From 52ae97b6aad982b51a121566282ccfc3d56452df Mon Sep 17 00:00:00 2001 From: Jingning Han Date: Thu, 1 May 2014 18:34:46 -0700 Subject: SSSE3 implementation of full inverse 8x8 2D-DCT This commit enables SSSE3 version full inverse 8x8 2D-DCT and reconstruction. It makes the runtime of vp9_idct8x8_64_add down from 256 cycles (SSE2) to 246 cycles. Change-Id: I0600feac894d6a443a3c9d18daf34156d4e225c3 --- vp9/common/vp9_rtcd_defs.pl | 2 +- vp9/common/x86/vp9_idct_ssse3.asm | 162 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 163 insertions(+), 1 deletion(-) create mode 100644 vp9/common/x86/vp9_idct_ssse3.asm (limited to 'vp9/common') diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl index d4c306507..954d75dd1 100644 --- a/vp9/common/vp9_rtcd_defs.pl +++ b/vp9/common/vp9_rtcd_defs.pl @@ -310,7 +310,7 @@ add_proto qw/void vp9_idct8x8_1_add/, "const int16_t *input, uint8_t *dest, int specialize qw/vp9_idct8x8_1_add sse2 neon dspr2/; add_proto qw/void vp9_idct8x8_64_add/, "const int16_t *input, uint8_t *dest, int dest_stride"; -specialize qw/vp9_idct8x8_64_add sse2 neon dspr2/; +specialize qw/vp9_idct8x8_64_add sse2 neon dspr2/, "$ssse3_x86_64"; add_proto qw/void vp9_idct8x8_10_add/, "const int16_t *input, uint8_t *dest, int dest_stride"; specialize qw/vp9_idct8x8_10_add sse2 neon dspr2/; diff --git a/vp9/common/x86/vp9_idct_ssse3.asm b/vp9/common/x86/vp9_idct_ssse3.asm new file mode 100644 index 000000000..f2a120f04 --- /dev/null +++ b/vp9/common/x86/vp9_idct_ssse3.asm @@ -0,0 +1,162 @@ +; +; Copyright (c) 2014 The WebM project authors. All Rights Reserved. +; +; Use of this source code is governed by a BSD-style license +; that can be found in the LICENSE file in the root of the source +; tree. An additional intellectual property rights grant can be found +; in the file PATENTS. All contributing project authors may +; be found in the AUTHORS file in the root of the source tree. +; +%include "third_party/x86inc/x86inc.asm" + +; This file provides SSSE3 version of the inverse transformation. Part +; of the functions are originally derived from the ffmpeg project. +; Note that the current version applies to x86 64-bit only. + +SECTION_RODATA + +pw_11585x2: times 8 dw 23170 +pd_8192: times 4 dd 8192 +pw_16: times 8 dw 16 + +%macro TRANSFORM_COEFFS 2 +pw_%1_%2: dw %1, %2, %1, %2, %1, %2, %1, %2 +pw_m%2_%1: dw -%2, %1, -%2, %1, -%2, %1, -%2, %1 +%endmacro + +TRANSFORM_COEFFS 6270, 15137 +TRANSFORM_COEFFS 3196, 16069 +TRANSFORM_COEFFS 13623, 9102 + +SECTION .text + +%if ARCH_X86_64 +%macro SUM_SUB 3 + psubw m%3, m%1, m%2 + paddw m%1, m%2 + SWAP %2, %3 +%endmacro + +; butterfly operation +%macro MUL_ADD_2X 6 ; dst1, dst2, src, round, coefs1, coefs2 + pmaddwd m%1, m%3, %5 + pmaddwd m%2, m%3, %6 + paddd m%1, %4 + paddd m%2, %4 + psrad m%1, 14 + psrad m%2, 14 +%endmacro + +%macro BUTTERFLY_4X 7 ; dst1, dst2, coef1, coef2, round, tmp1, tmp2 + punpckhwd m%6, m%2, m%1 + MUL_ADD_2X %7, %6, %6, %5, [pw_m%4_%3], [pw_%3_%4] + punpcklwd m%2, m%1 + MUL_ADD_2X %1, %2, %2, %5, [pw_m%4_%3], [pw_%3_%4] + packssdw m%1, m%7 + packssdw m%2, m%6 +%endmacro + +; matrix transpose +%macro INTERLEAVE_2X 4 + punpckh%1 m%4, m%2, m%3 + punpckl%1 m%2, m%3 + SWAP %3, %4 +%endmacro + +%macro TRANSPOSE8X8 9 + INTERLEAVE_2X wd, %1, %2, %9 + INTERLEAVE_2X wd, %3, %4, %9 + INTERLEAVE_2X wd, %5, %6, %9 + INTERLEAVE_2X wd, %7, %8, %9 + + INTERLEAVE_2X dq, %1, %3, %9 + INTERLEAVE_2X dq, %2, %4, %9 + INTERLEAVE_2X dq, %5, %7, %9 + INTERLEAVE_2X dq, %6, %8, %9 + + INTERLEAVE_2X qdq, %1, %5, %9 + INTERLEAVE_2X qdq, %3, %7, %9 + INTERLEAVE_2X qdq, %2, %6, %9 + INTERLEAVE_2X qdq, %4, %8, %9 + + SWAP %2, %5 + SWAP %4, %7 +%endmacro + +%macro IDCT8_1D 0 + SUM_SUB 0, 4, 9 + BUTTERFLY_4X 2, 6, 6270, 15137, m8, 9, 10 + pmulhrsw m0, m12 + pmulhrsw m4, m12 + BUTTERFLY_4X 1, 7, 3196, 16069, m8, 9, 10 + BUTTERFLY_4X 5, 3, 13623, 9102, m8, 9, 10 + + SUM_SUB 1, 5, 9 + SUM_SUB 7, 3, 9 + SUM_SUB 0, 6, 9 + SUM_SUB 4, 2, 9 + SUM_SUB 3, 5, 9 + pmulhrsw m3, m12 + pmulhrsw m5, m12 + + SUM_SUB 0, 7, 9 + SUM_SUB 4, 3, 9 + SUM_SUB 2, 5, 9 + SUM_SUB 6, 1, 9 + + SWAP 3, 6 + SWAP 1, 4 +%endmacro + +; This macro handles 8 pixels per line +%macro ADD_STORE_8P_2X 5; src1, src2, tmp1, tmp2, zero + paddw m%1, m11 + paddw m%2, m11 + psraw m%1, 5 + psraw m%2, 5 + + movh m%3, [outputq] + movh m%4, [outputq + strideq] + punpcklbw m%3, m%5 + punpcklbw m%4, m%5 + paddw m%3, m%1 + paddw m%4, m%2 + packuswb m%3, m%5 + packuswb m%4, m%5 + movh [outputq], m%3 + movh [outputq + strideq], m%4 +%endmacro + +INIT_XMM ssse3 +cglobal idct8x8_64_add, 3, 5, 13, input, output, stride + mova m8, [pd_8192] + mova m11, [pw_16] + mova m12, [pw_11585x2] + + lea r3, [2 * strideq] + + mova m0, [inputq + 0] + mova m1, [inputq + 16] + mova m2, [inputq + 32] + mova m3, [inputq + 48] + mova m4, [inputq + 64] + mova m5, [inputq + 80] + mova m6, [inputq + 96] + mova m7, [inputq + 112] + + TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9 + IDCT8_1D + TRANSPOSE8X8 0, 1, 2, 3, 4, 5, 6, 7, 9 + IDCT8_1D + + pxor m12, m12 + ADD_STORE_8P_2X 0, 1, 9, 10, 12 + lea outputq, [outputq + r3] + ADD_STORE_8P_2X 2, 3, 9, 10, 12 + lea outputq, [outputq + r3] + ADD_STORE_8P_2X 4, 5, 9, 10, 12 + lea outputq, [outputq + r3] + ADD_STORE_8P_2X 6, 7, 9, 10, 12 + + RET +%endif -- cgit v1.2.3