[FFmpeg-cvslog] avcodec/ac3: Implement ac3_extract_exponents for aarch64 NEON
Geoff Hill
git at videolan.org
Mon Apr 8 13:39:15 EEST 2024
ffmpeg | branch: master | Geoff Hill <geoff at geoffhill.org> | Sat Apr 6 07:26:01 2024 -0700| [69cb34f8859ed62fd1c46d3612912a334542fe28] | committer: Martin Storsjö
avcodec/ac3: Implement ac3_extract_exponents for aarch64 NEON
Signed-off-by: Geoff Hill <geoff at geoffhill.org>
Signed-off-by: Martin Storsjö <martin at martin.st>
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=69cb34f8859ed62fd1c46d3612912a334542fe28
---
libavcodec/aarch64/ac3dsp_init_aarch64.c | 2 ++
libavcodec/aarch64/ac3dsp_neon.S | 14 ++++++++++++
tests/checkasm/ac3dsp.c | 38 ++++++++++++++++++++++++++++++++
3 files changed, 54 insertions(+)
diff --git a/libavcodec/aarch64/ac3dsp_init_aarch64.c b/libavcodec/aarch64/ac3dsp_init_aarch64.c
index 8874b41393..1bdc215b51 100644
--- a/libavcodec/aarch64/ac3dsp_init_aarch64.c
+++ b/libavcodec/aarch64/ac3dsp_init_aarch64.c
@@ -26,6 +26,7 @@
#include "config.h"
void ff_ac3_exponent_min_neon(uint8_t *exp, int num_reuse_blocks, int nb_coefs);
+void ff_ac3_extract_exponents_neon(uint8_t *exp, int32_t *coef, int nb_coefs);
void ff_float_to_fixed24_neon(int32_t *dst, const float *src, size_t len);
av_cold void ff_ac3dsp_init_aarch64(AC3DSPContext *c)
@@ -34,5 +35,6 @@ av_cold void ff_ac3dsp_init_aarch64(AC3DSPContext *c)
if (!have_neon(cpu_flags)) return;
c->ac3_exponent_min = ff_ac3_exponent_min_neon;
+ c->extract_exponents = ff_ac3_extract_exponents_neon;
c->float_to_fixed24 = ff_float_to_fixed24_neon;
}
diff --git a/libavcodec/aarch64/ac3dsp_neon.S b/libavcodec/aarch64/ac3dsp_neon.S
index f916c32538..c350c1f173 100644
--- a/libavcodec/aarch64/ac3dsp_neon.S
+++ b/libavcodec/aarch64/ac3dsp_neon.S
@@ -37,6 +37,20 @@ function ff_ac3_exponent_min_neon, export=1
3: ret
endfunc
+function ff_ac3_extract_exponents_neon, export=1
+ movi v1.4s, #8
+1: ld1 {v0.4s}, [x1], #16
+ abs v0.4s, v0.4s
+ clz v0.4s, v0.4s
+ sub v0.4s, v0.4s, v1.4s
+ xtn v0.4h, v0.4s
+ xtn v0.8b, v0.8h
+ st1 {v0.s}[0], [x0], #4
+ subs w2, w2, #4
+ b.gt 1b
+ ret
+endfunc
+
function ff_float_to_fixed24_neon, export=1
1: ld1 {v0.4s, v1.4s}, [x1], #32
fcvtzs v0.4s, v0.4s, #24
diff --git a/tests/checkasm/ac3dsp.c b/tests/checkasm/ac3dsp.c
index 06f31339f9..dc1b169e68 100644
--- a/tests/checkasm/ac3dsp.c
+++ b/tests/checkasm/ac3dsp.c
@@ -19,6 +19,7 @@
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
+#include <stdint.h>
#include <string.h>
#include "libavutil/mem.h"
@@ -36,6 +37,16 @@
} \
} while (0)
+#define randomize_i24(buf, len) \
+ do { \
+ int i; \
+ for (i = 0; i < len; i++) { \
+ int32_t v = (int32_t)rnd(); \
+ int32_t u = (v & 0xFFFFFF); \
+ buf[i] = (v < 0) ? -u : u; \
+ } \
+ } while (0)
+
#define randomize_float(buf, len) \
do { \
int i; \
@@ -77,6 +88,32 @@ static void check_ac3_exponent_min(AC3DSPContext *c) {
report("ac3_exponent_min");
}
+static void check_ac3_extract_exponents(AC3DSPContext *c) {
+#define MAX_EXPS 3072
+ LOCAL_ALIGNED_16(int32_t, src, [MAX_EXPS]);
+ LOCAL_ALIGNED_16(uint8_t, v1, [MAX_EXPS]);
+ LOCAL_ALIGNED_16(uint8_t, v2, [MAX_EXPS]);
+ int n;
+
+ declare_func(void, uint8_t *, int32_t *, int);
+
+ for (n = 512; n <= MAX_EXPS; n += 256) {
+ if (check_func(c->extract_exponents, "ac3_extract_exponents_n%d", n)) {
+ randomize_i24(src, n);
+
+ call_ref(v1, src, n);
+ call_new(v2, src, n);
+
+ if (memcmp(v1, v2, n) != 0)
+ fail();
+
+ bench_new(v1, src, n);
+ }
+ }
+
+ report("ac3_extract_exponents");
+}
+
static void check_float_to_fixed24(AC3DSPContext *c) {
#define BUF_SIZE 1024
LOCAL_ALIGNED_32(float, src, [BUF_SIZE]);
@@ -108,5 +145,6 @@ void checkasm_check_ac3dsp(void)
ff_ac3dsp_init(&c);
check_ac3_exponent_min(&c);
+ check_ac3_extract_exponents(&c);
check_float_to_fixed24(&c);
}
More information about the ffmpeg-cvslog
mailing list