package/libs/wolfssl/patches/001-Maths-x86-asm-change-asm-snippets-to-get-compiling.patch

   1 From fa8f23284d4689c2a737204b337b58d966dcbd8c Mon Sep 17 00:00:00 2001
   2 From: Sean Parkinson <sean@wolfssl.com>
   3 Date: Fri, 20 Aug 2021 10:23:38 +1000
   4 Subject: [PATCH] Maths x86 asm: change asm snippets to get compiling
   5
   6 TFM:
   7   Use register or memory for c0, c1, c2 in SQRADD and SQRADD2.
   8 SP:
   9   Use register or memory for vl, vh, vo in SP_ASM_MUL_ADD,
  10 SP_ASM_MUL_ADD2 and SP_ASM_SQR_ADD.
  11 ---
  12  wolfcrypt/src/asm.c    | 29 ++++++++++++++++++++---------
  13  wolfcrypt/src/sp_int.c |  6 +++---
  14  2 files changed, 23 insertions(+), 12 deletions(-)
  15
  16 --- a/wolfcrypt/src/asm.c
  17 +++ b/wolfcrypt/src/asm.c
  18 @@ -698,33 +698,39 @@ __asm__(                             \
  19
  20  #define SQRADD(i, j)                                      \
  21  __asm__(                                                  \
  22 -     "movl  %6,%%eax     \n\t"                            \
  23 +     "movl  %3,%%eax     \n\t"                            \
  24       "mull  %%eax        \n\t"                            \
  25       "addl  %%eax,%0     \n\t"                            \
  26       "adcl  %%edx,%1     \n\t"                            \
  27       "adcl  $0,%2        \n\t"                            \
  28 -     :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i) :"%eax","%edx","cc");
  29 +     :"+rm"(c0), "+rm"(c1), "+rm"(c2)                     \
  30 +     : "m"(i)                                             \
  31 +     :"%eax","%edx","cc");
  32
  33  #define SQRADD2(i, j)                                     \
  34  __asm__(                                                  \
  35 -     "movl  %6,%%eax     \n\t"                            \
  36 -     "mull  %7           \n\t"                            \
  37 +     "movl  %3,%%eax     \n\t"                            \
  38 +     "mull  %4           \n\t"                            \
  39       "addl  %%eax,%0     \n\t"                            \
  40       "adcl  %%edx,%1     \n\t"                            \
  41       "adcl  $0,%2        \n\t"                            \
  42       "addl  %%eax,%0     \n\t"                            \
  43       "adcl  %%edx,%1     \n\t"                            \
  44       "adcl  $0,%2        \n\t"                            \
  45 -     :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i), "m"(j)  :"%eax","%edx", "cc");
  46 +     :"+rm"(c0), "+rm"(c1), "+rm"(c2)                     \
  47 +     : "m"(i), "m"(j)                                     \
  48 +     :"%eax","%edx", "cc");
  49
  50  #define SQRADDSC(i, j)                                    \
  51 -__asm__(                                                     \
  52 +__asm__(                                                  \
  53       "movl  %3,%%eax     \n\t"                            \
  54       "mull  %4           \n\t"                            \
  55       "movl  %%eax,%0     \n\t"                            \
  56       "movl  %%edx,%1     \n\t"                            \
  57       "xorl  %2,%2        \n\t"                            \
  58 -     :"=r"(sc0), "=r"(sc1), "=r"(sc2): "g"(i), "g"(j) :"%eax","%edx","cc");
  59 +     :"=r"(sc0), "=r"(sc1), "=r"(sc2)                     \
  60 +     : "g"(i), "g"(j)                                     \
  61 +     :"%eax","%edx","cc");
  62
  63  #define SQRADDAC(i, j)                                    \
  64  __asm__(                                                  \
  65 @@ -733,7 +739,9 @@ __asm__(
  66       "addl  %%eax,%0     \n\t"                            \
  67       "adcl  %%edx,%1     \n\t"                            \
  68       "adcl  $0,%2        \n\t"                            \
  69 -     :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "g"(i), "g"(j) :"%eax","%edx","cc");
  70 +     :"=r"(sc0), "=r"(sc1), "=r"(sc2)                     \
  71 +     : "0"(sc0), "1"(sc1), "2"(sc2), "g"(i), "g"(j)       \
  72 +     :"%eax","%edx","cc");
  73
  74  #define SQRADDDB                                          \
  75  __asm__(                                                  \
  76 @@ -743,7 +751,10 @@ __asm__(
  77       "addl %6,%0         \n\t"                            \
  78       "adcl %7,%1         \n\t"                            \
  79       "adcl %8,%2         \n\t"                            \
  80 -     :"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(sc0), "r"(sc1), "r"(sc2) : "cc");
  81 +     :"=r"(c0), "=r"(c1), "=r"(c2)                        \
  82 +     : "0"(c0), "1"(c1), "2"(c2), "r"(sc0), "r"(sc1),     \
  83 +       "r"(sc2)                                           \
  84 +     : "cc");
  85
  86  #elif defined(TFM_X86_64)
  87  /* x86-64 optimized */
  88 --- a/wolfcrypt/src/sp_int.c
  89 +++ b/wolfcrypt/src/sp_int.c
  90 @@ -476,7 +476,7 @@ static WC_INLINE sp_int_digit sp_div_wor
  91          "addl  %%eax, %[l]     \n\t"                    \
  92          "adcl  %%edx, %[h]     \n\t"                    \
  93          "adcl  $0   , %[o]     \n\t"                    \
  94 -        : [l] "+r" (vl), [h] "+r" (vh), [o] "+r" (vo)    \
  95 +        : [l] "+rm" (vl), [h] "+rm" (vh), [o] "+rm" (vo) \
  96          : [a] "r" (va), [b] "r" (vb)                     \
  97          : "eax", "edx", "cc"                             \
  98      )
  99 @@ -502,7 +502,7 @@ static WC_INLINE sp_int_digit sp_div_wor
 100          "addl  %%eax, %[l]     \n\t"                    \
 101          "adcl  %%edx, %[h]     \n\t"                    \
 102          "adcl  $0   , %[o]     \n\t"                    \
 103 -        : [l] "+r" (vl), [h] "+r" (vh), [o] "+r" (vo)    \
 104 +        : [l] "+rm" (vl), [h] "+rm" (vh), [o] "+rm" (vo) \
 105          : [a] "r" (va), [b] "r" (vb)                     \
 106          : "eax", "edx", "cc"                             \
 107      )
 108 @@ -541,7 +541,7 @@ static WC_INLINE sp_int_digit sp_div_wor
 109          "addl  %%eax, %[l]     \n\t"                    \
 110          "adcl  %%edx, %[h]     \n\t"                    \
 111          "adcl  $0   , %[o]     \n\t"                    \
 112 -        : [l] "+r" (vl), [h] "+r" (vh), [o] "+r" (vo)    \
 113 +        : [l] "+rm" (vl), [h] "+rm" (vh), [o] "+rm" (vo) \
 114          : [a] "m" (va)                                   \
 115          : "eax", "edx", "cc"                             \
 116      )