Re: [PATCH] kbuild: add -fno-discard-value-names to cmd_cc_ll_c

[Nick Desaulniers]

On Thu, Oct 13, 2022 at 1:41 PM Fangrui Song <maskray@xxxxxxxxxx> wrote:
>
> On 2022-10-13, Nathan Chancellor wrote:
> >On Fri, Oct 07, 2022 at 01:32:36PM -0700, Nick Desaulniers wrote:
> >> When debugging LLVM IR, it can be handy for clang to not discard value
> >> names used for local variables and parameters. Compare the generated IR.
> >>
> >> Implicit Default (-fdiscard-value-names):
> >>   define i32 @core_sys_select(i32 %0, ptr %1, ptr %2, ptr %3, ptr %4) {
> >>     %6 = alloca i64
> >>     %7 = alloca %struct.poll_wqueues
> >>     %8 = alloca [64 x i32]
> >>
> >> Explicit -fno-discard-value-names:
> >>   define i32 @core_sys_select(i32 %n, ptr %inp, ptr %outp, ptr %exp,
> >>                               ptr %end_time) {
> >>     %expire.i = alloca i64
> >>     %table.i = alloca %struct.poll_wqueues
> >>     %stack_fds = alloca [64 x i32]
> >>
> >> The rule for generating human readable LLVM IR (.ll) is only useful as a
> >> debugging feature:
> >>
> >> $ make LLVM=1 fs/select.ll
> >>
> >> Clang defaults to -fdiscard-value-names to save memory when generating
> >> LLVM IR. For debugging purposes, the improvement in readability at a
> >> cost of more verbose IR is a cost we're happy to pay.
>
> A LLVM_ENABLE_ASSEERTIONS=off build of Clang defaults to -fdiscard-value-names.
>
> A LLVM_ENABLE_ASSEERTIONS=on build of Clang defaults to -fno-discard-value-names.
>
> See clang/lib/Driver/ToolChains/Clang.cpp:5030

Thanks!

I thought I recalled this; I was literally just rereading through
https://github.com/ClangBuiltLinux/linux/issues/1467#issuecomment-1046071624
again to understand.

That would be useful to include in the commit message, and is more
precise than "Clang defaults to -fdiscard-value-names to save memory
when generating LLVM IR."  Masahiro, would you mind including or
replacing that, or shall I send a v2?

I was surprised to find later in the day that I was running an
assertions-disabled (ie. release) build of clang, which I almost never
do!  Non-llvm-developers are more likely to be using release builds
than assertions-enabled builds of llvm, so this patch still very much
has merit.

>
> >> Signed-off-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> >
> >Thanks for the patch!
> >
> >Link: https://github.com/ClangBuiltLinux/linux/issues/1467
> >Reviewed-by: Nathan Chancellor <nathan@xxxxxxxxxx>
>
> Reviewed-by: Fangrui Song <maskray@xxxxxxxxxx>
>
> >> ---
> >>  scripts/Makefile.build | 2 +-
> >>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/scripts/Makefile.build b/scripts/Makefile.build
> >> index 27be77c0d6d8..d0e4f476dfee 100644
> >> --- a/scripts/Makefile.build
> >> +++ b/scripts/Makefile.build
> >> @@ -140,7 +140,7 @@ $(obj)/%.symtypes : $(src)/%.c FORCE
> >>  # LLVM assembly
> >>  # Generate .ll files from .c
> >>  quiet_cmd_cc_ll_c = CC $(quiet_modtag)  $@
> >> -      cmd_cc_ll_c = $(CC) $(c_flags) -emit-llvm -S -o $@ $<
> >> +      cmd_cc_ll_c = $(CC) $(c_flags) -emit-llvm -S -fno-discard-value-names -o $@ $<
> >>
> >>  $(obj)/%.ll: $(src)/%.c FORCE
> >>      $(call if_changed_dep,cc_ll_c)
> >>
> >> base-commit: 93ed07a23fd08b8613f64cf0a15d7fbdaca010fd
> >> --
> >> 2.38.0.rc2.412.g84df46c1b4-goog
> >>
> >



-- 
Thanks,
~Nick Desaulniers