Date: Fri, 15 Jul 2016 01:27:43 +0000 (UTC) From: Mikhail Teterin <mi@FreeBSD.org> To: ports-committers@freebsd.org, svn-ports-all@freebsd.org, svn-ports-head@freebsd.org Subject: svn commit: r418567 - in head/graphics: . lepton lepton/files Message-ID: <201607150127.u6F1RhaS087043@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mi Date: Fri Jul 15 01:27:43 2016 New Revision: 418567 URL: https://svnweb.freebsd.org/changeset/ports/418567 Log: Add port of "lepton" -- a tool for manipulating files in LEP-format, which was recently open-sourced by Dropbox. The patch Added: head/graphics/lepton/ head/graphics/lepton/Makefile (contents, props changed) head/graphics/lepton/distinfo (contents, props changed) head/graphics/lepton/files/ head/graphics/lepton/files/patch-base-dependencies (contents, props changed) head/graphics/lepton/files/patch-bsd (contents, props changed) head/graphics/lepton/files/patch-sse2 (contents, props changed) head/graphics/lepton/pkg-descr (contents, props changed) Modified: head/graphics/Makefile Modified: head/graphics/Makefile ============================================================================== --- head/graphics/Makefile Fri Jul 15 01:10:15 2016 (r418566) +++ head/graphics/Makefile Fri Jul 15 01:27:43 2016 (r418567) @@ -460,6 +460,7 @@ SUBDIR += lcms2 SUBDIR += leafpak SUBDIR += lensfun + SUBDIR += lepton SUBDIR += leptonica SUBDIR += lfview SUBDIR += lib3ds Added: head/graphics/lepton/Makefile ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/graphics/lepton/Makefile Fri Jul 15 01:27:43 2016 (r418567) @@ -0,0 +1,30 @@ +# Created by: Mikhail T. <mi@aldan.algebra.com> +# $FreeBSD$ + +PORTNAME= lepton +PORTVERSION= 0.2016.07.16 +CATEGORIES= graphics devel + +MAINTAINER= mi@aldan.algebra.com +COMMENT= Tool for manipulating files in LEP image format + +LICENSE= APACHE20 + +USE_GITHUB= yes +GH_ACCOUNT= dropbox +GH_TAGNAME= a34ee2f4b0a6454eff8ebe334750dd008d57de35 + +USES= autoreconf +GNU_CONFIGURE= yes +TEST_TARGET= check +CFLAGS+= -DBSD +.ifndef WITH_DEBUG +CFLAGS+= -DNDEBUG +.endif +EXTRACT_AFTER_ARGS=--no-same-owner --no-same-permissions --exclude dependencies +PLIST_FILES= bin/lepton + +do-install: + ${INSTALL_PROGRAM} ${WRKSRC}/lepton ${STAGEDIR}${PREFIX}/bin + +.include <bsd.port.mk> Added: head/graphics/lepton/distinfo ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/graphics/lepton/distinfo Fri Jul 15 01:27:43 2016 (r418567) @@ -0,0 +1,3 @@ +TIMESTAMP = 1468524984 +SHA256 (dropbox-lepton-0.2016.07.16-a34ee2f4b0a6454eff8ebe334750dd008d57de35_GH0.tar.gz) = 9e1941c7cc72b50a20f4f7a9495df42f7978d4cd1b9764c74787af997bd386f9 +SIZE (dropbox-lepton-0.2016.07.16-a34ee2f4b0a6454eff8ebe334750dd008d57de35_GH0.tar.gz) = 50216342 Added: head/graphics/lepton/files/patch-base-dependencies ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/graphics/lepton/files/patch-base-dependencies Fri Jul 15 01:27:43 2016 (r418567) @@ -0,0 +1,99 @@ +Do not use the zlib and md5 implementations bundled by +the author(s). Use base-system's... + + -mi + ++++ src/io/ZlibCompression.hh +@@ -28,7 +28,7 @@ + */ + + #include "Reader.hh" +-#include "../../dependencies/zlib/zlib.h" ++#include <zlib.h> + namespace Sirikata{ + class SIRIKATA_EXPORT ZlibDecoderDecompressionReader : public DecoderReader { + protected: ++++ src/lepton/bitops.cc +@@ -38,7 +38,7 @@ reading and writing of arrays + #include <algorithm> + #include <assert.h> + extern "C" { +-#include "../../dependencies/md5/md5.h" ++#include <openssl/md5.h> + } + #include "bitops.hh" + ++++ src/io/ioutil.cc +@@ -12,5 +12,5 @@ + #include "Reader.hh" + #include "ioutil.hh" +-#include "../../dependencies/md5/md5.h" ++#include <openssl/md5.h> + #ifdef _WIN32 + #include <Windows.h> ++++ Makefile.am +@@ -1,13 +1,13 @@ +-includes = -I$(srcdir)/dependencies/xz/src/liblzma/api -I$(srcdir)/src/vp8/util -I$(srcdir)/src/vp8/model -I$(srcdir)/src/vp8/decoder -I$(srcdir)/src/vp8/encoder -I$(srcdir)/dependencies -I$(srcdir)/dependencies -I$(srcdir)/dependencies/xz/src/liblzma/common -I$(srcdir)/dependencies/xz/src/common -I$(srcdir)/dependencies/xz/src/liblzma/lzma -I$(srcdir)/dependencies/xz/src/liblzma/lz -I$(srcdir)/dependencies/xz/src/liblzma/check -I$(srcdir)/dependencies/xz/src/liblzma/rangecoder -I$(srcdir)/dependencies/xz/src/liblzma/api -I$(srcdir)/dependencies/xz/src/liblzma/simple -I$(srcdir)/dependencies/xz/src/liblzma/delta -I$(srcdir)/dependencies/xz/src/liblzma ++includes = -I$(srcdir)/src/vp8/util -I$(srcdir)/src/vp8/model -I$(srcdir)/src/vp8/decoder -I$(srcdir)/src/vp8/encoder + +-AM_CXXFLAGS = $(CXX11_FLAGS) $(PICKY_CXXFLAGS) $(ARCH_FLAGS) $(SANITIZE_FLAGS) $(NODEBUG_CXXFLAGS) -DGIT_REVISION=\"$(shell git describe --dirty --always)\" $(includes) ++AM_CXXFLAGS = $(CXX11_FLAGS) $(PICKY_CXXFLAGS) $(ARCH_FLAGS) $(SANITIZE_FLAGS) $(NODEBUG_CXXFLAGS) $(includes) + + AM_CFLAGS = $(C99_FLAGS) $(PICKY_CFLAGS) $(ARCH_FLAGS) $(SANITIZE_FLAGS) $(NODEBUG_CXXFLAGS) + +-noinst_LIBRARIES = liblocalzlib.a libtestdriver.a liblocalmd5.a ++noinst_LIBRARIES = libtestdriver.a + + bin_PROGRAMS = lepton test_suite/test_invariants + +-lepton_LDADD = liblocalmd5.a liblocalzlib.a -lpthread ++lepton_LDADD = -lz -lcrypto -lpthread + + lepton_SOURCES = \ +@@ -80,5 +80,4 @@ + src/vp8/model/model.cc \ + src/vp8/model/model.hh \ +- src/vp8/model/numeric.cc \ + src/vp8/model/numeric.hh \ + src/vp8/model/jpeg_meta.hh \ +@@ -94,33 +93,4 @@ + src/vp8/decoder/vpx_bool_reader.hh + +-liblocalmd5_a_SOURCES = dependencies/md5/md5.c +- +-liblocalzlib_a_SOURCES = dependencies/zlib/inflate.c \ +-dependencies/zlib/inflate.h \ +-dependencies/zlib/gzguts.h \ +-dependencies/zlib/infback.c \ +-dependencies/zlib/trees.c \ +-dependencies/zlib/adler32.c \ +-dependencies/zlib/gzclose.c \ +-dependencies/zlib/inftrees.h \ +-dependencies/zlib/zconf.h \ +-dependencies/zlib/compress.c \ +-dependencies/zlib/crc32.c \ +-dependencies/zlib/crc32.h \ +-dependencies/zlib/trees.h \ +-dependencies/zlib/inftrees.c \ +-dependencies/zlib/zutil.c \ +-dependencies/zlib/gzwrite.c \ +-dependencies/zlib/zutil.h \ +-dependencies/zlib/zlib.h \ +-dependencies/zlib/inffixed.h \ +-dependencies/zlib/deflate.c \ +-dependencies/zlib/inffast.h \ +-dependencies/zlib/inffast.c \ +-dependencies/zlib/uncompr.c \ +-dependencies/zlib/gzread.c \ +-dependencies/zlib/deflate.h \ +-dependencies/zlib/gzlib.c +- + libtestdriver_a_SOURCES = test_suite/timing_driver.cc + +@@ -142,5 +112,5 @@ + src/io/DecoderPlatform.hh + +-test_suite_test_invariants_LDADD = liblocalzlib.a liblocalmd5.a ++test_suite_test_invariants_LDADD = -lz -lcrypto + + check_PROGRAMS = test_suite/test_recode_memory_bound test_suite/test_truncate_lowmem test_suite/test_android_lowmem test_suite/test_invariants test_suite/test_baseline_ujg test_suite/test_baseline test_suite/test_misc test_suite/test_iphone test_suite/test_phone_outdoor test_suite/test_truncate_ujg test_suite/test_truncate test_suite/test_SLR test_suite/test_progressive_ujg test_suite/test_progressive_disallowed test_suite/test_progressive test_suite/test_arithmetic_failfast test_suite/test_hq test_suite/test_baseline_unjailed test_suite/test_baseline_unjailed_thread test_suite/test_baseline_unjailed_decode test_suite/test_baseline_unjailed_decode_thread test_suite/test_seccomp_encode_main test_suite/test_seccomp_encode_thread test_suite/test_seccomp_decode_main test_suite/test_seccomp_decode_thread test_suite/test_nofsync test_suite/test_colorswap test_suite/test_odd_rst test_suite/test_trailing_header test_suite/test_trailing_rst test_suite/test_gray2sf test_suite/test_truncated _zero_run test_suite/test_bad_zero_run Added: head/graphics/lepton/files/patch-bsd ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/graphics/lepton/files/patch-bsd Fri Jul 15 01:27:43 2016 (r418567) @@ -0,0 +1,148 @@ +Make buildable on BSD-systems other than Apple. Mostly this +simply replaces #ifdef __APPLE__ with #ifdef BSD, but not +only... + + -mi + ++++ src/io/DecoderPlatform.hh +@@ -26,7 +26,7 @@ + * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ +-#if defined (__linux) || defined (__APPLE__) ++#if defined (__linux) || defined (BSD) + #define SIRIKATA_FUNCTION_EXPORT __attribute__ ((visibility("default"))) + #define SIRIKATA_EXPORT __attribute__ ((visibility("default"))) + #define SIRIKATA_PLUGIN_EXPORT __attribute__ ((visibility("default"))) ++++ src/io/MemMgrAllocator.cc +@@ -41,7 +41,7 @@ + #include <cstdint> + #include "DecoderPlatform.hh" + #include "MemMgrAllocator.hh" +-#if (defined(__APPLE__) || __cplusplus <= 199711L) && !defined(_WIN32) ++#if (defined(BSD) || __cplusplus <= 199711L) && !defined(_WIN32) + #define THREAD_LOCAL_STORAGE __thread + #else + #include <atomic> +@@ -151,7 +151,7 @@ void setup_memmgr(MemMgrState& memmgr, u + } + void memmgr_init(size_t main_thread_pool_size, size_t worker_thread_pool_size, size_t num_workers, size_t x_min_pool_alloc_quantas, bool needs_huge_pages) + { +-#ifdef __APPLE__ ++#ifdef BSD + // in apple, the thread_local storage winds up different when destroying the thread + num_workers *= 2; + #endif ++++ src/lepton/fork_serve.cc +@@ -10,7 +10,7 @@ + #include <fcntl.h> + #include <unistd.h> + #include <algorithm> +-#ifndef __APPLE__ ++#ifndef BSD + #include <wait.h> + #else + #include <sys/wait.h> ++++ src/lepton/socket_serve.cc +@@ -11,7 +11,7 @@ + #include <algorithm> + #include <netinet/in.h> + #include <sys/time.h> +-#ifndef __APPLE__ ++#ifndef BSD + #include <sys/signalfd.h> + #include <wait.h> + #else +@@ -127,7 +127,7 @@ int should_wait_bitmask(size_t children_ + + int make_sigchld_fd() { + int fd = -1; +-#ifndef __APPLE__ ++#ifndef BSD + sigset_t sigset; + int err = sigemptyset(&sigset); + always_assert(err == 0); +@@ -233,7 +233,7 @@ void serving_loop(int unix_domain_socket + if (fds[i].revents & POLLIN) { + fds[i].revents = 0; + if (fds[i].fd == sigchild_fd) { +-#ifndef __APPLE__ ++#ifndef BSD + struct signalfd_siginfo info; + ssize_t ignore = read(fds[i].fd, &info, sizeof(info)); + (void)ignore; ++++ src/vp8/model/jpeg_meta.hh +@@ -3,7 +3,7 @@ + + #include <vector> + #include <type_traits> +-#ifndef __APPLE__ ++#ifndef BSD + #ifndef _WIN32 + #include <endian.h> + #endif ++++ src/vp8/util/memory.cc +@@ -13,7 +13,7 @@ + #ifdef _WIN32 + #define USE_STANDARD_MEMORY_ALLOCATORS + #endif +-#if defined(__APPLE__) || (__cplusplus <= 199711L && !defined(_WIN32)) ++#if defined(BSD) || (__cplusplus <= 199711L && !defined(_WIN32)) + #define THREAD_LOCAL_STORAGE __thread + #else + #define THREAD_LOCAL_STORAGE thread_local ++++ src/vp8/util/vpx_config.hh +@@ -44,6 +44,6 @@ + #ifndef _BSD_SOURCE + #define _BSD_SOURCE /* See feature_test_macros(7) */ + #endif +-#include <endian.h> ++#include <sys/endian.h> + #endif + #endif ++++ test_suite/timing_driver.cc +@@ -6,11 +6,13 @@ + #include <unistd.h> + #include <assert.h> + #include <errno.h> ++#include <stdlib.h> + #include <string.h> + #include <sys/types.h> + #include <sys/wait.h> + #include <fcntl.h> + #include <sys/select.h> ++#include <sys/signal.h> + #include <sys/time.h> + #include "smalljpg.hh" + ++++ src/lepton/jpgcoder.cc +@@ -77,11 +77,8 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBI + int g_argc = 0; + const char** g_argv = NULL; + #ifndef GIT_REVISION +-#include "version.hh" +-#ifndef GIT_REVISION + #define GIT_REVISION "unknown" + #endif +-#endif + bool fast_exit = true; + #ifdef SKIP_VALIDATION + bool g_skip_validation = true; ++++ src/io/ioutil.cc +@@ -325,13 +325,13 @@ + int input_tee_flags = 0; + int copy_to_storage_flags = 0; +-#ifndef __APPLE__ ++#ifndef BSD + input_tee_flags = fcntl(input_tee, F_GETFL, 0); + #endif + fcntl(input_tee, F_SETFL, input_tee_flags | O_NONBLOCK); +-#ifndef __APPLE__ ++#ifndef BSD + copy_to_input_tee_flags = fcntl(copy_to_input_tee, F_GETFL, 0); + #endif + fcntl(copy_to_input_tee, F_SETFL, copy_to_input_tee_flags | O_NONBLOCK); +-#ifndef __APPLE__ ++#ifndef BSD + copy_to_storage_flags = fcntl(copy_to_storage, F_GETFL, 0); + #endif Added: head/graphics/lepton/files/patch-sse2 ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/graphics/lepton/files/patch-sse2 Fri Jul 15 01:27:43 2016 (r418567) @@ -0,0 +1,209 @@ +Make -- or attempt to -- the code work on CPUs with only SSE2 +instruction set... + + -mi + +--- src/lepton/idct.cc ++++ src/lepton/idct.cc +@@ -1,8 +1,6 @@ + /* -*-mode:c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */ +-#include <emmintrin.h> +-#include <smmintrin.h> +-#include <immintrin.h> + #include "../vp8/util/aligned_block.hh" ++#include "../vp8/util/mm_mullo_epi32.hh" + + namespace idct_local{ + enum { +@@ -23,7 +21,10 @@ enum { + r2 = 181 // 256/sqrt(2) + }; + } +-void idct_scalar(const AlignedBlock &block, const uint16_t q[64], int16_t outp[64], bool ignore_dc) { ++ ++#ifndef __SSE2__ ++static void ++idct_scalar(const AlignedBlock &block, const uint16_t q[64], int16_t outp[64], bool ignore_dc) { + int32_t intermed[64]; + using namespace idct_local; + // Horizontal 1-D IDCT. +@@ -149,6 +150,8 @@ void idct_scalar(const AlignedBlock &blo + //outp[i]>>=3; + } + } ++#else /* At least SSE2 is available { */ ++ + template<int which_vec, int offset, int stride> __m128i vget_raster(const AlignedBlock&block) { + return _mm_set_epi32(block.coefficients_raster(which_vec + 3 * stride + offset), + block.coefficients_raster(which_vec + 2 * stride + offset), +@@ -162,8 +165,8 @@ template<int offset, int stride> __m128i + q[which_vec + offset])); + } + +- +-__m128i epi32l_to_epi16(__m128i lowvec) { ++static __m128i ++epi32l_to_epi16(__m128i lowvec) { + return _mm_shuffle_epi8(lowvec, _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1, + 0xd, 0xc, 0x9, 0x8, 0x5, 0x4, 0x1, 0x0)); + } +@@ -181,9 +184,8 @@ __m128i epi32l_to_epi16(__m128i lowvec) + }while(0) + + +- +- +-void idct_sse(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { ++static void ++idct_sse(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { + + char vintermed_storage[64 * sizeof(int32_t) + 16]; + // align intermediate storage to 16 bytes +@@ -202,7 +204,12 @@ void idct_sse(const AlignedBlock &block, + xv6 = vget_raster<0, 5, 8>(block); + xv7 = vget_raster<0, 3, 8>(block); + if (__builtin_expect(ignore_dc, true)) { ++#ifdef __SSE4_1__ + xv0 = _mm_insert_epi32(xv0, 0, 0); ++#else ++// See http://stackoverflow.com/questions/38384520/is-there-a-sse2-equivalent-for-mm-insert-epi32 ++ xv0 = _mm_insert_epi16(_mm_insert_epi16(xv0, 0, 0), 0, 1); ++#endif + } + } else { + xv0 = vget_raster<32, 0, 8>(block); +@@ -378,7 +385,8 @@ __m128i m256_to_epi16(__m256i vec) { + + }*/ + #if __AVX2__ +-void idct_avx(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { ++static void ++idct_avx(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { + // align intermediate storage to 16 bytes + using namespace idct_local; + // Horizontal 1-D IDCT. +@@ -589,11 +597,16 @@ void idct_avx(const AlignedBlock &block, + #endif + } + } +-#else +-void idct_avx(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { +- idct_sse(block, q, voutp, ignore_dc); +-} + #endif +-void idct(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { ++#endif /* } SSE2 or higher is available */ ++ ++void ++idct(const AlignedBlock &block, const uint16_t q[64], int16_t voutp[64], bool ignore_dc) { ++#ifdef __AVX2__ + idct_avx(block, q, voutp, ignore_dc); ++#elif __SSE2__ ++ idct_sse(block, q, voutp, ignore_dc); ++#else ++ idct_scalar(block, q, voutp, ignore_dc); ++#endif + } +--- src/lepton/vp8_encoder.cc ++++ src/lepton/vp8_encoder.cc +@@ -150,29 +150,34 @@ void VP8ComponentEncoder::process_row(Pr + } + } + uint32_t aligned_block_cost(const AlignedBlock &block) { +- uint32_t cost = 16; // .25 cost for zeros +- if (VECTORIZE) { +- for (int i = 0; i < 64; i+= 8) { +- __m128i val = _mm_abs_epi16(_mm_load_si128((const __m128i*)(const char*)(block.raw_data() + i))); +- __m128i v_cost = _mm_set1_epi16(0); +- while (!_mm_test_all_zeros(val, val)) { +- __m128i mask = _mm_cmpgt_epi16(val, _mm_setzero_si128()); +- v_cost = _mm_add_epi16(v_cost, _mm_and_si128(mask, _mm_set1_epi16(2))); +- val = _mm_srli_epi16(val, 1); +- } +- __m128i sum = _mm_add_epi16(v_cost, _mm_srli_si128(v_cost, 8)); +- sum = _mm_add_epi16(sum ,_mm_srli_si128(sum, 4)); +- sum = _mm_add_epi16(sum, _mm_srli_si128(sum, 2)); +- cost += _mm_extract_epi16(sum, 0); +- } +- } else { +- uint32_t scost = 0; +- for (int i = 0; i < 64; ++i) { +- scost += 1 + 2 * uint16bit_length(abs(block.raw_data()[i])); ++#ifdef __SSE2__ /* SSE2 or higher instruction set available { */ ++ const __m128i zero = _mm_setzero_si128(); ++ __m128i v_cost; ++ for (int i = 0; i < 64; i+= 8) { ++ __m128i val = _mm_abs_epi16(_mm_load_si128((const __m128i*)(const char*)(block.raw_data() + i))); ++ v_cost = _mm_set1_epi16(0); ++#ifndef __SSE4_1__ ++ while (_mm_movemask_epi8(_mm_cmpeq_epi32(val, zero)) != 0xFFFF) ++#else ++ while (!_mm_test_all_zeros(val, val)) ++#endif ++ { ++ __m128i mask = _mm_cmpgt_epi16(val, zero); ++ v_cost = _mm_add_epi16(v_cost, _mm_and_si128(mask, _mm_set1_epi16(2))); ++ val = _mm_srli_epi16(val, 1); + } +- cost = scost; ++ v_cost = _mm_add_epi16(v_cost, _mm_srli_si128(v_cost, 8)); ++ v_cost = _mm_add_epi16(v_cost ,_mm_srli_si128(v_cost, 4)); ++ v_cost = _mm_add_epi16(v_cost, _mm_srli_si128(v_cost, 2)); + } +- return cost; ++ return 16 + _mm_extract_epi16(v_cost, 0); ++#else /* } No SSE2 instructions { */ ++ uint32_t scost = 0; ++ for (int i = 0; i < 64; ++i) { ++ scost += 1 + 2 * uint16bit_length(abs(block.raw_data()[i])); ++ } ++ return scost; ++#endif /* } */ + } + + #ifdef ALLOW_FOUR_COLORS +--- src/vp8/model/model.hh ++++ src/vp8/model/model.hh +@@ -11,9 +11,7 @@ + #include "branch.hh" + #include "../util/aligned_block.hh" + #include "../util/block_based_image.hh" +-#include <smmintrin.h> +-#include <immintrin.h> +-#include <emmintrin.h> ++#include "../util/mm_mullo_epi32.hh" + + class BoolEncoder; + constexpr bool advanced_dc_prediction = true; +--- src/vp8/model/numeric.hh ++++ src/vp8/model/numeric.hh +@@ -8,8 +8,8 @@ + // for std::min + #include <algorithm> + #include <assert.h> +-#include <smmintrin.h> +-#include <emmintrin.h> ++#include <immintrin.h> ++#include "../util/mm_mullo_epi32.hh" + + #ifdef _WIN32 + #include <intrin.h> +--- src/vp8/util/mm_mullo_epi32.hh ++++ src/vp8/util/mm_mullo_epi32.hh +@@ -0,0 +1,16 @@ ++#if defined(__SSE2__) && !defined(__SSE4_1__) && !defined(MM_MULLO_EPI32_H) ++#define MM_MULLO_EPI32_H ++#include <immintrin.h> ++// See: http://stackoverflow.com/questions/10500766/sse-multiplication-of-4-32-bit-integers ++// and https://software.intel.com/en-us/forums/intel-c-compiler/topic/288768 ++static inline __m128i ++_mm_mullo_epi32(const __m128i &a, const __m128i &b) ++{ ++ __m128i tmp1 = _mm_mul_epu32(a,b); /* mul 2,0*/ ++ __m128i tmp2 = _mm_mul_epu32(_mm_srli_si128(a,4), ++ _mm_srli_si128(b,4)); /* mul 3,1 */ ++ return _mm_unpacklo_epi32( /* shuffle results to [63..0] and pack */ ++ _mm_shuffle_epi32(tmp1, _MM_SHUFFLE (0,0,2,0)), ++ _mm_shuffle_epi32(tmp2, _MM_SHUFFLE (0,0,2,0))); ++} ++#endif Added: head/graphics/lepton/pkg-descr ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ head/graphics/lepton/pkg-descr Fri Jul 15 01:27:43 2016 (r418567) @@ -0,0 +1,9 @@ +A tool for creating and manipulating images in Lepton format. + +Developed by Dropbox it is currently available under Apache-2.0 +license. + +Compared to JPEG, the format averages 22% lower file-sizes without +any (additional) loss of quality. + +WWW: https://github.com/dropbox/lepton/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201607150127.u6F1RhaS087043>