Commit b71962fd authored by Timothy B. Terriberry's avatar Timothy B. Terriberry Committed by Johann
Browse files

Add runtime CPU detection support for ARM.

The primary goal is to allow a binary to be built which supports
 NEON, but can fall back to non-NEON routines, since some Android
 devices do not have NEON, even if they are otherwise ARMv7 (e.g.,
 Tegra).
The configure-generated flags HAVE_ARMV7, etc., are used to decide
 which versions of each function to build, and when
 CONFIG_RUNTIME_CPU_DETECT is enabled, the correct version is chosen
 at run time.
In order for this to work, the CFLAGS must be set to something
 appropriate (e.g., without -mfpu=neon for ARMv7, and with
 appropriate -march and -mcpu for even earlier configurations), or
 the native C code will not be able to run.
The ASFLAGS must remain set for the most advanced instruction set
 required at build time, since the ARM assembler will refuse to emit
 them otherwise.
I have not attempted to make any changes to configure to do this
 automatically.
Doing so will probably require the addition of new configure options.

Many of the hooks for RTCD on ARM were already there, but a lot of
 the code had bit-rotted, and a good deal of the ARM-specific code
 is not integrated into the RTCD structs at all.
I did not try to resolve the latter, merely to add the minimal amount
 of protection around them to allow RTCD to work.
Those functions that were called based on an ifdef at the calling
 site were expanded to check the RTCD flags at that site, but they
 should be added to an RTCD struct somewhere in the future.
The functions invoked with global function pointers still are, but
 these should be moved into an RTCD struct for thread safety (I
 believe every platform currently supported has atomic pointer
 stores, but this is not guaranteed).

The encoder's boolhuff functions did not even have _c and armv7
 suffixes, and the correct version was resolved at link time.
The token packing functions did have appropriate suffixes, but the
 version was selected with a define, with no associated RTCD struct.
However, for both of these, the only armv7 instruction they actually
 used was rbit, and this was completely superfluous, so I reworked
 them to avoid it.
The only non-ARMv4 instruction remaining in them is clz, which is
 ARMv5 (not even ARMv5TE is required).
Considering that there are no ARM-specific configs which are not at
 least ARMv5TE, I did not try to detect these at runtime, and simply
 enable them for ARMv5 and above.

Finally, the NEON register saving code was completely non-reentrant,
 since it saved the registers to a global, static variable.
I moved the storage for this onto the stack.
A single binary built with this code was tested on an ARM11 (ARMv6)
 and a Cortex A8 (ARMv7 w/NEON), for both the encoder and decoder,
 and produced identical output, while using the correct accelerated
 functions on each.
I did not test on any earlier processors.

Change-Id: I45cbd63a614f4554c3b325c45d46c0806f009eaa
parent e81e30c2
......@@ -93,6 +93,7 @@ CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/x86.h
CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/x86_abi_support.asm
CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/x86_cpuid.c
endif
CODEC_SRCS-$(ARCH_ARM) += vpx_ports/arm_cpudetect.c
CODEC_SRCS-$(ARCH_ARM) += $(BUILD_PFX)vpx_config.asm
CODEC_EXPORTS-$(BUILD_LIBVPX) += vpx/exports_com
CODEC_EXPORTS-$(CONFIG_ENCODERS) += vpx/exports_enc
......
/*
* Copyright (c) 2010 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 "vpx_ports/config.h"
#include "vpx_ports/arm.h"
#include "g_common.h"
#include "pragmas.h"
#include "subpixel.h"
#include "loopfilter.h"
#include "recon.h"
#include "idct.h"
#include "onyxc_int.h"
extern void (*vp8_build_intra_predictors_mby_ptr)(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby_neon(MACROBLOCKD *x);
extern void (*vp8_build_intra_predictors_mby_s_ptr)(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby_s_neon(MACROBLOCKD *x);
void vp8_arch_arm_common_init(VP8_COMMON *ctx)
{
#if CONFIG_RUNTIME_CPU_DETECT
VP8_COMMON_RTCD *rtcd = &ctx->rtcd;
int flags = arm_cpu_caps();
int has_edsp = flags & HAS_EDSP;
int has_media = flags & HAS_MEDIA;
int has_neon = flags & HAS_NEON;
rtcd->flags = flags;
/* Override default functions with fastest ones for this CPU. */
#if HAVE_ARMV6
if (has_media)
{
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_armv6;
rtcd->subpix.sixtap8x8 = vp8_sixtap_predict8x8_armv6;
rtcd->subpix.sixtap8x4 = vp8_sixtap_predict8x4_armv6;
rtcd->subpix.sixtap4x4 = vp8_sixtap_predict_armv6;
rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_armv6;
rtcd->subpix.bilinear8x8 = vp8_bilinear_predict8x8_armv6;
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_armv6;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_armv6;
rtcd->idct.idct1 = vp8_short_idct4x4llm_1_v6;
rtcd->idct.idct16 = vp8_short_idct4x4llm_v6_dual;
rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_v6;
rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_v6;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_armv6;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_armv6;
rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_armv6;
rtcd->loopfilter.normal_b_h = vp8_loop_filter_bh_armv6;
rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_armv6;
rtcd->loopfilter.simple_b_v = vp8_loop_filter_bvs_armv6;
rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_armv6;
rtcd->loopfilter.simple_b_h = vp8_loop_filter_bhs_armv6;
rtcd->recon.copy16x16 = vp8_copy_mem16x16_v6;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_v6;
rtcd->recon.copy8x4 = vp8_copy_mem8x4_v6;
rtcd->recon.recon = vp8_recon_b_armv6;
rtcd->recon.recon2 = vp8_recon2b_armv6;
rtcd->recon.recon4 = vp8_recon4b_armv6;
}
#endif
#if HAVE_ARMV7
if (has_neon)
{
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_neon;
rtcd->subpix.sixtap8x8 = vp8_sixtap_predict8x8_neon;
rtcd->subpix.sixtap8x4 = vp8_sixtap_predict8x4_neon;
rtcd->subpix.sixtap4x4 = vp8_sixtap_predict_neon;
rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_neon;
rtcd->subpix.bilinear8x8 = vp8_bilinear_predict8x8_neon;
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_neon;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_neon;
rtcd->idct.idct1 = vp8_short_idct4x4llm_1_neon;
rtcd->idct.idct16 = vp8_short_idct4x4llm_neon;
rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_neon;
rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_neon;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_neon;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_neon;
rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_neon;
rtcd->loopfilter.normal_b_h = vp8_loop_filter_bh_neon;
rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_neon;
rtcd->loopfilter.simple_b_v = vp8_loop_filter_bvs_neon;
rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_neon;
rtcd->loopfilter.simple_b_h = vp8_loop_filter_bhs_neon;
rtcd->recon.copy16x16 = vp8_copy_mem16x16_neon;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_neon;
rtcd->recon.copy8x4 = vp8_copy_mem8x4_neon;
rtcd->recon.recon = vp8_recon_b_neon;
rtcd->recon.recon2 = vp8_recon2b_neon;
rtcd->recon.recon4 = vp8_recon4b_neon;
}
#endif
#endif
#if HAVE_ARMV6
#if CONFIG_RUNTIME_CPU_DETECT
if (has_media)
#endif
{
vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby;
vp8_build_intra_predictors_mby_s_ptr = vp8_build_intra_predictors_mby_s;
}
#endif
#if HAVE_ARMV7
#if CONFIG_RUNTIME_CPU_DETECT
if (has_neon)
#endif
{
vp8_build_intra_predictors_mby_ptr =
vp8_build_intra_predictors_mby_neon;
vp8_build_intra_predictors_mby_s_ptr =
vp8_build_intra_predictors_mby_s_neon;
}
#endif
}
......@@ -19,6 +19,7 @@ extern prototype_idct_scalar_add(vp8_dc_only_idct_add_v6);
extern prototype_second_order(vp8_short_inv_walsh4x4_1_v6);
extern prototype_second_order(vp8_short_inv_walsh4x4_v6);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_idct_idct1
#define vp8_idct_idct1 vp8_short_idct4x4llm_1_v6
......@@ -34,6 +35,7 @@ extern prototype_second_order(vp8_short_inv_walsh4x4_v6);
#undef vp8_idct_iwalsh16
#define vp8_idct_iwalsh16 vp8_short_inv_walsh4x4_v6
#endif
#endif
#if HAVE_ARMV7
extern prototype_idct(vp8_short_idct4x4llm_1_neon);
......@@ -42,6 +44,7 @@ extern prototype_idct_scalar_add(vp8_dc_only_idct_add_neon);
extern prototype_second_order(vp8_short_inv_walsh4x4_1_neon);
extern prototype_second_order(vp8_short_inv_walsh4x4_neon);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_idct_idct1
#define vp8_idct_idct1 vp8_short_idct4x4llm_1_neon
......@@ -57,5 +60,6 @@ extern prototype_second_order(vp8_short_inv_walsh4x4_neon);
#undef vp8_idct_iwalsh16
#define vp8_idct_iwalsh16 vp8_short_inv_walsh4x4_neon
#endif
#endif
#endif
......@@ -22,6 +22,7 @@ extern prototype_loopfilter_block(vp8_loop_filter_bvs_armv6);
extern prototype_loopfilter_block(vp8_loop_filter_mbhs_armv6);
extern prototype_loopfilter_block(vp8_loop_filter_bhs_armv6);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_lf_normal_mb_v
#define vp8_lf_normal_mb_v vp8_loop_filter_mbv_armv6
......@@ -46,6 +47,7 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_armv6);
#undef vp8_lf_simple_b_h
#define vp8_lf_simple_b_h vp8_loop_filter_bhs_armv6
#endif
#endif
#if HAVE_ARMV7
extern prototype_loopfilter_block(vp8_loop_filter_mbv_neon);
......@@ -57,6 +59,7 @@ extern prototype_loopfilter_block(vp8_loop_filter_bvs_neon);
extern prototype_loopfilter_block(vp8_loop_filter_mbhs_neon);
extern prototype_loopfilter_block(vp8_loop_filter_bhs_neon);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_lf_normal_mb_v
#define vp8_lf_normal_mb_v vp8_loop_filter_mbv_neon
......@@ -81,5 +84,6 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_neon);
#undef vp8_lf_simple_b_h
#define vp8_lf_simple_b_h vp8_loop_filter_bhs_neon
#endif
#endif
#endif
......@@ -21,6 +21,7 @@ extern prototype_copy_block(vp8_copy_mem8x8_v6);
extern prototype_copy_block(vp8_copy_mem8x4_v6);
extern prototype_copy_block(vp8_copy_mem16x16_v6);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_recon_recon
#define vp8_recon_recon vp8_recon_b_armv6
......@@ -39,6 +40,7 @@ extern prototype_copy_block(vp8_copy_mem16x16_v6);
#undef vp8_recon_copy16x16
#define vp8_recon_copy16x16 vp8_copy_mem16x16_v6
#endif
#endif
#if HAVE_ARMV7
extern prototype_recon_block(vp8_recon_b_neon);
......@@ -49,6 +51,7 @@ extern prototype_copy_block(vp8_copy_mem8x8_neon);
extern prototype_copy_block(vp8_copy_mem8x4_neon);
extern prototype_copy_block(vp8_copy_mem16x16_neon);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_recon_recon
#define vp8_recon_recon vp8_recon_b_neon
......@@ -67,5 +70,6 @@ extern prototype_copy_block(vp8_copy_mem16x16_neon);
#undef vp8_recon_copy16x16
#define vp8_recon_copy16x16 vp8_copy_mem16x16_neon
#endif
#endif
#endif
......@@ -22,6 +22,7 @@ extern prototype_subpixel_predict(vp8_bilinear_predict8x8_armv6);
extern prototype_subpixel_predict(vp8_bilinear_predict8x4_armv6);
extern prototype_subpixel_predict(vp8_bilinear_predict4x4_armv6);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_subpix_sixtap16x16
#define vp8_subpix_sixtap16x16 vp8_sixtap_predict16x16_armv6
......@@ -46,6 +47,7 @@ extern prototype_subpixel_predict(vp8_bilinear_predict4x4_armv6);
#undef vp8_subpix_bilinear4x4
#define vp8_subpix_bilinear4x4 vp8_bilinear_predict4x4_armv6
#endif
#endif
#if HAVE_ARMV7
extern prototype_subpixel_predict(vp8_sixtap_predict16x16_neon);
......@@ -57,6 +59,7 @@ extern prototype_subpixel_predict(vp8_bilinear_predict8x8_neon);
extern prototype_subpixel_predict(vp8_bilinear_predict8x4_neon);
extern prototype_subpixel_predict(vp8_bilinear_predict4x4_neon);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_subpix_sixtap16x16
#define vp8_subpix_sixtap16x16 vp8_sixtap_predict16x16_neon
......@@ -81,5 +84,6 @@ extern prototype_subpixel_predict(vp8_bilinear_predict4x4_neon);
#undef vp8_subpix_bilinear4x4
#define vp8_subpix_bilinear4x4 vp8_bilinear_predict4x4_neon
#endif
#endif
#endif
/*
* Copyright (c) 2010 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 "vpx_ports/config.h"
#include "g_common.h"
#include "pragmas.h"
#include "subpixel.h"
#include "loopfilter.h"
#include "recon.h"
#include "idct.h"
#include "onyxc_int.h"
void (*vp8_build_intra_predictors_mby_ptr)(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby_neon(MACROBLOCKD *x);
void (*vp8_build_intra_predictors_mby_s_ptr)(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby_s_neon(MACROBLOCKD *x);
void vp8_machine_specific_config(VP8_COMMON *ctx)
{
#if CONFIG_RUNTIME_CPU_DETECT
VP8_COMMON_RTCD *rtcd = &ctx->rtcd;
#if HAVE_ARMV7
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_neon;
rtcd->subpix.sixtap8x8 = vp8_sixtap_predict8x8_neon;
rtcd->subpix.sixtap8x4 = vp8_sixtap_predict8x4_neon;
rtcd->subpix.sixtap4x4 = vp8_sixtap_predict_neon;
rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_neon;
rtcd->subpix.bilinear8x8 = vp8_bilinear_predict8x8_neon;
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_neon;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_neon;
rtcd->idct.idct1 = vp8_short_idct4x4llm_1_neon;
rtcd->idct.idct16 = vp8_short_idct4x4llm_neon;
rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_neon;
rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_neon;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_neon;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_neon;
rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_neon;
rtcd->loopfilter.normal_b_h = vp8_loop_filter_bh_neon;
rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_neon;
rtcd->loopfilter.simple_b_v = vp8_loop_filter_bvs_neon;
rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_neon;
rtcd->loopfilter.simple_b_h = vp8_loop_filter_bhs_neon;
rtcd->recon.copy16x16 = vp8_copy_mem16x16_neon;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_neon;
rtcd->recon.copy8x4 = vp8_copy_mem8x4_neon;
rtcd->recon.recon = vp8_recon_b_neon;
rtcd->recon.recon2 = vp8_recon2b_neon;
rtcd->recon.recon4 = vp8_recon4b_neon;
#elif HAVE_ARMV6
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_armv6;
rtcd->subpix.sixtap8x8 = vp8_sixtap_predict8x8_armv6;
rtcd->subpix.sixtap8x4 = vp8_sixtap_predict8x4_armv6;
rtcd->subpix.sixtap4x4 = vp8_sixtap_predict_armv6;
rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_armv6;
rtcd->subpix.bilinear8x8 = vp8_bilinear_predict8x8_armv6;
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_armv6;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_armv6;
rtcd->idct.idct1 = vp8_short_idct4x4llm_1_v6;
rtcd->idct.idct16 = vp8_short_idct4x4llm_v6_dual;
rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_armv6;
rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_armv6;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_armv6;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_armv6;
rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_armv6;
rtcd->loopfilter.normal_b_h = vp8_loop_filter_bh_armv6;
rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_armv6;
rtcd->loopfilter.simple_b_v = vp8_loop_filter_bvs_armv6;
rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_armv6;
rtcd->loopfilter.simple_b_h = vp8_loop_filter_bhs_armv6;
rtcd->recon.copy16x16 = vp8_copy_mem16x16_v6;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_v6;
rtcd->recon.copy8x4 = vp8_copy_mem8x4_v6;
rtcd->recon.recon = vp8_recon_b_armv6;
rtcd->recon.recon2 = vp8_recon2b_armv6;
rtcd->recon.recon4 = vp8_recon4b_armv6;
#else
//pure c
rtcd->idct.idct1 = vp8_short_idct4x4llm_1_c;
rtcd->idct.idct16 = vp8_short_idct4x4llm_c;
rtcd->idct.idct1_scalar = vp8_dc_only_idct_c;
rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_c;
rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_c;
rtcd->recon.copy16x16 = vp8_copy_mem16x16_c;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_c;
rtcd->recon.copy8x4 = vp8_copy_mem8x4_c;
rtcd->recon.recon = vp8_recon_b_c;
rtcd->recon.recon2 = vp8_recon2b_c;
rtcd->recon.recon4 = vp8_recon4b_c;
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_c;
rtcd->subpix.sixtap8x8 = vp8_sixtap_predict8x8_c;
rtcd->subpix.sixtap8x4 = vp8_sixtap_predict8x4_c;
rtcd->subpix.sixtap4x4 = vp8_sixtap_predict_c;
rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_c;
rtcd->subpix.bilinear8x8 = vp8_bilinear_predict8x8_c;
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_c;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_c;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_c;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_c;
rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_c;
rtcd->loopfilter.normal_b_h = vp8_loop_filter_bh_c;
rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_c;
rtcd->loopfilter.simple_b_v = vp8_loop_filter_bvs_c;
rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_c;
rtcd->loopfilter.simple_b_h = vp8_loop_filter_bhs_c;
#endif
#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_PSNR)
rtcd->postproc.down = vp8_mbpost_proc_down_c;
rtcd->postproc.across = vp8_mbpost_proc_across_ip_c;
rtcd->postproc.downacross = vp8_post_proc_down_and_across_c;
rtcd->postproc.addnoise = vp8_plane_add_noise_c;
#endif
#endif
#if HAVE_ARMV7
vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby_neon;
vp8_build_intra_predictors_mby_s_ptr = vp8_build_intra_predictors_mby_s_neon;
#elif HAVE_ARMV6
vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby;
vp8_build_intra_predictors_mby_s_ptr = vp8_build_intra_predictors_mby_s;
#else
vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby;
vp8_build_intra_predictors_mby_s_ptr = vp8_build_intra_predictors_mby_s;
#endif
}
......@@ -18,6 +18,7 @@
#include "onyxc_int.h"
extern void vp8_arch_x86_common_init(VP8_COMMON *ctx);
extern void vp8_arch_arm_common_init(VP8_COMMON *ctx);
void (*vp8_build_intra_predictors_mby_ptr)(MACROBLOCKD *x);
extern void vp8_build_intra_predictors_mby(MACROBLOCKD *x);
......@@ -77,4 +78,8 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
vp8_arch_x86_common_init(ctx);
#endif
#if ARCH_ARM
vp8_arch_arm_common_init(ctx);
#endif
}
......@@ -74,6 +74,7 @@ typedef struct VP8_COMMON_RTCD
vp8_subpix_rtcd_vtable_t subpix;
vp8_loopfilter_rtcd_vtable_t loopfilter;
vp8_postproc_rtcd_vtable_t postproc;
int flags;
#else
int unused;
#endif
......
/*
* Copyright (c) 2010 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 "vpx_ports/config.h"
#include "vpx_ports/arm.h"
#include "blockd.h"
#include "pragmas.h"
#include "postproc.h"
#include "dboolhuff.h"
#include "dequantize.h"
#include "onyxd_int.h"
void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
{
#if CONFIG_RUNTIME_CPU_DETECT
int flags = pbi->common.rtcd.flags;
int has_edsp = flags & HAS_EDSP;
int has_media = flags & HAS_MEDIA;
int has_neon = flags & HAS_NEON;
#if HAVE_ARMV6
if (has_media)
{
pbi->dequant.block = vp8_dequantize_b_v6;
pbi->dequant.idct_add = vp8_dequant_idct_add_v6;
pbi->dequant.dc_idct_add = vp8_dequant_dc_idct_add_v6;
pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_v6;
pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_v6;
pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_v6;
#if 0 //For use with RTCD, when implemented
pbi->dboolhuff.start = vp8dx_start_decode_c;
pbi->dboolhuff.fill = vp8dx_bool_decoder_fill_c;
pbi->dboolhuff.debool = vp8dx_decode_bool_c;
pbi->dboolhuff.devalue = vp8dx_decode_value_c;
#endif
}
#endif
#if HAVE_ARMV7
if (has_neon)
{
pbi->dequant.block = vp8_dequantize_b_neon;
pbi->dequant.idct_add = vp8_dequant_idct_add_neon;
/*This is not used: NEON always dequants two blocks at once.
pbi->dequant.dc_idct_add = vp8_dequant_dc_idct_add_neon;*/
pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_neon;
pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_neon;
pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_neon;
#if 0 //For use with RTCD, when implemented
pbi->dboolhuff.start = vp8dx_start_decode_c;
pbi->dboolhuff.fill = vp8dx_bool_decoder_fill_c;
pbi->dboolhuff.debool = vp8dx_decode_bool_c;
pbi->dboolhuff.devalue = vp8dx_decode_value_c;
#endif
}
#endif
#endif
}
......@@ -20,6 +20,7 @@ extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_v6)
extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_v6);
extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_v6);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_dequant_block
#define vp8_dequant_block vp8_dequantize_b_v6
......@@ -38,6 +39,7 @@ extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_v6);
#undef vp8_dequant_idct_add_uv_block
#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_v6
#endif
#endif
#if HAVE_ARMV7
extern prototype_dequant_block(vp8_dequantize_b_neon);
......@@ -47,6 +49,7 @@ extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_neo
extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_neon);
extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_neon);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_dequant_block
#define vp8_dequant_block vp8_dequantize_b_neon
......@@ -65,5 +68,6 @@ extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_neon);
#undef vp8_dequant_idct_add_uv_block
#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_neon
#endif
#endif
#endif
......@@ -14,6 +14,7 @@
#include "onyxd_int.h"
extern void vp8_arch_x86_decode_init(VP8D_COMP *pbi);
extern void vp8_arch_arm_decode_init(VP8D_COMP *pbi);
void vp8_dmachine_specific_config(VP8D_COMP *pbi)
{
......@@ -37,4 +38,8 @@ void vp8_dmachine_specific_config(VP8D_COMP *pbi)
#if ARCH_X86 || ARCH_X86_64
vp8_arch_x86_decode_init(pbi);
#endif
#if ARCH_ARM
vp8_arch_arm_decode_init(pbi);
#endif
}
......@@ -30,6 +30,9 @@
#include "systemdependent.h"
#include "vpx_ports/vpx_timer.h"
#include "detokenize.h"
#if ARCH_ARM
#include "vpx_ports/arm.h"
#endif
extern void vp8_init_loop_filter(VP8_COMMON *cm);
extern void vp8cx_init_de_quantizer(VP8D_COMP *pbi);
......@@ -224,7 +227,6 @@ int vp8dx_set_reference(VP8D_PTR ptr, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_C
#if HAVE_ARMV7
extern void vp8_push_neon(INT64 *store);
extern void vp8_pop_neon(INT64 *store);
static INT64 dx_store_reg[8];
#endif
static int get_free_fb (VP8_COMMON *cm)
......@@ -312,6 +314,9 @@ static int swap_frame_buffers (VP8_COMMON *cm)
int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsigned char *source, INT64 time_stamp)
{
#if HAVE_ARMV7
INT64 dx_store_reg[8];
#endif
VP8D_COMP *pbi = (VP8D_COMP *) ptr;
VP8_COMMON *cm = &pbi->common;
int retcode = 0;
......@@ -327,10 +332,27 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign
pbi->common.error.error_code = VPX_CODEC_OK;
#if HAVE_ARMV7
#if CONFIG_RUNTIME_CPU_DETECT
if (cm->rtcd.flags & HAS_NEON)
#endif
{
vp8_push_neon(dx_store_reg);
}
#endif
cm->new_fb_idx = get_free_fb (cm);
if (setjmp(pbi->common.error.jmp))
{
#if HAVE_ARMV7
#if CONFIG_RUNTIME_CPU_DETECT
if (cm->rtcd.flags & HAS_NEON)
#endif
{
vp8_pop_neon(dx_store_reg);
}
#endif
pbi->common.error.setjmp = 0;
if (cm->fb_idx_ref_cnt[cm->new_fb_idx] > 0)
cm->fb_idx_ref_cnt[cm->new_fb_idx]--;
......@@ -339,10 +361,6 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign