[FFmpeg-cvslog] aarch64/hevcdsp_idct_neon: Add implementation for idct dc 12

[Zhao Zhili]

ffmpeg | branch: master | Zhao Zhili <zhilizhao at tencent.com> | Thu Feb 20 00:50:31 2025 +0800| [3e9777dc7546425651bc265a49187d6bbb42bdf4] | committer: Zhao Zhili

aarch64/hevcdsp_idct_neon: Add implementation for idct dc 12

Reduce binary size at the same time. The performance compared to clang -O3
is the same.

Reviewed-by: Martin Storsjö <martin at martin.st>
Signed-off-by: Zhao Zhili <zhilizhao at tencent.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3e9777dc7546425651bc265a49187d6bbb42bdf4
---

 libavcodec/aarch64/hevcdsp_idct_neon.S    | 43 +++++++++++++++++++------------
 libavcodec/aarch64/hevcdsp_init_aarch64.c |  8 ++++++
 2 files changed, 35 insertions(+), 16 deletions(-)

diff --git a/libavcodec/aarch64/hevcdsp_idct_neon.S b/libavcodec/aarch64/hevcdsp_idct_neon.S
index 2337ed8ba3..954ce2407a 100644
--- a/libavcodec/aarch64/hevcdsp_idct_neon.S
+++ b/libavcodec/aarch64/hevcdsp_idct_neon.S
@@ -901,15 +901,33 @@ endfunc
 .endm
 
 // void ff_hevc_idct_NxN_dc_DEPTH_neon(int16_t *coeffs)
-.macro idct_dc size, bitdepth
-function ff_hevc_idct_\size\()x\size\()_dc_\bitdepth\()_neon, export=1
+.macro idct_dc size
+function ff_hevc_idct_\size\()x\size\()_dc_10_neon, export=1
         ldrsh           w1, [x0]
         add             w1, w1, #1
         asr             w1, w1, #1
-        add             w1, w1, #(1 << (13 - \bitdepth))
-        asr             w1, w1, #(14 - \bitdepth)
-        dup             v0.8h, w1
+        add             w1, w1, #(1 << (13 - 10))
+        asr             w1, w1, #(14 - 10)
+        b               2f
+endfunc
+
+function ff_hevc_idct_\size\()x\size\()_dc_12_neon, export=1
+        ldrsh           w1, [x0]
+        add             w1, w1, #1
+        asr             w1, w1, #1
+        add             w1, w1, #(1 << (13 - 12))
+        asr             w1, w1, #(14 - 12)
+        b               2f
+endfunc
 
+function ff_hevc_idct_\size\()x\size\()_dc_8_neon, export=1
+        ldrsh           w1, [x0]
+        add             w1, w1, #1
+        asr             w1, w1, #1
+        add             w1, w1, #(1 << (13 - 8))
+        asr             w1, w1, #(14 - 8)
+2:
+        dup             v0.8h, w1
 .if \size < 8
         stp             q0, q0, [x0]
 .elseif \size < 16
@@ -928,14 +946,7 @@ function ff_hevc_idct_\size\()x\size\()_dc_\bitdepth\()_neon, export=1
 endfunc
 .endm
 
-idct_dc 4, 8
-idct_dc 4, 10
-
-idct_dc 8, 8
-idct_dc 8, 10
-
-idct_dc 16, 8
-idct_dc 16, 10
-
-idct_dc 32, 8
-idct_dc 32, 10
+idct_dc 4
+idct_dc 8
+idct_dc 16
+idct_dc 32
diff --git a/libavcodec/aarch64/hevcdsp_init_aarch64.c b/libavcodec/aarch64/hevcdsp_init_aarch64.c
index 386d7c59c8..5dd470baaa 100644
--- a/libavcodec/aarch64/hevcdsp_init_aarch64.c
+++ b/libavcodec/aarch64/hevcdsp_init_aarch64.c
@@ -91,6 +91,10 @@ void ff_hevc_idct_4x4_dc_10_neon(int16_t *coeffs);
 void ff_hevc_idct_8x8_dc_10_neon(int16_t *coeffs);
 void ff_hevc_idct_16x16_dc_10_neon(int16_t *coeffs);
 void ff_hevc_idct_32x32_dc_10_neon(int16_t *coeffs);
+void ff_hevc_idct_4x4_dc_12_neon(int16_t *coeffs);
+void ff_hevc_idct_8x8_dc_12_neon(int16_t *coeffs);
+void ff_hevc_idct_16x16_dc_12_neon(int16_t *coeffs);
+void ff_hevc_idct_32x32_dc_12_neon(int16_t *coeffs);
 void ff_hevc_transform_luma_4x4_neon_8(int16_t *coeffs);
 
 #define NEON8_FNASSIGN(member, v, h, fn, ext) \
@@ -267,5 +271,9 @@ av_cold void ff_hevc_dsp_init_aarch64(HEVCDSPContext *c, const int bit_depth)
         c->add_residual[1]             = ff_hevc_add_residual_8x8_12_neon;
         c->add_residual[2]             = ff_hevc_add_residual_16x16_12_neon;
         c->add_residual[3]             = ff_hevc_add_residual_32x32_12_neon;
+        c->idct_dc[0]                  = ff_hevc_idct_4x4_dc_12_neon;
+        c->idct_dc[1]                  = ff_hevc_idct_8x8_dc_12_neon;
+        c->idct_dc[2]                  = ff_hevc_idct_16x16_dc_12_neon;
+        c->idct_dc[3]                  = ff_hevc_idct_32x32_dc_12_neon;
     }
 }