Merge "Begin to restrict x86inc.asm usage"

configure

@@ -257,6 +257,7 @@ CONFIG_LIST="
     install_bins
     install_libs
     install_srcs
+    force_x86inc
     debug
     gprof
     gcov

libs.mk

@@ -57,6 +57,19 @@ CLEAN-OBJS += $$(BUILD_PFX)$(1).h
 RTCD += $$(BUILD_PFX)$(1).h
 endef
 
+# x86inc.asm is not compatible with pic 32bit builds. Restrict
+# files which use it to 64bit builds or 32bit without pic
+USE_X86INC = no
+ifneq ($(CONFIG_PIC),yes)
+  USE_X86INC = yes
+endif
+ifeq ($(ARCH_X86_64),yes)
+  USE_X86INC = yes
+endif
+ifeq ($(CONFIG_FORCE_X86INC),yes)
+  USE_X86INC = yes
+endif
+
 CODEC_SRCS-yes += CHANGELOG
 CODEC_SRCS-yes += libs.mk
 

test/convolve_test.cc

@@ -527,9 +527,9 @@ INSTANTIATE_TEST_CASE_P(C, ConvolveTest, ::testing::Values(
 
 #if HAVE_SSSE3
 const ConvolveFunctions convolve8_ssse3(
-    vp9_convolve8_horiz_ssse3, vp9_convolve8_avg_horiz_c,
-    vp9_convolve8_vert_ssse3, vp9_convolve8_avg_vert_c,
-    vp9_convolve8_ssse3, vp9_convolve8_avg_c);
+    vp9_convolve8_horiz_ssse3, vp9_convolve8_avg_horiz_ssse3,
+    vp9_convolve8_vert_ssse3, vp9_convolve8_avg_vert_ssse3,
+    vp9_convolve8_ssse3, vp9_convolve8_avg_ssse3);
 
 INSTANTIATE_TEST_CASE_P(SSSE3, ConvolveTest, ::testing::Values(
     make_tuple(4, 4, &convolve8_ssse3),

vp9/common/vp9_rtcd_defs.sh

@@ -20,7 +20,11 @@ EOF
 }
 forward_decls vp9_common_forward_decls
 
-[ $arch = "x86_64" ] && mmx_x86_64=mmx && sse2_x86_64=sse2
+# x86inc.asm doesn't work if pic is enabled on 32 bit platforms so no assembly.
+[ $arch = "x86_64" -o ! "$CONFIG_PIC" = "yes" -o $CONFIG_FORCE_X86INC == "yes"] && mmx_x86inc=mmx && sse2_x86inc=sse2  && ssse3_x86inc=ssse3
+
+# this variable is for functions that are 64 bit only.
+[ $arch = "x86_64" ] && mmx_x86_64=mmx && sse2_x86_64=sse2  && ssse3_x86_64=ssse3
 
 #
 # Dequant
@@ -263,10 +267,10 @@ specialize vp9_blend_b
 # Sub Pixel Filters
 #
 prototype void vp9_convolve_copy "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve_copy sse2
+specialize vp9_convolve_copy $sse2_x86inc
 
 prototype void vp9_convolve_avg "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve_avg sse2
+specialize vp9_convolve_avg $sse2_x86inc
 
 prototype void vp9_convolve8 "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8 ssse3 neon
@@ -702,8 +706,6 @@ specialize vp9_block_error sse2
 prototype void vp9_subtract_block "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride"
 specialize vp9_subtract_block sse2
 
-[ $arch = "x86_64" ] && ssse3_x86_64=ssse3
-
 prototype void vp9_quantize_b "int16_t *coeff_ptr, intptr_t n_coeffs, int skip_block, int16_t *zbin_ptr, int16_t *round_ptr, int16_t *quant_ptr, int16_t *quant_shift_ptr, int16_t *qcoeff_ptr, int16_t *dqcoeff_ptr, int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"
 specialize vp9_quantize_b $ssse3_x86_64
 
@@ -714,13 +716,11 @@ specialize vp9_quantize_b_32x32 $ssse3_x86_64
 # Structured Similarity (SSIM)
 #
 if [ "$CONFIG_INTERNAL_STATS" = "yes" ]; then
-    [ $arch = "x86_64" ] && sse2_on_x86_64=sse2
-
     prototype void vp9_ssim_parms_8x8 "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
-    specialize vp9_ssim_parms_8x8 $sse2_on_x86_64
+    specialize vp9_ssim_parms_8x8 $sse2_x86_64
 
     prototype void vp9_ssim_parms_16x16 "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
-    specialize vp9_ssim_parms_16x16 $sse2_on_x86_64
+    specialize vp9_ssim_parms_16x16 $sse2_x86_64
 fi
 
 # fdct functions

vp9/vp9_common.mk

@@ -74,7 +74,6 @@ VP9_COMMON_SRCS-$(ARCH_X86)$(ARCH_X86_64) += common/x86/vp9_loopfilter_intrin_ss
 VP9_COMMON_SRCS-$(CONFIG_POSTPROC) += common/vp9_postproc.h
 VP9_COMMON_SRCS-$(CONFIG_POSTPROC) += common/vp9_postproc.c
 VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_loopfilter_mmx.asm
-VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_copy_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_intrapred_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_intrapred_ssse3.asm
 VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_subpixel_8t_ssse3.asm
@@ -83,6 +82,10 @@ VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_postproc_mmx.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_postproc_sse2.asm
 endif
 
+ifeq ($(USE_X86INC),yes)
+VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_copy_sse2.asm
+endif
+
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_idct_intrin_sse2.c
 
 VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_convolve_neon.c