diff --git a/vp9/common/vp9_findnearmv.c b/vp9/common/vp9_findnearmv.c
index ed8a99af2235a75f7ad4be598efd048123db16ea..02c28fbdee0298ad1629492670539311f4c395e0 100644
--- a/vp9/common/vp9_findnearmv.c
+++ b/vp9/common/vp9_findnearmv.c
@@ -136,7 +136,9 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
unsigned char *above_ref;
unsigned char *left_ref;
unsigned int score;
+#if CONFIG_SUBPELREFMV
unsigned int sse;
+#endif
unsigned int ref_scores[MAX_MV_REFS] = {0};
int_mv sorted_mvs[MAX_MV_REFS];
int zero_seen = FALSE;
diff --git a/vp9/common/vp9_idctllm.c b/vp9/common/vp9_idctllm.c
index 9d559ed228086d4c5b0aae0a29cb91d93c53fc9d..9622dfdee1335b596afa7b88f53b60b77b630903 100644
--- a/vp9/common/vp9_idctllm.c
+++ b/vp9/common/vp9_idctllm.c
@@ -279,14 +279,6 @@ static const int16_t iadst_i16[256] = {
3936, -3526, 3084, -2614, 2120, -1607, 1080, -542
};
-/* For test */
-#define TEST_INT 1
-#if TEST_INT
-#define vp9_ihtllm_int_c vp9_ihtllm_c
-#else
-#define vp9_ihtllm_float_c vp9_ihtllm_c
-#endif
-
void vp9_ihtllm_float_c(const int16_t *input, int16_t *output, int pitch,
TX_TYPE tx_type, int tx_dim) {
vp9_clear_system_state(); // Make it simd safe : __asm emms;
@@ -411,7 +403,7 @@ void vp9_ihtllm_float_c(const int16_t *input, int16_t *output, int pitch,
#define VERTICAL_ROUNDING ((1 << (VERTICAL_SHIFT - 1)) - 1)
#define HORIZONTAL_SHIFT 17 // 15
#define HORIZONTAL_ROUNDING ((1 << (HORIZONTAL_SHIFT - 1)) - 1)
-void vp9_ihtllm_int_c(const int16_t *input, int16_t *output, int pitch,
+void vp9_ihtllm_c(const int16_t *input, int16_t *output, int pitch,
TX_TYPE tx_type, int tx_dim) {
int i, j, k;
int16_t imbuf[256];
diff --git a/vp9/common/vp9_invtrans.c b/vp9/common/vp9_invtrans.c
index 627a62bf74c942e2f23574f85df5dbb15ba29cb3..c78f1ad3c6ab26801d54627c7ad206e08787922e 100644
--- a/vp9/common/vp9_invtrans.c
+++ b/vp9/common/vp9_invtrans.c
@@ -51,7 +51,7 @@ void vp9_inverse_transform_mby_4x4(MACROBLOCKD *xd) {
for (i = 0; i < 16; i++) {
TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[i]);
if (tx_type != DCT_DCT) {
- vp9_ihtllm_c(xd->block[i].dqcoeff, xd->block[i].diff, 32,
+ vp9_ihtllm(xd->block[i].dqcoeff, xd->block[i].diff, 32,
tx_type, 4);
} else {
vp9_inverse_transform_b_4x4(xd, i, 32);
@@ -91,7 +91,7 @@ void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd) {
for (i = 0; i < 9; i += 8) {
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[i]);
if (tx_type != DCT_DCT) {
- vp9_ihtllm_c(xd->block[i].dqcoeff, xd->block[i].diff, 32, tx_type, 8);
+ vp9_ihtllm(xd->block[i].dqcoeff, xd->block[i].diff, 32, tx_type, 8);
} else {
vp9_inverse_transform_b_8x8(&blockd[i].dqcoeff[0],
&blockd[i].diff[0], 32);
@@ -100,7 +100,7 @@ void vp9_inverse_transform_mby_8x8(MACROBLOCKD *xd) {
for (i = 2; i < 11; i += 8) {
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[i]);
if (tx_type != DCT_DCT) {
- vp9_ihtllm_c(xd->block[i + 2].dqcoeff, xd->block[i].diff, 32, tx_type, 8);
+ vp9_ihtllm(xd->block[i + 2].dqcoeff, xd->block[i].diff, 32, tx_type, 8);
} else {
vp9_inverse_transform_b_8x8(&blockd[i + 2].dqcoeff[0],
&blockd[i].diff[0], 32);
@@ -132,7 +132,7 @@ void vp9_inverse_transform_mby_16x16(MACROBLOCKD *xd) {
BLOCKD *bd = &xd->block[0];
TX_TYPE tx_type = get_tx_type_16x16(xd, bd);
if (tx_type != DCT_DCT) {
- vp9_ihtllm_c(bd->dqcoeff, bd->diff, 32, tx_type, 16);
+ vp9_ihtllm(bd->dqcoeff, bd->diff, 32, tx_type, 16);
} else {
vp9_inverse_transform_b_16x16(&xd->block[0].dqcoeff[0],
&xd->block[0].diff[0], 32);
diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
index ee9a66cdca7208f59ab97f1356d46e554031ae1f..810b693714bbc439e68635dc8a2e8ab80bbe05b6 100644
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -366,6 +366,9 @@ specialize vp9_short_idct16x16
prototype void vp9_short_idct10_16x16 "short *input, short *output, int pitch"
specialize vp9_short_idct10_16x16
+prototype void vp9_ihtllm "const short *input, short *output, int pitch, int tx_type, int tx_dim"
+specialize vp9_ihtllm
+
#
# 2nd order
#
diff --git a/vp9/decoder/vp9_dequantize.c b/vp9/decoder/vp9_dequantize.c
index b119418e8a42314d877f7fbd5139cbf95c247de6..79114d58ca1a3cfb2e9ee426f5d43887f7410650 100644
--- a/vp9/decoder/vp9_dequantize.c
+++ b/vp9/decoder/vp9_dequantize.c
@@ -83,7 +83,7 @@ void vp9_ht_dequant_idct_add_c(TX_TYPE tx_type, int16_t *input,
input[i] = dq[i] * input[i];
}
- vp9_ihtllm_c(input, output, 4 << 1, tx_type, 4);
+ vp9_ihtllm(input, output, 4 << 1, tx_type, 4);
vpx_memset(input, 0, 32);
@@ -103,7 +103,7 @@ void vp9_ht_dequant_idct_add_8x8_c(TX_TYPE tx_type, int16_t *input,
input[i] = dq[1] * input[i];
}
- vp9_ihtllm_c(input, output, 16, tx_type, 8);
+ vp9_ihtllm(input, output, 16, tx_type, 8);
vpx_memset(input, 0, 128);
@@ -281,7 +281,7 @@ void vp9_ht_dequant_idct_add_16x16_c(TX_TYPE tx_type, int16_t *input,
input[i] = input[i] * dq[1];
// inverse hybrid transform
- vp9_ihtllm_c(input, output, 32, tx_type, 16);
+ vp9_ihtllm(input, output, 32, tx_type, 16);
// the idct halves ( >> 1) the pitch
// vp9_short_idct16x16_c(input, output, 32);
diff --git a/vp9/encoder/vp9_encodeintra.c b/vp9/encoder/vp9_encodeintra.c
index 17a3a952b6f4d561b2d6cb4da9973f1c8a282bfd..466ce16565d92ff2812e4185dd9fedc13bbe8b6b 100644
--- a/vp9/encoder/vp9_encodeintra.c
+++ b/vp9/encoder/vp9_encodeintra.c
@@ -70,7 +70,7 @@ void vp9_encode_intra4x4block(MACROBLOCK *x, int ib) {
if (tx_type != DCT_DCT) {
vp9_fht(be->src_diff, 32, be->coeff, tx_type, 4);
vp9_ht_quantize_b_4x4(be, b, tx_type);
- vp9_ihtllm_c(b->dqcoeff, b->diff, 32, tx_type, 4);
+ vp9_ihtllm(b->dqcoeff, b->diff, 32, tx_type, 4);
} else {
x->vp9_short_fdct4x4(be->src_diff, be->coeff, 32);
x->quantize_b_4x4(be, b) ;
@@ -191,7 +191,7 @@ void vp9_encode_intra8x8(MACROBLOCK *x, int ib) {
vp9_fht(be->src_diff, 32, (x->block + idx)->coeff,
tx_type, 8);
x->quantize_b_8x8(x->block + idx, xd->block + idx);
- vp9_ihtllm_c(xd->block[idx].dqcoeff, xd->block[ib].diff, 32,
+ vp9_ihtllm(xd->block[idx].dqcoeff, xd->block[ib].diff, 32,
tx_type, 8);
} else {
x->vp9_short_fdct8x8(be->src_diff, (x->block + idx)->coeff, 32);
@@ -206,7 +206,7 @@ void vp9_encode_intra8x8(MACROBLOCK *x, int ib) {
if (tx_type != DCT_DCT) {
vp9_fht_c(be->src_diff, 32, be->coeff, tx_type, 4);
vp9_ht_quantize_b_4x4(be, b, tx_type);
- vp9_ihtllm_c(b->dqcoeff, b->diff, 32, tx_type, 4);
+ vp9_ihtllm(b->dqcoeff, b->diff, 32, tx_type, 4);
} else {
x->vp9_short_fdct4x4(be->src_diff, be->coeff, 32);
x->quantize_b_4x4(be, b);
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index ce9a8ad62b137db197125de9a160b6a7a9eeec19..9004511fc93b755905da7290281ccc80c4746de7 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1125,7 +1125,7 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, BLOCK *be,
// inverse transform
if (best_tx_type != DCT_DCT)
- vp9_ihtllm_c(best_dqcoeff, b->diff, 32, best_tx_type, 4);
+ vp9_ihtllm(best_dqcoeff, b->diff, 32, best_tx_type, 4);
else
xd->inv_xform4x4_x8(best_dqcoeff, b->diff, 32);