geemili
/
SDL
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

SDL_stretch.c: compilation NEON on Windows 10

main
Sylvain Becker 2021-01-03 15:24:47 +01:00
parent acccc71f27
commit fea49a3e2f
1 changed files with 23 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
src/video/SDL_stretch.c
View File

@ -610,12 +610,16 @@ scale_mat(const Uint32 *src, int src_w, int src_h, int src_pitch,
#if defined(__ARM_NEON) #if defined(__ARM_NEON)
# define HAVE_NEON_INTRINSICS 1 # define HAVE_NEON_INTRINSICS 1
# define CAST_uint8x8_t (uint8x8_t)
# define CAST_uint32x2_t (uint32x2_t)
#endif #endif
/* TODO: this didn't compile on Window10 universal package last time I tried .. */
#if defined(__WINRT__) || defined(_MSC_VER) #if defined(__WINRT__) || defined(_MSC_VER)
# if defined(HAVE_NEON_INTRINSICS) # if defined(HAVE_NEON_INTRINSICS)
# undef HAVE_NEON_INTRINSICS # undef CAST_uint8x8_t
# undef CAST_uint32x2_t
# define CAST_uint8x8_t
# define CAST_uint32x2_t
# endif # endif
#endif #endif
@ -824,8 +828,8 @@ INTERPOL_BILINEAR_NEON(const Uint32 *s0, const Uint32 *s1, int frac_w, uint8x8_t
uint16x8_t d0; uint16x8_t d0;
uint8x8_t e0; uint8x8_t e0;
x_00_01 = (uint8x8_t)vld1_u32(s0); /* Load 2 pixels */ x_00_01 = CAST_uint8x8_t vld1_u32(s0); /* Load 2 pixels */
x_10_11 = (uint8x8_t)vld1_u32(s1); x_10_11 = CAST_uint8x8_t vld1_u32(s1);
/* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */ k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */
@ -846,7 +850,7 @@ INTERPOL_BILINEAR_NEON(const Uint32 *s0, const Uint32 *s1, int frac_w, uint8x8_t
e0 = vmovn_u16(d0); e0 = vmovn_u16(d0);
/* Store 1 pixel */ /* Store 1 pixel */
*dst = vget_lane_u32((uint32x2_t)e0, 0); *dst = vget_lane_u32(CAST_uint32x2_t e0, 0);
} }
static int static int
@ -911,14 +915,14 @@ scale_mat_NEON(const Uint32 *src, int src_w, int src_h, int src_pitch, Uint32 *d
s_16_17 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_3); s_16_17 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_3);
/* Interpolation vertical */ /* Interpolation vertical */
x_00_01 = (uint8x8_t)vld1_u32(s_00_01); /* Load 2 pixels */ x_00_01 = CAST_uint8x8_t vld1_u32(s_00_01); /* Load 2 pixels */
x_02_03 = (uint8x8_t)vld1_u32(s_02_03); x_02_03 = CAST_uint8x8_t vld1_u32(s_02_03);
x_04_05 = (uint8x8_t)vld1_u32(s_04_05); x_04_05 = CAST_uint8x8_t vld1_u32(s_04_05);
x_06_07 = (uint8x8_t)vld1_u32(s_06_07); x_06_07 = CAST_uint8x8_t vld1_u32(s_06_07);
x_10_11 = (uint8x8_t)vld1_u32(s_10_11); x_10_11 = CAST_uint8x8_t vld1_u32(s_10_11);
x_12_13 = (uint8x8_t)vld1_u32(s_12_13); x_12_13 = CAST_uint8x8_t vld1_u32(s_12_13);
x_14_15 = (uint8x8_t)vld1_u32(s_14_15); x_14_15 = CAST_uint8x8_t vld1_u32(s_14_15);
x_16_17 = (uint8x8_t)vld1_u32(s_16_17); x_16_17 = CAST_uint8x8_t vld1_u32(s_16_17);
/* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */ k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */
@ -970,7 +974,7 @@ scale_mat_NEON(const Uint32 *src, int src_w, int src_h, int src_pitch, Uint32 *d
/* Narrow again */ /* Narrow again */
e1 = vmovn_u16(d1); e1 = vmovn_u16(d1);
f0 = vcombine_u32((uint32x2_t)e0, (uint32x2_t)e1); f0 = vcombine_u32(CAST_uint32x2_t e0, CAST_uint32x2_t e1);
/* Store 4 pixels */ /* Store 4 pixels */
vst1q_u32(dst, f0); vst1q_u32(dst, f0);
@ -1009,10 +1013,10 @@ scale_mat_NEON(const Uint32 *src, int src_w, int src_h, int src_pitch, Uint32 *d
s_12_13 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_1); s_12_13 = (const Uint32 *)((const Uint8 *)src_h1 + index_w_1);
/* Interpolation vertical */ /* Interpolation vertical */
x_00_01 = (uint8x8_t)vld1_u32(s_00_01);/* Load 2 pixels */ x_00_01 = CAST_uint8x8_t vld1_u32(s_00_01);/* Load 2 pixels */
x_02_03 = (uint8x8_t)vld1_u32(s_02_03); x_02_03 = CAST_uint8x8_t vld1_u32(s_02_03);
x_10_11 = (uint8x8_t)vld1_u32(s_10_11); x_10_11 = CAST_uint8x8_t vld1_u32(s_10_11);
x_12_13 = (uint8x8_t)vld1_u32(s_12_13); x_12_13 = CAST_uint8x8_t vld1_u32(s_12_13);
/* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */ /* Interpolated == x0 + frac * (x1 - x0) == x0 * (1 - frac) + x1 * frac */
k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */ k0 = vmull_u8(x_00_01, v_frac_h1); /* k0 := x0 * (1 - frac) */
@ -1043,7 +1047,7 @@ scale_mat_NEON(const Uint32 *src, int src_w, int src_h, int src_pitch, Uint32 *d
e0 = vmovn_u16(d0); e0 = vmovn_u16(d0);
/* Store 2 pixels */ /* Store 2 pixels */
vst1_u32(dst, (uint32x2_t)e0); vst1_u32(dst, CAST_uint32x2_t e0);
dst += 2; dst += 2;
} }