Add optimized crc32 for Power 8+ processors

.gitignore

@@ -13,6 +13,9 @@
 *.gcno
 *.gcov
 
+/crc32_test
+/crc32_test64
+/crc32_testsh
 /example
 /example64
 /examplesh

CMakeLists.txt

@@ -5,6 +5,8 @@ project(zlib C)
 
 set(VERSION "1.3")
 
+option(POWER "Enable building power implementation")
+
 set(INSTALL_BIN_DIR "${CMAKE_INSTALL_PREFIX}/bin" CACHE PATH "Installation directory for executables")
 set(INSTALL_LIB_DIR "${CMAKE_INSTALL_PREFIX}/lib" CACHE PATH "Installation directory for libraries")
 set(INSTALL_INC_DIR "${CMAKE_INSTALL_PREFIX}/include" CACHE PATH "Installation directory for headers")
@@ -126,6 +128,76 @@ if(NOT MINGW)
     )
 endif()
 
+if(CMAKE_COMPILER_IS_GNUCC)
+
+    # test to see if we can use a GNU indirect function to detect and load optimized code at runtime
+    CHECK_C_SOURCE_COMPILES("
+    static int test_ifunc_native(void)
+    {
+       return 1;
+    }
+    static int (*(check_ifunc_native(void)))(void)
+    {
+       return test_ifunc_native;
+    }
+    int test_ifunc(void) __attribute__ ((ifunc (\"check_ifunc_native\")));
+    int main(void)
+    {
+       return 0;
+    }
+    " HAS_C_ATTR_IFUNC)
+
+    if(HAS_C_ATTR_IFUNC)
+        add_definitions(-DHAVE_IFUNC)
+        set(ZLIB_PRIVATE_HDRS ${ZLIB_PRIVATE_HDRS} contrib/gcc/zifunc.h)
+    endif()
+
+    if(POWER)
+        # Test to see if we can use the optimizations for Power
+        CHECK_C_SOURCE_COMPILES("
+        #ifndef _ARCH_PPC
+            #error \"Target is not Power\"
+        #endif
+        #ifndef __BUILTIN_CPU_SUPPORTS__
+            #error \"Target doesn't support __builtin_cpu_supports()\"
+        #endif
+        int main() { return 0; }
+        " HAS_POWER_SUPPORT)
+
+        if(HAS_POWER_SUPPORT AND HAS_C_ATTR_IFUNC)
+            add_definitions(-DZ_POWER_OPT)
+
+            set(CMAKE_REQUIRED_FLAGS -mcpu=power8)
+            CHECK_C_SOURCE_COMPILES("int main(void){return 0;}" POWER8)
+
+            if(POWER8)
+                add_definitions(-DZ_POWER8)
+                set(ZLIB_POWER8
+                  contrib/power/crc32_z_power8.c)
+
+                set_source_files_properties(
+                    ${ZLIB_POWER8}
+                    PROPERTIES COMPILE_FLAGS -mcpu=power8)
+            endif()
+
+            set(CMAKE_REQUIRED_FLAGS -mcpu=power9)
+            CHECK_C_SOURCE_COMPILES("int main(void){return 0;}" POWER9)
+
+            if(POWER9)
+                add_definitions(-DZ_POWER9)
+                set(ZLIB_POWER9 )
+
+                set_source_files_properties(
+                    ${ZLIB_POWER9}
+                    PROPERTIES COMPILE_FLAGS -mcpu=power9)
+            endif()
+
+            set(ZLIB_PRIVATE_HDRS ${ZLIB_PRIVATE_HDRS} contrib/power/power.h)
+            set(ZLIB_SRCS ${ZLIB_SRCS} ${ZLIB_POWER8} ${ZLIB_POWER9})
+        endif()
+    endif()
+endif()
+
 # parse the full version number from zlib.h and include in ZLIB_FULL_VERSION
 file(READ ${CMAKE_CURRENT_SOURCE_DIR}/zlib.h _zlib_h_contents)
 string(REGEX REPLACE ".*#define[ \t]+ZLIB_VERSION[ \t]+\"([-0-9A-Za-z.]+)\".*"
@@ -198,6 +270,10 @@ add_executable(example test/example.c)
 target_link_libraries(example zlib)
 add_test(example example)
 
+add_executable(crc32_test test/crc32_test.c)
+target_link_libraries(crc32_test zlib)
+add_test(crc32_test crc32_test)
+
 add_executable(minigzip test/minigzip.c)
 target_link_libraries(minigzip zlib)
 

Makefile.in

@@ -71,19 +71,19 @@ PIC_OBJS = $(PIC_OBJC) $(PIC_OBJA)
 
 all: static shared
 
-static: example$(EXE) minigzip$(EXE)
+static: crc32_test$(EXE) example$(EXE) minigzip$(EXE)
 
-shared: examplesh$(EXE) minigzipsh$(EXE)
+shared: crc32_testsh$(EXE) examplesh$(EXE) minigzipsh$(EXE)
 
-all64: example64$(EXE) minigzip64$(EXE)
+all64: crc32_test64$(EXE) example64$(EXE) minigzip64$(EXE)
 
 check: test
 
 test: all teststatic testshared
 
 teststatic: static
 	@TMPST=tmpst_$$; \
-	if echo hello world | ${QEMU_RUN} ./minigzip | ${QEMU_RUN} ./minigzip -d && ${QEMU_RUN} ./example $$TMPST ; then \
+	if echo hello world | ${QEMU_RUN} ./minigzip | ${QEMU_RUN} ./minigzip -d && ${QEMU_RUN} ./example $$TMPST && ${QEMU_RUN} ./crc32_test; then \
 	  echo '		*** zlib test OK ***'; \
 	else \
 	  echo '		*** zlib test FAILED ***'; false; \
@@ -96,7 +96,7 @@ testshared: shared
 	DYLD_LIBRARY_PATH=`pwd`:$(DYLD_LIBRARY_PATH) ; export DYLD_LIBRARY_PATH; \
 	SHLIB_PATH=`pwd`:$(SHLIB_PATH) ; export SHLIB_PATH; \
 	TMPSH=tmpsh_$$; \
-	if echo hello world | ${QEMU_RUN} ./minigzipsh | ${QEMU_RUN} ./minigzipsh -d && ${QEMU_RUN} ./examplesh $$TMPSH; then \
+	if echo hello world | ${QEMU_RUN} ./minigzipsh | ${QEMU_RUN} ./minigzipsh -d && ${QEMU_RUN} ./examplesh $$TMPSH && ${QEMU_RUN} ./crc32_testsh; then \
 	  echo '		*** zlib shared test OK ***'; \
 	else \
 	  echo '		*** zlib shared test FAILED ***'; false; \
@@ -105,7 +105,7 @@ testshared: shared
 
 test64: all64
 	@TMP64=tmp64_$$; \
-	if echo hello world | ${QEMU_RUN} ./minigzip64 | ${QEMU_RUN} ./minigzip64 -d && ${QEMU_RUN} ./example64 $$TMP64; then \
+	if echo hello world | ${QEMU_RUN} ./minigzip64 | ${QEMU_RUN} ./minigzip64 -d && ${QEMU_RUN} ./example64 $$TMP64 && ${QEMU_RUN} ./crc32_test64; then \
 	  echo '		*** zlib 64-bit test OK ***'; \
 	else \
 	  echo '		*** zlib 64-bit test FAILED ***'; false; \
@@ -139,12 +139,18 @@ match.lo: match.S
 	mv _match.o match.lo
 	rm -f _match.s
 
+crc32_test.o: $(SRCDIR)test/crc32_test.c $(SRCDIR)zlib.h zconf.h
+	$(CC) $(CFLAGS) $(ZINCOUT) -c -o $@ $(SRCDIR)test/crc32_test.c
+
 example.o: $(SRCDIR)test/example.c $(SRCDIR)zlib.h zconf.h
 	$(CC) $(CFLAGS) $(ZINCOUT) -c -o $@ $(SRCDIR)test/example.c
 
 minigzip.o: $(SRCDIR)test/minigzip.c $(SRCDIR)zlib.h zconf.h
 	$(CC) $(CFLAGS) $(ZINCOUT) -c -o $@ $(SRCDIR)test/minigzip.c
 
+crc32_test64.o: $(SRCDIR)test/crc32_test.c $(SRCDIR)zlib.h zconf.h
+	$(CC) $(CFLAGS) $(ZINCOUT) -D_FILE_OFFSET_BITS=64 -c -o $@ $(SRCDIR)test/crc32_test.c
+
 example64.o: $(SRCDIR)test/example.c $(SRCDIR)zlib.h zconf.h
 	$(CC) $(CFLAGS) $(ZINCOUT) -D_FILE_OFFSET_BITS=64 -c -o $@ $(SRCDIR)test/example.c
 
@@ -158,6 +164,9 @@ adler32.o: $(SRCDIR)adler32.c
 crc32.o: $(SRCDIR)crc32.c
 	$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)crc32.c
 
+crc32_z_power8.o: $(SRCDIR)contrib/power/crc32_z_power8.c
+	$(CC) $(CFLAGS) -mcpu=power8 $(ZINC) -c -o $@ $(SRCDIR)contrib/power/crc32_z_power8.c
+
 deflate.o: $(SRCDIR)deflate.c
 	$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)deflate.c
 
@@ -208,6 +217,11 @@ crc32.lo: $(SRCDIR)crc32.c
 	$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/crc32.o $(SRCDIR)crc32.c
 	-@mv objs/crc32.o $@
 
+crc32_z_power8.lo: $(SRCDIR)contrib/power/crc32_z_power8.c
+	-@mkdir objs 2>/dev/null || test -d objs
+	$(CC) $(SFLAGS) -mcpu=power8 $(ZINC) -DPIC -c -o objs/crc32_z_power8.o $(SRCDIR)contrib/power/crc32_z_power8.c
+	-@mv objs/crc32_z_power8.o $@
+
 deflate.lo: $(SRCDIR)deflate.c
 	-@mkdir objs 2>/dev/null || test -d objs
 	$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/deflate.o $(SRCDIR)deflate.c
@@ -281,18 +295,27 @@ placebo $(SHAREDLIBV): $(PIC_OBJS) libz.a
 	ln -s $@ $(SHAREDLIBM)
 	-@rmdir objs
 
+crc32_test$(EXE): crc32_test.o $(STATICLIB)
+	$(CC) $(CFLAGS) -o $@ crc32_test.o $(TEST_LDFLAGS)
+
 example$(EXE): example.o $(STATICLIB)
 	$(CC) $(CFLAGS) -o $@ example.o $(TEST_LDFLAGS)
 
 minigzip$(EXE): minigzip.o $(STATICLIB)
 	$(CC) $(CFLAGS) -o $@ minigzip.o $(TEST_LDFLAGS)
 
+crc32_testsh$(EXE): crc32_test.o $(SHAREDLIBV)
+	$(CC) $(CFLAGS) -o $@ crc32_test.o -L. $(SHAREDLIBV)
+
 examplesh$(EXE): example.o $(SHAREDLIBV)
 	$(CC) $(CFLAGS) -o $@ example.o $(LDFLAGS) -L. $(SHAREDLIBV)
 
 minigzipsh$(EXE): minigzip.o $(SHAREDLIBV)
 	$(CC) $(CFLAGS) -o $@ minigzip.o $(LDFLAGS) -L. $(SHAREDLIBV)
 
+crc32_test64$(EXE): crc32_test64.o $(STATICLIB)
+	$(CC) $(CFLAGS) -o $@ crc32_test64.o $(TEST_LDFLAGS)
+
 example64$(EXE): example64.o $(STATICLIB)
 	$(CC) $(CFLAGS) -o $@ example64.o $(TEST_LDFLAGS)
 
@@ -368,8 +391,8 @@ minizip-clean:
 mostlyclean: clean
 clean: minizip-clean
 	rm -f *.o *.lo *~ \
-	   example$(EXE) minigzip$(EXE) examplesh$(EXE) minigzipsh$(EXE) \
-	   example64$(EXE) minigzip64$(EXE) \
+	   crc32_test$(EXE) example$(EXE) minigzip$(EXE) crc32_testsh$(EXE) examplesh$(EXE) minigzipsh$(EXE) \
+	   crc32_test64$(EXE) example64$(EXE) minigzip64$(EXE) \
 	   infcover \
 	   libz.* foo.gz so_locations \
 	   _match.s maketree contrib/infback9/*.o
@@ -391,7 +414,7 @@ tags:
 
 adler32.o zutil.o: $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h
 gzclose.o gzlib.o gzread.o gzwrite.o: $(SRCDIR)zlib.h zconf.h $(SRCDIR)gzguts.h
-compress.o example.o minigzip.o uncompr.o: $(SRCDIR)zlib.h zconf.h
+compress.o crc32_test.o example.o minigzip.o uncompr.o: $(SRCDIR)zlib.h zconf.h
 crc32.o: $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h $(SRCDIR)crc32.h
 deflate.o: $(SRCDIR)deflate.h $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h
 infback.o inflate.o: $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h $(SRCDIR)inftrees.h $(SRCDIR)inflate.h $(SRCDIR)inffast.h $(SRCDIR)inffixed.h
@@ -401,7 +424,7 @@ trees.o: $(SRCDIR)deflate.h $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h $(SRCDIR)tr
 
 adler32.lo zutil.lo: $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h
 gzclose.lo gzlib.lo gzread.lo gzwrite.lo: $(SRCDIR)zlib.h zconf.h $(SRCDIR)gzguts.h
-compress.lo example.lo minigzip.lo uncompr.lo: $(SRCDIR)zlib.h zconf.h
+compress.lo crc32_test.lo example.lo minigzip.lo uncompr.lo: $(SRCDIR)zlib.h zconf.h
 crc32.lo: $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h $(SRCDIR)crc32.h
 deflate.lo: $(SRCDIR)deflate.h $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h
 infback.lo inflate.lo: $(SRCDIR)zutil.h $(SRCDIR)zlib.h zconf.h $(SRCDIR)inftrees.h $(SRCDIR)inflate.h $(SRCDIR)inffast.h $(SRCDIR)inffixed.h

configure

@@ -834,6 +834,75 @@ EOF
   fi
 fi
 
+# test to see if we can use a gnu indirection function to detect and load optimized code at runtime
+echo >> configure.log
+cat > $test.c <<EOF
+static int test_ifunc_native(void)
+{
+  return 1;
+}
+
+static int (*(check_ifunc_native(void)))(void)
+{
+  return test_ifunc_native;
+}
+
+int test_ifunc(void) __attribute__ ((ifunc ("check_ifunc_native")));
+EOF
+
+if tryboth $CC -c $CFLAGS $test.c; then
+  SFLAGS="${SFLAGS} -DHAVE_IFUNC"
+  CFLAGS="${CFLAGS} -DHAVE_IFUNC"
+  echo "Checking for attribute(ifunc) support... Yes." | tee -a configure.log
+else
+  echo "Checking for attribute(ifunc) support... No." | tee -a configure.log
+fi
+
+# Test to see if we can use the optimizations for Power
+echo >> configure.log
+cat > $test.c <<EOF
+#ifndef _ARCH_PPC
+  #error "Target is not Power"
+#endif
+#if !(defined(__PPC64__) || defined(__powerpc64__))
+  #error "Target is not 64 bits"
+#endif
+#ifndef HAVE_IFUNC
+  #error "Target doesn't support ifunc"
+#endif
+#ifndef __BUILTIN_CPU_SUPPORTS__
+  #error "Target doesn't support __builtin_cpu_supports()"
+#endif
+EOF
+
+if tryboth $CC -c $CFLAGS $test.c; then
+  echo "int main(void){return 0;}" > $test.c
+
+  if tryboth $CC -c $CFLAGS -mcpu=power8 $test.c; then
+    POWER8="-DZ_POWER8"
+    PIC_OBJC="${PIC_OBJC} crc32_z_power8.lo"
+    OBJC="${OBJC} crc32_z_power8.o"
+    echo "Checking for -mcpu=power8 support... Yes." | tee -a configure.log
+  else
+    echo "Checking for -mcpu=power8 support... No." | tee -a configure.log
+  fi
+
+  if tryboth $CC -c $CFLAGS -mcpu=power9 $test.c; then
+    POWER9="-DZ_POWER9"
+    PIC_OBJC="${PIC_OBJC}"
+    OBJC="${OBJC}"
+    echo "Checking for -mcpu=power9 support... Yes." | tee -a configure.log
+  else
+    echo "Checking for -mcpu=power9 support... No." | tee -a configure.log
+  fi
+
+  SFLAGS="${SFLAGS} ${POWER8} ${POWER9} -DZ_POWER_OPT"
+  CFLAGS="${CFLAGS} ${POWER8} ${POWER9} -DZ_POWER_OPT"
+  echo "Checking for Power optimizations support... Yes." | tee -a configure.log
+else
+  echo "Checking for Power optimizations support... No." | tee -a configure.log
+fi
+
 # show the results in the log
 echo >> configure.log
 echo ALL = $ALL >> configure.log

contrib/README.contrib

@@ -11,6 +11,10 @@ ada/        by Dmitriy Anisimkov <[email protected]>
 blast/      by Mark Adler <[email protected]>
         Decompressor for output of PKWare Data Compression Library (DCL)
 
+gcc/        by Matheus Castanho <[email protected]>
+            and Rogerio Alves <[email protected]>
+        Optimization helpers using GCC-specific extensions
+
 delphi/     by Cosmin Truta <[email protected]>
         Support for Delphi and C++ Builder
 
@@ -42,6 +46,11 @@ minizip/    by Gilles Vollant <[email protected]>
 pascal/     by Bob Dellaca <[email protected]> et al.
         Support for Pascal
 
+power/      by Daniel Black <[email protected]>
+               Matheus Castanho <[email protected]>
+            and Rogerio Alves <[email protected]>
+        Optimized functions for Power processors
+
 puff/       by Mark Adler <[email protected]>
         Small, low memory usage inflate.  Also serves to provide an
         unambiguous description of the deflate format.