aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rwxr-xr-xbuild/make/rtcd.pl2
-rw-r--r--test/convolve_test.cc4
-rw-r--r--vp9/common/vp9_rtcd_defs.pl13
-rw-r--r--vp9/common/x86/vp9_asm_stubs.c20
4 files changed, 23 insertions, 16 deletions
diff --git a/build/make/rtcd.pl b/build/make/rtcd.pl
index 5b0cefa45..40bcb33d0 100755
--- a/build/make/rtcd.pl
+++ b/build/make/rtcd.pl
@@ -49,7 +49,7 @@ open CONFIG_FILE, $opts{config} or
my %config = ();
while (<CONFIG_FILE>) {
- next if !/^CONFIG_/;
+ next if !/^(?:CONFIG_|HAVE_)/;
chomp;
my @pair = split /=/;
$config{$pair[0]} = $pair[1];
diff --git a/test/convolve_test.cc b/test/convolve_test.cc
index 1724db36f..2e3659ff3 100644
--- a/test/convolve_test.cc
+++ b/test/convolve_test.cc
@@ -645,7 +645,7 @@ INSTANTIATE_TEST_CASE_P(SSSE3, ConvolveTest, ::testing::Values(
make_tuple(64, 64, &convolve8_ssse3)));
#endif
-#if HAVE_AVX2
+#if HAVE_AVX2 && HAVE_SSSE3
const ConvolveFunctions convolve8_avx2(
vp9_convolve8_horiz_avx2, vp9_convolve8_avg_horiz_ssse3,
vp9_convolve8_vert_avx2, vp9_convolve8_avg_vert_ssse3,
@@ -665,7 +665,7 @@ INSTANTIATE_TEST_CASE_P(AVX2, ConvolveTest, ::testing::Values(
make_tuple(64, 32, &convolve8_avx2),
make_tuple(32, 64, &convolve8_avx2),
make_tuple(64, 64, &convolve8_avx2)));
-#endif
+#endif // HAVE_AVX2 && HAVE_SSSE3
#if HAVE_NEON_ASM
const ConvolveFunctions convolve8_neon(
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl
index 667e057b6..c2724dd08 100644
--- a/vp9/common/vp9_rtcd_defs.pl
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -45,6 +45,13 @@ if ($opts{arch} eq "x86_64") {
$avx_x86_64 = $avx2_x86_64 = '';
}
+# optimizations which depend on multiple features
+if ((vpx_config("HAVE_AVX2") eq "yes") && (vpx_config("HAVE_SSSE3") eq "yes")) {
+ $avx2_ssse3 = 'avx2';
+} else {
+ $avx2_ssse3 = '';
+}
+
#
# RECON
#
@@ -296,15 +303,15 @@ specialize qw/vp9_convolve_avg neon_asm dspr2/, "$sse2_x86inc";
$vp9_convolve_avg_neon_asm=vp9_convolve_avg_neon;
add_proto qw/void vp9_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
-specialize qw/vp9_convolve8 sse2 ssse3 neon_asm dspr2 avx2/;
+specialize qw/vp9_convolve8 sse2 ssse3 neon_asm dspr2/, "$avx2_ssse3";
$vp9_convolve8_neon_asm=vp9_convolve8_neon;
add_proto qw/void vp9_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
-specialize qw/vp9_convolve8_horiz sse2 ssse3 neon_asm dspr2 avx2/;
+specialize qw/vp9_convolve8_horiz sse2 ssse3 neon_asm dspr2/, "$avx2_ssse3";
$vp9_convolve8_horiz_neon_asm=vp9_convolve8_horiz_neon;
add_proto qw/void vp9_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
-specialize qw/vp9_convolve8_vert sse2 ssse3 neon_asm dspr2 avx2/;
+specialize qw/vp9_convolve8_vert sse2 ssse3 neon_asm dspr2/, "$avx2_ssse3";
$vp9_convolve8_vert_neon_asm=vp9_convolve8_vert_neon;
add_proto qw/void vp9_convolve8_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";
diff --git a/vp9/common/x86/vp9_asm_stubs.c b/vp9/common/x86/vp9_asm_stubs.c
index 1b4904c39..b6847b92e 100644
--- a/vp9/common/x86/vp9_asm_stubs.c
+++ b/vp9/common/x86/vp9_asm_stubs.c
@@ -139,25 +139,25 @@ void vp9_convolve8_##avg##opt(const uint8_t *src, ptrdiff_t src_stride, \
filter_x, x_step_q4, filter_y, y_step_q4, w, h); \
} \
}
-#if HAVE_AVX2
+#if HAVE_AVX2 && HAVE_SSSE3
filter8_1dfunction vp9_filter_block1d16_v8_avx2;
filter8_1dfunction vp9_filter_block1d16_h8_avx2;
filter8_1dfunction vp9_filter_block1d4_v8_ssse3;
-#if (ARCH_X86_64)
+#if ARCH_X86_64
filter8_1dfunction vp9_filter_block1d8_v8_intrin_ssse3;
filter8_1dfunction vp9_filter_block1d8_h8_intrin_ssse3;
filter8_1dfunction vp9_filter_block1d4_h8_intrin_ssse3;
#define vp9_filter_block1d8_v8_avx2 vp9_filter_block1d8_v8_intrin_ssse3
#define vp9_filter_block1d8_h8_avx2 vp9_filter_block1d8_h8_intrin_ssse3
#define vp9_filter_block1d4_h8_avx2 vp9_filter_block1d4_h8_intrin_ssse3
-#else
+#else // ARCH_X86
filter8_1dfunction vp9_filter_block1d8_v8_ssse3;
filter8_1dfunction vp9_filter_block1d8_h8_ssse3;
filter8_1dfunction vp9_filter_block1d4_h8_ssse3;
#define vp9_filter_block1d8_v8_avx2 vp9_filter_block1d8_v8_ssse3
#define vp9_filter_block1d8_h8_avx2 vp9_filter_block1d8_h8_ssse3
#define vp9_filter_block1d4_h8_avx2 vp9_filter_block1d4_h8_ssse3
-#endif
+#endif // ARCH_X86_64 / ARCH_X86
filter8_1dfunction vp9_filter_block1d16_v2_ssse3;
filter8_1dfunction vp9_filter_block1d16_h2_ssse3;
filter8_1dfunction vp9_filter_block1d8_v2_ssse3;
@@ -190,9 +190,9 @@ FUN_CONV_1D(vert, y_step_q4, filter_y, v, src - src_stride * 3, , avx2);
// const int16_t *filter_y, int y_step_q4,
// int w, int h);
FUN_CONV_2D(, avx2);
-#endif
+#endif // HAVE_AX2 && HAVE_SSSE3
#if HAVE_SSSE3
-#if (ARCH_X86_64)
+#if ARCH_X86_64
filter8_1dfunction vp9_filter_block1d16_v8_intrin_ssse3;
filter8_1dfunction vp9_filter_block1d16_h8_intrin_ssse3;
filter8_1dfunction vp9_filter_block1d8_v8_intrin_ssse3;
@@ -204,14 +204,14 @@ filter8_1dfunction vp9_filter_block1d4_h8_intrin_ssse3;
#define vp9_filter_block1d8_v8_ssse3 vp9_filter_block1d8_v8_intrin_ssse3
#define vp9_filter_block1d8_h8_ssse3 vp9_filter_block1d8_h8_intrin_ssse3
#define vp9_filter_block1d4_h8_ssse3 vp9_filter_block1d4_h8_intrin_ssse3
-#else
+#else // ARCH_X86
filter8_1dfunction vp9_filter_block1d16_v8_ssse3;
filter8_1dfunction vp9_filter_block1d16_h8_ssse3;
filter8_1dfunction vp9_filter_block1d8_v8_ssse3;
filter8_1dfunction vp9_filter_block1d8_h8_ssse3;
filter8_1dfunction vp9_filter_block1d4_v8_ssse3;
filter8_1dfunction vp9_filter_block1d4_h8_ssse3;
-#endif
+#endif // ARCH_X86_64 / ARCH_X86
filter8_1dfunction vp9_filter_block1d16_v8_avg_ssse3;
filter8_1dfunction vp9_filter_block1d16_h8_avg_ssse3;
filter8_1dfunction vp9_filter_block1d8_v8_avg_ssse3;
@@ -270,7 +270,7 @@ FUN_CONV_1D(avg_vert, y_step_q4, filter_y, v, src - src_stride * 3, avg_,
// int w, int h);
FUN_CONV_2D(, ssse3);
FUN_CONV_2D(avg_ , ssse3);
-#endif
+#endif // HAVE_SSSE3
#if HAVE_SSE2
filter8_1dfunction vp9_filter_block1d16_v8_sse2;
@@ -336,4 +336,4 @@ FUN_CONV_1D(avg_vert, y_step_q4, filter_y, v, src - src_stride * 3, avg_, sse2);
// int w, int h);
FUN_CONV_2D(, sse2);
FUN_CONV_2D(avg_ , sse2);
-#endif
+#endif // HAVE_SSE2
generated by cgit v1.2.3-70-g09d2 (git 2.50.0) at 2025-08-05 20:36:20 +0000