dma: SG elem lists replaced with arrays

[mmaka1]

Multiple alloctions of SG elements linked as a list
replaced with single allocations of arrays.

• the code flows (esp. cleanup-on-error) are simplified,

• number of 64bytes chunk allocations reduced from 18 to 12 for
  example very basic topology, no increase in other areas
  (2-period sgls still fit),

• more efficient allocation for host page tables on legacy platforms;
  instead of going from a desc array to the list that is copied to
  another list, there is just a single allocation of array moved
  to the destination component.

Signed-off-by: Marcin Maka marcin.maka@linux.intel.com

[lgirdwood]

@mmaka1 looks like host build failed.

depbase=`echo testbench.o | sed 's|[^/]*$|.deps/&|;s|\.o$||'`;\
gcc -DHAVE_CONFIG_H -I. -I../../src/include  -I /home/sof/work/sof.git/host-lib/include/sof -I /home/sof/work/sof.git/src/library/include -I /home/sof/work/sof.git/src/include -I /home/sof/work/sof.git/src/arch/host/include -DSOF_INC=\"/home/sof/work/sof.git/host-lib/include/sof\"  -O3 -Wall -Werror -Wl,-EL -Wmissing-prototypes -Wimplicit-fallthrough=3 -g -Wall -g -O2 -MT testbench.o -MD -MP -MF $depbase.Tpo -c -o testbench.o testbench.c &&\
mv -f $depbase.Tpo $depbase.Po
/bin/bash ../../libtool  --tag=CC   --mode=link gcc -O3 -Wall -Werror -Wl,-EL -Wmissing-prototypes -Wimplicit-fallthrough=3 -g -Wall -g -O2 -lpthread -L../ipc -L../audio/.libs  -o testbench testbench.o -ldl -lm -lsof_ipc libtb_common.a -lsof -lcrypto 
libtool: link: gcc -O3 -Wall -Werror -Wl,-EL -Wmissing-prototypes -Wimplicit-fallthrough=3 -g -Wall -g -O2 -o .libs/testbench testbench.o  -lpthread -L../ipc -L../audio/.libs -ldl -lm /home/sof/work/sof.git/src/ipc/.libs/libsof_ipc.so libtb_common.a /home/sof/work/sof.git/src/audio/.libs/libsof.so -lcrypto -Wl,-rpath -Wl,/home/sof/work/sof.git/host-lib/lib
/home/sof/work/sof.git/src/audio/.libs/libsof.so: undefined reference to `dma_sg_free'
/home/sof/work/sof.git/src/audio/.libs/libsof.so: undefined reference to `dma_sg_alloc'
collect2: error: ld returned 1 exit status
Makefile:452: recipe for target 'testbench' failed
make[2]: *** [testbench] Error 1
make[2]: Leaving directory '/home/sof/work/sof.git/src/host'
Makefile:383: recipe for target 'all-recursive' failed
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory '/home/sof/work/sof.git/src'
Makefile:436: recipe for target 'all-recursive' failed

[mmaka1]

@lgirdwood I tried to compile the host locally but it failed with: ../../src/include/host/topology.h:34:10: fatal error: sound/asoc.h: No such file or directory. Are some dependency checks missing in the config routine?

[lgirdwood]

@mmaka1 oh it could be you are missing the linux kernel headers on your dev machine. These can be installed via the package manager or are alternatively missing alsa-lib development package.

[mmaka1]

@lrgirdwood Yes, I have the headers available in /usr/include/alsa on my machine. And assumed this should satisfy the build environment requirements. I just could not find anything like -I pointing there when followed the scripts/host-build-all.sh. That is why I asked about missing dependency checks (and compiler settings) in the configure.
Could not find anything like optional --with for host build to configure the location documented in the README.md too.

[tlauda]

@mmaka1 @lgirdwood Had the same problem. It looks like the gcc searches for the alsa headers in /usr/include/sound not in /usr/include/alsa/sound. Maybe the path has changed with the different versions of alsa. I've just copied the headers and it worked, but I don't know what should be the right solution.

[lgirdwood]

@mmaka1 @tlauda can you try this

diff --git a/configure.ac b/configure.ac
index 02c2579a..ba2cbf16 100644
--- a/configure.ac
+++ b/configure.ac
@@ -137,6 +137,8 @@ case "$with_arch" in
 
        ARCH="host"
        AC_SUBST(ARCH)
+
+       AC_CHECK_HEADERS([sound/asoc.h])
     ;;
     *)
        if test "$have_rimage" = "no" && test "$have_doc" = "no" ; then

This should hopefully detect and add the include path to your -I flags

[mmaka1]

@lgirdwood CI restarted and passed.

[lgirdwood]

@mmaka1 I will fix the header issue in a separate patch, looks a little more complex than my patch above.