Merge pull request #2871 from vbystricky:oclopt_integral

modules/imgproc/src/opencl/integral_sum.cl

@@ -61,115 +61,64 @@
 #define GET_CONFLICT_OFFSET(lid) ((lid) >> LOG_NUM_BANKS)
 
 #if sdepth == 4
+#define sumT               int
+#define vecSumT            int4
+#define convertToSum4      convert_int4
+#elif sdepth == 5
+#define sumT               float
+#define vecSumT            float4
+#define convertToSum4      convert_float4
+#endif
+
 
-kernel void integral_sum_cols(__global uchar4 *src, __global int *sum,
-                              int src_offset, int pre_invalid, int rows, int cols, int src_step, int dst_step)
+kernel void integral_sum_cols(__global const uchar4 *src, __global uchar *sum_ptr,
+                              int src_offset, int rows, int cols, int src_step, int dst_step)
 {
+    __global sumT *sum = (__global sumT *)sum_ptr;
     int lid = get_local_id(0);
     int gid = get_group_id(0);
-    int4 src_t[2], sum_t[2];
-    __local int4 lm_sum[2][LSIZE + LOG_LSIZE];
-    __local int* sum_p;
+    vecSumT src_t[2], sum_t[2];
+    __local vecSumT lm_sum[2][LSIZE + LOG_LSIZE];
+    __local sumT* sum_p;
     src_step = src_step >> 2;
     gid = gid << 1;
-    for(int i = 0; i < rows; i =i + LSIZE_1)
+    int lid_prim = ((lid & 127) << 1) + 1;
+    for (int i = 0; i < rows; i += LSIZE_1)
     {
-        src_t[0] = (i + lid < rows ? convert_int4(src[src_offset + (lid+i) * src_step + gid]) : 0);
-        src_t[1] = (i + lid < rows ? convert_int4(src[src_offset + (lid+i) * src_step + gid + 1]) : 0);
-
-        sum_t[0] =  (i == 0 ? 0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
-        sum_t[1] =  (i == 0 ? 0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
-        barrier(CLK_LOCAL_MEM_FENCE);
-
-        int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
-        lm_sum[0][bf_loc] = src_t[0];
-
-        lm_sum[1][bf_loc] = src_t[1];
-
-        int offset = 1;
-        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        if (i + lid < rows)
         {
-            barrier(CLK_LOCAL_MEM_FENCE);
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
-            ai += GET_CONFLICT_OFFSET(ai);
-            bi += GET_CONFLICT_OFFSET(bi);
-
-            if((lid & 127) < d)
-            {
-                lm_sum[lid >> 7][bi]  +=  lm_sum[lid >> 7][ai];
-            }
-            offset <<= 1;
+            int src_index = mad24((lid+i), src_step, gid + src_offset);
+            src_t[0] = convertToSum4(src[src_index]);
+            src_t[1] = convertToSum4(src[src_index + 1]);
         }
-        barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid < 2)
+        else
         {
-            lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+            src_t[0] = (vecSumT)0;
+            src_t[1] = (vecSumT)0;
         }
-        for(int d = 1;  d < LSIZE; d <<= 1)
-        {
-            barrier(CLK_LOCAL_MEM_FENCE);
-            offset >>= 1;
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
-            ai += GET_CONFLICT_OFFSET(ai);
-            bi += GET_CONFLICT_OFFSET(bi);
 
-            if((lid & 127) < d)
-            {
-                lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
-                lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
-            }
+        if (i == 0)
+        {
+            sum_t[0] = (vecSumT)0;
+            sum_t[1] = (vecSumT)0;
         }
-        barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid > 0 && (i+lid) <= rows)
+        else
         {
-            int loc_s0 = gid * dst_step + i + lid - 1 - pre_invalid * dst_step / 4, loc_s1 = loc_s0 + dst_step ;
-            lm_sum[0][bf_loc] += sum_t[0];
-            lm_sum[1][bf_loc] += sum_t[1];
-            sum_p = (__local int*)(&(lm_sum[0][bf_loc]));
-            for(int k = 0; k < 4; k++)
-            {
-                if(gid * 4 + k >= cols + pre_invalid || gid * 4 + k < pre_invalid) continue;
-                sum[loc_s0 + k * dst_step / 4] = sum_p[k];
-            }
-            sum_p = (__local int*)(&(lm_sum[1][bf_loc]));
-            for(int k = 0; k < 4; k++)
-            {
-                if(gid * 4 + k + 4 >= cols + pre_invalid) break;
-                sum[loc_s1 + k * dst_step / 4] = sum_p[k];
-            }
+            sum_t[0] =  lm_sum[0][LSIZE_2 + LOG_LSIZE];
+            sum_t[1] =  lm_sum[1][LSIZE_2 + LOG_LSIZE];
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-    }
-}
-
-kernel void integral_sum_rows(__global int4 *srcsum, __global int *sum,
-                              int rows, int cols, int src_step, int sum_step, int sum_offset)
-{
-    int lid = get_local_id(0);
-    int gid = get_group_id(0);
-    int4 src_t[2], sum_t[2];
-    __local int4 lm_sum[2][LSIZE + LOG_LSIZE];
-    __local int *sum_p;
-    src_step = src_step >> 4;
-    for(int i = 0; i < rows; i =i + LSIZE_1)
-    {
-        src_t[0] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2] : 0;
-        src_t[1] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2 + 1] : 0;
-
-        sum_t[0] =  (i == 0 ? 0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
-        sum_t[1] =  (i == 0 ? 0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
-        barrier(CLK_LOCAL_MEM_FENCE);
 
         int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
-        lm_sum[0][bf_loc] = src_t[0];
 
+        lm_sum[0][bf_loc] = src_t[0];
         lm_sum[1][bf_loc] = src_t[1];
 
         int offset = 1;
-        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        for (int d = LSIZE >> 1 ;  d > 0; d>>=1)
         {
             barrier(CLK_LOCAL_MEM_FENCE);
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            int ai = offset * lid_prim - 1,bi = ai + offset;
             ai += GET_CONFLICT_OFFSET(ai);
             bi += GET_CONFLICT_OFFSET(bi);
 
@@ -180,15 +129,15 @@ kernel void integral_sum_rows(__global int4 *srcsum, __global int *sum,
             offset <<= 1;
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid < 2)
+        if (lid < 2)
         {
             lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
         }
-        for(int d = 1;  d < LSIZE; d <<= 1)
+        for (int d = 1;  d < LSIZE; d <<= 1)
         {
             barrier(CLK_LOCAL_MEM_FENCE);
             offset >>= 1;
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            int ai = offset * lid_prim - 1,bi = ai + offset;
             ai += GET_CONFLICT_OFFSET(ai);
             bi += GET_CONFLICT_OFFSET(bi);
 
@@ -199,152 +148,78 @@ kernel void integral_sum_rows(__global int4 *srcsum, __global int *sum,
             }
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-        if(gid == 0 && (i + lid) <= rows)
-        {
-            sum[sum_offset + i + lid] = 0;
-        }
-        if(i + lid == 0)
-        {
-            int loc0 = gid * 2 * sum_step;
-            for(int k = 1; k <= 8; k++)
-            {
-                if(gid * 8 + k > cols) break;
-                sum[sum_offset + loc0 + k * sum_step / 4] = 0;
-            }
-        }
-
-        if(lid > 0 && (i+lid) <= rows)
+        if (lid > 0 && (i+lid) <= rows)
         {
-            int loc_s0 = sum_offset + gid * 2 * sum_step + sum_step / 4 + i + lid, loc_s1 = loc_s0 + sum_step ;
+            int loc_s0 = mad24(gid, dst_step, i + lid - 1), loc_s1 = loc_s0 + dst_step;
             lm_sum[0][bf_loc] += sum_t[0];
             lm_sum[1][bf_loc] += sum_t[1];
-            sum_p = (__local int*)(&(lm_sum[0][bf_loc]));
-            for(int k = 0; k < 4; k++)
+            sum_p = (__local sumT*)(&(lm_sum[0][bf_loc]));
+            for (int k = 0; k < 4; k++)
             {
-                if(gid * 8 + k >= cols) break;
-                sum[loc_s0 + k * sum_step / 4] = sum_p[k];
+                if (gid * 4 + k >= cols)
+                    break;
+                sum[loc_s0 + k * dst_step / 4] = sum_p[k];
             }
-            sum_p = (__local int*)(&(lm_sum[1][bf_loc]));
-            for(int k = 0; k < 4; k++)
+            sum_p = (__local sumT*)(&(lm_sum[1][bf_loc]));
+            for (int k = 0; k < 4; k++)
             {
-                if(gid * 8 + 4 + k >= cols) break;
-                sum[loc_s1 + k * sum_step / 4] = sum_p[k];
+                if (gid * 4 + k + 4 >= cols)
+                    break;
+                sum[loc_s1 + k * dst_step / 4] = sum_p[k];
             }
         }
         barrier(CLK_LOCAL_MEM_FENCE);
     }
 }
 
-#elif sdepth == 5
 
-kernel void integral_sum_cols(__global uchar4 *src, __global float *sum,
-                              int src_offset, int pre_invalid, int rows, int cols, int src_step, int dst_step)
+kernel void integral_sum_rows(__global const uchar *srcsum_ptr, __global uchar *sum_ptr,
+                              int rows, int cols, int src_step, int sum_step, int sum_offset)
 {
+    __global const vecSumT *srcsum = (__global const vecSumT *)srcsum_ptr;
+    __global sumT *sum = (__global sumT *)sum_ptr;
     int lid = get_local_id(0);
     int gid = get_group_id(0);
-    float4 src_t[2], sum_t[2];
-    __local float4 lm_sum[2][LSIZE + LOG_LSIZE];
-    __local float* sum_p;
-    src_step = src_step >> 2;
-    gid = gid << 1;
-    for(int i = 0; i < rows; i =i + LSIZE_1)
+    vecSumT src_t[2], sum_t[2];
+    __local vecSumT lm_sum[2][LSIZE + LOG_LSIZE];
+    __local sumT *sum_p;
+    src_step = src_step >> 4;
+    int lid_prim = ((lid & 127) << 1) + 1;
+    for (int i = 0; i < rows; i += LSIZE_1)
     {
-        src_t[0] = (i + lid < rows ? convert_float4(src[src_offset + (lid+i) * src_step + gid]) : (float4)0);
-        src_t[1] = (i + lid < rows ? convert_float4(src[src_offset + (lid+i) * src_step + gid + 1]) : (float4)0);
-
-        sum_t[0] =  (i == 0 ? (float4)0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
-        sum_t[1] =  (i == 0 ? (float4)0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
-        barrier(CLK_LOCAL_MEM_FENCE);
-
-        int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
-        lm_sum[0][bf_loc] = src_t[0];
-
-        lm_sum[1][bf_loc] = src_t[1];
-
-        int offset = 1;
-        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        if (i + lid < rows)
         {
-            barrier(CLK_LOCAL_MEM_FENCE);
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
-            ai += GET_CONFLICT_OFFSET(ai);
-            bi += GET_CONFLICT_OFFSET(bi);
-
-            if((lid & 127) < d)
-            {
-                lm_sum[lid >> 7][bi]  +=  lm_sum[lid >> 7][ai];
-            }
-            offset <<= 1;
+            int sum_idx = mad24(lid + i, src_step, gid * 2);
+            src_t[0] = srcsum[sum_idx];
+            src_t[1] = srcsum[sum_idx + 1];
         }
-        barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid < 2)
+        else
         {
-            lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
+            src_t[0] = 0;
+            src_t[1] = 0;
         }
-        for(int d = 1;  d < LSIZE; d <<= 1)
+        if (i == 0)
         {
-            barrier(CLK_LOCAL_MEM_FENCE);
-            offset >>= 1;
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
-            ai += GET_CONFLICT_OFFSET(ai);
-            bi += GET_CONFLICT_OFFSET(bi);
-
-            if((lid & 127) < d)
-            {
-                lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
-                lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
-            }
+            sum_t[0] =  0;
+            sum_t[1] =  0;
         }
-        barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid > 0 && (i+lid) <= rows)
+        else
         {
-            int loc_s0 = gid * dst_step + i + lid - 1 - pre_invalid * dst_step / 4, loc_s1 = loc_s0 + dst_step ;
-            lm_sum[0][bf_loc] += sum_t[0];
-            lm_sum[1][bf_loc] += sum_t[1];
-            sum_p = (__local float*)(&(lm_sum[0][bf_loc]));
-            for(int k = 0; k < 4; k++)
-            {
-                if(gid * 4 + k >= cols + pre_invalid || gid * 4 + k < pre_invalid) continue;
-                sum[loc_s0 + k * dst_step / 4] = sum_p[k];
-            }
-            sum_p = (__local float*)(&(lm_sum[1][bf_loc]));
-            for(int k = 0; k < 4; k++)
-            {
-                if(gid * 4 + k + 4 >= cols + pre_invalid) break;
-                sum[loc_s1 + k * dst_step / 4] = sum_p[k];
-            }
+            sum_t[0] =  lm_sum[0][LSIZE_2 + LOG_LSIZE];
+            sum_t[1] =  lm_sum[1][LSIZE_2 + LOG_LSIZE];
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-    }
-}
-
-kernel void integral_sum_rows(__global float4 *srcsum, __global float *sum,
-                              int rows, int cols, int src_step, int sum_step, int sum_offset)
-{
-    int lid = get_local_id(0);
-    int gid = get_group_id(0);
-    float4 src_t[2], sum_t[2];
-    __local float4 lm_sum[2][LSIZE + LOG_LSIZE];
-    __local float *sum_p;
-    src_step = src_step >> 4;
-    for(int i = 0; i < rows; i =i + LSIZE_1)
-    {
-        src_t[0] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2] : (float4)0;
-        src_t[1] = i + lid < rows ? srcsum[(lid+i) * src_step + gid * 2 + 1] : (float4)0;
-
-        sum_t[0] =  (i == 0 ? (float4)0 : lm_sum[0][LSIZE_2 + LOG_LSIZE]);
-        sum_t[1] =  (i == 0 ? (float4)0 : lm_sum[1][LSIZE_2 + LOG_LSIZE]);
-        barrier(CLK_LOCAL_MEM_FENCE);
 
         int bf_loc = lid + GET_CONFLICT_OFFSET(lid);
-        lm_sum[0][bf_loc] = src_t[0];
 
+        lm_sum[0][bf_loc] = src_t[0];
         lm_sum[1][bf_loc] = src_t[1];
 
         int offset = 1;
-        for(int d = LSIZE >> 1 ;  d > 0; d>>=1)
+        for (int d = LSIZE >> 1 ;  d > 0; d>>=1)
         {
             barrier(CLK_LOCAL_MEM_FENCE);
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            int ai = offset * lid_prim - 1, bi = ai + offset;
             ai += GET_CONFLICT_OFFSET(ai);
             bi += GET_CONFLICT_OFFSET(bi);
 
@@ -355,59 +230,60 @@ kernel void integral_sum_rows(__global float4 *srcsum, __global float *sum,
             offset <<= 1;
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-        if(lid < 2)
+        if (lid < 2)
         {
             lm_sum[lid][LSIZE_2 + LOG_LSIZE] = 0;
         }
-        for(int d = 1;  d < LSIZE; d <<= 1)
+        for (int d = 1;  d < LSIZE; d <<= 1)
         {
             barrier(CLK_LOCAL_MEM_FENCE);
             offset >>= 1;
-            int ai = offset * (((lid & 127)<<1) +1) - 1,bi = ai + offset;
+            int ai = offset * lid_prim - 1,bi = ai + offset;
             ai += GET_CONFLICT_OFFSET(ai);
             bi += GET_CONFLICT_OFFSET(bi);
 
-            if((lid & 127) < d)
+            if ((lid & 127) < d)
             {
                 lm_sum[lid >> 7][bi] += lm_sum[lid >> 7][ai];
                 lm_sum[lid >> 7][ai] = lm_sum[lid >> 7][bi] - lm_sum[lid >> 7][ai];
             }
         }
         barrier(CLK_LOCAL_MEM_FENCE);
-        if(gid == 0 && (i + lid) <= rows)
+        if (gid == 0 && (i + lid) <= rows)
         {
             sum[sum_offset + i + lid] = 0;
         }
-        if(i + lid == 0)
+        if (i + lid == 0)
         {
             int loc0 = gid * 2 * sum_step;
             for(int k = 1; k <= 8; k++)
             {
-                if(gid * 8 + k > cols) break;
+                if (gid * 8 + k > cols)
+                    break;
                 sum[sum_offset + loc0 + k * sum_step / 4] = 0;
             }
         }
 
-        if(lid > 0 && (i+lid) <= rows)
+        if (lid > 0 && (i+lid) <= rows)
         {
             int loc_s0 = sum_offset + gid * 2 * sum_step + sum_step / 4 + i + lid, loc_s1 = loc_s0 + sum_step ;
             lm_sum[0][bf_loc] += sum_t[0];
             lm_sum[1][bf_loc] += sum_t[1];
-            sum_p = (__local float*)(&(lm_sum[0][bf_loc]));
+            sum_p = (__local sumT*)(&(lm_sum[0][bf_loc]));
             for(int k = 0; k < 4; k++)
             {
-                if(gid * 8 + k >= cols) break;
+                if (gid * 8 + k >= cols)
+                    break;
                 sum[loc_s0 + k * sum_step / 4] = sum_p[k];
             }
-            sum_p = (__local float*)(&(lm_sum[1][bf_loc]));
+            sum_p = (__local sumT*)(&(lm_sum[1][bf_loc]));
             for(int k = 0; k < 4; k++)
             {
-                if(gid * 8 + 4 + k >= cols) break;
+                if (gid * 8 + 4 + k >= cols)
+                    break;
                 sum[loc_s1 + k * sum_step / 4] = sum_p[k];
             }
         }
         barrier(CLK_LOCAL_MEM_FENCE);
     }
 }
-
-#endif

modules/imgproc/src/sumpixels.cpp

@@ -254,19 +254,19 @@ static bool ocl_integral( InputArray _src, OutputArray _sum, int sdepth )
     UMat src = _src.getUMat(), t_sum(t_size, sdepth), sum = _sum.getUMat();
     t_sum = t_sum(Range::all(), Range(0, size.height));
 
-    int offset = (int)src.offset / vlen, pre_invalid = (int)src.offset % vlen;
-    int vcols = (pre_invalid + src.cols + vlen - 1) / vlen;
+    int offset = (int)src.offset / vlen;
+    int vcols = (src.cols + vlen - 1) / vlen;
     int sum_offset = (int)sum.offset / vlen;
 
     k1.args(ocl::KernelArg::PtrReadOnly(src), ocl::KernelArg::PtrWriteOnly(t_sum),
-            offset, pre_invalid, src.rows, src.cols, (int)src.step, (int)t_sum.step);
+            offset, src.rows, src.cols, (int)src.step, (int)t_sum.step);
     size_t gt = ((vcols + 1) / 2) * 256, lt = 256;
     if (!k1.run(1, &gt, &lt, false))
         return false;
 
     ocl::Kernel k2("integral_sum_rows", ocl::imgproc::integral_sum_oclsrc,
                    format("-D sdepth=%d", sdepth));
-    k2.args(ocl::KernelArg::PtrReadWrite(t_sum), ocl::KernelArg::PtrWriteOnly(sum),
+    k2.args(ocl::KernelArg::PtrReadOnly(t_sum), ocl::KernelArg::PtrWriteOnly(sum),
             t_sum.rows, t_sum.cols, (int)t_sum.step, (int)sum.step, sum_offset);
 
     size_t gt2 = t_sum.cols  * 32, lt2 = 256;