 +;; Linux preserves: RDI +;; ----------------------------------------------------------- +;; Clobbers ZMM0-31 + +%include "include/os.asm" +;%define DO_DBGPRINT +%include "include/dbgprint.asm" +%include "mb_mgr_datastruct.asm" +%include "include/transpose_avx512.asm" +%include "include/reg_sizes.asm" + +; re-use K256 from sha256_oct_avx2.asm +extern K256 + +;; code to compute x16 SHA256 using AVX512 + +%define APPEND(a,b) a %+ b + +; Define Stack Layout +START_FIELDS +;;; name size align +FIELD _DIGEST_SAVE, 8*64, 64 +FIELD _rsp, 8, 8 +%assign STACK_SPACE _FIELD_OFFSET + +%ifdef LINUX +; Linux register definitions + %define arg1 rdi + %define arg2 rsi + %define arg3 rcx + %define arg4 rdx +%else +; Windows definitions + %define arg1 rcx + %define arg2 rdx + %define arg3 rsi + %define arg4 rdi +%endif + +%define STATE arg1 +%define INP_SIZE arg2 +%define IDX arg3 +%define TBL arg4 + +%define A zmm0 +%define B zmm1 +%define C zmm2 +%define D zmm3 +%define E zmm4 +%define F zmm5 +%define G zmm6 +%define H zmm7 +%define T1 zmm8 +%define TMP0 zmm9 +%define TMP1 zmm10 +%define TMP2 zmm11 +%define TMP3 zmm12 +%define TMP4 zmm13 +%define TMP5 zmm14 +%define TMP6 zmm15 + +%define W0 zmm16 +%define W1 zmm17 +%define W2 zmm18 +%define W3 zmm19 +%define W4 zmm20 +%define W5 zmm21 +%define W6 zmm22 +%define W7 zmm23 +%define W8 zmm24 +%define W9 zmm25 +%define W10 zmm26 +%define W11 zmm27 +%define W12 zmm28 +%define W13 zmm29 +%define W14 zmm30 +%define W15 zmm31 + +%define inp0 r9 +%define inp1 r10 +%define inp2 r11 +%define inp3 r12 +%define inp4 r13 +%define inp5 r14 +%define inp6 r15 +%define inp7 rax + +%macro ROTATE_ARGS 0 +%xdefine TMP_ H +%xdefine H G +%xdefine G F +%xdefine F E +%xdefine E D +%xdefine D C +%xdefine C B +%xdefine B A +%xdefine A TMP_ +%endm + +;; CH(A, B, C) = (A&B) ^ (~A&C) +;; MAJ(E, F, G) = (E&F) ^ (E&G) ^ (F&G) +;; SIGMA0 = ROR_2 ^ ROR_13 ^ ROR_22 +;; SIGMA1 = ROR_6 ^ ROR_11 ^ ROR_25 +;; sigma0 = ROR_7 ^ ROR_18 ^ SHR_3 +;; sigma1 = ROR_17 ^ ROR_19 ^ SHR_10 + +; Main processing loop per round +%macro PROCESS_LOOP 2 +%define %%WT %1 +%define %%ROUND %2 + ;; T1 = H + SIGMA1(E) + CH(E, F, G) + Kt + Wt + ;; T2 = SIGMA0(A) + MAJ(A, B, C) + ;; H=G, G=F, F=E, E=D+T1, D=C, C=B, B=A, A=T1+T2 + + ;; H becomes T2, then add T1 for A + ;; D becomes D + T1 for E + + vpaddd T1, H, TMP3 ; T1 = H + Kt + vmovdqa32 TMP0, E + vprord TMP1, E, 6 ; ROR_6(E) + vprord TMP2, E, 11 ; ROR_11(E) + vprord TMP3, E, 25 ; ROR_25(E) + vpternlogd TMP0, F, G, 0xCA ; TMP0 = CH(E,F,G) + vpaddd T1, T1, %%WT ; T1 = T1 + Wt + vpternlogd TMP1, TMP2, TMP3, 0x96 ; TMP1 = SIGMA1(E) + vpaddd T1, T1, TMP0 ; T1 = T1 + CH(E,F,G) + vpaddd T1, T1, TMP1 ; T1 = T1 + SIGMA1(E) + vpaddd D, D, T1 ; D = D + T1 + + vprord H, A, 2 ; ROR_2(A) + vprord TMP2, A, 13 ; ROR_13(A) + vprord TMP3, A, 22 ; ROR_22(A) + vmovdqa32 TMP0, A + vpternlogd TMP0, B, C, 0xE8 ; TMP0 = MAJ(A,B,C) + vpternlogd H, TMP2, TMP3, 0x96 ; H(T2) = SIGMA0(A) + vpaddd H, H, TMP0 ; H(T2) = SIGMA0(A) + MAJ(A,B,C) + vpaddd H, H, T1 ; H(A) = H(T2) + T1 + + vmovdqa32 TMP3, [TBL + ((%%ROUND+1)*64)] ; Next Kt + + ;; Rotate the args A-H (rotation of names associated with regs) + ROTATE_ARGS +%endmacro + +; This is supposed to be SKL optimized assuming: +; vpternlog, vpaddd ports 5,8 +; vprord ports 1,8 +; However, vprord is only working on port 8 +; +; Main processing loop per round +; Get the msg schedule word 16 from the current, now unneccessary word +%macro PROCESS_LOOP_00_47 5 +%define %%WT %1 +%define %%ROUND %2 +%define %%WTp1 %3 +%define %%WTp9 %4 +%define %%WTp14 %5 + ;; T1 = H + SIGMA1(E) + CH(E, F, G) + Kt + Wt + ;; T2 = SIGMA0(A) + MAJ(A, B, C) + ;; H=G, G=F, F=E, E=D+T1, D=C, C=B, B=A, A=T1+T2 + + ;; H becomes T2, then add T1 for A + ;; D becomes D + T1 for E + + ;; For next value in msg schedule + ;; Wt+16 = sigma1(Wt+14) + Wt+9 + sigma0(Wt+1) + Wt + + vmovdqa32 TMP0, E + vprord TMP1, E, 6 ; ROR_6(E) + vprord TMP2, E, 11 ; ROR_11(E) + vprord TMP3, E, 25 ; ROR_25(E) + vpternlogd TMP0, F, G, 0xCA ; TMP0 = CH(E,F,G) + vpaddd T1, H, %%WT ; T1 = H + Wt + vpternlogd TMP1, TMP2, TMP3, 0x96 ; TMP1 = SIGMA1(E) + vpaddd T1, T1, TMP6 ; T1 = T1 + Kt + vprord H, A, 2 ; ROR_2(A) + vpaddd T1, T1, TMP0 ; T1 = T1 + CH(E,F,G) + vprord TMP2, A, 13 ; ROR_13(A) + vmovdqa32 TMP0, A + vprord TMP3, A, 22 ; ROR_22(A) + vpaddd T1, T1, TMP1 ; T1 = T1 + SIGMA1(E) + vpternlogd TMP0, B, C, 0xE8 ; TMP0 = MAJ(A,B,C) + vpaddd D, D, T1 ; D = D + T1 + vpternlogd H, TMP2, TMP3, 0x96 ; H(T2) = SIGMA0(A) + vprord TMP4, %%WTp14, 17 ; ROR_17(Wt-2) + vpaddd H, H, TMP0 ; H(T2) = SIGMA0(A) + MAJ(A,B,C) + vprord TMP5, %%WTp14, 19 ; ROR_19(Wt-2) + vpsrld TMP6, %%WTp14, 10 ; SHR_10(Wt-2) + vpaddd H, H, T1 ; H(A) = H(T2) + T1 + vpternlogd TMP4, TMP5, TMP6, 0x96 ; TMP4 = sigma1(Wt-2) + vpaddd %%WT, %%WT, TMP4 ; Wt = Wt-16 + sigma1(Wt-2) + vprord TMP4, %%WTp1, 7 ; ROR_7(Wt-15) + vprord TMP5, %%WTp1, 18 ; ROR_18(Wt-15) + vpaddd %%WT, %%WT, %%WTp9 ; Wt = Wt-16 + sigma1(Wt-2) + Wt-7 + vpsrld TMP6, %%WTp1, 3 ; SHR_3(Wt-15) + vpternlogd TMP4, TMP5, TMP6, 0x96 ; TMP4 = sigma0(Wt-15) + vpaddd %%WT, %%WT, TMP4 ; Wt = Wt-16 + sigma1(Wt-2) + + ; Wt-7 + sigma0(Wt-15) + + + vmovdqa32 TMP6, [TBL + ((%%ROUND+1)*64)] ; Next Kt + + ;; Rotate the args A-H (rotation of names associated with regs) + ROTATE_ARGS +%endmacro + +%macro MSG_SCHED_ROUND_16_63 4 +%define %%WT %1 +%define %%WTp1 %2 +%define %%WTp9 %3 +%define %%WTp14 %4 + vprord TMP4, %%WTp14, 17 ; ROR_17(Wt-2) + vprord TMP5, %%WTp14, 19 ; ROR_19(Wt-2) + vpsrld TMP6, %%WTp14, 10 ; SHR_10(Wt-2) + vpternlogd TMP4, TMP5, TMP6, 0x96 ; TMP4 = sigma1(Wt-2) + + vpaddd %%WT, %%WT, TMP4 ; Wt = Wt-16 + sigma1(Wt-2) + vpaddd %%WT, %%WT, %%WTp9 ; Wt = Wt-16 + sigma1(Wt-2) + Wt-7 + + vprord TMP4, %%WTp1, 7 ; ROR_7(Wt-15) + vprord TMP5, %%WTp1, 18 ; ROR_18(Wt-15) + vpsrld TMP6, %%WTp1, 3 ; SHR_3(Wt-15) + vpternlogd TMP4, TMP5, TMP6, 0x96 ; TMP4 = sigma0(Wt-15) + + vpaddd %%WT, %%WT, TMP4 ; Wt = Wt-16 + sigma1(Wt-2) + + ; Wt-7 + sigma0(Wt-15) + +%endmacro + +; Note this is reading in two blocks of data from each lane, +; in preparation for the upcoming needed transpose to build msg schedule. +; Each register will contain 32 bytes from one lane plus 32 bytes +; from another lane. +; The first 8 registers will contain the first 32 bytes of all lanes, +; where register X (0 <= X <= 7) will contain bytes 0-31 from lane X in the first half +; and 0-31 bytes from lane X+8 in the second half. +; The last 8 registers will contain the last 32 bytes of all lanes, +; where register Y (8 <= Y <= 15) wil contain bytes 32-63 from lane Y-8 in the first half +; and 32-63 bytes from lane Y in the second half. +; This method helps reducing the number of shuffles required to transpose the data. +%macro MSG_SCHED_ROUND_00_15 6 +%define %%Wt %1 ; [out] zmm register to load the next block +%define %%LANE_IDX %2 ; [in] lane index (0-15) +%define %%BASE_PTR %3 ; [in] base address of the input data +%define %%OFFSET_PTR %4 ; [in] offset to get next block of data from the lane +%define %%TMP1 %5 ; [clobbered] temporary gp register +%define %%TMP2 %6 ; [clobbered] temporary gp register +%if (%%LANE_IDX < 8) + mov %%TMP1, [%%BASE_PTR + %%LANE_IDX*PTR_SZ] + mov %%TMP2, [%%BASE_PTR + (%%LANE_IDX+8)*PTR_SZ] + vmovups YWORD(%%Wt), [%%TMP1+%%OFFSET_PTR] + vinserti64x4 %%Wt, %%Wt, [%%TMP2+%%OFFSET_PTR], 0x01 +%else + mov %%TMP1, [%%BASE_PTR + (%%LANE_IDX-8)*PTR_SZ] + mov %%TMP2, [%%BASE_PTR + %%LANE_IDX*PTR_SZ] + vmovups YWORD(%%Wt), [%%TMP1+%%OFFSET_PTR+32] + vinserti64x4 %%Wt, %%Wt, [%%TMP2+%%OFFSET_PTR+32], 0x01 +%endif +%endmacro + + section .data +default rel +align 64 +TABLE: + dq 0x428a2f98428a2f98, 0x428a2f98428a2f98 + dq 0x428a2f98428a2f98, 0x428a2f98428a2f98 + dq 0x428a2f98428a2f98, 0x428a2f98428a2f98 + dq 0x428a2f98428a2f98, 0x428a2f98428a2f98 + dq 0x7137449171374491, 0x7137449171374491 + dq 0x7137449171374491, 0x7137449171374491 + dq 0x7137449171374491, 0x7137449171374491 + dq 0x7137449171374491, 0x7137449171374491 + dq 0xb5c0fbcfb5c0fbcf, 0xb5c0fbcfb5c0fbcf + dq 0xb5c0fbcfb5c0fbcf, 0xb5c0fbcfb5c0fbcf + dq 0xb5c0fbcfb5c0fbcf, 0xb5c0fbcfb5c0fbcf + dq 0xb5c0fbcfb5c0fbcf, 0xb5c0fbcfb5c0fbcf + dq 0xe9b5dba5e9b5dba5, 0xe9b5dba5e9b5dba5 + dq 0xe9b5dba5e9b5dba5, 0xe9b5dba5e9b5dba5 + dq 0xe9b5dba5e9b5dba5, 0xe9b5dba5e9b5dba5 + dq 0xe9b5dba5e9b5dba5, 0xe9b5dba5e9b5dba5 + dq 0x3956c25b3956c25b, 0x3956c25b3956c25b + dq 0x3956c25b3956c25b, 0x3956c25b3956c25b + dq 0x3956c25b3956c25b, 0x3956c25b3956c25b + dq 0x3956c25b3956c25b, 0x3956c25b3956c25b + dq 0x59f111f159f111f1, 0x59f111f159f111f1 + dq 0x59f111f159f111f1, 0x59f111f159f111f1 + dq 0x59f111f159f111f1, 0x59f111f159f111f1 + dq 0x59f111f159f111f1, 0x59f111f159f111f1 + dq 0x923f82a4923f82a4, 0x923f82a4923f82a4 + dq 0x923f82a4923f82a4, 0x923f82a4923f82a4 + dq 0x923f82a4923f82a4, 0x923f82a4923f82a4 + dq 0x923f82a4923f82a4, 0x923f82a4923f82a4 + dq 0xab1c5ed5ab1c5ed5, 0xab1c5ed5ab1c5ed5 + dq 0xab1c5ed5ab1c5ed5, 0xab1c5ed5ab1c5ed5 + dq 0xab1c5ed5ab1c5ed5, 0xab1c5ed5ab1c5ed5 + dq 0xab1c5ed5ab1c5ed5, 0xab1c5ed5ab1c5ed5 + dq 0xd807aa98d807aa98, 0xd807aa98d807aa98 + dq 0xd807aa98d807aa98, 0xd807aa98d807aa98 + dq 0xd807aa98d807aa98, 0xd807aa98d807aa98 + dq 0xd807aa98d807aa98, 0xd807aa98d807aa98 + dq 0x12835b0112835b01, 0x12835b0112835b01 + dq 0x12835b0112835b01, 0x12835b0112835b01 + dq 0x12835b0112835b01, 0x12835b0112835b01 + dq 0x12835b0112835b01, 0x12835b0112835b01 + dq 0x243185be243185be, 0x243185be243185be + dq 0x243185be243185be, 0x243185be243185be + dq 0x243185be243185be, 0x243185be243185be + dq 0x243185be243185be, 0x243185be243185be + dq 0x550c7dc3550c7dc3, 0x550c7dc3550c7dc3 + dq 0x550c7dc3550c7dc3, 0x550c7dc3550c7dc3 + dq 0x550c7dc3550c7dc3, 0x550c7dc3550c7dc3 + dq 0x550c7dc3550c7dc3, 0x550c7dc3550c7dc3 + dq 0x72be5d7472be5d74, 0x72be5d7472be5d74 + dq 0x72be5d7472be5d74, 0x72be5d7472be5d74 + dq 0x72be5d7472be5d74, 0x72be5d7472be5d74 + dq 0x72be5d7472be5d74, 0x72be5d7472be5d74 + dq 0x80deb1fe80deb1fe, 0x80deb1fe80deb1fe + dq 0x80deb1fe80deb1fe, 0x80deb1fe80deb1fe + dq 0x80deb1fe80deb1fe, 0x80deb1fe80deb1fe + dq 0x80deb1fe80deb1fe, 0x80deb1fe80deb1fe + dq 0x9bdc06a79bdc06a7, 0x9bdc06a79bdc06a7 + dq 0x9bdc06a79bdc06a7, 0x9bdc06a79bdc06a7 + dq 0x9bdc06a79bdc06a7, 0x9bdc06a79bdc06a7 + dq 0x9bdc06a79bdc06a7, 0x9bdc06a79bdc06a7 + dq 0xc19bf174c19bf174, 0xc19bf174c19bf174 + dq 0xc19bf174c19bf174, 0xc19bf174c19bf174 + dq 0xc19bf174c19bf174, 0xc19bf174c19bf174 + dq 0xc19bf174c19bf174, 0xc19bf174c19bf174 + dq 0xe49b69c1e49b69c1, 0xe49b69c1e49b69c1 + dq 0xe49b69c1e49b69c1, 0xe49b69c1e49b69c1 + dq 0xe49b69c1e49b69c1, 0xe49b69c1e49b69c1 + dq 0xe49b69c1e49b69c1, 0xe49b69c1e49b69c1 + dq 0xefbe4786efbe4786, 0xefbe4786efbe4786 + dq 0xefbe4786efbe4786, 0xefbe4786efbe4786 + dq 0xefbe4786efbe4786, 0xefbe4786efbe4786 + dq 0xefbe4786efbe4786, 0xefbe4786efbe4786 + dq 0x0fc19dc60fc19dc6, 0x0fc19dc60fc19dc6 + dq 0x0fc19dc60fc19dc6, 0x0fc19dc60fc19dc6 + dq 0x0fc19dc60fc19dc6, 0x0fc19dc60fc19dc6 + dq 0x0fc19dc60fc19dc6, 0x0fc19dc60fc19dc6 + dq 0x240ca1cc240ca1cc, 0x240ca1cc240ca1cc + dq 0x240ca1cc240ca1cc, 0x240ca1cc240ca1cc + dq 0x240ca1cc240ca1cc, 0x240ca1cc240ca1cc + dq 0x240ca1cc240ca1cc, 0x240ca1cc240ca1cc + dq 0x2de92c6f2de92c6f, 0x2de92c6f2de92c6f + dq 0x2de92c6f2de92c6f, 0x2de92c6f2de92c6f + dq 0x2de92c6f2de92c6f, 0x2de92c6f2de92c6f + dq 0x2de92c6f2de92c6f, 0x2de92c6f2de92c6f + dq 0x4a7484aa4a7484aa, 0x4a7484aa4a7484aa + dq 0x4a7484aa4a7484aa, 0x4a7484aa4a7484aa + dq 0x4a7484aa4a7484aa, 0x4a7484aa4a7484aa + dq 0x4a7484aa4a7484aa, 0x4a7484aa4a7484aa + dq 0x5cb0a9dc5cb0a9dc, 0x5cb0a9dc5cb0a9dc + dq 0x5cb0a9dc5cb0a9dc, 0x5cb0a9dc5cb0a9dc + dq 0x5cb0a9dc5cb0a9dc, 0x5cb0a9dc5cb0a9dc + dq 0x5cb0a9dc5cb0a9dc, 0x5cb0a9dc5cb0a9dc + dq 0x76f988da76f988da, 0x76f988da76f988da + dq 0x76f988da76f988da, 0x76f988da76f988da + dq 0x76f988da76f988da, 0x76f988da76f988da + dq 0x76f988da76f988da, 0x76f988da76f988da + dq 0x983e5152983e5152, 0x983e5152983e5152 + dq 0x983e5152983e5152, 0x983e5152983e5152 + dq 0x983e5152983e5152, 0x983e5152983e5152 + dq 0x983e5152983e5152, 0x983e5152983e5152 + dq 0xa831c66da831c66d, 0xa831c66da831c66d + dq 0xa831c66da831c66d, 0xa831c66da831c66d + dq 0xa831c66da831c66d, 0xa831c66da831c66d + dq 0xa831c66da831c66d, 0xa831c66da831c66d + dq 0xb00327c8b00327c8, 0xb00327c8b00327c8 + dq 0xb00327c8b00327c8, 0xb00327c8b00327c8 + dq 0xb00327c8b00327c8, 0xb00327c8b00327c8 + dq 0xb00327c8b00327c8, 0xb00327c8b00327c8 + dq 0xbf597fc7bf597fc7, 0xbf597fc7bf597fc7 + dq 0xbf597fc7bf597fc7, 0xbf597fc7bf597fc7 + dq 0xbf597fc7bf597fc7, 0xbf597fc7bf597fc7 + dq 0xbf597fc7bf597fc7, 0xbf597fc7bf597fc7 + dq 0xc6e00bf3c6e00bf3, 0xc6e00bf3c6e00bf3 + dq 0xc6e00bf3c6e00bf3, 0xc6e00bf3c6e00bf3 + dq 0xc6e00bf3c6e00bf3, 0xc6e00bf3c6e00bf3 + dq 0xc6e00bf3c6e00bf3, 0xc6e00bf3c6e00bf3 + dq 0xd5a79147d5a79147, 0xd5a79147d5a79147 + dq 0xd5a79147d5a79147, 0xd5a79147d5a79147 + dq 0xd5a79147d5a79147, 0xd5a79147d5a79147 + dq 0xd5a79147d5a79147, 0xd5a79147d5a79147 + dq 0x06ca635106ca6351, 0x06ca635106ca6351 + dq 0x06ca635106ca6351, 0x06ca635106ca6351 + dq 0x06ca635106ca6351, 0x06ca635106ca6351 + dq 0x06ca635106ca6351, 0x06ca635106ca6351 + dq 0x1429296714292967, 0x1429296714292967 + dq 0x1429296714292967, 0x1429296714292967 + dq 0x1429296714292967, 0x1429296714292967 + dq 0x1429296714292967, 0x1429296714292967 + dq 0x27b70a8527b70a85, 0x27b70a8527b70a85 + dq 0x27b70a8527b70a85, 0x27b70a8527b70a85 + dq 0x27b70a8527b70a85, 0x27b70a8527b70a85 + dq 0x27b70a8527b70a85, 0x27b70a8527b70a85 + dq 0x2e1b21382e1b2138, 0x2e1b21382e1b2138 + dq 0x2e1b21382e1b2138, 0x2e1b21382e1b2138 + dq 0x2e1b21382e1b2138, 0x2e1b21382e1b2138 + dq 0x2e1b21382e1b2138, 0x2e1b21382e1b2138 + dq 0x4d2c6dfc4d2c6dfc, 0x4d2c6dfc4d2c6dfc + dq 0x4d2c6dfc4d2c6dfc, 0x4d2c6dfc4d2c6dfc + dq 0x4d2c6dfc4d2c6dfc, 0x4d2c6dfc4d2c6dfc + dq 0x4d2c6dfc4d2c6dfc, 0x4d2c6dfc4d2c6dfc + dq 0x53380d1353380d13, 0x53380d1353380d13 + dq 0x53380d1353380d13, 0x53380d1353380d13 + dq 0x53380d1353380d13, 0x53380d1353380d13 + dq 0x53380d1353380d13, 0x53380d1353380d13 + dq 0x650a7354650a7354, 0x650a7354650a7354 + dq 0x650a7354650a7354, 0x650a7354650a7354 + dq 0x650a7354650a7354, 0x650a7354650a7354 + dq 0x650a7354650a7354, 0x650a7354650a7354 + dq 0x766a0abb766a0abb, 0x766a0abb766a0abb + dq 0x766a0abb766a0abb, 0x766a0abb766a0abb + dq 0x766a0abb766a0abb, 0x766a0abb766a0abb + dq 0x766a0abb766a0abb, 0x766a0abb766a0abb + dq 0x81c2c92e81c2c92e, 0x81c2c92e81c2c92e + dq 0x81c2c92e81c2c92e, 0x81c2c92e81c2c92e + dq 0x81c2c92e81c2c92e, 0x81c2c92e81c2c92e + dq 0x81c2c92e81c2c92e, 0x81c2c92e81c2c92e + dq 0x92722c8592722c85, 0x92722c8592722c85 + dq 0x92722c8592722c85, 0x92722c8592722c85 + dq 0x92722c8592722c85, 0x92722c8592722c85 + dq 0x92722c8592722c85, 0x92722c8592722c85 + dq 0xa2bfe8a1a2bfe8a1, 0xa2bfe8a1a2bfe8a1 + dq 0xa2bfe8a1a2bfe8a1, 0xa2bfe8a1a2bfe8a1 + dq 0xa2bfe8a1a2bfe8a1, 0xa2bfe8a1a2bfe8a1 + dq 0xa2bfe8a1a2bfe8a1, 0xa2bfe8a1a2bfe8a1 + dq 0xa81a664ba81a664b, 0xa81a664ba81a664b + dq 0xa81a664ba81a664b, 0xa81a664ba81a664b + dq 0xa81a664ba81a664b, 0xa81a664ba81a664b + dq 0xa81a664ba81a664b, 0xa81a664ba81a664b + dq 0xc24b8b70c24b8b70, 0xc24b8b70c24b8b70 + dq 0xc24b8b70c24b8b70, 0xc24b8b70c24b8b70 + dq 0xc24b8b70c24b8b70, 0xc24b8b70c24b8b70 + dq 0xc24b8b70c24b8b70, 0xc24b8b70c24b8b70 + dq 0xc76c51a3c76c51a3, 0xc76c51a3c76c51a3 + dq 0xc76c51a3c76c51a3, 0xc76c51a3c76c51a3 + dq 0xc76c51a3c76c51a3, 0xc76c51a3c76c51a3 + dq 0xc76c51a3c76c51a3, 0xc76c51a3c76c51a3 + dq 0xd192e819d192e819, 0xd192e819d192e819 + dq 0xd192e819d192e819, 0xd192e819d192e819 + dq 0xd192e819d192e819, 0xd192e819d192e819 + dq 0xd192e819d192e819, 0xd192e819d192e819 + dq 0xd6990624d6990624, 0xd6990624d6990624 + dq 0xd6990624d6990624, 0xd6990624d6990624 + dq 0xd6990624d6990624, 0xd6990624d6990624 + dq 0xd6990624d6990624, 0xd6990624d6990624 + dq 0xf40e3585f40e3585, 0xf40e3585f40e3585 + dq 0xf40e3585f40e3585, 0xf40e3585f40e3585 + dq 0xf40e3585f40e3585, 0xf40e3585f40e3585 + dq 0xf40e3585f40e3585, 0xf40e3585f40e3585 + dq 0x106aa070106aa070, 0x106aa070106aa070 + dq 0x106aa070106aa070, 0x106aa070106aa070 + dq 0x106aa070106aa070, 0x106aa070106aa070 + dq 0x106aa070106aa070, 0x106aa070106aa070 + dq 0x19a4c11619a4c116, 0x19a4c11619a4c116 + dq 0x19a4c11619a4c116, 0x19a4c11619a4c116 + dq 0x19a4c11619a4c116, 0x19a4c11619a4c116 + dq 0x19a4c11619a4c116, 0x19a4c11619a4c116 + dq 0x1e376c081e376c08, 0x1e376c081e376c08 + dq 0x1e376c081e376c08, 0x1e376c081e376c08 + dq 0x1e376c081e376c08, 0x1e376c081e376c08 + dq 0x1e376c081e376c08, 0x1e376c081e376c08 + dq 0x2748774c2748774c, 0x2748774c2748774c + dq 0x2748774c2748774c, 0x2748774c2748774c + dq 0x2748774c2748774c, 0x2748774c2748774c + dq 0x2748774c2748774c, 0x2748774c2748774c + dq 0x34b0bcb534b0bcb5, 0x34b0bcb534b0bcb5 + dq 0x34b0bcb534b0bcb5, 0x34b0bcb534b0bcb5 + dq 0x34b0bcb534b0bcb5, 0x34b0bcb534b0bcb5 + dq 0x34b0bcb534b0bcb5, 0x34b0bcb534b0bcb5 + dq 0x391c0cb3391c0cb3, 0x391c0cb3391c0cb3 + dq 0x391c0cb3391c0cb3, 0x391c0cb3391c0cb3 + dq 0x391c0cb3391c0cb3, 0x391c0cb3391c0cb3 + dq 0x391c0cb3391c0cb3, 0x391c0cb3391c0cb3 + dq 0x4ed8aa4a4ed8aa4a, 0x4ed8aa4a4ed8aa4a + dq 0x4ed8aa4a4ed8aa4a, 0x4ed8aa4a4ed8aa4a + dq 0x4ed8aa4a4ed8aa4a, 0x4ed8aa4a4ed8aa4a + dq 0x4ed8aa4a4ed8aa4a, 0x4ed8aa4a4ed8aa4a + dq 0x5b9cca4f5b9cca4f, 0x5b9cca4f5b9cca4f + dq 0x5b9cca4f5b9cca4f, 0x5b9cca4f5b9cca4f + dq 0x5b9cca4f5b9cca4f, 0x5b9cca4f5b9cca4f + dq 0x5b9cca4f5b9cca4f, 0x5b9cca4f5b9cca4f + dq 0x682e6ff3682e6ff3, 0x682e6ff3682e6ff3 + dq 0x682e6ff3682e6ff3, 0x682e6ff3682e6ff3 + dq 0x682e6ff3682e6ff3, 0x682e6ff3682e6ff3 + dq 0x682e6ff3682e6ff3, 0x682e6ff3682e6ff3 + dq 0x748f82ee748f82ee, 0x748f82ee748f82ee + dq 0x748f82ee748f82ee, 0x748f82ee748f82ee + dq 0x748f82ee748f82ee, 0x748f82ee748f82ee + dq 0x748f82ee748f82ee, 0x748f82ee748f82ee + dq 0x78a5636f78a5636f, 0x78a5636f78a5636f + dq 0x78a5636f78a5636f, 0x78a5636f78a5636f + dq 0x78a5636f78a5636f, 0x78a5636f78a5636f + dq 0x78a5636f78a5636f, 0x78a5636f78a5636f + dq 0x84c8781484c87814, 0x84c8781484c87814 + dq 0x84c8781484c87814, 0x84c8781484c87814 + dq 0x84c8781484c87814, 0x84c8781484c87814 + dq 0x84c8781484c87814, 0x84c8781484c87814 + dq 0x8cc702088cc70208, 0x8cc702088cc70208 + dq 0x8cc702088cc70208, 0x8cc702088cc70208 + dq 0x8cc702088cc70208, 0x8cc702088cc70208 + dq 0x8cc702088cc70208, 0x8cc702088cc70208 + dq 0x90befffa90befffa, 0x90befffa90befffa + dq 0x90befffa90befffa, 0x90befffa90befffa + dq 0x90befffa90befffa, 0x90befffa90befffa + dq 0x90befffa90befffa, 0x90befffa90befffa + dq 0xa4506ceba4506ceb, 0xa4506ceba4506ceb + dq 0xa4506ceba4506ceb, 0xa4506ceba4506ceb + dq 0xa4506ceba4506ceb, 0xa4506ceba4506ceb + dq 0xa4506ceba4506ceb, 0xa4506ceba4506ceb + dq 0xbef9a3f7bef9a3f7, 0xbef9a3f7bef9a3f7 + dq 0xbef9a3f7bef9a3f7, 0xbef9a3f7bef9a3f7 + dq 0xbef9a3f7bef9a3f7, 0xbef9a3f7bef9a3f7 + dq 0xbef9a3f7bef9a3f7, 0xbef9a3f7bef9a3f7 + dq 0xc67178f2c67178f2, 0xc67178f2c67178f2 + dq 0xc67178f2c67178f2, 0xc67178f2c67178f2 + dq 0xc67178f2c67178f2, 0xc67178f2c67178f2 + dq 0xc67178f2c67178f2, 0xc67178f2c67178f2 + +PSHUFFLE_BYTE_FLIP_MASK: + ;ddq 0x0c0d0e0f08090a0b0405060700010203 + dq 0x0405060700010203, 0x0c0d0e0f08090a0b + ;ddq 0x0c0d0e0f08090a0b0405060700010203 + dq 0x0405060700010203, 0x0c0d0e0f08090a0b + ;ddq 0x0c0d0e0f08090a0b0405060700010203 + dq 0x0405060700010203, 0x0c0d0e0f08090a0b + ;ddq 0x0c0d0e0f08090a0b0405060700010203 + dq 0x0405060700010203, 0x0c0d0e0f08090a0b + +section .text + +;; void sha256_x16_avx512(void **input_data, UINT128 *digest[16], UINT64 size) +;; arg 1 : pointer to SHA256 args structure +;; arg 2 : size (in blocks) ;; assumed to be >= 1 +;; arg 1 : rcx : pointer to array of pointers to input data +;; arg 2 : rdx : pointer to array of pointers to digest +;; arg 3 : r8 : size of input in bytes +MKGLOBAL(sha256_x16_avx512,function,internal) +align 64 +sha256_x16_avx512: + mov rax, rsp + sub rsp, STACK_SPACE + and rsp, ~63 ; align stack to multiple of 64 + mov [rsp + _rsp], rax + + ;; Initialize digests + vmovdqu32 A, [STATE + 0*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 B, [STATE + 1*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 C, [STATE + 2*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 D, [STATE + 3*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 E, [STATE + 4*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 F, [STATE + 5*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 G, [STATE + 6*SHA256_DIGEST_ROW_SIZE] + vmovdqu32 H, [STATE + 7*SHA256_DIGEST_ROW_SIZE] + + lea TBL, [rel TABLE] + + ; Do we need to transpose digests??? + ; SHA1 does not, but SHA256 has been + + xor IDX, IDX + + ;; Load first blocks of data into ZMM registers before + ;; performing a 16x16 32-bit transpose. + ;; To speed up the transpose, data is loaded in chunks of 32 bytes, + ;; interleaving data between lane X and lane X+8. + ;; This way, final shuffles between top half and bottom half + ;; of the matrix are avoided. + mov inp0, [STATE + _data_ptr_sha256 + 0*PTR_SZ] + mov inp1, [STATE + _data_ptr_sha256 + 1*PTR_SZ] + mov inp2, [STATE + _data_ptr_sha256 + 2*PTR_SZ] + mov inp3, [STATE + _data_ptr_sha256 + 3*PTR_SZ] + mov inp4, [STATE + _data_ptr_sha256 + 4*PTR_SZ] + mov inp5, [STATE + _data_ptr_sha256 + 5*PTR_SZ] + mov inp6, [STATE + _data_ptr_sha256 + 6*PTR_SZ] + mov inp7, [STATE + _data_ptr_sha256 + 7*PTR_SZ] + + TRANSPOSE16_U32_LOAD_FIRST8 W0, W1, W2, W3, W4, W5, W6, W7, \ + W8, W9, W10, W11, W12, W13, W14, W15, \ + inp0, inp1, inp2, inp3, inp4, inp5, \ + inp6, inp7, IDX + + mov inp0, [STATE + _data_ptr_sha256 + 8*PTR_SZ] + mov inp1, [STATE + _data_ptr_sha256 + 9*PTR_SZ] + mov inp2, [STATE + _data_ptr_sha256 +10*PTR_SZ] + mov inp3, [STATE + _data_ptr_sha256 +11*PTR_SZ] + mov inp4, [STATE + _data_ptr_sha256 +12*PTR_SZ] + mov inp5, [STATE + _data_ptr_sha256 +13*PTR_SZ] + mov inp6, [STATE + _data_ptr_sha256 +14*PTR_SZ] + mov inp7, [STATE + _data_ptr_sha256 +15*PTR_SZ] + + TRANSPOSE16_U32_LOAD_LAST8 W0, W1, W2, W3, W4, W5, W6, W7, \ + W8, W9, W10, W11, W12, W13, W14, W15, \ + inp0, inp1, inp2, inp3, inp4, inp5, \ + inp6, inp7, IDX + + align 32 +lloop: + vmovdqa32 TMP2, [rel PSHUFFLE_BYTE_FLIP_MASK] + + vmovdqa32 TMP3, [TBL] ; First K + + ; Save digests for later addition + vmovdqa32 [rsp + _DIGEST_SAVE + 64*0], A + vmovdqa32 [rsp + _DIGEST_SAVE + 64*1], B + vmovdqa32 [rsp + _DIGEST_SAVE + 64*2], C + vmovdqa32 [rsp + _DIGEST_SAVE + 64*3], D + vmovdqa32 [rsp + _DIGEST_SAVE + 64*4], E + vmovdqa32 [rsp + _DIGEST_SAVE + 64*5], F + vmovdqa32 [rsp + _DIGEST_SAVE + 64*6], G + vmovdqa32 [rsp + _DIGEST_SAVE + 64*7], H + + add IDX, 64 + + TRANSPOSE16_U32 W0, W1, W2, W3, W4, W5, W6, W7, W8, W9, W10, W11, W12, W13, W14, W15, TMP0, TMP1, TMP4, TMP5 + +%assign I 0 +%rep 16 + vpshufb APPEND(W,I), APPEND(W,I), TMP2 +%assign I (I+1) +%endrep + + ; MSG Schedule for W0-W15 is now complete in registers + ; Process first 48 rounds + ; Calculate next Wt+16 after processing is complete and Wt is unneeded + + ; PROCESS_LOOP_00_47 APPEND(W,J), I, APPEND(W,K), APPEND(W,L), APPEND(W,M) +%assign I 0 +%assign J 0 +%assign K 1 +%assign L 9 +%assign M 14 +%rep 48 + PROCESS_LOOP APPEND(W,J), I + MSG_SCHED_ROUND_16_63 APPEND(W,J), APPEND(W,K), APPEND(W,L), APPEND(W,M) +%assign I (I+1) +%assign J ((J+1)% 16) +%assign K ((K+1)% 16) +%assign L ((L+1)% 16) +%assign M ((M+1)% 16) +%endrep + + ; Check is this is the last block + sub INP_SIZE, 1 + je lastLoop + + ; Process last 16 rounds + ; Read in next block msg data for use in first 16 words of msg sched +%assign I 48 +%assign J 0 +%rep 16 + PROCESS_LOOP APPEND(W,J), I + MSG_SCHED_ROUND_00_15 APPEND(W,J), J, STATE + _data_ptr_sha256, IDX, inp0, inp1 +%assign I (I+1) +%assign J (J+1) +%endrep + + ; Add old digest + vpaddd A, A, [rsp + _DIGEST_SAVE + 64*0] + vpaddd B, B, [rsp + _DIGEST_SAVE + 64*1] + vpaddd C, C, [rsp + _DIGEST_SAVE + 64*2] + vpaddd D, D, [rsp + _DIGEST_SAVE + 64*3] + vpaddd E, E, [rsp + _DIGEST_SAVE + 64*4] + vpaddd F, F, [rsp + _DIGEST_SAVE + 64*5] + vpaddd G, G, [rsp + _DIGEST_SAVE + 64*6] + vpaddd H, H, [rsp + _DIGEST_SAVE + 64*7] + + jmp lloop + +lastLoop: + ; Process last 16 rounds +%assign I 48 +%assign J 0 +%rep 16 + PROCESS_LOOP APPEND(W,J), I +%assign I (I+1) +%assign J (J+1) +%endrep + + ; Add old digest + vpaddd A, A, [rsp + _DIGEST_SAVE + 64*0] + vpaddd B, B, [rsp + _DIGEST_SAVE + 64*1] + vpaddd C, C, [rsp + _DIGEST_SAVE + 64*2] + vpaddd D, D, [rsp + _DIGEST_SAVE + 64*3] + vpaddd E, E, [rsp + _DIGEST_SAVE + 64*4] + vpaddd F, F, [rsp + _DIGEST_SAVE + 64*5] + vpaddd G, G, [rsp + _DIGEST_SAVE + 64*6] + vpaddd H, H, [rsp + _DIGEST_SAVE + 64*7] + + ; Write out digest + ; Do we need to untranspose digests??? + vmovdqu32 [STATE + 0*SHA256_DIGEST_ROW_SIZE], A + vmovdqu32 [STATE + 1*SHA256_DIGEST_ROW_SIZE], B + vmovdqu32 [STATE + 2*SHA256_DIGEST_ROW_SIZE], C + vmovdqu32 [STATE + 3*SHA256_DIGEST_ROW_SIZE], D + vmovdqu32 [STATE + 4*SHA256_DIGEST_ROW_SIZE], E + vmovdqu32 [STATE + 5*SHA256_DIGEST_ROW_SIZE], F + vmovdqu32 [STATE + 6*SHA256_DIGEST_ROW_SIZE], G + vmovdqu32 [STATE + 7*SHA256_DIGEST_ROW_SIZE], H + + ; update input pointers +%assign I 0 +%rep 16 + add [STATE + _data_ptr_sha256 + I*PTR_SZ], IDX +%assign I (I+1) +%endrep + +%ifdef SAFE_DATA + ;; Clear stack frame (8*64 bytes) + vpxorq zmm0, zmm0 +%assign i 0 +%rep 8 + vmovdqa64 [rsp + i*64], zmm0 +%assign i (i+1) +%endrep +%endif + + mov rsp, [rsp + _rsp] + ret + +%ifdef LINUX +section .note.GNU-stack noalloc noexec nowrite progbits +%endif -- cgit v1.2.3