hevc: Add NEON add_residual for bitdepth 10

Signed-off-by: Martin Storsjö <martin@martin.st>

libavcodec/arm/hevc_idct.S

@@ -30,6 +30,13 @@ const trans, align=4
         .short 57, 43, 25, 9
 endconst
 
+.macro clip10 in1, in2, c1, c2
+        vmax.s16        \in1, \in1, \c1
+        vmax.s16        \in2, \in2, \c1
+        vmin.s16        \in1, \in1, \c2
+        vmin.s16        \in2, \in2, \c2
+.endm
+
 function ff_hevc_add_residual_4x4_8_neon, export=1
         vld1.16         {q0-q1}, [r1, :128]
         vld1.32         d4[0], [r0, :32], r2
@@ -50,6 +57,25 @@ function ff_hevc_add_residual_4x4_8_neon, export=1
         bx              lr
 endfunc
 
+function ff_hevc_add_residual_4x4_10_neon, export=1
+        mov             r12, r0
+        vld1.16         {q0-q1}, [r1, :128]
+        vld1.16         d4, [r12, :64], r2
+        vld1.16         d5, [r12, :64], r2
+        vld1.16         d6, [r12, :64], r2
+        vqadd.s16       q0, q2
+        vld1.16         d7, [r12, :64], r2
+        vmov.s16        q12, #0
+        vqadd.s16       q1, q3
+        vmvn.s16        q13, #0xFC00 @ vmov.s16 #0x3FF
+        clip10          q0, q1, q12, q13
+        vst1.16         d0, [r0, :64], r2
+        vst1.16         d1, [r0, :64], r2
+        vst1.16         d2, [r0, :64], r2
+        vst1.16         d3, [r0, :64], r2
+        bx              lr
+endfunc
+
 function ff_hevc_add_residual_8x8_8_neon, export=1
         add             r12, r0, r2
         add             r2,  r2, r2
@@ -70,6 +96,25 @@ function ff_hevc_add_residual_8x8_8_neon, export=1
         bx              lr
 endfunc
 
+function ff_hevc_add_residual_8x8_10_neon, export=1
+        add             r12, r0, r2
+        add             r2,  r2, r2
+        mov             r3,  #8
+        vmov.s16        q12, #0
+        vmvn.s16        q13, #0xFC00 @ vmov.s16 #0x3FF
+1:      subs            r3,  #2
+        vld1.16         {q0-q1}, [r1, :128]!
+        vld1.16         {q8},    [r0, :128]
+        vqadd.s16       q0, q8
+        vld1.16         {q9},    [r12, :128]
+        vqadd.s16       q1, q9
+        clip10          q0, q1, q12, q13
+        vst1.16         {q0}, [r0, :128], r2
+        vst1.16         {q1}, [r12, :128], r2
+        bne             1b
+        bx              lr
+endfunc
+
 function ff_hevc_add_residual_16x16_8_neon, export=1
         mov             r3,  #16
         add             r12, r0, r2
@@ -97,6 +142,29 @@ function ff_hevc_add_residual_16x16_8_neon, export=1
         bx              lr
 endfunc
 
+function ff_hevc_add_residual_16x16_10_neon, export=1
+        mov             r3,  #16
+        vmov.s16        q12, #0
+        vmvn.s16        q13, #0xFC00 @ vmov.s16 #0x3FF
+        add             r12, r0, r2
+        add             r2,  r2, r2
+1:      subs            r3,  #2
+        vld1.16         {q8-q9},   [r0, :128]
+        vld1.16         {q0, q1},  [r1, :128]!
+        vqadd.s16       q0, q8
+        vld1.16         {q10-q11}, [r12, :128]
+        vqadd.s16       q1, q9
+        vld1.16         {q2, q3},  [r1, :128]!
+        vqadd.s16       q2, q10
+        vqadd.s16       q3, q11
+        clip10          q0, q1, q12, q13
+        clip10          q2, q3, q12, q13
+        vst1.16         {q0-q1},   [r0, :128], r2
+        vst1.16         {q2-q3},   [r12, :128], r2
+        bne             1b
+        bx              lr
+endfunc
+
 function ff_hevc_add_residual_32x32_8_neon, export=1
         vpush           {q4-q7}
         add             r12, r0, r2
@@ -137,6 +205,27 @@ function ff_hevc_add_residual_32x32_8_neon, export=1
         bx              lr
 endfunc
 
+function ff_hevc_add_residual_32x32_10_neon, export=1
+        mov             r3,  #32
+        add             r12, r0, #32
+        vmov.s16        q12, #0
+        vmvn.s16        q13, #0xFC00 @ vmov.s16 #0x3FF
+1:      subs            r3,  #1
+        vldm            r1!, {q0-q3}
+        vld1.16         {q8, q9},   [r0, :128]
+        vld1.16         {q10, q11}, [r12, :128]
+        vqadd.s16       q0, q8
+        vqadd.s16       q1, q9
+        vqadd.s16       q2, q10
+        vqadd.s16       q3, q11
+        clip10          q0, q1, q12, q13
+        clip10          q2, q3, q12, q13
+        vst1.16         {q0-q1},   [r0, :128], r2
+        vst1.16         {q2-q3},   [r12, :128], r2
+        bne             1b
+        bx              lr
+endfunc
+
 .macro idct_4x4_dc bitdepth
 function ff_hevc_idct_4x4_dc_\bitdepth\()_neon, export=1
         ldrsh           r1, [r0]

libavcodec/arm/hevcdsp_init_arm.c

@@ -28,12 +28,20 @@
 
 void ff_hevc_add_residual_4x4_8_neon(uint8_t *_dst, int16_t *coeffs,
                                      ptrdiff_t stride);
+void ff_hevc_add_residual_4x4_10_neon(uint8_t *_dst, int16_t *coeffs,
+                                      ptrdiff_t stride);
 void ff_hevc_add_residual_8x8_8_neon(uint8_t *_dst, int16_t *coeffs,
                                      ptrdiff_t stride);
+void ff_hevc_add_residual_8x8_10_neon(uint8_t *_dst, int16_t *coeffs,
+                                      ptrdiff_t stride);
 void ff_hevc_add_residual_16x16_8_neon(uint8_t *_dst, int16_t *coeffs,
                                        ptrdiff_t stride);
+void ff_hevc_add_residual_16x16_10_neon(uint8_t *_dst, int16_t *coeffs,
+                                        ptrdiff_t stride);
 void ff_hevc_add_residual_32x32_8_neon(uint8_t *_dst, int16_t *coeffs,
                                        ptrdiff_t stride);
+void ff_hevc_add_residual_32x32_10_neon(uint8_t *_dst, int16_t *coeffs,
+                                        ptrdiff_t stride);
 
 void ff_hevc_idct_4x4_dc_8_neon(int16_t *coeffs);
 void ff_hevc_idct_8x8_dc_8_neon(int16_t *coeffs);
@@ -72,6 +80,11 @@ av_cold void ff_hevc_dsp_init_arm(HEVCDSPContext *c, int bit_depth)
             c->idct[2] = ff_hevc_idct_16x16_8_neon;
         }
         if (bit_depth == 10) {
+            c->add_residual[0] = ff_hevc_add_residual_4x4_10_neon;
+            c->add_residual[1] = ff_hevc_add_residual_8x8_10_neon;
+            c->add_residual[2] = ff_hevc_add_residual_16x16_10_neon;
+            c->add_residual[3] = ff_hevc_add_residual_32x32_10_neon;
+
             c->idct_dc[0] = ff_hevc_idct_4x4_dc_10_neon;
             c->idct_dc[1] = ff_hevc_idct_8x8_dc_10_neon;
             c->idct_dc[2] = ff_hevc_idct_16x16_dc_10_neon;