shithub: libvpx

Info  •  Files  •  Log  •  Branches

sad + miscellaneous updates

Enable use_x86inc as a commandline option.  Fix Bug with sse2 when
x86inc is disabled. Adds Sad asm protection to x86inc protection

Change-Id: Iee0f9dd235ea10e8ace512eb362ba9bebe8c9df6

--- a/configure

+++ b/configure

@@ -314,6 +314,7 @@

     gprof

     gcov

     pic

+    use_x86inc

     optimizations

     ccache

     runtime_cpu_detect

--- a/test/sad_test.cc

+++ b/test/sad_test.cc

@@ -428,6 +428,7 @@

 #if HAVE_SSE

 #if CONFIG_VP9_ENCODER

+#if CONFIG_USE_X86INC

 const sad_m_by_n_fn_t sad_4x4_sse_vp9 = vp9_sad4x4_sse;

 const sad_m_by_n_fn_t sad_4x8_sse_vp9 = vp9_sad4x8_sse;

 INSTANTIATE_TEST_CASE_P(SSE, SADTest, ::testing::Values(

@@ -441,6 +442,7 @@

                         make_tuple(4, 4, sad_4x4x4d_sse)));

 #endif

 #endif

+#endif

 #if HAVE_SSE2

 #if CONFIG_VP8_ENCODER

@@ -451,6 +453,7 @@

 const sad_m_by_n_fn_t sad_4x4_wmt = vp8_sad4x4_wmt;

 #endif

 #if CONFIG_VP9_ENCODER

+#if CONFIG_USE_X86INC

 const sad_m_by_n_fn_t sad_64x64_sse2_vp9 = vp9_sad64x64_sse2;

 const sad_m_by_n_fn_t sad_64x32_sse2_vp9 = vp9_sad64x32_sse2;

 const sad_m_by_n_fn_t sad_32x64_sse2_vp9 = vp9_sad32x64_sse2;

@@ -463,6 +466,7 @@

 const sad_m_by_n_fn_t sad_8x8_sse2_vp9 = vp9_sad8x8_sse2;

 const sad_m_by_n_fn_t sad_8x4_sse2_vp9 = vp9_sad8x4_sse2;

 #endif

+#endif

 const sad_m_by_n_test_param_t sse2_tests[] = {

 #if CONFIG_VP8_ENCODER

   make_tuple(16, 16, sad_16x16_wmt),

@@ -472,6 +476,7 @@

   make_tuple(4, 4, sad_4x4_wmt),

 #endif

 #if CONFIG_VP9_ENCODER

+#if CONFIG_USE_X86INC

   make_tuple(64, 64, sad_64x64_sse2_vp9),

   make_tuple(64, 32, sad_64x32_sse2_vp9),

   make_tuple(32, 64, sad_32x64_sse2_vp9),

@@ -483,6 +488,7 @@

   make_tuple(8, 16, sad_8x16_sse2_vp9),

   make_tuple(8, 8, sad_8x8_sse2_vp9),

   make_tuple(8, 4, sad_8x4_sse2_vp9),

+#endif

 #endif

 };

 INSTANTIATE_TEST_CASE_P(SSE2, SADTest, ::testing::ValuesIn(sse2_tests));

--- a/vp9/common/vp9_rtcd_defs.sh

+++ b/vp9/common/vp9_rtcd_defs.sh

@@ -480,82 +480,82 @@

 specialize vp9_sub_pixel_avg_variance4x4 sse ssse3

 prototype unsigned int vp9_sad64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad64x64 sse2

+specialize vp9_sad64x64 $sse2_x86inc

 prototype unsigned int vp9_sad32x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"

-specialize vp9_sad32x64 sse2

+specialize vp9_sad32x64 $sse2_x86inc

 prototype unsigned int vp9_sad64x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"

-specialize vp9_sad64x32 sse2

+specialize vp9_sad64x32 $sse2_x86inc

 prototype unsigned int vp9_sad32x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"

-specialize vp9_sad32x16 sse2

+specialize vp9_sad32x16 $sse2_x86inc

 prototype unsigned int vp9_sad16x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"

-specialize vp9_sad16x32 sse2

+specialize vp9_sad16x32 $sse2_x86inc

 prototype unsigned int vp9_sad32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad32x32 sse2

+specialize vp9_sad32x32 $sse2_x86inc

 prototype unsigned int vp9_sad16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad16x16 mmx sse2

+specialize vp9_sad16x16 mmx $sse2_x86inc

 prototype unsigned int vp9_sad16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad16x8 mmx sse2

+specialize vp9_sad16x8 mmx $sse2_x86inc

 prototype unsigned int vp9_sad8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad8x16 mmx sse2

+specialize vp9_sad8x16 mmx $sse2_x86inc

 prototype unsigned int vp9_sad8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad8x8 mmx sse2

+specialize vp9_sad8x8 mmx $sse2_x86inc

 prototype unsigned int vp9_sad8x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"

-specialize vp9_sad8x4 sse2

+specialize vp9_sad8x4 $sse2_x86inc

 prototype unsigned int vp9_sad4x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"

-specialize vp9_sad4x8 sse

+specialize vp9_sad4x8 $sse_x86inc

 prototype unsigned int vp9_sad4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"

-specialize vp9_sad4x4 mmx sse

+specialize vp9_sad4x4 mmx $sse_x86inc

 prototype unsigned int vp9_sad64x64_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad64x64_avg sse2

+specialize vp9_sad64x64_avg $sse2_x86inc

 prototype unsigned int vp9_sad32x64_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad32x64_avg sse2

+specialize vp9_sad32x64_avg $sse2_x86inc

 prototype unsigned int vp9_sad64x32_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad64x32_avg sse2

+specialize vp9_sad64x32_avg $sse2_x86inc

 prototype unsigned int vp9_sad32x16_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad32x16_avg sse2

+specialize vp9_sad32x16_avg $sse2_x86inc

 prototype unsigned int vp9_sad16x32_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad16x32_avg sse2

+specialize vp9_sad16x32_avg $sse2_x86inc

 prototype unsigned int vp9_sad32x32_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad32x32_avg sse2

+specialize vp9_sad32x32_avg $sse2_x86inc

 prototype unsigned int vp9_sad16x16_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad16x16_avg sse2

+specialize vp9_sad16x16_avg $sse2_x86inc

 prototype unsigned int vp9_sad16x8_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad16x8_avg sse2

+specialize vp9_sad16x8_avg $sse2_x86inc

 prototype unsigned int vp9_sad8x16_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad8x16_avg sse2

+specialize vp9_sad8x16_avg $sse2_x86inc

 prototype unsigned int vp9_sad8x8_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad8x8_avg sse2

+specialize vp9_sad8x8_avg $sse2_x86inc

 prototype unsigned int vp9_sad8x4_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad8x4_avg sse2

+specialize vp9_sad8x4_avg $sse2_x86inc

 prototype unsigned int vp9_sad4x8_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad4x8_avg sse

+specialize vp9_sad4x8_avg $sse_x86inc

 prototype unsigned int vp9_sad4x4_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"

-specialize vp9_sad4x4_avg sse

+specialize vp9_sad4x4_avg $sse_x86inc

 prototype unsigned int vp9_variance_halfpixvar16x16_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"

 specialize vp9_variance_halfpixvar16x16_h sse2

--- a/vp9/vp9cx.mk

+++ b/vp9/vp9cx.mk

@@ -80,7 +80,6 @@

 VP9_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp9_sad_mmx.asm

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_variance_sse2.c

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_variance_impl_sse2.asm

-VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_sad_sse2.asm

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_sad4d_sse2.asm

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_subpel_variance.asm

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_subpel_variance_impl_sse2.asm

@@ -89,6 +88,7 @@

 ifeq ($(USE_X86INC),yes)

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_error_sse2.asm

+VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_sad_sse2.asm

 VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_subtract_sse2.asm

 endif