ceph/src/crypto/isa-l/isa-l_crypto/aes/cbc_dec_192_x8_avx.asm

   1 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
   2 ;  Copyright(c) 2011-2016 Intel Corporation All rights reserved.
   3 ;
   4 ;  Redistribution and use in source and binary forms, with or without
   5 ;  modification, are permitted provided that the following conditions
   6 ;  are met:
   7 ;    * Redistributions of source code must retain the above copyright
   8 ;      notice, this list of conditions and the following disclaimer.
   9 ;    * Redistributions in binary form must reproduce the above copyright
  10 ;      notice, this list of conditions and the following disclaimer in
  11 ;      the documentation and/or other materials provided with the
  12 ;      distribution.
  13 ;    * Neither the name of Intel Corporation nor the names of its
  14 ;      contributors may be used to endorse or promote products derived
  15 ;      from this software without specific prior written permission.
  16 ;
  17 ;  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  18 ;  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  19 ;  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  20 ;  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  21 ;  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  22 ;  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  23 ;  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  24 ;  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  25 ;  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  26 ;  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  27 ;  OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  28 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
  29
  30 ; routine to do AES192 CBC decrypt
  31
  32 %include "reg_sizes.asm"
  33
  34 %ifidn __OUTPUT_FORMAT__, elf64
  35 %define IN              rdi
  36 %define IV              rsi
  37 %define KEYS            rdx
  38 %define OUT             rcx
  39 %define LEN             r8
  40 %define func(x) x:
  41 %define FUNC_SAVE
  42 %define FUNC_RESTORE
  43 %endif
  44
  45 %ifidn __OUTPUT_FORMAT__, win64
  46 %define IN              rcx
  47 %define IV              rdx
  48 %define KEYS            r8
  49 %define OUT             r9
  50 %define LEN             r10
  51 %define PS              8
  52 %define stack_size      10*16 + 1*8     ; must be an odd multiple of 8
  53 %define arg(x)          [rsp + stack_size + PS + PS*x]
  54
  55 %define func(x) proc_frame x
  56 %macro FUNC_SAVE 0
  57         alloc_stack     stack_size
  58         save_xmm128     xmm6, 0*16
  59         save_xmm128     xmm7, 1*16
  60         save_xmm128     xmm8, 2*16
  61         save_xmm128     xmm9, 3*16
  62         save_xmm128     xmm10, 4*16
  63         save_xmm128     xmm11, 5*16
  64         save_xmm128     xmm12, 6*16
  65         save_xmm128     xmm13, 7*16
  66         save_xmm128     xmm14, 8*16
  67         save_xmm128     xmm15, 9*16
  68         end_prolog
  69         mov     LEN, arg(4)
  70 %endmacro
  71
  72 %macro FUNC_RESTORE 0
  73         movdqa  xmm6, [rsp + 0*16]
  74         movdqa  xmm7, [rsp + 1*16]
  75         movdqa  xmm8, [rsp + 2*16]
  76         movdqa  xmm9, [rsp + 3*16]
  77         movdqa  xmm10, [rsp + 4*16]
  78         movdqa  xmm11, [rsp + 5*16]
  79         movdqa  xmm12, [rsp + 6*16]
  80         movdqa  xmm13, [rsp + 7*16]
  81         movdqa  xmm14, [rsp + 8*16]
  82         movdqa  xmm15, [rsp + 9*16]
  83         add     rsp, stack_size
  84 %endmacro
  85 %endif
  86
  87 ; configuration paramaters for AES-CBC
  88 %define KEY_ROUNDS 13
  89 %define XMM_USAGE    (16)
  90 %define EARLY_BLOCKS (4)
  91 %define PARALLEL_BLOCKS (11)
  92 %define IV_CNT          (1)
  93
  94 ; instruction set specific operation definitions
  95 %define MOVDQ         vmovdqu
  96 %macro PXOR 2
  97    vpxor %1, %1, %2
  98 %endm
  99
 100 %macro AES_DEC 2
 101   vaesdec %1, %1, %2
 102 %endm
 103
 104 %macro AES_DEC_LAST 2
 105   vaesdeclast %1, %1, %2
 106 %endm
 107
 108 %include "cbc_common.asm"
 109
 110
 111 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 112 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 113 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 114 ;; aes_cbc_dec_192_avx(void *in, void *IV, void *keys, void *out, UINT64 num_bytes)
 115 global aes_cbc_dec_192_avx:function
 116 func(aes_cbc_dec_192_avx)
 117         FUNC_SAVE
 118
 119         FILL_KEY_CACHE CKEY_CNT, FIRST_CKEY, KEYS, MOVDQ
 120
 121         MOVDQ  reg(IV_IDX), [IV]         ; Load IV for next round of block decrypt
 122         mov IDX, 0
 123         cmp     LEN, PARALLEL_BLOCKS*16
 124         jge     main_loop                ; if enough data blocks remain enter main_loop
 125         jmp  partials
 126
 127 main_loop:
 128         CBC_DECRYPT_BLOCKS KEY_ROUNDS, PARALLEL_BLOCKS, EARLY_BLOCKS, MOVDQ, PXOR, AES_DEC, AES_DEC_LAST, CKEY_CNT, TMP, TMP_CNT, FIRST_CKEY, KEYS, FIRST_XDATA, IN, OUT, IDX, LEN
 129         cmp     LEN, PARALLEL_BLOCKS*16
 130         jge     main_loop                ; enough blocks to do another full parallel set
 131         jz  done
 132
 133 partials:                                ; fewer than 'PARALLEL_BLOCKS' left do in groups of 4, 2 or 1
 134         cmp     LEN, 0
 135         je      done
 136         cmp     LEN, 4*16
 137         jge     initial_4
 138         cmp     LEN, 2*16
 139         jge     initial_2
 140
 141 initial_1:
 142         CBC_DECRYPT_BLOCKS KEY_ROUNDS, 1, EARLY_BLOCKS, MOVDQ, PXOR, AES_DEC, AES_DEC_LAST, CKEY_CNT, TMP, TMP_CNT, FIRST_CKEY, KEYS, FIRST_XDATA, IN, OUT, IDX, LEN
 143         jmp     done
 144
 145 initial_2:
 146         CBC_DECRYPT_BLOCKS KEY_ROUNDS, 2, EARLY_BLOCKS, MOVDQ, PXOR, AES_DEC, AES_DEC_LAST, CKEY_CNT, TMP, TMP_CNT, FIRST_CKEY, KEYS, FIRST_XDATA, IN, OUT, IDX, LEN
 147         jz  done
 148         jmp     partials
 149
 150 initial_4:
 151         CBC_DECRYPT_BLOCKS KEY_ROUNDS, 4, EARLY_BLOCKS, MOVDQ, PXOR, AES_DEC, AES_DEC_LAST, CKEY_CNT, TMP, TMP_CNT, FIRST_CKEY, KEYS, FIRST_XDATA, IN, OUT, IDX, LEN
 152         jnz     partials
 153 done:
 154         FUNC_RESTORE
 155         ret
 156
 157 endproc_frame