diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index 22d827fe3e30c3b8f264e22e6bc4ddbbd4f9fe2a..dcdefa45e8fc2b37ac7f3808430fbdacff19014a 100644
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -699,7 +699,7 @@ if (vpx_config("CONFIG_VPX_HIGHBITDEPTH") eq "yes") {
specialize qw/vpx_iwht4x4_1_add/;
add_proto qw/void vpx_iwht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vpx_iwht4x4_16_add/;
+ specialize qw/vpx_iwht4x4_16_add/, "$sse2_x86inc";
add_proto qw/void vpx_highbd_idct4x4_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct4x4_1_add/;
@@ -762,7 +762,7 @@ if (vpx_config("CONFIG_VPX_HIGHBITDEPTH") eq "yes") {
add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1_add/;
-
+
add_proto qw/void vpx_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
specialize qw/vpx_highbd_idct4x4_16_add/;
@@ -785,10 +785,10 @@ if (vpx_config("CONFIG_VPX_HIGHBITDEPTH") eq "yes") {
specialize qw/vpx_idct4x4_1_add sse2/;
add_proto qw/void vpx_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vpx_idct8x8_64_add sse2/;
+ specialize qw/vpx_idct8x8_64_add sse2/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vpx_idct8x8_12_add sse2/;
+ specialize qw/vpx_idct8x8_12_add sse2/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct8x8_1_add sse2/;
@@ -803,14 +803,15 @@ if (vpx_config("CONFIG_VPX_HIGHBITDEPTH") eq "yes") {
specialize qw/vpx_idct16x16_1_add sse2/;
add_proto qw/void vpx_idct32x32_1024_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vpx_idct32x32_1024_add sse2/;
+ specialize qw/vpx_idct32x32_1024_add sse2/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_idct32x32_135_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vpx_idct32x32_135_add sse2/;
+ specialize qw/vpx_idct32x32_135_add sse2/, "$ssse3_x86_64_x86inc";
+ # Need to add 135 eob idct32x32 implementations.
$vpx_idct32x32_135_add_sse2=vpx_idct32x32_1024_add_sse2;
add_proto qw/void vpx_idct32x32_34_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
- specialize qw/vpx_idct32x32_34_add sse2/;
+ specialize qw/vpx_idct32x32_34_add sse2/, "$ssse3_x86_64_x86inc";
add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
specialize qw/vpx_idct32x32_1_add sse2/;
diff --git a/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm b/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
index 80a330b7b9bad33590a48526968a63d4c934d775..a835161ea02de3d2a9db9bcf5a72392402239091 100644
--- a/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
+++ b/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
@@ -220,7 +220,24 @@ cglobal idct8x8_64_add, 3, 5, 13, input, output, stride
mova m12, [pw_11585x2]
lea r3, [2 * strideq]
-
+%if CONFIG_VPX_HIGHBITDEPTH
+ mova m0, [inputq + 0]
+ packssdw m0, [inputq + 16]
+ mova m1, [inputq + 32]
+ packssdw m1, [inputq + 48]
+ mova m2, [inputq + 64]
+ packssdw m2, [inputq + 80]
+ mova m3, [inputq + 96]
+ packssdw m3, [inputq + 112]
+ mova m4, [inputq + 128]
+ packssdw m4, [inputq + 144]
+ mova m5, [inputq + 160]
+ packssdw m5, [inputq + 176]
+ mova m6, [inputq + 192]
+ packssdw m6, [inputq + 208]
+ mova m7, [inputq + 224]
+ packssdw m7, [inputq + 240]
+%else
mova m0, [inputq + 0]
mova m1, [inputq + 16]
mova m2, [inputq + 32]
@@ -229,7 +246,7 @@ cglobal idct8x8_64_add, 3, 5, 13, input, output, stride
mova m5, [inputq + 80]
mova m6, [inputq + 96]
mova m7, [inputq + 112]
-
+%endif
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
IDCT8_1D
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
@@ -254,10 +271,21 @@ cglobal idct8x8_12_add, 3, 5, 13, input, output, stride
lea r3, [2 * strideq]
+%if CONFIG_VPX_HIGHBITDEPTH
+ mova m0, [inputq + 0]
+ packssdw m0, [inputq + 16]
+ mova m1, [inputq + 32]
+ packssdw m1, [inputq + 48]
+ mova m2, [inputq + 64]
+ packssdw m2, [inputq + 80]
+ mova m3, [inputq + 96]
+ packssdw m3, [inputq + 112]
+%else
mova m0, [inputq + 0]
mova m1, [inputq + 16]
mova m2, [inputq + 32]
mova m3, [inputq + 48]
+%endif
punpcklwd m0, m1
punpcklwd m2, m3
@@ -765,6 +793,24 @@ idct32x32_34:
lea r4, [rsp + transposed_in]
idct32x32_34_transpose:
+%if CONFIG_VPX_HIGHBITDEPTH
+ mova m0, [r3 + 0]
+ packssdw m0, [r3 + 16]
+ mova m1, [r3 + 32 * 4]
+ packssdw m1, [r3 + 32 * 4 + 16]
+ mova m2, [r3 + 32 * 8]
+ packssdw m2, [r3 + 32 * 8 + 16]
+ mova m3, [r3 + 32 * 12]
+ packssdw m3, [r3 + 32 * 12 + 16]
+ mova m4, [r3 + 32 * 16]
+ packssdw m4, [r3 + 32 * 16 + 16]
+ mova m5, [r3 + 32 * 20]
+ packssdw m5, [r3 + 32 * 20 + 16]
+ mova m6, [r3 + 32 * 24]
+ packssdw m6, [r3 + 32 * 24 + 16]
+ mova m7, [r3 + 32 * 28]
+ packssdw m7, [r3 + 32 * 28 + 16]
+%else
mova m0, [r3 + 0]
mova m1, [r3 + 16 * 4]
mova m2, [r3 + 16 * 8]
@@ -773,6 +819,7 @@ idct32x32_34_transpose:
mova m5, [r3 + 16 * 20]
mova m6, [r3 + 16 * 24]
mova m7, [r3 + 16 * 28]
+%endif
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
@@ -1176,6 +1223,24 @@ idct32x32_135:
mov r7, 2
idct32x32_135_transpose:
+%if CONFIG_VPX_HIGHBITDEPTH
+ mova m0, [r3 + 0]
+ packssdw m0, [r3 + 16]
+ mova m1, [r3 + 32 * 4]
+ packssdw m1, [r3 + 32 * 4 + 16]
+ mova m2, [r3 + 32 * 8]
+ packssdw m2, [r3 + 32 * 8 + 16]
+ mova m3, [r3 + 32 * 12]
+ packssdw m3, [r3 + 32 * 12 + 16]
+ mova m4, [r3 + 32 * 16]
+ packssdw m4, [r3 + 32 * 16 + 16]
+ mova m5, [r3 + 32 * 20]
+ packssdw m5, [r3 + 32 * 20 + 16]
+ mova m6, [r3 + 32 * 24]
+ packssdw m6, [r3 + 32 * 24 + 16]
+ mova m7, [r3 + 32 * 28]
+ packssdw m7, [r3 + 32 * 28 + 16]
+%else
mova m0, [r3 + 0]
mova m1, [r3 + 16 * 4]
mova m2, [r3 + 16 * 8]
@@ -1184,7 +1249,7 @@ idct32x32_135_transpose:
mova m5, [r3 + 16 * 20]
mova m6, [r3 + 16 * 24]
mova m7, [r3 + 16 * 28]
-
+%endif
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
mova [r4 + 0], m0
@@ -1196,14 +1261,22 @@ idct32x32_135_transpose:
mova [r4 + 16 * 6], m6
mova [r4 + 16 * 7], m7
+%if CONFIG_VPX_HIGHBITDEPTH
+ add r3, 32
+%else
add r3, 16
+%endif
add r4, 16 * 8
dec r7
jne idct32x32_135_transpose
IDCT32X32_135 16*0, 16*32, 16*64, 16*96
lea stp, [stp + 16 * 8]
+%if CONFIG_VPX_HIGHBITDEPTH
+ lea inputq, [inputq + 32 * 32]
+%else
lea inputq, [inputq + 16 * 32]
+%endif
dec r6
jnz idct32x32_135
@@ -1614,6 +1687,24 @@ idct32x32_1024:
mov r7, 4
idct32x32_1024_transpose:
+%if CONFIG_VPX_HIGHBITDEPTH
+ mova m0, [r3 + 0]
+ packssdw m0, [r3 + 16]
+ mova m1, [r3 + 32 * 4]
+ packssdw m1, [r3 + 32 * 4 + 16]
+ mova m2, [r3 + 32 * 8]
+ packssdw m2, [r3 + 32 * 8 + 16]
+ mova m3, [r3 + 32 * 12]
+ packssdw m3, [r3 + 32 * 12 + 16]
+ mova m4, [r3 + 32 * 16]
+ packssdw m4, [r3 + 32 * 16 + 16]
+ mova m5, [r3 + 32 * 20]
+ packssdw m5, [r3 + 32 * 20 + 16]
+ mova m6, [r3 + 32 * 24]
+ packssdw m6, [r3 + 32 * 24 + 16]
+ mova m7, [r3 + 32 * 28]
+ packssdw m7, [r3 + 32 * 28 + 16]
+%else
mova m0, [r3 + 0]
mova m1, [r3 + 16 * 4]
mova m2, [r3 + 16 * 8]
@@ -1622,6 +1713,7 @@ idct32x32_1024_transpose:
mova m5, [r3 + 16 * 20]
mova m6, [r3 + 16 * 24]
mova m7, [r3 + 16 * 28]
+%endif
TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9
@@ -1633,8 +1725,11 @@ idct32x32_1024_transpose:
mova [r4 + 16 * 5], m5
mova [r4 + 16 * 6], m6
mova [r4 + 16 * 7], m7
-
+%if CONFIG_VPX_HIGHBITDEPTH
+ add r3, 32
+%else
add r3, 16
+%endif
add r4, 16 * 8
dec r7
jne idct32x32_1024_transpose
@@ -1642,7 +1737,11 @@ idct32x32_1024_transpose:
IDCT32X32_1024 16*0, 16*32, 16*64, 16*96
lea stp, [stp + 16 * 8]
+%if CONFIG_VPX_HIGHBITDEPTH
+ lea inputq, [inputq + 32 * 32]
+%else
lea inputq, [inputq + 16 * 32]
+%endif
dec r6
jnz idct32x32_1024
diff --git a/vpx_dsp/x86/inv_wht_sse2.asm b/vpx_dsp/x86/inv_wht_sse2.asm
index df6f4692b48dcc07f43e46330641f792625e12db..eec504755016fb229a4cb97dd89f23d6f4b143bd 100644
--- a/vpx_dsp/x86/inv_wht_sse2.asm
+++ b/vpx_dsp/x86/inv_wht_sse2.asm
@@ -82,9 +82,15 @@ SECTION .text
INIT_XMM sse2
cglobal iwht4x4_16_add, 3, 3, 7, input, output, stride
+%if CONFIG_VPX_HIGHBITDEPTH
+ mova m0, [inputq + 0]
+ packssdw m0, [inputq + 16]
+ mova m1, [inputq + 32]
+ packssdw m1, [inputq + 48]
+%else
mova m0, [inputq + 0]
mova m1, [inputq + 16]
-
+%endif
psraw m0, 2
psraw m1, 2