From 12059868a5c9192bb15aa08e8d25562d52d389dd Mon Sep 17 00:00:00 2001 From: kui zheng Date: Wed, 16 Nov 2011 17:08:45 +0800 Subject: [PATCH] generic: Add NEON version of Dacc_modulate_argb Signed-off-by: kui zheng --- src/gfx/generic/generic.c | 5 ++ src/gfx/generic/generic_neon.h | 93 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 98 insertions(+), 0 deletions(-) diff --git a/src/gfx/generic/generic.c b/src/gfx/generic/generic.c index 461da55..a697155 100644 --- a/src/gfx/generic/generic.c +++ b/src/gfx/generic/generic.c @@ -9668,6 +9668,11 @@ static void gInit_NEON( void ) /********************************* Xacc_blend *********************************/ Xacc_blend[DSBF_INVSRCALPHA-1] = Xacc_blend_invsrcalpha_NEON; Xacc_blend[DSBF_SRCALPHA-1] = Xacc_blend_srcalpha_NEON; +/********************************* Dacc_modulation ****************************/ + Dacc_modulation[DSBLIT_BLEND_ALPHACHANNEL | + DSBLIT_BLEND_COLORALPHA | + DSBLIT_COLORIZE] = Dacc_modulate_argb_NEON; + } #endif diff --git a/src/gfx/generic/generic_neon.h b/src/gfx/generic/generic_neon.h index e601b8f..79088e6 100644 --- a/src/gfx/generic/generic_neon.h +++ b/src/gfx/generic/generic_neon.h @@ -884,3 +884,96 @@ static void Xacc_blend_srcalpha_NEON( GenefxState *gfxs ) else Xacc_blend_srcalpha_Sacc0_NEON ( gfxs ); } + +/* + * NEON version of Dacc_modulate_argb_NEON. + */ +static void Dacc_modulate_argb_NEON( GenefxState *gfxs ) +{ + int w = gfxs->length; + GenefxAccumulator *D = gfxs->Dacc; + GenefxAccumulator Cacc = gfxs->Cacc; + unsigned int loop = w >> 3; + unsigned int single = w & 0x7; + u16 maska = 0xF000; + + while (single){ + if (!(D->RGB.a & 0xF000)) { + D->RGB.a = (Cacc.RGB.a * D->RGB.a) >> 8; + D->RGB.r = (Cacc.RGB.r * D->RGB.r) >> 8; + D->RGB.g = (Cacc.RGB.g * D->RGB.g) >> 8; + D->RGB.b = (Cacc.RGB.b * D->RGB.b) >> 8; + } + D++; + single--; + } + + if (loop) { + __asm__ __volatile__ ( + "mov r4, %[D] \n\t" + "mov r5, %[D] \n\t" + "vdup.16 q3, %[Cacc_a] \n\t" + "vdup.16 q2, %[Cacc_r] \n\t" + "vdup.16 q1, %[Cacc_g] \n\t" + "vdup.16 q0, %[Cacc_b] \n\t" + "vdup.16 q8, %[maska] \n\t" + "1: \n\t" + "pld [r4, #0xC0] \n\t" + "pld [r4, #0x100] \n\t" + /* vload q4:b, q5:g, q6:r, q7:a */ + "vld4.16 {d8, d10, d12, d14}, [r4]! \n\t" + "vld4.16 {d9, d11, d13, d15}, [r4]! \n\t" + + "vand q9, q7, q8 \n\t" + "vceq.i16 q9, q9, #0 \n\t" + /* b:q10 q0, q4 */ + "vmull.u16 q14, d8, d0 \n\t" + "vshrn.i32 d20, q14, #8 \n\t" + "vmull.u16 q14, d9, d1 \n\t" + "vshrn.i32 d21, q14, #8 \n\t" + /* g:q11, q1, q5 */ + "vmull.u16 q14, d10, d2 \n\t" + "vshrn.i32 d22, q14, #8 \n\t" + "vmull.u16 q14, d11, d3 \n\t" + "vshrn.i32 d23, q14, #8 \n\t" + /* r:q12, q2, q6 */ + "vmull.u16 q14, d12, d4 \n\t" + "vshrn.i32 d24, q14, #8 \n\t" + "vmull.u16 q14, d13, d5 \n\t" + "vshrn.i32 d25, q14, #8 \n\t" + /* a:q13, q3, q7 */ + "vmull.u16 q14, d14, d6 \n\t" + "vshrn.i32 d26, q14, #8 \n\t" + "vmull.u16 q14, d15, d7 \n\t" + "vshrn.i32 d27, q14, #8 \n\t" + /* if (!(D->RGB.a & 0xF000)) */ + "vand q10, q9, q10 \n\t" + "vand q11, q9, q11 \n\t" + "vand q12, q9, q12 \n\t" + "vand q13, q9, q13 \n\t" + /* if ((D->RGB.a & 0xF000)) */ + "vceq.i16 q9, q9, #0 \n\t" + "vand q4, q9, q4 \n\t" + "vand q5, q9, q5 \n\t" + "vand q6, q9, q6 \n\t" + "vand q7, q9, q7 \n\t" + /* Dacc: q4(b), q5(g), q6(r), q7(a) */ + "vorr q4, q4, q10 \n\t" + "vorr q5, q5, q11 \n\t" + "vorr q6, q6, q12 \n\t" + "vorr q7, q7, q13 \n\t" + "vst4.16 {d8, d10, d12, d14}, [r5]! \n\t" + "vst4.16 {d9, d11, d13, d15}, [r5]! \n\t" + "subs %[loop], %[loop], #1 \n\t" + "bne 1b " + : + : [Cacc_a] "r" (Cacc.RGB.a), [Cacc_r] "r" (Cacc.RGB.r), + [Cacc_g] "r" (Cacc.RGB.g), [Cacc_b] "r" (Cacc.RGB.b), + [D] "r" (D), [maska] "r" (maska), [loop] "r" (loop) + : "memory", "r4", "r5", "d0", "d1", "d2", "d3", "d4", "d5", "d6", "d7", + "d8", "d9", "d10", "d11", "d12", "d13", "d14", "d15", "d16", "d17", "d18", + "d19", "d20", "d21", "d22", "d23", "d24", "d25", "d26", "d27", "d28", "d29" + ); + } + +} -- 1.7.1