cpu: conv: bwd_d: fix r_overflow1 bug

Author: kwiersch

src/cpu/jit_avx2_conv_kernel_f32.cpp

@@ -757,8 +757,8 @@ void jit_avx2_conv_bwd_data_kernel_f32::generate() {
     int l_overflow = nstl::max(0, (jcp.kw - 1 - jcp.l_pad) / jcp.stride_w);
     int r_overflow = nstl::max(0, (jcp.kw - 1
                     - nstl::max(0, jcp.r_pad)) / jcp.stride_w);
-    int r_overflow1 = nstl::max(0, (jcp.kw - 1
-                    - nstl::max(0, jcp.r_pad) - jcp.ur_w_tail) / jcp.stride_w);
+    int r_overflow1 = nstl::max(
+            0, (jcp.kw - 1 - jcp.r_pad - jcp.ur_w_tail) / jcp.stride_w);
 
     int n_oi = jcp.iw / jcp.ur_w;
     if (r_overflow1 > 0)
@@ -946,13 +946,16 @@ status_t jit_avx2_conv_bwd_data_kernel_f32::init_conf(jit_conv_conf_t &jcp,
 
     jcp.ur_w_tail = jcp.iw % jcp.ur_w;
 
-    int r_overflow_no_tail = nstl::max(0, (jcp.kw - 1 - jcp.ur_w_tail
-                    - nstl::max(0, jcp.r_pad) - jcp.ur_w_tail) / jcp.stride_w);
-    /* maximum 1 ur_w block with r_overflow so far */
-    if (r_overflow_no_tail * jcp.stride_w > jcp.ur_w)
-        return status::unimplemented;
-
-    if ((jcp.iw > jcp.ur_w) && (jcp.ur_w % jcp.stride_w != 0))
+    int r_overflow_no_tail = nstl::max(
+            0, (jcp.kw - 1 - jcp.r_pad - jcp.ur_w_tail) / jcp.stride_w);
+    bool tails_not_ok = false
+            /* maximum 1 ur_w block with r_overflow so far */
+            || r_overflow_no_tail * jcp.stride_w > jcp.ur_w
+            /* ur_w must be a multiple of stride */
+            || ((jcp.iw > jcp.ur_w) && (jcp.ur_w % jcp.stride_w != 0))
+            /* r_pad must not extend beyond ur_w_tail */
+            || ((jcp.iw > jcp.ur_w) && (jcp.r_pad + jcp.ur_w_tail < 0));
+    if (tails_not_ok)
         return status::unimplemented;
 
     return status::success;

src/cpu/jit_avx512_common_conv_kernel.cpp

@@ -2478,8 +2478,8 @@ void jit_avx512_common_conv_bwd_data_kernel_f32::generate()
     int l_overflow = nstl::max(0, ((kw - 1) * dilate_w - jcp.l_pad) / stride_w);
     int r_overflow = nstl::max(0, ((kw - 1) * dilate_w
                     - nstl::max(0, jcp.r_pad)) / stride_w);
-    int r_overflow1 = nstl::max(0, ((kw - 1) * dilate_w
-                    - nstl::max(0, jcp.r_pad) - ur_w_tail) / stride_w);
+    int r_overflow1 = nstl::max(
+            0, ((kw - 1) * dilate_w - jcp.r_pad - ur_w_tail) / stride_w);
 
     int n_oi = iw / ur_w;
     if (r_overflow1 > 0) n_oi--;
@@ -2639,8 +2639,9 @@ status_t jit_avx512_common_conv_bwd_data_kernel_f32::init_conf(
     }
     int l_overflow = nstl::max(0, ((jcp.kw - 1) * (jcp.dilate_w + 1)
                     - jcp.l_pad) / jcp.stride_w);
-    int r_overflow1 = nstl::max(0, ((jcp.kw - 1) * (jcp.dilate_w + 1)
-                    - nstl::max(0, jcp.r_pad) - jcp.iw % jcp.ur_w) / jcp.stride_w);
+    int r_overflow1 = nstl::max(0,
+            ((jcp.kw - 1) * (jcp.dilate_w + 1) - jcp.r_pad - jcp.iw % jcp.ur_w)
+                    / jcp.stride_w);
     int n_oi = jcp.iw / jcp.ur_w;
     if (r_overflow1 > 0) n_oi--;
 
@@ -2782,11 +2783,17 @@ status_t jit_avx512_common_conv_bwd_data_kernel_f32::init_conf(
 
     if (l_overflow * jcp.stride_w > jcp.ur_w)
         return status::unimplemented;
-    int r_overflow_no_tail = nstl::max(0, ((jcp.kw - 1) * (jcp.dilate_w + 1)
-                    - nstl::max(0, jcp.r_pad) - jcp.ur_w_tail) / jcp.stride_w);
-    if (r_overflow_no_tail * jcp.stride_w > jcp.ur_w)
-        return status::unimplemented;
-    if ((jcp.iw > jcp.ur_w) && (jcp.ur_w % jcp.stride_w != 0))
+    int r_overflow_no_tail = nstl::max(0,
+            ((jcp.kw - 1) * (jcp.dilate_w + 1) - jcp.r_pad - jcp.ur_w_tail)
+                    / jcp.stride_w);
+    bool tails_not_ok = false
+            /* maximum 1 ur_w block with r_overflow so far */
+            || r_overflow_no_tail * jcp.stride_w > jcp.ur_w
+            /* ur_w must be a multiple of stride */
+            || ((jcp.iw > jcp.ur_w) && (jcp.ur_w % jcp.stride_w != 0))
+            /* r_pad must not extend beyond ur_w_tail */
+            || ((jcp.iw > jcp.ur_w) && (jcp.r_pad + jcp.ur_w_tail < 0));
+    if (tails_not_ok)
         return status::unimplemented;
 
     pick_loop_order(jcp);

src/cpu/jit_avx512_core_bf16_conv_kernel.cpp

@@ -950,8 +950,8 @@ void jit_avx512_core_bf16_bwd_data_kernel::generate()
     int l_overflow = nstl::max(0, ((kw - 1) * dilate_w - jcp.l_pad) / stride_w);
     int r_overflow = nstl::max(0, ((kw - 1) * dilate_w
                     - nstl::max(0, jcp.r_pad)) / stride_w);
-    int r_overflow1 = nstl::max(0, ((kw - 1) * dilate_w
-                    - nstl::max(0, jcp.r_pad) - ur_w_tail) / stride_w);
+    int r_overflow1 = nstl::max(
+            0, ((kw - 1) * dilate_w - jcp.r_pad - ur_w_tail) / stride_w);
 
     int n_oi = iw / ur_w;
     if (r_overflow1 > 0) n_oi--;
@@ -1095,8 +1095,9 @@ status_t jit_avx512_core_bf16_bwd_data_kernel::init_conf(
                                                   reserved */
     int l_overflow = nstl::max(0, ((jcp.kw - 1) * (jcp.dilate_w + 1)
                     - jcp.l_pad) / jcp.stride_w);
-    int r_overflow1 = nstl::max(0, ((jcp.kw - 1) * (jcp.dilate_w + 1)
-                    - nstl::max(0, jcp.r_pad) - jcp.iw % jcp.ur_w) / jcp.stride_w);
+    int r_overflow1 = nstl::max(0,
+            ((jcp.kw - 1) * (jcp.dilate_w + 1) - jcp.r_pad - jcp.iw % jcp.ur_w)
+                    / jcp.stride_w);
     int n_oi = jcp.iw / jcp.ur_w;
     if (r_overflow1 > 0) n_oi--;
 
@@ -1153,11 +1154,17 @@ status_t jit_avx512_core_bf16_bwd_data_kernel::init_conf(
 
     if (l_overflow * jcp.stride_w > jcp.ur_w)
         return status::unimplemented;
-    int r_overflow_no_tail = nstl::max(0, ((jcp.kw - 1) * (jcp.dilate_w + 1)
-                    - nstl::max(0, jcp.r_pad) - jcp.ur_w_tail) / jcp.stride_w);
-    if (r_overflow_no_tail * jcp.stride_w > jcp.ur_w)
-        return status::unimplemented;
-    if ((jcp.iw > jcp.ur_w) && (jcp.ur_w % jcp.stride_w != 0))
+    int r_overflow_no_tail = nstl::max(0,
+            ((jcp.kw - 1) * (jcp.dilate_w + 1) - jcp.r_pad - jcp.ur_w_tail)
+                    / jcp.stride_w);
+    bool tails_not_ok = false
+            /* maximum 1 ur_w block with r_overflow so far */
+            || r_overflow_no_tail * jcp.stride_w > jcp.ur_w
+            /* ur_w must be a multiple of stride */
+            || ((jcp.iw > jcp.ur_w) && (jcp.ur_w % jcp.stride_w != 0))
+            /* r_pad must not extend beyond ur_w_tail */
+            || ((jcp.iw > jcp.ur_w) && (jcp.r_pad + jcp.ur_w_tail < 0));
+    if (tails_not_ok)
         return status::unimplemented;
 
     pick_loop_order(jcp);

tests/benchdnn/inputs/test_conv_regression_general

@@ -98,3 +98,8 @@ mb1_g2ic6oc16_ih5oh5kh1sh1dh0ph0_iw5ow5kw1sw1dw0pw0
 
 # MFDNN-1968 1x1 conv asymmetrical strides with stride_h=1
 --reset --dir=BWD_D mb1ic2ih10iw8oc4oh10ow4kh1kw1sh1sw2ph0pw0dh0dw0
+
+# github #542
+# r_overflow1 corner case
+--reset --dir=bwd_d mb1_ic9oc3_ih32oh10kh5sh3dh0ph1
+