diff options
author | Johann <johannkoenig@google.com> | 2017-02-03 15:57:28 -0800 |
---|---|---|
committer | Johann <johannkoenig@google.com> | 2017-02-16 07:40:55 -0800 |
commit | ac3996a6d1bf1f3696cffbf1102528643391ec19 (patch) | |
tree | fdbb9cfdb6cfbf15ac1b21cb03c7024e646c6731 /vp9/encoder/x86 | |
parent | 44600442dca48a5586105b01403ac03fad54d05b (diff) | |
download | libvpx-ac3996a6d1bf1f3696cffbf1102528643391ec19.tar libvpx-ac3996a6d1bf1f3696cffbf1102528643391ec19.tar.gz libvpx-ac3996a6d1bf1f3696cffbf1102528643391ec19.tar.bz2 libvpx-ac3996a6d1bf1f3696cffbf1102528643391ec19.zip |
quantize_fp highbd sse2: use tran_low_t for coeff
Change-Id: Id96a8df33354a7987ce890a3d6798c7375ffa4aa
Diffstat (limited to 'vp9/encoder/x86')
-rw-r--r-- | vp9/encoder/x86/vp9_quantize_sse2.c | 53 |
1 files changed, 28 insertions, 25 deletions
diff --git a/vp9/encoder/x86/vp9_quantize_sse2.c b/vp9/encoder/x86/vp9_quantize_sse2.c index 3f8ee5f24..954f8cb82 100644 --- a/vp9/encoder/x86/vp9_quantize_sse2.c +++ b/vp9/encoder/x86/vp9_quantize_sse2.c @@ -13,14 +13,16 @@ #include "./vp9_rtcd.h" #include "vpx/vpx_integer.h" +#include "vpx_dsp/vpx_dsp_common.h" +#include "vpx_dsp/x86/bitdepth_conversion_sse2.h" -void vp9_quantize_fp_sse2(const int16_t *coeff_ptr, intptr_t n_coeffs, +void vp9_quantize_fp_sse2(const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, - const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr, - int16_t *dqcoeff_ptr, const int16_t *dequant_ptr, - uint16_t *eob_ptr, const int16_t *scan_ptr, - const int16_t *iscan_ptr) { + const int16_t *quant_shift_ptr, + tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, + const int16_t *dequant_ptr, uint16_t *eob_ptr, + const int16_t *scan_ptr, const int16_t *iscan_ptr) { __m128i zero; __m128i thr; int16_t nzflag; @@ -53,8 +55,8 @@ void vp9_quantize_fp_sse2(const int16_t *coeff_ptr, intptr_t n_coeffs, __m128i qcoeff0, qcoeff1; __m128i qtmp0, qtmp1; // Do DC and first 15 AC - coeff0 = _mm_load_si128((const __m128i *)(coeff_ptr + n_coeffs)); - coeff1 = _mm_load_si128((const __m128i *)(coeff_ptr + n_coeffs) + 1); + coeff0 = load_tran_low(coeff_ptr + n_coeffs); + coeff1 = load_tran_low(coeff_ptr + n_coeffs + 8); // Poor man's sign extract coeff0_sign = _mm_srai_epi16(coeff0, 15); @@ -77,15 +79,15 @@ void vp9_quantize_fp_sse2(const int16_t *coeff_ptr, intptr_t n_coeffs, qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign); qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), qcoeff0); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, qcoeff1); + store_tran_low(qcoeff0, qcoeff_ptr + n_coeffs); + store_tran_low(qcoeff1, qcoeff_ptr + n_coeffs + 8); coeff0 = _mm_mullo_epi16(qcoeff0, dequant); dequant = _mm_unpackhi_epi64(dequant, dequant); coeff1 = _mm_mullo_epi16(qcoeff1, dequant); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), coeff0); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, coeff1); + store_tran_low(coeff0, dqcoeff_ptr + n_coeffs); + store_tran_low(coeff1, dqcoeff_ptr + n_coeffs + 8); } { @@ -120,8 +122,8 @@ void vp9_quantize_fp_sse2(const int16_t *coeff_ptr, intptr_t n_coeffs, __m128i qcoeff0, qcoeff1; __m128i qtmp0, qtmp1; - coeff0 = _mm_load_si128((const __m128i *)(coeff_ptr + n_coeffs)); - coeff1 = _mm_load_si128((const __m128i *)(coeff_ptr + n_coeffs) + 1); + coeff0 = load_tran_low(coeff_ptr + n_coeffs); + coeff1 = load_tran_low(coeff_ptr + n_coeffs + 8); // Poor man's sign extract coeff0_sign = _mm_srai_epi16(coeff0, 15); @@ -146,20 +148,20 @@ void vp9_quantize_fp_sse2(const int16_t *coeff_ptr, intptr_t n_coeffs, qcoeff0 = _mm_sub_epi16(qcoeff0, coeff0_sign); qcoeff1 = _mm_sub_epi16(qcoeff1, coeff1_sign); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), qcoeff0); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, qcoeff1); + store_tran_low(qcoeff0, qcoeff_ptr + n_coeffs); + store_tran_low(qcoeff1, qcoeff_ptr + n_coeffs + 8); coeff0 = _mm_mullo_epi16(qcoeff0, dequant); coeff1 = _mm_mullo_epi16(qcoeff1, dequant); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), coeff0); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, coeff1); + store_tran_low(coeff0, dqcoeff_ptr + n_coeffs); + store_tran_low(coeff1, dqcoeff_ptr + n_coeffs + 8); } else { - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), zero); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, zero); + store_zero_tran_low(qcoeff_ptr + n_coeffs); + store_zero_tran_low(qcoeff_ptr + n_coeffs + 8); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), zero); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, zero); + store_zero_tran_low(dqcoeff_ptr + n_coeffs); + store_zero_tran_low(dqcoeff_ptr + n_coeffs + 8); } } @@ -199,10 +201,11 @@ void vp9_quantize_fp_sse2(const int16_t *coeff_ptr, intptr_t n_coeffs, } } else { do { - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs), zero); - _mm_store_si128((__m128i *)(dqcoeff_ptr + n_coeffs) + 1, zero); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs), zero); - _mm_store_si128((__m128i *)(qcoeff_ptr + n_coeffs) + 1, zero); + store_zero_tran_low(qcoeff_ptr + n_coeffs); + store_zero_tran_low(qcoeff_ptr + n_coeffs + 8); + + store_zero_tran_low(dqcoeff_ptr + n_coeffs); + store_zero_tran_low(dqcoeff_ptr + n_coeffs + 8); n_coeffs += 8 * 2; } while (n_coeffs < 0); *eob_ptr = 0; |