在 2020年7月21日 +0800 PM2:55，Bin.Cheng <amker.ch...@gmail.com>，写道：
> On Tue, Jul 21, 2020 at 11:14 AM Jojo <jiejie_r...@c-sky.com> wrote:
> >
> > gcc/ChangeLog:
> >
> > * genemit.c (main): Print 'split line'.
> > * Makefile.in (insn-emit.c): Define split count and file
> >
>
> Thanks for working one this, following comments are based on the
> assumption that the approach is feasible after your investigation.
>
> It's great to accelerate compilation time, do you have any number
> showing how much this can achieve, on a typical machine with
> reasonable parallelism?
>
Here are time collections from machine:
20 Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz,
Size of insn-emit.c (du -sh insn-emit.c) : 121M
Not split (same like j1) : 195 second
Split 20, -j1 : 213 second
Split 20, -j4: 76 second
Split 20, -j8: 44 second
Split 20, -j12: 44 second
Split 20, -j16: 44 second
Split 20, -j20: 44 second
Also, litter RAM will be consumed by litter code :)
> > ---
> > gcc/Makefile.in | 10 ++++++
> > gcc/genemit.c | 86 +++++++++++++++++++++++++++----------------------
> > 2 files changed, 58 insertions(+), 38 deletions(-)
> >
> > diff --git a/gcc/Makefile.in b/gcc/Makefile.in
> > index 2ba76656dbf..f805050a119 100644
> > --- a/gcc/Makefile.in
> > +++ b/gcc/Makefile.in
> > @@ -1253,6 +1253,13 @@ ANALYZER_OBJS = \
> > # We put the *-match.o and insn-*.o files first so that a parallel make
> > # will build them sooner, because they are large and otherwise tend to be
> > # the last objects to finish building.
> > +
> > +insn-generated-split-num = 15
> Hardcode number 15 looks strange here, how shall we know it's enough?
> Or one step further, can we use some kind of general match "*" for
> writing make rules here?
I have researched the detail of Makefile rules, As far as I know, ‘Make’
analyze the dependent objects before
some object running it’s shell command, so we need to static define the
‘insn-emit.c’ counts for creating ‘insn-emit.o’.
We can define split counts from processor numbers from machine, it should be
nice ?
It’s fixed in v3 patch.
> > +
> > +insn-emit-split-c = $(foreach o, $(shell for i in
> > {1..$(insn-generated-split-num)}; do echo $$i; done), insn-emit$(o).c)
> > +insn-emit-split-obj = $(patsubst %.c,%.o, $(insn-emit-split-c))
> > +$(insn-emit-split-c): insn-emit.c
> > +
> > OBJS = \
> > gimple-match.o \
> > generic-match.o \
> > @@ -1260,6 +1267,7 @@ OBJS = \
> > insn-automata.o \
> > insn-dfatab.o \
> > insn-emit.o \
> > + $(insn-emit-split-obj) \
> > insn-extract.o \
> > insn-latencytab.o \
> > insn-modes.o \
> > @@ -2367,6 +2375,8 @@ $(simple_generated_c:insn-%.c=s-%): s-%:
> > build/gen%$(build_exeext)
> > $(RUN_GEN) build/gen$*$(build_exeext) $(md_file) \
> > $(filter insn-conditions.md,$^) > tmp-$*.c
> > $(SHELL) $(srcdir)/../move-if-change tmp-$*.c insn-$*.c
> > + -csplit insn-$*.c /i\ am\ split\ line/ -k -s
> > {$(insn-generated-split-num)} -f insn-$* -b "%d.c"
> > + -( [ ! -s insn-$*0.c ] && for i in {1..$(insn-generated-split-num)}; do
> > touch insn-$*$$i.c; done && echo "" > insn-$*.c)
> Not sure if this is the first time that csplit/coreutils is used,
> shall we mention it here
> https://gcc.gnu.org/install/prerequisites.html, even check it in
> configure?
I grep coreutils from gcc source code and much more modules have checked the
coreutils
like libcc1, libssp and so on.
> > $(STAMP) s-$*
> >
> > # gencheck doesn't read the machine description, and the file produced
> > diff --git a/gcc/genemit.c b/gcc/genemit.c
> > index 84d07d388ee..fd60cdeeb96 100644
> > --- a/gcc/genemit.c
> > +++ b/gcc/genemit.c
> > @@ -847,6 +847,45 @@ handle_overloaded_gen (overloaded_name *oname)
> > }
> > }
> >
> > +#define printf_include() \
> > + printf ("/* Generated automatically by the program `genemit'\n\
> > +from the machine description file `md'. */\n\n"); \
> > + printf ("#define IN_TARGET_CODE 1\n"); \
> > + printf ("#include \"config.h\"\n"); \
> > + printf ("#include \"system.h\"\n"); \
> > + printf ("#include \"coretypes.h\"\n"); \
> > + printf ("#include \"backend.h\"\n"); \
> > + printf ("#include \"predict.h\"\n"); \
> > + printf ("#include \"tree.h\"\n"); \
> > + printf ("#include \"rtl.h\"\n"); \
> > + printf ("#include \"alias.h\"\n"); \
> > + printf ("#include \"varasm.h\"\n"); \
> > + printf ("#include \"stor-layout.h\"\n"); \
> > + printf ("#include \"calls.h\"\n"); \
> > + printf ("#include \"memmodel.h\"\n"); \
> > + printf ("#include \"tm_p.h\"\n"); \
> > + printf ("#include \"flags.h\"\n"); \
> > + printf ("#include \"insn-config.h\"\n"); \
> > + printf ("#include \"expmed.h\"\n"); \
> > + printf ("#include \"dojump.h\"\n"); \
> > + printf ("#include \"explow.h\"\n"); \
> > + printf ("#include \"emit-rtl.h\"\n"); \
> > + printf ("#include \"stmt.h\"\n"); \
> > + printf ("#include \"expr.h\"\n"); \
> > + printf ("#include \"insn-codes.h\"\n"); \
> > + printf ("#include \"optabs.h\"\n"); \
> > + printf ("#include \"dfp.h\"\n"); \
> > + printf ("#include \"output.h\"\n"); \
> > + printf ("#include \"recog.h\"\n"); \
> > + printf ("#include \"df.h\"\n"); \
> > + printf ("#include \"resource.h\"\n"); \
> > + printf ("#include \"reload.h\"\n"); \
> > + printf ("#include \"diagnostic-core.h\"\n"); \
> > + printf ("#include \"regs.h\"\n"); \
> > + printf ("#include \"tm-constrs.h\"\n"); \
> > + printf ("#include \"ggc.h\"\n"); \
> > + printf ("#include \"target.h\"\n\n"); \
>
> Can you use do {} while(0) style for defining this code block macro?
> The trailing '\' is also strange here.
It’s fixed in v3 patch.
> > +
> > int
> > main (int argc, const char **argv)
> > {
> > @@ -862,49 +901,19 @@ main (int argc, const char **argv)
> > /* Assign sequential codes to all entries in the machine description
> > in parallel with the tables in insn-output.c. */
> >
> > - printf ("/* Generated automatically by the program `genemit'\n\
> > -from the machine description file `md'. */\n\n");
> > -
> > - printf ("#define IN_TARGET_CODE 1\n");
> > - printf ("#include \"config.h\"\n");
> > - printf ("#include \"system.h\"\n");
> > - printf ("#include \"coretypes.h\"\n");
> > - printf ("#include \"backend.h\"\n");
> > - printf ("#include \"predict.h\"\n");
> > - printf ("#include \"tree.h\"\n");
> > - printf ("#include \"rtl.h\"\n");
> > - printf ("#include \"alias.h\"\n");
> > - printf ("#include \"varasm.h\"\n");
> > - printf ("#include \"stor-layout.h\"\n");
> > - printf ("#include \"calls.h\"\n");
> > - printf ("#include \"memmodel.h\"\n");
> > - printf ("#include \"tm_p.h\"\n");
> > - printf ("#include \"flags.h\"\n");
> > - printf ("#include \"insn-config.h\"\n");
> > - printf ("#include \"expmed.h\"\n");
> > - printf ("#include \"dojump.h\"\n");
> > - printf ("#include \"explow.h\"\n");
> > - printf ("#include \"emit-rtl.h\"\n");
> > - printf ("#include \"stmt.h\"\n");
> > - printf ("#include \"expr.h\"\n");
> > - printf ("#include \"insn-codes.h\"\n");
> > - printf ("#include \"optabs.h\"\n");
> > - printf ("#include \"dfp.h\"\n");
> > - printf ("#include \"output.h\"\n");
> > - printf ("#include \"recog.h\"\n");
> > - printf ("#include \"df.h\"\n");
> > - printf ("#include \"resource.h\"\n");
> > - printf ("#include \"reload.h\"\n");
> > - printf ("#include \"diagnostic-core.h\"\n");
> > - printf ("#include \"regs.h\"\n");
> > - printf ("#include \"tm-constrs.h\"\n");
> > - printf ("#include \"ggc.h\"\n");
> > - printf ("#include \"target.h\"\n\n");
> > + long long read_count = 0;
> >
> > /* Read the machine description. */
> >
> > md_rtx_info info;
> > while (read_md_rtx (&info))
> > + {
> > + if (!(read_count++ % 10000))
> > + {
> > + printf ("/* i am split line %d :) */\n\n", read_count);
> I would prefer a better comment explaining this line is to split file
> into separate compilation units for parallel compilation.
It’s fixed in v3 patch.
> > + printf_include()
> Together with above comment about macro, it's better to have a semicolon here?
It’s fixed in v3 patch.
>
> Thanks,
> bin
> > + }
> > +
> > switch (GET_CODE (info.def))
> > {
> > case DEFINE_INSN:
> > @@ -929,6 +938,7 @@ from the machine description file `md'. */\n\n");
> > default:
> > break;
> > }
> > + }
> >
> > /* Write out the routines to add CLOBBERs to a pattern and say whether they
> > clobber a hard reg. */
> > --
> > 2.24.3 (Apple Git-128)
