diff --git a/libavcodec/arm/hevcdsp_idct_neon.S b/libavcodec/arm/hevcdsp_idct_neon.S index b59496ec25..8f831a37f2 100644 --- a/libavcodec/arm/hevcdsp_idct_neon.S +++ b/libavcodec/arm/hevcdsp_idct_neon.S @@ -29,82 +29,6 @@ const trans, align=4 .short 57, 43, 25, 9 endconst -function ff_hevc_idct_4x4_dc_neon_8, export=1 - ldrsh r1, [r0] - ldr r2, =0x20 - add r1, #1 - asr r1, #1 - add r1, r2 - asr r1, #6 - vdup.16 q0, r1 - vdup.16 q1, r1 - vst1.16 {q0, q1}, [r0] - bx lr -endfunc - -function ff_hevc_idct_8x8_dc_neon_8, export=1 - ldrsh r1, [r0] - ldr r2, =0x20 - add r1, #1 - asr r1, #1 - add r1, r2 - asr r1, #6 - vdup.16 q8, r1 - vdup.16 q9, r1 - vmov.16 q10, q8 - vmov.16 q11, q8 - vmov.16 q12, q8 - vmov.16 q13, q8 - vmov.16 q14, q8 - vmov.16 q15, q8 - vstm r0, {q8-q15} - bx lr -endfunc - -function ff_hevc_idct_16x16_dc_neon_8, export=1 - ldrsh r1, [r0] - ldr r2, =0x20 - add r1, #1 - asr r1, #1 - add r1, r2 - asr r1, #6 - vdup.16 q8, r1 - vdup.16 q9, r1 - vmov.16 q10, q8 - vmov.16 q11, q8 - vmov.16 q12, q8 - vmov.16 q13, q8 - vmov.16 q14, q8 - vmov.16 q15, q8 - vstm r0!, {q8-q15} - vstm r0!, {q8-q15} - vstm r0!, {q8-q15} - vstm r0, {q8-q15} - bx lr -endfunc - -function ff_hevc_idct_32x32_dc_neon_8, export=1 - ldrsh r1, [r0] - ldr r2, =0x20 - add r1, #1 - asr r1, #1 - add r1, r2 - asr r1, #6 - mov r3, #16 - vdup.16 q8, r1 - vdup.16 q9, r1 - vmov.16 q10, q8 - vmov.16 q11, q8 - vmov.16 q12, q8 - vmov.16 q13, q8 - vmov.16 q14, q8 - vmov.16 q15, q8 -1: subs r3, #1 - vstm r0!, {q8-q15} - bne 1b - bx lr -endfunc - function ff_hevc_add_residual_4x4_neon_8, export=1 vldm r1, {q0-q1} vld1.32 d4[0], [r0], r2 @@ -235,6 +159,82 @@ function ff_hevc_transform_luma_4x4_neon_8, export=1 bx lr endfunc +function ff_hevc_idct_4x4_dc_8_neon, export=1 + ldrsh r1, [r0] + ldr r2, =0x20 + add r1, #1 + asr r1, #1 + add r1, r2 + asr r1, #6 + vdup.16 q0, r1 + vdup.16 q1, r1 + vst1.16 {q0, q1}, [r0, :128] + bx lr +endfunc + +function ff_hevc_idct_8x8_dc_8_neon, export=1 + ldrsh r1, [r0] + ldr r2, =0x20 + add r1, #1 + asr r1, #1 + add r1, r2 + asr r1, #6 + vdup.16 q8, r1 + vdup.16 q9, r1 + vmov.16 q10, q8 + vmov.16 q11, q8 + vmov.16 q12, q8 + vmov.16 q13, q8 + vmov.16 q14, q8 + vmov.16 q15, q8 + vstm r0, {q8-q15} + bx lr +endfunc + +function ff_hevc_idct_16x16_dc_8_neon, export=1 + ldrsh r1, [r0] + ldr r2, =0x20 + add r1, #1 + asr r1, #1 + add r1, r2 + asr r1, #6 + vdup.16 q8, r1 + vdup.16 q9, r1 + vmov.16 q10, q8 + vmov.16 q11, q8 + vmov.16 q12, q8 + vmov.16 q13, q8 + vmov.16 q14, q8 + vmov.16 q15, q8 + vstm r0!, {q8-q15} + vstm r0!, {q8-q15} + vstm r0!, {q8-q15} + vstm r0, {q8-q15} + bx lr +endfunc + +function ff_hevc_idct_32x32_dc_8_neon, export=1 + ldrsh r1, [r0] + ldr r2, =0x20 + add r1, #1 + asr r1, #1 + add r1, r2 + asr r1, #6 + mov r3, #16 + vdup.16 q8, r1 + vdup.16 q9, r1 + vmov.16 q10, q8 + vmov.16 q11, q8 + vmov.16 q12, q8 + vmov.16 q13, q8 + vmov.16 q14, q8 + vmov.16 q15, q8 +1: subs r3, #1 + vstm r0!, {q8-q15} + bne 1b + bx lr +endfunc + .macro sum_sub out, in, c, op .ifc \op, + vmlal.s16 \out, \in, \c diff --git a/libavcodec/arm/hevcdsp_init_neon.c b/libavcodec/arm/hevcdsp_init_neon.c index 562ebf96d1..9537561513 100644 --- a/libavcodec/arm/hevcdsp_init_neon.c +++ b/libavcodec/arm/hevcdsp_init_neon.c @@ -27,16 +27,16 @@ void ff_hevc_v_loop_filter_luma_neon(uint8_t *_pix, ptrdiff_t _stride, int _beta void ff_hevc_h_loop_filter_luma_neon(uint8_t *_pix, ptrdiff_t _stride, int _beta, int *_tc, uint8_t *_no_p, uint8_t *_no_q); void ff_hevc_v_loop_filter_chroma_neon(uint8_t *_pix, ptrdiff_t _stride, int *_tc, uint8_t *_no_p, uint8_t *_no_q); void ff_hevc_h_loop_filter_chroma_neon(uint8_t *_pix, ptrdiff_t _stride, int *_tc, uint8_t *_no_p, uint8_t *_no_q); +void ff_hevc_idct_4x4_dc_8_neon(int16_t *coeffs); +void ff_hevc_idct_8x8_dc_8_neon(int16_t *coeffs); +void ff_hevc_idct_16x16_dc_8_neon(int16_t *coeffs); +void ff_hevc_idct_32x32_dc_8_neon(int16_t *coeffs); void ff_hevc_idct_4x4_8_neon(int16_t *coeffs, int col_limit); void ff_hevc_idct_8x8_8_neon(int16_t *coeffs, int col_limit); void ff_hevc_idct_16x16_8_neon(int16_t *coeffs, int col_limit); void ff_hevc_idct_4x4_10_neon(int16_t *coeffs, int col_limit); void ff_hevc_idct_8x8_10_neon(int16_t *coeffs, int col_limit); void ff_hevc_idct_16x16_10_neon(int16_t *coeffs, int col_limit); -void ff_hevc_idct_4x4_dc_neon_8(int16_t *coeffs); -void ff_hevc_idct_8x8_dc_neon_8(int16_t *coeffs); -void ff_hevc_idct_16x16_dc_neon_8(int16_t *coeffs); -void ff_hevc_idct_32x32_dc_neon_8(int16_t *coeffs); void ff_hevc_transform_luma_4x4_neon_8(int16_t *coeffs); void ff_hevc_add_residual_4x4_neon_8(uint8_t *_dst, int16_t *coeffs, ptrdiff_t stride); @@ -154,13 +154,13 @@ av_cold void ff_hevc_dsp_init_neon(HEVCDSPContext *c, const int bit_depth) c->hevc_h_loop_filter_luma = ff_hevc_h_loop_filter_luma_neon; c->hevc_v_loop_filter_chroma = ff_hevc_v_loop_filter_chroma_neon; c->hevc_h_loop_filter_chroma = ff_hevc_h_loop_filter_chroma_neon; + c->idct_dc[0] = ff_hevc_idct_4x4_dc_8_neon; + c->idct_dc[1] = ff_hevc_idct_8x8_dc_8_neon; + c->idct_dc[2] = ff_hevc_idct_16x16_dc_8_neon; + c->idct_dc[3] = ff_hevc_idct_32x32_dc_8_neon; c->idct[0] = ff_hevc_idct_4x4_8_neon; c->idct[1] = ff_hevc_idct_8x8_8_neon; c->idct[2] = ff_hevc_idct_16x16_8_neon; - c->idct_dc[0] = ff_hevc_idct_4x4_dc_neon_8; - c->idct_dc[1] = ff_hevc_idct_8x8_dc_neon_8; - c->idct_dc[2] = ff_hevc_idct_16x16_dc_neon_8; - c->idct_dc[3] = ff_hevc_idct_32x32_dc_neon_8; c->add_residual[0] = ff_hevc_add_residual_4x4_neon_8; c->add_residual[1] = ff_hevc_add_residual_8x8_neon_8; c->add_residual[2] = ff_hevc_add_residual_16x16_neon_8;