Merge pull request aous72#190 from aous72/issue_188

Fixed crashes when avx512 encoder is employed.
It is happening with clang only, but it is a bug for any compiler.

Author: aous72

src/core/coding/ojph_block_encoder_avx512.cpp

@@ -539,7 +539,7 @@ static void proc_pixel(__m512i *src_vec, ui32 p,
         /*   s[0] = --val + (t >> 31); //v_n = 2(\mu_p-1) + s_n */
         val_vec[i] = _mm512_mask_sub_epi32(ZERO, val_mask[i], val_vec[i], ONE);
         _s_vec[i] = _mm512_mask_srli_epi32(ZERO, val_mask[i], src_vec[i], 31);
-        _s_vec[i] = 
+        _s_vec[i] =
           _mm512_mask_add_epi32(ZERO, val_mask[i], _s_vec[i], val_vec[i]);
         /* } */
     }
@@ -571,18 +571,18 @@ static void proc_pixel(__m512i *src_vec, ui32 p,
         ui32 o_idx = i & 0x1;
 
         eq_vec[i] = _mm512_permutexvar_epi32(idx[e_idx], _eq_vec[o_idx]);
-        eq_vec[i] = _mm512_mask_permutexvar_epi32(eq_vec[i], 0xFF00, 
-                                                  idx[e_idx], 
+        eq_vec[i] = _mm512_mask_permutexvar_epi32(eq_vec[i], 0xFF00,
+                                                  idx[e_idx],
                                                   _eq_vec[o_idx + 2]);
 
         s_vec[i] = _mm512_permutexvar_epi32(idx[e_idx], _s_vec[o_idx]);
         s_vec[i] = _mm512_mask_permutexvar_epi32(s_vec[i], 0xFF00,
-                                                 idx[e_idx], 
+                                                 idx[e_idx],
                                                  _s_vec[o_idx + 2]);
 
         _rho_vec[i] = _mm512_permutexvar_epi32(idx[e_idx], val_vec[o_idx]);
         _rho_vec[i] = _mm512_mask_permutexvar_epi32(_rho_vec[i], 0xFF00,
-                                                    idx[e_idx], 
+                                                    idx[e_idx],
                                                     val_vec[o_idx + 2]);
         _rho_vec[i] = _mm512_slli_epi32(_rho_vec[i], i);
 
@@ -695,11 +695,11 @@ static void proc_ms_encode(ms_struct *msp,
         /* cwd = s[i * 4 + 0] & ((1U << m) - 1)
          * cwd_len = m
          */
-        _mm512_store_epi32(cwd_len, m_vec[i]);
+        _mm512_storeu_epi32(cwd_len, m_vec[i]);
         tmp = _mm512_sllv_epi32(ONE, m_vec[i]);
         tmp = _mm512_sub_epi32(tmp, ONE);
         tmp = _mm512_and_epi32(tmp, s_vec[i]);
-        _mm512_store_epi32(cwd, tmp);
+        _mm512_storeu_epi32(cwd, tmp);
 
         for (ui32 j = 0; j < 8; ++j) {
             ui32 idx = j * 2;
@@ -712,7 +712,7 @@ static void proc_ms_encode(ms_struct *msp,
     }
 }
 
-static __m512i cal_eps_vec(__m512i *eq_vec, __m512i &u_q_vec, 
+static __m512i cal_eps_vec(__m512i *eq_vec, __m512i &u_q_vec,
                            __m512i &e_qmax_vec)
 {
     /* if (u_q[i] > 0) {
@@ -755,7 +755,7 @@ static void update_lep(ui32 x, __m512i &prev_e_val_vec,
      */
     auto tmp = _mm512_mask_permutexvar_epi32(prev_e_val_vec, 0xFFFE,
                                              left_shift, eq_vec[3]);
-    prev_e_val_vec = _mm512_mask_permutexvar_epi32(ZERO, 0x1, left_shift, 
+    prev_e_val_vec = _mm512_mask_permutexvar_epi32(ZERO, 0x1, left_shift,
                                                    eq_vec[3]);
     e_val_vec[x] = _mm512_max_epi32(eq_vec[1], tmp);
 }
@@ -769,9 +769,9 @@ static void update_lcxp(ui32 x, __m512i &prev_cx_val_vec,
      * lcxp[0] = (ui8)((rho[0] & 8) >> 3);
      * Or (rho[0] & 2) and (rho[0] of the previous round & 8).
      */
-    auto tmp = _mm512_mask_permutexvar_epi32(prev_cx_val_vec, 0xFFFE, 
+    auto tmp = _mm512_mask_permutexvar_epi32(prev_cx_val_vec, 0xFFFE,
                                              left_shift, rho_vec);
-    prev_cx_val_vec = _mm512_mask_permutexvar_epi32(ZERO, 0x1, left_shift, 
+    prev_cx_val_vec = _mm512_mask_permutexvar_epi32(ZERO, 0x1, left_shift,
                                                     rho_vec);
 
     tmp = _mm512_and_epi32(tmp, _mm512_set1_epi32(8));
@@ -793,7 +793,7 @@ static __m512i cal_tuple(__m512i &cq_vec, __m512i &rho_vec,
     return _mm512_i32gather_epi32(tmp, vlc_tbl, 4);
 }
 
-static __m512i proc_cq1(ui32 x, __m512i *cx_val_vec, __m512i &rho_vec, 
+static __m512i proc_cq1(ui32 x, __m512i *cx_val_vec, __m512i &rho_vec,
                         const __m512i right_shift)
 {
     ojph_unused(x);
@@ -809,8 +809,8 @@ static __m512i proc_cq1(ui32 x, __m512i *cx_val_vec, __m512i &rho_vec,
 static __m512i proc_cq2(ui32 x, __m512i *cx_val_vec, __m512i &rho_vec,
                         const __m512i right_shift)
 {
-    // c_q[i + 1] = (lcxp[i + 1] + (lcxp[i + 2] << 2)) 
-    //            | (((rho[i] & 4) >> 1) | ((rho[i] & 8) >> 2)); 
+    // c_q[i + 1] = (lcxp[i + 1] + (lcxp[i + 2] << 2))
+    //            | (((rho[i] & 4) >> 1) | ((rho[i] & 8) >> 2));
     auto lcxp1_vec = _mm512_permutexvar_epi32(right_shift, cx_val_vec[x]);
     auto lcxp2_vec = _mm512_permutexvar_epi32(right_shift, cx_val_vec[x + 1]);
     auto tmp = _mm512_permutexvar_epi32(right_shift, lcxp1_vec);
@@ -831,7 +831,7 @@ static __m512i proc_cq2(ui32 x, __m512i *cx_val_vec, __m512i &rho_vec,
 
 using fn_proc_cq = __m512i (*)(ui32, __m512i *, __m512i &, const __m512i);
 
-static void proc_mel_encode1(mel_struct *melp, __m512i &cq_vec, 
+static void proc_mel_encode1(mel_struct *melp, __m512i &cq_vec,
                              __m512i &rho_vec, __m512i u_q_vec, ui32 ignore,
                              const __m512i right_shift)
 {
@@ -849,7 +849,7 @@ static void proc_mel_encode1(mel_struct *melp, __m512i &cq_vec,
 
     /* if (u_q[i] > 0 && u_q[i + 1] > 0) { } */
     auto mel_need_encode2 = (ui16)_mm512_cmpgt_epi32_mask(u_q_vec, ZERO);
-    mel_need_encode2 = 
+    mel_need_encode2 =
       mel_need_encode2 & (ui16)_mm512_cmpgt_epi32_mask(tmp, ZERO);
 
     ui32 i_max = 16 - (ignore / 2);
@@ -873,7 +873,7 @@ static void proc_mel_encode1(mel_struct *melp, __m512i &cq_vec,
     }
 }
 
-static void proc_mel_encode2(mel_struct *melp, __m512i &cq_vec, 
+static void proc_mel_encode2(mel_struct *melp, __m512i &cq_vec,
                              __m512i &rho_vec, __m512i u_q_vec, ui32 ignore,
                              const __m512i right_shift)
 {
@@ -897,7 +897,7 @@ static void proc_mel_encode2(mel_struct *melp, __m512i &cq_vec,
     }
 }
 
-using fn_proc_mel_encode = void (*)(mel_struct *, __m512i &, __m512i &, 
+using fn_proc_mel_encode = void (*)(mel_struct *, __m512i &, __m512i &,
                                     __m512i, ui32, const __m512i);
 
 static void proc_vlc_encode1(vlc_struct_avx512 *vlcp, ui32 *tuple,
@@ -1006,8 +1006,8 @@ static void proc_vlc_encode2(vlc_struct_avx512 *vlcp, ui32 *tuple,
 
 using fn_proc_vlc_encode = void (*)(vlc_struct_avx512 *, ui32 *, ui32 *, ui32);
 
-void ojph_encode_codeblock_avx512(ui32* buf, ui32 missing_msbs, 
-                                  ui32 num_passes, ui32 _width, ui32 height, 
+void ojph_encode_codeblock_avx512(ui32* buf, ui32 missing_msbs,
+                                  ui32 num_passes, ui32 _width, ui32 height,
                                   ui32 stride, ui32* lengths,
                                   ojph::mem_elastic_allocator *elastic,
                                   ojph::coded_lists *& coded)
@@ -1111,7 +1111,7 @@ void ojph_encode_codeblock_avx512(ui32* buf, ui32 missing_msbs,
 
             if (y + 1 < height) {
                 src_vec[1] = _mm512_maskz_loadu_epi32(load_mask0, sp + stride);
-                src_vec[3] = 
+                src_vec[3] =
                   _mm512_maskz_loadu_epi32(load_mask1, sp + 16 + stride);
             } else {
                 src_vec[1] = ZERO;
@@ -1148,7 +1148,7 @@ void ojph_encode_codeblock_avx512(ui32* buf, ui32 missing_msbs,
             tmp = proc_cq(x, cx_val_vec, rho_vec, right_shift);
             auto cq_vec = _mm512_mask_permutexvar_epi32(prev_cq_vec, 0xFFFE,
                                                         left_shift, tmp);
-            prev_cq_vec = _mm512_mask_permutexvar_epi32(ZERO, 0x1, left_shift, 
+            prev_cq_vec = _mm512_mask_permutexvar_epi32(ZERO, 0x1, left_shift,
                                                         tmp);
 
             update_lep(x, prev_e_val_vec, eq_vec, e_val_vec, left_shift);
@@ -1163,7 +1163,7 @@ void ojph_encode_codeblock_avx512(ui32* buf, ui32 missing_msbs,
             __m512i tuple_vec = cal_tuple(cq_vec, rho_vec, eps_vec, vlc_tbl);
             ui32 _ignore = ((n_loop - 1) == x) ? ignore : 0;
 
-            proc_mel_encode(&mel, cq_vec, rho_vec, u_q_vec, _ignore, 
+            proc_mel_encode(&mel, cq_vec, rho_vec, u_q_vec, _ignore,
                             right_shift);
 
             proc_ms_encode(&ms, tuple_vec, uq_vec, rho_vec, s_vec);
@@ -1177,8 +1177,8 @@ void ojph_encode_codeblock_avx512(ui32* buf, ui32 missing_msbs,
              * So in the vlc_encode, the tuple will only be scaled by 2.
              */
             tuple_vec = _mm512_srli_epi32(tuple_vec, 4);
-            _mm512_store_epi32(tuple, tuple_vec);
-            _mm512_store_epi32(u_q, u_q_vec);
+            _mm512_storeu_epi32(tuple, tuple_vec);
+            _mm512_storeu_epi32(u_q, u_q_vec);
             proc_vlc_encode(&vlc, tuple, u_q, _ignore);
         }