diff mbox series

[FFmpeg-devel,1/3] sws/input: R-V V 32-bit RGB to Y

Message ID 20240606182617.39636-1-remi@remlab.net
State New
Headers show
Series [FFmpeg-devel,1/3] sws/input: R-V V 32-bit RGB to Y | expand

Checks

Context Check Description
andriy/configure_x86 warning Failed to apply patch

Commit Message

Rémi Denis-Courmont June 6, 2024, 6:26 p.m. UTC 
T-Head C908:
abgr_to_y_8_c:            2.5
abgr_to_y_8_rvv_i32:      2.2
abgr_to_y_128_c:         37.0
abgr_to_y_128_rvv_i32:    8.5
abgr_to_y_1080_c:       327.0
abgr_to_y_1080_rvv_i32:  69.5
abgr_to_y_1920_c:       552.0
abgr_to_y_1920_rvv_i32: 122.2
bgra_to_y_8_c:            2.5
bgra_to_y_8_rvv_i32:      2.2
bgra_to_y_128_c:         37.2
bgra_to_y_128_rvv_i32:    8.5
bgra_to_y_1080_c:       310.2
bgra_to_y_1080_rvv_i32:  69.5
bgra_to_y_1920_c:       568.2
bgra_to_y_1920_rvv_i32: 122.5

SpacemiT X60:
abgr_to_y_8_c:            2.5
abgr_to_y_8_rvv_i32:      2.0
abgr_to_y_128_c:         33.0
abgr_to_y_128_rvv_i32:    3.7
abgr_to_y_1080_c:       276.0
abgr_to_y_1080_rvv_i32:  31.5
abgr_to_y_1920_c:       493.7
abgr_to_y_1920_rvv_i32:  55.5
bgra_to_y_8_c:            2.2
bgra_to_y_8_rvv_i32:      2.0
bgra_to_y_128_c:         33.0
bgra_to_y_128_rvv_i32:    3.7
bgra_to_y_1080_c:       276.0
bgra_to_y_1080_rvv_i32:  31.5
bgra_to_y_1920_c:       490.7
bgra_to_y_1920_rvv_i32:  55.5
---
 libswscale/riscv/input.S   | 45 +++++++++++++++++++++++++++++++++++++
 libswscale/riscv/swscale.c | 46 ++++++++++++++++++++++++++------------
 2 files changed, 77 insertions(+), 14 deletions(-)
diff mbox series

Patch

diff --git a/libswscale/riscv/input.S b/libswscale/riscv/input.S
index f84f79899b..8dfd012a9d 100644
--- a/libswscale/riscv/input.S
+++ b/libswscale/riscv/input.S
@@ -149,3 +149,48 @@  func ff_rgb24ToUV_half_rvv, zve32x
 
         ret
 endfunc
+
+        .macro  rgba_input chr0, chr1, high
+func ff_\chr1\()ToY_rvv, zve32x
+        lw      t1, 8(a5) # BY
+        lw      t3, 0(a5) # RY
+        j       1f
+endfunc
+
+func ff_\chr0\()ToY_rvv, zve32x
+        lw      t1, 0(a5) # RY
+        lw      t3, 8(a5) # BY
+1:
+        lw      t2, 4(a5) # GY
+        li      t4, (32 << (15 - 1)) + (1 << (15 - 7))
+        li      t5, 0xff
+2:
+        vsetvli   t0, a4, e32, m8, ta, ma
+        vle32.v   v0, (a1)
+        sub       a4, a4, t0
+        .if     \high
+        vsrl.vi   v8, v0, 24
+        .else
+        vand.vx   v8, v0, t5
+        .endif
+        sh2add    a1, t0, a1
+        vsrl.vi   v16, v0, 8 * (1 + \high)
+        vmul.vx   v24, v8, t1
+        vand.vx   v16, v16, t5
+        vsrl.vi   v8, v0, 8 * (2 - \high)
+        vmacc.vx  v24, t2, v16
+        vand.vx   v8, v8, t5
+        vadd.vx   v24, v24, t4
+        vmacc.vx  v24, t3, v8
+        vsetvli   zero, zero, e16, m4, ta, ma
+        vnsra.wi  v0, v24, 15 - 6
+        vse16.v   v0, (a0)
+        sh1add    a0, t0, a0
+        bnez      a4, 2b
+
+        ret
+endfunc
+        .endm
+
+rgba_input rgba32, bgra32, 0
+rgba_input abgr32, argb32, 1
diff --git a/libswscale/riscv/swscale.c b/libswscale/riscv/swscale.c
index bae626f64a..fcb338b295 100644
--- a/libswscale/riscv/swscale.c
+++ b/libswscale/riscv/swscale.c
@@ -21,20 +21,22 @@ 
 #include "libavutil/riscv/cpu.h"
 #include "libswscale/swscale_internal.h"
 
-void ff_bgr24ToY_rvv(uint8_t *dst, const uint8_t *src, const uint8_t *,
-                     const uint8_t *, int width, uint32_t *coeffs, void *);
-void ff_bgr24ToUV_rvv(uint8_t *, uint8_t *, const uint8_t *, const uint8_t *,
-                      const uint8_t *, int width, uint32_t *coeffs, void *);
-void ff_bgr24ToUV_half_rvv(uint8_t *, uint8_t *, const uint8_t *,
-                           const uint8_t *, const uint8_t *, int width,
-                           uint32_t *coeffs, void *);
-void ff_rgb24ToY_rvv(uint8_t *dst, const uint8_t *src, const uint8_t *,
-                     const uint8_t *, int width, uint32_t *coeffs, void *);
-void ff_rgb24ToUV_rvv(uint8_t *, uint8_t *, const uint8_t *, const uint8_t *,
-                      const uint8_t *, int width, uint32_t *coeffs, void *);
-void ff_rgb24ToUV_half_rvv(uint8_t *, uint8_t *, const uint8_t *,
-                           const uint8_t *, const uint8_t *, int width,
-                           uint32_t *coeffs, void *);
+#define RVV_INPUT(name) \
+void ff_##name##ToY_rvv(uint8_t *dst, const uint8_t *src, const uint8_t *, \
+                        const uint8_t *, int w, uint32_t *coeffs, void *); \
+void ff_##name##ToUV_rvv(uint8_t *, uint8_t *, const uint8_t *, \
+                         const uint8_t *, const uint8_t *, int w, \
+                         uint32_t *coeffs, void *); \
+void ff_##name##ToUV_half_rvv(uint8_t *, uint8_t *, const uint8_t *, \
+                              const uint8_t *, const uint8_t *, int w, \
+                              uint32_t *coeffs, void *)
+
+RVV_INPUT(abgr32);
+RVV_INPUT(argb32);
+RVV_INPUT(bgr24);
+RVV_INPUT(bgra32);
+RVV_INPUT(rgb24);
+RVV_INPUT(rgba32);
 
 av_cold void ff_sws_init_swscale_riscv(SwsContext *c)
 {
@@ -43,6 +45,14 @@  av_cold void ff_sws_init_swscale_riscv(SwsContext *c)
 
     if ((flags & AV_CPU_FLAG_RVV_I32) && (flags & AV_CPU_FLAG_RVB_ADDR)) {
         switch (c->srcFormat) {
+            case AV_PIX_FMT_ABGR:
+                c->lumToYV12 = ff_abgr32ToY_rvv;
+                break;
+
+            case AV_PIX_FMT_ARGB:
+                c->lumToYV12 = ff_argb32ToY_rvv;
+                break;
+
             case AV_PIX_FMT_BGR24:
                 c->lumToYV12 = ff_bgr24ToY_rvv;
                 if (c->chrSrcHSubSample)
@@ -51,6 +61,10 @@  av_cold void ff_sws_init_swscale_riscv(SwsContext *c)
                     c->chrToYV12 = ff_bgr24ToUV_rvv;
                 break;
 
+            case AV_PIX_FMT_BGRA:
+                c->lumToYV12 = ff_bgra32ToY_rvv;
+                break;
+
             case AV_PIX_FMT_RGB24:
                 c->lumToYV12 = ff_rgb24ToY_rvv;
                 if (c->chrSrcHSubSample)
@@ -58,6 +72,10 @@  av_cold void ff_sws_init_swscale_riscv(SwsContext *c)
                 else
                     c->chrToYV12 = ff_rgb24ToUV_rvv;
                 break;
+
+            case AV_PIX_FMT_RGBA:
+                c->lumToYV12 = ff_rgba32ToY_rvv;
+                break;
         }
     }
 #endif