ref: e022ce84acf7604e151ab7ef07203e1aa9d6ea61
parent: d48be6ad734890e0093d25df61b2943d1f402c0d
author: Johann <johannkoenig@google.com>
date: Fri Aug 11 06:44:36 EDT 2017
Rename vp8 quantize file BUG=webm:1457 Change-Id: Ie8fae018ad8417724fde087055b90228850d631d
--- a/vp8/encoder/x86/quantize_ssse3.c
+++ /dev/null
@@ -1,114 +1,0 @@
-/*
- * Copyright (c) 2012 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 <tmmintrin.h> /* SSSE3 */
-
-#include "./vp8_rtcd.h"
-#include "vp8/encoder/block.h"
-
-/* bitscan reverse (bsr) */
-#if defined(_MSC_VER)
-#include <intrin.h>
-#pragma intrinsic(_BitScanReverse)
-static int bsr(int mask) {- unsigned long eob;
- _BitScanReverse(&eob, mask);
- eob++;
- if (mask == 0) eob = 0;
- return eob;
-}
-#else
-static int bsr(int mask) {- int eob;
-#if defined(__GNUC__) && __GNUC__
- __asm__ __volatile__("bsr %1, %0" : "=r"(eob) : "r"(mask) : "flags");-#elif defined(__SUNPRO_C) || defined(__SUNPRO_CC)
- asm volatile("bsr %1, %0" : "=r"(eob) : "r"(mask) : "flags");-#endif
- eob++;
- if (mask == 0) eob = 0;
- return eob;
-}
-#endif
-
-void vp8_fast_quantize_b_ssse3(BLOCK *b, BLOCKD *d) {- int eob, mask;
-
- __m128i z0 = _mm_load_si128((__m128i *)(b->coeff));
- __m128i z1 = _mm_load_si128((__m128i *)(b->coeff + 8));
- __m128i round0 = _mm_load_si128((__m128i *)(b->round));
- __m128i round1 = _mm_load_si128((__m128i *)(b->round + 8));
- __m128i quant_fast0 = _mm_load_si128((__m128i *)(b->quant_fast));
- __m128i quant_fast1 = _mm_load_si128((__m128i *)(b->quant_fast + 8));
- __m128i dequant0 = _mm_load_si128((__m128i *)(d->dequant));
- __m128i dequant1 = _mm_load_si128((__m128i *)(d->dequant + 8));
-
- __m128i sz0, sz1, x, x0, x1, y0, y1, zeros, abs0, abs1;
-
- DECLARE_ALIGNED(16, const uint8_t, pshufb_zig_zag_mask[16]) = {- 0, 1, 4, 8, 5, 2, 3, 6, 9, 12, 13, 10, 7, 11, 14, 15
- };
- __m128i zig_zag = _mm_load_si128((const __m128i *)pshufb_zig_zag_mask);
-
- /* sign of z: z >> 15 */
- sz0 = _mm_srai_epi16(z0, 15);
- sz1 = _mm_srai_epi16(z1, 15);
-
- /* x = abs(z) */
- x0 = _mm_abs_epi16(z0);
- x1 = _mm_abs_epi16(z1);
-
- /* x += round */
- x0 = _mm_add_epi16(x0, round0);
- x1 = _mm_add_epi16(x1, round1);
-
- /* y = (x * quant) >> 16 */
- y0 = _mm_mulhi_epi16(x0, quant_fast0);
- y1 = _mm_mulhi_epi16(x1, quant_fast1);
-
- /* ASM saves Y for EOB */
- /* I think we can ignore that because adding the sign doesn't change anything
- * and multiplying 0 by dequant is OK as well */
- abs0 = y0;
- abs1 = y1;
-
- /* Restore the sign bit. */
- y0 = _mm_xor_si128(y0, sz0);
- y1 = _mm_xor_si128(y1, sz1);
- x0 = _mm_sub_epi16(y0, sz0);
- x1 = _mm_sub_epi16(y1, sz1);
-
- /* qcoeff = x */
- _mm_store_si128((__m128i *)(d->qcoeff), x0);
- _mm_store_si128((__m128i *)(d->qcoeff + 8), x1);
-
- /* x * dequant */
- x0 = _mm_mullo_epi16(x0, dequant0);
- x1 = _mm_mullo_epi16(x1, dequant1);
-
- /* dqcoeff = x * dequant */
- _mm_store_si128((__m128i *)(d->dqcoeff), x0);
- _mm_store_si128((__m128i *)(d->dqcoeff + 8), x1);
-
- zeros = _mm_setzero_si128();
-
- x0 = _mm_cmpgt_epi16(abs0, zeros);
- x1 = _mm_cmpgt_epi16(abs1, zeros);
-
- x = _mm_packs_epi16(x0, x1);
-
- x = _mm_shuffle_epi8(x, zig_zag);
-
- mask = _mm_movemask_epi8(x);
-
- eob = bsr(mask);
-
- *d->eob = 0xFF & eob;
-}
--- /dev/null
+++ b/vp8/encoder/x86/vp8_quantize_ssse3.c
@@ -1,0 +1,114 @@
+/*
+ * Copyright (c) 2012 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 <tmmintrin.h> /* SSSE3 */
+
+#include "./vp8_rtcd.h"
+#include "vp8/encoder/block.h"
+
+/* bitscan reverse (bsr) */
+#if defined(_MSC_VER)
+#include <intrin.h>
+#pragma intrinsic(_BitScanReverse)
+static int bsr(int mask) {+ unsigned long eob;
+ _BitScanReverse(&eob, mask);
+ eob++;
+ if (mask == 0) eob = 0;
+ return eob;
+}
+#else
+static int bsr(int mask) {+ int eob;
+#if defined(__GNUC__) && __GNUC__
+ __asm__ __volatile__("bsr %1, %0" : "=r"(eob) : "r"(mask) : "flags");+#elif defined(__SUNPRO_C) || defined(__SUNPRO_CC)
+ asm volatile("bsr %1, %0" : "=r"(eob) : "r"(mask) : "flags");+#endif
+ eob++;
+ if (mask == 0) eob = 0;
+ return eob;
+}
+#endif
+
+void vp8_fast_quantize_b_ssse3(BLOCK *b, BLOCKD *d) {+ int eob, mask;
+
+ __m128i z0 = _mm_load_si128((__m128i *)(b->coeff));
+ __m128i z1 = _mm_load_si128((__m128i *)(b->coeff + 8));
+ __m128i round0 = _mm_load_si128((__m128i *)(b->round));
+ __m128i round1 = _mm_load_si128((__m128i *)(b->round + 8));
+ __m128i quant_fast0 = _mm_load_si128((__m128i *)(b->quant_fast));
+ __m128i quant_fast1 = _mm_load_si128((__m128i *)(b->quant_fast + 8));
+ __m128i dequant0 = _mm_load_si128((__m128i *)(d->dequant));
+ __m128i dequant1 = _mm_load_si128((__m128i *)(d->dequant + 8));
+
+ __m128i sz0, sz1, x, x0, x1, y0, y1, zeros, abs0, abs1;
+
+ DECLARE_ALIGNED(16, const uint8_t, pshufb_zig_zag_mask[16]) = {+ 0, 1, 4, 8, 5, 2, 3, 6, 9, 12, 13, 10, 7, 11, 14, 15
+ };
+ __m128i zig_zag = _mm_load_si128((const __m128i *)pshufb_zig_zag_mask);
+
+ /* sign of z: z >> 15 */
+ sz0 = _mm_srai_epi16(z0, 15);
+ sz1 = _mm_srai_epi16(z1, 15);
+
+ /* x = abs(z) */
+ x0 = _mm_abs_epi16(z0);
+ x1 = _mm_abs_epi16(z1);
+
+ /* x += round */
+ x0 = _mm_add_epi16(x0, round0);
+ x1 = _mm_add_epi16(x1, round1);
+
+ /* y = (x * quant) >> 16 */
+ y0 = _mm_mulhi_epi16(x0, quant_fast0);
+ y1 = _mm_mulhi_epi16(x1, quant_fast1);
+
+ /* ASM saves Y for EOB */
+ /* I think we can ignore that because adding the sign doesn't change anything
+ * and multiplying 0 by dequant is OK as well */
+ abs0 = y0;
+ abs1 = y1;
+
+ /* Restore the sign bit. */
+ y0 = _mm_xor_si128(y0, sz0);
+ y1 = _mm_xor_si128(y1, sz1);
+ x0 = _mm_sub_epi16(y0, sz0);
+ x1 = _mm_sub_epi16(y1, sz1);
+
+ /* qcoeff = x */
+ _mm_store_si128((__m128i *)(d->qcoeff), x0);
+ _mm_store_si128((__m128i *)(d->qcoeff + 8), x1);
+
+ /* x * dequant */
+ x0 = _mm_mullo_epi16(x0, dequant0);
+ x1 = _mm_mullo_epi16(x1, dequant1);
+
+ /* dqcoeff = x * dequant */
+ _mm_store_si128((__m128i *)(d->dqcoeff), x0);
+ _mm_store_si128((__m128i *)(d->dqcoeff + 8), x1);
+
+ zeros = _mm_setzero_si128();
+
+ x0 = _mm_cmpgt_epi16(abs0, zeros);
+ x1 = _mm_cmpgt_epi16(abs1, zeros);
+
+ x = _mm_packs_epi16(x0, x1);
+
+ x = _mm_shuffle_epi8(x, zig_zag);
+
+ mask = _mm_movemask_epi8(x);
+
+ eob = bsr(mask);
+
+ *d->eob = 0xFF & eob;
+}
--- a/vp8/vp8cx.mk
+++ b/vp8/vp8cx.mk
@@ -85,7 +85,7 @@
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/dct_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/fwalsh_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp8_quantize_sse2.c
-VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/quantize_ssse3.c
+VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/vp8_quantize_ssse3.c
VP8_CX_SRCS-$(HAVE_SSE4_1) += encoder/x86/quantize_sse4.c
ifeq ($(CONFIG_TEMPORAL_DENOISING),yes)