[FFmpeg-devel] [PATCH 1/2] lavu/riscv: CPU flag for fast misaligned accesses

[FFmpeg-devel] [PATCH 1/2] lavu/riscv: CPU flag for fast misaligned accesses

[Rémi Denis-Courmont]

---
 libavutil/cpu.c           | 1 +
 libavutil/cpu.h           | 1 +
 libavutil/riscv/cpu.c     | 3 +++
 libavutil/tests/cpu.c     | 3 ++-
 tests/checkasm/checkasm.c | 1 +
 5 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/libavutil/cpu.c b/libavutil/cpu.c
index 396eeb38d6..9ac2f01c20 100644
--- a/libavutil/cpu.c
+++ b/libavutil/cpu.c
@@ -193,6 +193,7 @@ int av_parse_cpu_caps(unsigned *flags, const char *s)
         { "zba",      NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_RVB_ADDR },    .unit = "flags" },
         { "zbb",      NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_RVB_BASIC },   .unit = "flags" },
         { "zvbb",     NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_RV_ZVBB },   .unit = "flags" },
+        { "misaligned", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_RV_MISALIGNED },   .unit = "flags" },
 #endif
         { NULL },
     };
diff --git a/libavutil/cpu.h b/libavutil/cpu.h
index cc19828d4b..a25901433e 100644
--- a/libavutil/cpu.h
+++ b/libavutil/cpu.h
@@ -91,6 +91,7 @@
 #define AV_CPU_FLAG_RVB_BASIC    (1 << 7) ///< Basic bit-manipulations
 #define AV_CPU_FLAG_RVB_ADDR     (1 << 8) ///< Address bit-manipulations
 #define AV_CPU_FLAG_RV_ZVBB      (1 << 9) ///< Vector basic bit-manipulations
+#define AV_CPU_FLAG_RV_MISALIGNED (1 <<10) ///< Fast misaligned accesses
 
 /**
  * Return the flags which specify extensions supported by the CPU.
diff --git a/libavutil/riscv/cpu.c b/libavutil/riscv/cpu.c
index 6755f0df69..1fe1a397c4 100644
--- a/libavutil/riscv/cpu.c
+++ b/libavutil/riscv/cpu.c
@@ -52,6 +52,7 @@ int ff_get_cpu_flags_riscv(void)
     struct riscv_hwprobe pairs[] = {
         { RISCV_HWPROBE_KEY_BASE_BEHAVIOR, 0 },
         { RISCV_HWPROBE_KEY_IMA_EXT_0, 0 },
+        { RISCV_HWPROBE_KEY_CPUPERF_0, 0 },
     };
 
     if (__riscv_hwprobe(pairs, FF_ARRAY_ELEMS(pairs), 0, NULL, 0) == 0) {
@@ -76,6 +77,8 @@ int ff_get_cpu_flags_riscv(void)
         if (pairs[1].value & RISCV_HWPROBE_EXT_ZVBB)
             ret |= AV_CPU_FLAG_RV_ZVBB;
 #endif
+        if (pairs[2].value & RISCV_HWPROBE_MISALIGNED_FAST)
+            ret |= AV_CPU_FLAG_RV_MISALIGNED;
     } else
 #endif
 #if HAVE_GETAUXVAL
diff --git a/libavutil/tests/cpu.c b/libavutil/tests/cpu.c
index 10e620963b..02b98682e3 100644
--- a/libavutil/tests/cpu.c
+++ b/libavutil/tests/cpu.c
@@ -94,7 +94,8 @@ static const struct {
     { AV_CPU_FLAG_RVV_F32,   "zve32f"     },
     { AV_CPU_FLAG_RVV_I64,   "zve64x"     },
     { AV_CPU_FLAG_RVV_F64,   "zve64d"     },
-    { AV_CPU_FLAG_RV_ZVBB,   "zvbb"      },
+    { AV_CPU_FLAG_RV_ZVBB,   "zvbb"       },
+    { AV_CPU_FLAG_RV_MISALIGNED, "misaligned" },
 #endif
     { 0 }
 };
diff --git a/tests/checkasm/checkasm.c b/tests/checkasm/checkasm.c
index 04f94f9d09..c6dc0cfa77 100644
--- a/tests/checkasm/checkasm.c
+++ b/tests/checkasm/checkasm.c
@@ -286,6 +286,7 @@ static const struct {
     { "RVVi64",   "rvv_i64",  AV_CPU_FLAG_RVV_I64 },
     { "RVVf64",   "rvv_f64",  AV_CPU_FLAG_RVV_F64 },
     { "RV_Zvbb",  "rv_zvbb",  AV_CPU_FLAG_RV_ZVBB },
+    { "misaligned", "misaligned", AV_CPU_FLAG_RV_MISALIGNED },
 #elif ARCH_MIPS
     { "MMI",      "mmi",      AV_CPU_FLAG_MMI },
     { "MSA",      "msa",      AV_CPU_FLAG_MSA },
-- 
2.43.0

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

[FFmpeg-devel] [PATCH 2/2] lavc/vp8dsp: restrict RVI optimisations

[Rémi Denis-Courmont]

They are actually awfully slow if the CPU does not support misaligned
accesses natively, so only use them if misaligned accesses are fast.
---
 libavcodec/riscv/vp8dsp_init.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavcodec/riscv/vp8dsp_init.c b/libavcodec/riscv/vp8dsp_init.c
index dc3e087f01..fe4fa5b867 100644
--- a/libavcodec/riscv/vp8dsp_init.c
+++ b/libavcodec/riscv/vp8dsp_init.c
@@ -45,7 +45,7 @@ av_cold void ff_vp78dsp_init_riscv(VP8DSPContext *c)
 {
 #if HAVE_RV
     int flags = av_get_cpu_flags();
-    if (flags & AV_CPU_FLAG_RVI) {
+    if (flags & AV_CPU_FLAG_RV_MISALIGNED) {
 #if __riscv_xlen >= 64
         c->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvi;
         c->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvi;
-- 
2.43.0

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".