diff --git a/libavcodec/faandct.c b/libavcodec/faandct.c
index e78d49b47c..2a91184953 100644
--- a/libavcodec/faandct.c
+++ b/libavcodec/faandct.c
@@ -70,16 +70,13 @@ B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7,
 B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
 };
 
-void ff_faandct(DCTELEM * data)
+static always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
 {
     FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
     FLOAT tmp10, tmp11, tmp12, tmp13;
     FLOAT z1, z2, z3, z4, z5, z11, z13;
-    FLOAT temp[64];
     int i;
 
-    emms_c();
-
     for (i=0; i<8*8; i+=8) {
         tmp0= data[0 + i] + data[7 + i];
         tmp7= data[0 + i] - data[7 + i];
@@ -118,7 +115,20 @@ void ff_faandct(DCTELEM * data)
         temp[3 + i]= z13 - z2;
         temp[1 + i]= z11 + z4;
         temp[7 + i]= z11 - z4;
-    }
+    }    
+}
+
+void ff_faandct(DCTELEM * data)
+{
+    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+    FLOAT tmp10, tmp11, tmp12, tmp13;
+    FLOAT z1, z2, z3, z4, z5, z11, z13;
+    FLOAT temp[64];
+    int i;
+
+    emms_c();
+
+    row_fdct(temp, data);
 
     for (i=0; i<8; i++) {
         tmp0= temp[8*0 + i] + temp[8*7 + i];
@@ -165,51 +175,13 @@ void ff_faandct248(DCTELEM * data)
 {
     FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
     FLOAT tmp10, tmp11, tmp12, tmp13;
-    FLOAT z1, z2, z3, z4, z5, z11, z13;
+    FLOAT z1;
     FLOAT temp[64];
     int i;
 
     emms_c();
 
-    for (i=0; i<8*8; i+=8) {
-        tmp0= data[0 + i] + data[7 + i];
-        tmp7= data[0 + i] - data[7 + i];
-        tmp1= data[1 + i] + data[6 + i];
-        tmp6= data[1 + i] - data[6 + i];
-        tmp2= data[2 + i] + data[5 + i];
-        tmp5= data[2 + i] - data[5 + i];
-        tmp3= data[3 + i] + data[4 + i];
-        tmp4= data[3 + i] - data[4 + i];
-        
-        tmp10= tmp0 + tmp3;
-        tmp13= tmp0 - tmp3;
-        tmp11= tmp1 + tmp2;
-        tmp12= tmp1 - tmp2;
-        
-        temp[0 + i]= tmp10 + tmp11;
-        temp[4 + i]= tmp10 - tmp11;
-        
-        z1= (tmp12 + tmp13)*A1;
-        temp[2 + i]= tmp13 + z1;
-        temp[6 + i]= tmp13 - z1;
-        
-        tmp10= tmp4 + tmp5;
-        tmp11= tmp5 + tmp6;
-        tmp12= tmp6 + tmp7;
-
-        z5= (tmp10 - tmp12) * A5;
-        z2= tmp10*A2 + z5;
-        z4= tmp12*A4 + z5;
-        z3= tmp11*A1;
-
-        z11= tmp7 + z3;
-        z13= tmp7 - z3;
-
-        temp[5 + i]= z13 + z2;
-        temp[3 + i]= z13 - z2;
-        temp[1 + i]= z11 + z4;
-        temp[7 + i]= z11 - z4;
-    }
+    row_fdct(temp, data);
 
     for (i=0; i<8; i++) {
         tmp0 = temp[8*0 + i] + temp[8*1 + i];
diff --git a/libavcodec/jfdctfst.c b/libavcodec/jfdctfst.c
index acaf284cd2..a393c5ca81 100644
--- a/libavcodec/jfdctfst.c
+++ b/libavcodec/jfdctfst.c
@@ -112,17 +112,10 @@
 
 #define MULTIPLY(var,const)  ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
 
-
-/*
- * Perform the forward DCT on one block of samples.
- */
-
-GLOBAL(void)
-fdct_ifast (DCTELEM * data)
-{
-  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
-  DCTELEM tmp10, tmp11, tmp12, tmp13;
-  DCTELEM z1, z2, z3, z4, z5, z11, z13;
+static always_inline void row_fdct(DCTELEM * data){
+  int_fast16_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  int_fast16_t tmp10, tmp11, tmp12, tmp13;
+  int_fast16_t z1, z2, z3, z4, z5, z11, z13;
   DCTELEM *dataptr;
   int ctr;
   SHIFT_TEMPS
@@ -176,7 +169,24 @@ fdct_ifast (DCTELEM * data)
 
     dataptr += DCTSIZE;		/* advance pointer to next row */
   }
+}
 
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+fdct_ifast (DCTELEM * data)
+{
+  int_fast16_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  int_fast16_t tmp10, tmp11, tmp12, tmp13;
+  int_fast16_t z1, z2, z3, z4, z5, z11, z13;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  row_fdct(data);
+  
   /* Pass 2: process columns. */
 
   dataptr = data;
@@ -235,63 +245,15 @@ fdct_ifast (DCTELEM * data)
 GLOBAL(void)
 fdct_ifast248 (DCTELEM * data)
 {
-  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
-  DCTELEM tmp10, tmp11, tmp12, tmp13;
-  DCTELEM z1, z2, z3, z4, z5, z11, z13;
+  int_fast16_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  int_fast16_t tmp10, tmp11, tmp12, tmp13;
+  int_fast16_t z1;
   DCTELEM *dataptr;
   int ctr;
   SHIFT_TEMPS
 
-  /* Pass 1: process rows. */
-
-  dataptr = data;
-  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
-    tmp0 = dataptr[0] + dataptr[7];
-    tmp7 = dataptr[0] - dataptr[7];
-    tmp1 = dataptr[1] + dataptr[6];
-    tmp6 = dataptr[1] - dataptr[6];
-    tmp2 = dataptr[2] + dataptr[5];
-    tmp5 = dataptr[2] - dataptr[5];
-    tmp3 = dataptr[3] + dataptr[4];
-    tmp4 = dataptr[3] - dataptr[4];
-    
-    /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
-    tmp13 = tmp0 - tmp3;
-    tmp11 = tmp1 + tmp2;
-    tmp12 = tmp1 - tmp2;
-    
-    dataptr[0] = tmp10 + tmp11; /* phase 3 */
-    dataptr[4] = tmp10 - tmp11;
-    
-    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
-    dataptr[2] = tmp13 + z1;	/* phase 5 */
-    dataptr[6] = tmp13 - z1;
-    
-    /* Odd part */
-
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
-    tmp11 = tmp5 + tmp6;
-    tmp12 = tmp6 + tmp7;
-
-    /* The rotator is modified from fig 4-8 to avoid extra negations. */
-    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
-    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
-    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
-    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
-
-    z11 = tmp7 + z3;		/* phase 5 */
-    z13 = tmp7 - z3;
-
-    dataptr[5] = z13 + z2;	/* phase 6 */
-    dataptr[3] = z13 - z2;
-    dataptr[1] = z11 + z4;
-    dataptr[7] = z11 - z4;
-
-    dataptr += DCTSIZE;		/* advance pointer to next row */
-  }
-
+  row_fdct(data);
+   
   /* Pass 2: process columns. */
 
   dataptr = data;
diff --git a/libavcodec/jfdctint.c b/libavcodec/jfdctint.c
index d5797bddff..1fbd85b28c 100644
--- a/libavcodec/jfdctint.c
+++ b/libavcodec/jfdctint.c
@@ -148,16 +148,10 @@
 #endif
 
 
-/*
- * Perform the forward DCT on one block of samples.
- */
-
-GLOBAL(void)
-ff_jpeg_fdct_islow (DCTELEM * data)
-{
-  int32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
-  int32_t tmp10, tmp11, tmp12, tmp13;
-  int32_t z1, z2, z3, z4, z5;
+static always_inline void row_fdct(DCTELEM * data){
+  int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  int_fast32_t tmp10, tmp11, tmp12, tmp13;
+  int_fast32_t z1, z2, z3, z4, z5;
   DCTELEM *dataptr;
   int ctr;
   SHIFT_TEMPS
@@ -225,6 +219,23 @@ ff_jpeg_fdct_islow (DCTELEM * data)
     
     dataptr += DCTSIZE;		/* advance pointer to next row */
   }
+}
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+ff_jpeg_fdct_islow (DCTELEM * data)
+{
+  int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  int_fast32_t tmp10, tmp11, tmp12, tmp13;
+  int_fast32_t z1, z2, z3, z4, z5;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  row_fdct(data);
 
   /* Pass 2: process columns.
    * We remove the PASS1_BITS scaling, but leave the results scaled up
@@ -304,76 +315,14 @@ ff_jpeg_fdct_islow (DCTELEM * data)
 GLOBAL(void)
 ff_fdct248_islow (DCTELEM * data)
 {
-  int32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
-  int32_t tmp10, tmp11, tmp12, tmp13;
-  int32_t z1, z2, z3, z4, z5;
+  int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  int_fast32_t tmp10, tmp11, tmp12, tmp13;
+  int_fast32_t z1;
   DCTELEM *dataptr;
   int ctr;
   SHIFT_TEMPS
 
-  /* Pass 1: process rows. */
-  /* Note results are scaled up by sqrt(8) compared to a true DCT; */
-  /* furthermore, we scale the results by 2**PASS1_BITS. */
-
-  dataptr = data;
-  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
-    tmp0 = dataptr[0] + dataptr[7];
-    tmp7 = dataptr[0] - dataptr[7];
-    tmp1 = dataptr[1] + dataptr[6];
-    tmp6 = dataptr[1] - dataptr[6];
-    tmp2 = dataptr[2] + dataptr[5];
-    tmp5 = dataptr[2] - dataptr[5];
-    tmp3 = dataptr[3] + dataptr[4];
-    tmp4 = dataptr[3] - dataptr[4];
-    
-    /* Even part per LL&M figure 1 --- note that published figure is faulty;
-     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
-     */
-    
-    tmp10 = tmp0 + tmp3;
-    tmp13 = tmp0 - tmp3;
-    tmp11 = tmp1 + tmp2;
-    tmp12 = tmp1 - tmp2;
-    
-    dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
-    dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
-    
-    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
-    dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
-				   CONST_BITS-PASS1_BITS);
-    dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
-				   CONST_BITS-PASS1_BITS);
-    
-    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
-     * cK represents cos(K*pi/16).
-     * i0..i3 in the paper are tmp4..tmp7 here.
-     */
-    
-    z1 = tmp4 + tmp7;
-    z2 = tmp5 + tmp6;
-    z3 = tmp4 + tmp6;
-    z4 = tmp5 + tmp7;
-    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
-    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
-    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
-    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
-    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
-    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
-    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
-    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
-    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
-    z3 += z5;
-    z4 += z5;
-    
-    dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
-    dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
-    dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
-    dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
-    
-    dataptr += DCTSIZE;		/* advance pointer to next row */
-  }
+  row_fdct(data);
 
   /* Pass 2: process columns.
    * We remove the PASS1_BITS scaling, but leave the results scaled up