Rewrite vpx_idct16x16_{10,256}_add_sse2() and add case 38 function

BUG=webm:1412

Change-Id: I945f0fb6807b8948747243794dc7352b959221f7

test/partial_idct_test.cc

@@ -651,6 +651,15 @@ const PartialInvTxfmParam sse2_partial_idct_tests[] = {
   make_tuple(
       &vpx_highbd_fdct16x16_c, &highbd_wrapper<vpx_highbd_idct16x16_256_add_c>,
       &highbd_wrapper<vpx_highbd_idct16x16_256_add_sse2>, TX_16X16, 256, 12, 2),
+  make_tuple(
+      &vpx_highbd_fdct16x16_c, &highbd_wrapper<vpx_highbd_idct16x16_38_add_c>,
+      &highbd_wrapper<vpx_highbd_idct16x16_38_add_sse2>, TX_16X16, 38, 8, 2),
+  make_tuple(
+      &vpx_highbd_fdct16x16_c, &highbd_wrapper<vpx_highbd_idct16x16_38_add_c>,
+      &highbd_wrapper<vpx_highbd_idct16x16_38_add_sse2>, TX_16X16, 38, 10, 2),
+  make_tuple(
+      &vpx_highbd_fdct16x16_c, &highbd_wrapper<vpx_highbd_idct16x16_38_add_c>,
+      &highbd_wrapper<vpx_highbd_idct16x16_38_add_sse2>, TX_16X16, 38, 12, 2),
   make_tuple(
       &vpx_highbd_fdct16x16_c, &highbd_wrapper<vpx_highbd_idct16x16_10_add_c>,
       &highbd_wrapper<vpx_highbd_idct16x16_10_add_sse2>, TX_16X16, 10, 8, 2),

vpx_dsp/vpx_dsp_rtcd_defs.pl

@@ -658,7 +658,6 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
     specialize qw/vpx_highbd_idct8x8_12_add neon sse2 sse4_1/;
     specialize qw/vpx_highbd_idct16x16_256_add neon sse2/;
     specialize qw/vpx_highbd_idct16x16_38_add neon sse2/;
-    $vpx_highbd_idct16x16_38_add_sse2=vpx_highbd_idct16x16_256_add_sse2;
     specialize qw/vpx_highbd_idct16x16_10_add neon sse2/;
     specialize qw/vpx_highbd_idct32x32_1024_add neon/;
     specialize qw/vpx_highbd_idct32x32_135_add neon/;

vpx_dsp/x86/highbd_inv_txfm_sse2.h

@@ -115,7 +115,7 @@ static INLINE void highbd_multiplication_and_add_sse2(
     __m128i *const out0, __m128i *const out1) {
   const __m128i pair_c0 = pair_set_epi32(c0 << 2, 0);
   const __m128i pair_c1 = pair_set_epi32(c1 << 2, 0);
-  __m128i temp1[4], temp2[4], sign1[4], sign2[4];
+  __m128i temp1[4], temp2[4], sign1[2], sign2[2];
 
   abs_extend_64bit_sse2(in0, temp1, sign1);
   abs_extend_64bit_sse2(in1, temp2, sign2);
@@ -139,6 +139,29 @@ static INLINE void highbd_multiplication_and_add_sse2(
   *out1 = pack_4(temp2[0], temp2[1]);
 }
 
+// Note: c0 and c1 must be non negative.
+static INLINE void highbd_multiplication_sse2(const __m128i in, const int c0,
+                                              const int c1, __m128i *const out0,
+                                              __m128i *const out1) {
+  __m128i temp[2], sign[2];
+
+  abs_extend_64bit_sse2(in, temp, sign);
+  *out0 = multiplication_round_shift_sse2(temp, sign, c0);
+  *out1 = multiplication_round_shift_sse2(temp, sign, c1);
+}
+
+// Note: c0 and c1 must be non negative.
+static INLINE void highbd_multiplication_neg_sse2(const __m128i in,
+                                                  const int c0, const int c1,
+                                                  __m128i *const out0,
+                                                  __m128i *const out1) {
+  __m128i temp[2], sign[2];
+
+  abs_extend_64bit_sse2(in, temp, sign);
+  *out0 = multiplication_neg_round_shift_sse2(temp, sign, c1);
+  *out1 = multiplication_round_shift_sse2(temp, sign, c0);
+}
+
 static INLINE void highbd_idct8_stage4(const __m128i *const in,
                                        __m128i *const out) {
   out[0] = _mm_add_epi32(in[0], in[7]);