swscale/aarch64: add neon {lum,chr}ConvertRange

chrRangeFromJpeg_8_c: 29.2
chrRangeFromJpeg_8_neon: 19.5
chrRangeFromJpeg_24_c: 80.5
chrRangeFromJpeg_24_neon: 34.0
chrRangeFromJpeg_128_c: 413.7
chrRangeFromJpeg_128_neon: 156.0
chrRangeFromJpeg_144_c: 471.0
chrRangeFromJpeg_144_neon: 174.2
chrRangeFromJpeg_256_c: 842.0
chrRangeFromJpeg_256_neon: 305.5
chrRangeFromJpeg_512_c: 1699.0
chrRangeFromJpeg_512_neon: 608.0
chrRangeToJpeg_8_c: 51.7
chrRangeToJpeg_8_neon: 22.7
chrRangeToJpeg_24_c: 149.7
chrRangeToJpeg_24_neon: 38.0
chrRangeToJpeg_128_c: 761.7
chrRangeToJpeg_128_neon: 176.7
chrRangeToJpeg_144_c: 866.2
chrRangeToJpeg_144_neon: 198.7
chrRangeToJpeg_256_c: 1516.5
chrRangeToJpeg_256_neon: 348.7
chrRangeToJpeg_512_c: 3067.2
chrRangeToJpeg_512_neon: 692.7
lumRangeFromJpeg_8_c: 24.0
lumRangeFromJpeg_8_neon: 17.0
lumRangeFromJpeg_24_c: 56.7
lumRangeFromJpeg_24_neon: 21.0
lumRangeFromJpeg_128_c: 294.5
lumRangeFromJpeg_128_neon: 76.7
lumRangeFromJpeg_144_c: 332.5
lumRangeFromJpeg_144_neon: 86.7
lumRangeFromJpeg_256_c: 586.0
lumRangeFromJpeg_256_neon: 152.2
lumRangeFromJpeg_512_c: 1190.0
lumRangeFromJpeg_512_neon: 298.0
lumRangeToJpeg_8_c: 31.7
lumRangeToJpeg_8_neon: 19.5
lumRangeToJpeg_24_c: 83.5
lumRangeToJpeg_24_neon: 24.2
lumRangeToJpeg_128_c: 440.5
lumRangeToJpeg_128_neon: 91.0
lumRangeToJpeg_144_c: 504.2
lumRangeToJpeg_144_neon: 101.0
lumRangeToJpeg_256_c: 879.7
lumRangeToJpeg_256_neon: 177.2
lumRangeToJpeg_512_c: 1794.2
lumRangeToJpeg_512_neon: 354.0

5 files changed, 125 insertions, 1 deletions

diff --git a/libswscale/aarch64/Makefile b/libswscale/aarch64/Makefile
index adfd90a1b6..37ad960619 100644
--- a/libswscale/aarch64/Makefile
+++ b/libswscale/aarch64/Makefile
@@ -5,5 +5,6 @@ OBJS        += aarch64/rgb2rgb.o                \
 NEON-OBJS   += aarch64/hscale.o                 \
                aarch64/input.o                  \
                aarch64/output.o                 \
+               aarch64/range_convert_neon.o     \
                aarch64/rgb2rgb_neon.o           \
                aarch64/yuv2rgb_neon.o           \
diff --git a/libswscale/aarch64/range_convert_neon.S b/libswscale/aarch64/range_convert_neon.S
new file mode 100644
index 0000000000..30991ab2a6
--- /dev/null
+++ b/libswscale/aarch64/range_convert_neon.S
@@ -0,0 +1,99 @@
+/*
+ * Copyright (c) 2024 Ramiro Polla
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/aarch64/asm.S"
+
+.macro lumConvertRange name, max, mult, offset, shift
+function ff_\name, export=1
+.if \max != 0
+        mov             w3, #\max
+        dup             v24.8h, w3
+.endif
+        mov             w3, #\mult
+        dup             v25.4s, w3
+        movz            w3, #(\offset & 0xffff)
+        movk            w3, #((\offset >> 16) & 0xffff), lsl #16
+        dup             v26.4s, w3
+1:
+        ld1             {v0.8h}, [x0]
+.if \max != 0
+        smin            v0.8h, v0.8h, v24.8h
+.endif
+        mov             v16.16b, v26.16b
+        mov             v18.16b, v26.16b
+        sxtl            v20.4s, v0.4h
+        sxtl2           v22.4s, v0.8h
+        mla             v16.4s, v20.4s, v25.4s
+        mla             v18.4s, v22.4s, v25.4s
+        shrn            v0.4h, v16.4s, #\shift
+        shrn2           v0.8h, v18.4s, #\shift
+        subs            w1, w1, #8
+        st1             {v0.8h}, [x0], #16
+        b.gt            1b
+        ret
+endfunc
+.endm
+
+.macro chrConvertRange name, max, mult, offset, shift
+function ff_\name, export=1
+.if \max != 0
+        mov             w3, #\max
+        dup             v24.8h, w3
+.endif
+        mov             w3, #\mult
+        dup             v25.4s, w3
+        movz            w3, #(\offset & 0xffff)
+        movk            w3, #((\offset >> 16) & 0xffff), lsl #16
+        dup             v26.4s, w3
+1:
+        ld1             {v0.8h}, [x0]
+        ld1             {v1.8h}, [x1]
+.if \max != 0
+        smin            v0.8h, v0.8h, v24.8h
+        smin            v1.8h, v1.8h, v24.8h
+.endif
+        mov             v16.16b, v26.16b
+        mov             v17.16b, v26.16b
+        mov             v18.16b, v26.16b
+        mov             v19.16b, v26.16b
+        sxtl            v20.4s, v0.4h
+        sxtl            v21.4s, v1.4h
+        sxtl2           v22.4s, v0.8h
+        sxtl2           v23.4s, v1.8h
+        mla             v16.4s, v20.4s, v25.4s
+        mla             v17.4s, v21.4s, v25.4s
+        mla             v18.4s, v22.4s, v25.4s
+        mla             v19.4s, v23.4s, v25.4s
+        shrn            v0.4h, v16.4s, #\shift
+        shrn            v1.4h, v17.4s, #\shift
+        shrn2           v0.8h, v18.4s, #\shift
+        shrn2           v1.8h, v19.4s, #\shift
+        subs            w2, w2, #8
+        st1             {v0.8h}, [x0], #16
+        st1             {v1.8h}, [x1], #16
+        b.gt            1b
+        ret
+endfunc
+.endm
+
+lumConvertRange lumRangeToJpeg_neon,   30189, 19077, -39057361, 14
+chrConvertRange chrRangeToJpeg_neon,   30775,  4663,  -9289992, 12
+lumConvertRange lumRangeFromJpeg_neon,     0, 14071,  33561947, 14
+chrConvertRange chrRangeFromJpeg_neon,     0,  1799,   4081085, 11
diff --git a/libswscale/aarch64/swscale.c b/libswscale/aarch64/swscale.c
index 4c4ea39dc1..e4ea3309ba 100644
--- a/libswscale/aarch64/swscale.c
+++ b/libswscale/aarch64/swscale.c
@@ -215,6 +215,26 @@ void ff_rgb24ToUV_half_neon(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *unuse
                        const uint8_t *src2, int width, uint32_t *rgb2yuv,
                        void *opq);
 
+void ff_lumRangeFromJpeg_neon(int16_t *dst, int width);
+void ff_chrRangeFromJpeg_neon(int16_t *dstU, int16_t *dstV, int width);
+void ff_lumRangeToJpeg_neon(int16_t *dst, int width);
+void ff_chrRangeToJpeg_neon(int16_t *dstU, int16_t *dstV, int width);
+
+av_cold void ff_sws_init_range_convert_aarch64(SwsContext *c)
+{
+    if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {
+        if (c->dstBpc <= 14) {
+            if (c->srcRange) {
+                c->lumConvertRange = ff_lumRangeFromJpeg_neon;
+                c->chrConvertRange = ff_chrRangeFromJpeg_neon;
+            } else {
+                c->lumConvertRange = ff_lumRangeToJpeg_neon;
+                c->chrConvertRange = ff_chrRangeToJpeg_neon;
+            }
+        }
+    }
+}
+
 av_cold void ff_sws_init_swscale_aarch64(SwsContext *c)
 {
     int cpu_flags = av_get_cpu_flags();
@@ -237,5 +257,6 @@ av_cold void ff_sws_init_swscale_aarch64(SwsContext *c)
         default:
             break;
         }
+        ff_sws_init_range_convert_aarch64(c);
     }
 }
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index d5e7b5e71c..0818f50c7f 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -697,6 +697,7 @@ void ff_yuv2rgb_init_tables_ppc(SwsContext *c, const int inv_table[4],
 void ff_updateMMXDitherTables(SwsContext *c, int dstY);
 
 av_cold void ff_sws_init_range_convert(SwsContext *c);
+av_cold void ff_sws_init_range_convert_aarch64(SwsContext *c);
 av_cold void ff_sws_init_range_convert_loongarch(SwsContext *c);
 av_cold void ff_sws_init_range_convert_x86(SwsContext *c);
 
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 8dfa57b5ff..12dba712c1 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -1080,7 +1080,9 @@ int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
 
     if (need_reinit) {
         ff_sws_init_range_convert(c);
-#if ARCH_LOONGARCH64
+#if ARCH_AARCH64
+        ff_sws_init_range_convert_aarch64(c);
+#elif ARCH_LOONGARCH64
         ff_sws_init_range_convert_loongarch(c);
 #elif ARCH_X86
         ff_sws_init_range_convert_x86(c);