[Mesa-dev] [PATCH] nir: remove duplicate Makefile.sources

Rob Clark robdclark at gmail.com
Sat Apr 2 20:45:02 UTC 2016


On Sat, Apr 2, 2016 at 4:33 PM, Rob Clark <robdclark at gmail.com> wrote:
> On Sat, Apr 2, 2016 at 4:14 PM, Jason Ekstrand <jason at jlekstrand.net> wrote:
>> On Sat, Apr 2, 2016 at 11:35 AM, Rob Clark <robdclark at gmail.com> wrote:
>>>
>>> From: Rob Clark <robclark at freedesktop.org>
>>>
>>> Allegedly this was needed still by scons build.. but in practice it
>>> doesn't seem to be needed.  Removing it and running 'scons' results
>>> in no build errors.
>>
>>
>> Hrm... I'm not 100% sure what I'd like to see done here.  One option would
>> be to do what you've done and just remove the copies of Makefile.sources in
>> glsl and nir and fix scons to use the one in compiler.  Another would be to
>> make compiler/Makefile.am include the Makefile.sources from both glsl and
>> nir and do a $(foreach) in the makefile to add the "glsl/" or "nir/" prefix
>> to the filenames.  If that were a practical solution, I think it's what I
>> would prefer.
>
> hmm, src/compiler/glsl/Makefile.sources *does* actually seem to be
> used..  I didn't bother checking that one since I wasn't adding any
> glsl passes.. :-(

btw, if it wasn't clear, I think we should still at least push this
patch to drop src/compiler/nir/Makefile.sources since nir doesn't
build anyways under scons..

I'm not sure what the right thing to do w/ compiler/glsl but that can
come later..

BR,
-R

>> I did a bit of poking at the Android build and it looks like it uses
>> compiler/Makefile.sources for NIR and libcompiler but uses
>> compiler/glsl/Makefile.sources for GLSL stuff.
>
> ugg.. how do I always end up on things that touch build systems :-(
>
> BR,
> -R
>
>>>
>>> Signed-off-by: Rob Clark <robclark at freedesktop.org>
>>> ---
>>> So, afaict NIR is not even built w/ scons build (I'm just running
>>> 'scons' with no args, so let me know if I'm missing some build
>>> variant).  So at least if there is no scons variant that *does*
>>> build NIR, I think this is the right thing to do to reduce
>>> confusion.  But it brings up a bigger question of what to do
>>> with my patchset which adds NIR support in mesa/st, since that
>>> obviosly won't work with scons build as-is.
>>>
>>> I guess the two options are to try to add NIR into scons build
>>> (which involves some .py generated code, so maybe not trivial)
>>> or just #ifdef'ify all the mesa/st parts in my gallium-nir
>>> patchset which introduce dependencies on NIR.  Opinions?
>>>
>>> BR,
>>> -R
>>>
>>>  src/compiler/nir/Makefile.sources | 77
>>> ---------------------------------------
>>>  1 file changed, 77 deletions(-)
>>>  delete mode 100644 src/compiler/nir/Makefile.sources
>>>
>>> diff --git a/src/compiler/nir/Makefile.sources
>>> b/src/compiler/nir/Makefile.sources
>>> deleted file mode 100644
>>> index 3474302..0000000
>>> --- a/src/compiler/nir/Makefile.sources
>>> +++ /dev/null
>>> @@ -1,77 +0,0 @@
>>> -NIR_GENERATED_FILES = \
>>> -       nir_builder_opcodes.h \
>>> -       nir_constant_expressions.c \
>>> -       nir_opcodes.c \
>>> -       nir_opcodes.h \
>>> -       nir_opt_algebraic.c
>>> -
>>> -NIR_FILES = \
>>> -       glsl_to_nir.cpp \
>>> -       glsl_to_nir.h \
>>> -       nir.c \
>>> -       nir.h \
>>> -       nir_array.h \
>>> -       nir_builder.h \
>>> -       nir_clone.c \
>>> -       nir_constant_expressions.h \
>>> -       nir_control_flow.c \
>>> -       nir_control_flow.h \
>>> -       nir_control_flow_private.h \
>>> -       nir_dominance.c \
>>> -       nir_from_ssa.c \
>>> -       nir_gs_count_vertices.c \
>>> -       nir_inline_functions.c \
>>> -       nir_instr_set.c \
>>> -       nir_instr_set.h \
>>> -       nir_intrinsics.c \
>>> -       nir_intrinsics.h \
>>> -       nir_liveness.c \
>>> -       nir_lower_alu_to_scalar.c \
>>> -       nir_lower_atomics.c \
>>> -       nir_lower_clip.c \
>>> -       nir_lower_global_vars_to_local.c \
>>> -       nir_lower_gs_intrinsics.c \
>>> -       nir_lower_load_const_to_scalar.c \
>>> -       nir_lower_locals_to_regs.c \
>>> -       nir_lower_idiv.c \
>>> -       nir_lower_indirect_derefs.c \
>>> -       nir_lower_io.c \
>>> -       nir_lower_outputs_to_temporaries.c \
>>> -       nir_lower_phis_to_scalar.c \
>>> -       nir_lower_returns.c \
>>> -       nir_lower_samplers.c \
>>> -       nir_lower_system_values.c \
>>> -       nir_lower_tex.c \
>>> -       nir_lower_to_source_mods.c \
>>> -       nir_lower_two_sided_color.c \
>>> -       nir_lower_vars_to_ssa.c \
>>> -       nir_lower_var_copies.c \
>>> -       nir_lower_vec_to_movs.c \
>>> -       nir_metadata.c \
>>> -       nir_move_vec_src_uses_to_dest.c \
>>> -       nir_normalize_cubemap_coords.c \
>>> -       nir_opt_constant_folding.c \
>>> -       nir_opt_copy_propagate.c \
>>> -       nir_opt_cse.c \
>>> -       nir_opt_dce.c \
>>> -       nir_opt_dead_cf.c \
>>> -       nir_opt_gcm.c \
>>> -       nir_opt_global_to_local.c \
>>> -       nir_opt_peephole_select.c \
>>> -       nir_opt_remove_phis.c \
>>> -       nir_opt_undef.c \
>>> -       nir_phi_builder.c \
>>> -       nir_phi_builder.h \
>>> -       nir_print.c \
>>> -       nir_remove_dead_variables.c \
>>> -       nir_repair_ssa.c \
>>> -       nir_search.c \
>>> -       nir_search.h \
>>> -       nir_split_var_copies.c \
>>> -       nir_sweep.c \
>>> -       nir_to_ssa.c \
>>> -       nir_validate.c \
>>> -       nir_vla.h \
>>> -       nir_worklist.c \
>>> -       nir_worklist.h
>>> -
>>> --
>>> 2.5.5
>>>
>>


More information about the mesa-dev mailing list