dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2025-04-21 08:16:46 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

wasm: replace _mulhi_int16x8() with neon builtin

Browse Source 
BUG=webp:352

Change-Id: I5c21ee4c631fb7eccff36c7d3cd47a72badf1a89
This commit is contained in:
Scott LaVarnway 2017-08-22 11:44:16 -07:00 committed by James Zern
parent 0e8c3004be
commit 8ed24a564c
1 changed files with 13 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
src/dsp/dec_wasm.c
View File

@ -108,10 +108,19 @@ static WEBP_INLINE int32x4 _unpackhi_epi64(const int32x4 a, const int32x4 b) {
// #define ENABLE_X86_BUILTIN_MULHI_INT16X8 // #define ENABLE_X86_BUILTIN_MULHI_INT16X8
#endif #endif
#if defined(__ARM_NEON__) || defined(__aarch64__)
// #define ENABLE_NEON_BUILTIN_MULHI_INT16X8
#endif
static WEBP_INLINE int16x8 _mulhi_int16x8(const int16x8 in, const int32x4 k) { static WEBP_INLINE int16x8 _mulhi_int16x8(const int16x8 in, const int32x4 k) {
#if defined(ENABLE_X86_BUILTIN_MULHI_INT16X8) #if defined(ENABLE_X86_BUILTIN_MULHI_INT16X8)
const int16x8 k_16bit = splat_int16(k[0]); const int16x8 k_16bit = splat_int16(k[0]);
return (int16x8)__builtin_ia32_pmulhw128(in, k_16bit); return (int16x8)__builtin_ia32_pmulhw128(in, k_16bit);
#elif defined(ENABLE_NEON_BUILTIN_MULHI_INT16X8)
const int16x8 k_16bit = splat_int16(k[0]);
const int16x8 one = (int16x8){1, 1, 1, 1, 1, 1, 1, 1};
return ((int16x8)__builtin_neon_vqdmulhq_v((int8x16_t)in, (int8x16_t)k_16bit,
33)) >> one;
#else #else
const int16x8 zero = (int16x8){0, 0, 0, 0, 0, 0, 0, 0}; const int16x8 zero = (int16x8){0, 0, 0, 0, 0, 0, 0, 0};
const int32x4 sixteen = (int32x4){16, 16, 16, 16}; const int32x4 sixteen = (int32x4){16, 16, 16, 16};
@ -255,7 +264,8 @@ static void Transform(const int16_t* in, uint8_t* dst, int do_two) {
const int16x8 a = in0 + in2; const int16x8 a = in0 + in2;
const int16x8 b = in0 - in2; const int16x8 b = in0 - in2;
#if defined(ENABLE_X86_BUILTIN_MULHI_INT16X8) #if defined(ENABLE_X86_BUILTIN_MULHI_INT16X8) || \
defined(ENABLE_NEON_BUILTIN_MULHI_INT16X8)
// c = MUL(in1, K2) - MUL(in3, K1) = MUL(in1, k2) - MUL(in3, k1) + in1 - in3 // c = MUL(in1, K2) - MUL(in3, K1) = MUL(in1, k2) - MUL(in3, k1) + in1 - in3
const int16x8 c1 = _mulhi_int16x8(in1, k2); const int16x8 c1 = _mulhi_int16x8(in1, k2);
const int16x8 c2 = _mulhi_int16x8(in3, k1); const int16x8 c2 = _mulhi_int16x8(in3, k1);
@ -293,7 +303,8 @@ static void Transform(const int16_t* in, uint8_t* dst, int do_two) {
const int16x8 dc = T0 + four; const int16x8 dc = T0 + four;
const int16x8 a = dc + T2; const int16x8 a = dc + T2;
const int16x8 b = dc - T2; const int16x8 b = dc - T2;
#if defined(ENABLE_X86_BUILTIN_MULHI_INT16X8) #if defined(ENABLE_X86_BUILTIN_MULHI_INT16X8) || \
defined(ENABLE_NEON_BUILTIN_MULHI_INT16X8)
// c = MUL(T1, K2) - MUL(T3, K1) = MUL(T1, k2) - MUL(T3, k1) + T1 - T3 // c = MUL(T1, K2) - MUL(T3, K1) = MUL(T1, k2) - MUL(T3, k1) + T1 - T3
const int16x8 c1 = _mulhi_int16x8(T1, k2); const int16x8 c1 = _mulhi_int16x8(T1, k2);
const int16x8 c2 = _mulhi_int16x8(T3, k1); const int16x8 c2 = _mulhi_int16x8(T3, k1);