On 08/24/2016 12:57 PM, Ilia Mirkin wrote: > On Wed, Aug 24, 2016 at 3:40 PM, Ian Romanick <[email protected]> wrote: >> On 08/24/2016 10:40 AM, Ilia Mirkin wrote: >>> The GL_ARB_compute_shader spec does not make a mention of a #extension >>> enable, nor a #define with the ext name. It follows that local_size_* >>> should always be allowed in a compute shader stage, and the #extension >>> support removed. >>> >>> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97447#c7 >>> Signed-off-by: Ilia Mirkin <[email protected]> >>> Cc: [email protected] >>> --- >>> >>> The change to exposing the gl_MaxCompute* constants is a little dodgy, >>> arguably >>> it should be exposed everywhere when compute functionality is available, but >>> that is a little annoying to do, and is no worse than the previous state of >>> things. >>> >>> src/compiler/glsl/builtin_variables.cpp | 2 +- >>> src/compiler/glsl/glsl_lexer.ll | 4 ++-- >>> src/compiler/glsl/glsl_parser.yy | 5 ++--- >>> src/compiler/glsl/glsl_parser_extras.cpp | 1 - >>> src/compiler/glsl/glsl_parser_extras.h | 7 ------- >>> 5 files changed, 5 insertions(+), 14 deletions(-) >>> >>> diff --git a/src/compiler/glsl/builtin_variables.cpp >>> b/src/compiler/glsl/builtin_variables.cpp >>> index c9d8b1c..4f88170 100644 >>> --- a/src/compiler/glsl/builtin_variables.cpp >>> +++ b/src/compiler/glsl/builtin_variables.cpp >>> @@ -789,7 +789,7 @@ builtin_variable_generator::generate_constants() >>> } >>> } >>> >>> - if (state->is_version(430, 310) || state->ARB_compute_shader_enable) { >>> + if (state->is_version(430, 310) || state->stage == MESA_SHADER_COMPUTE) >>> { >>> add_const("gl_MaxComputeAtomicCounterBuffers", >>> state->Const.MaxComputeAtomicCounterBuffers); >>> add_const("gl_MaxComputeAtomicCounters", >>> diff --git a/src/compiler/glsl/glsl_lexer.ll >>> b/src/compiler/glsl/glsl_lexer.ll >>> index 7be2b89..3324cde 100644 >>> --- a/src/compiler/glsl/glsl_lexer.ll >>> +++ b/src/compiler/glsl/glsl_lexer.ll >>> @@ -410,7 +410,7 @@ writeonly KEYWORD_WITH_ALT(420, 300, 420, 310, >>> yyextra->ARB_shader_image_lo >>> >>> atomic_uint KEYWORD_WITH_ALT(420, 300, 420, 310, >>> yyextra->ARB_shader_atomic_counters_enable, ATOMIC_UINT); >>> >>> -shared KEYWORD_WITH_ALT(430, 310, 430, 310, >>> yyextra->ARB_compute_shader_enable, SHARED); >>> +shared KEYWORD_WITH_ALT(430, 310, 430, 310, yyextra->stage == >>> MESA_SHADER_COMPUTE, SHARED); >>> >>> struct return STRUCT; >>> void return VOID_TOK; >>> @@ -425,7 +425,7 @@ layout { >>> || yyextra->ARB_uniform_buffer_object_enable >>> || yyextra->ARB_fragment_coord_conventions_enable >>> || yyextra->ARB_shading_language_420pack_enable >>> - || yyextra->ARB_compute_shader_enable >>> + || yyextra->stage == MESA_SHADER_COMPUTE >>> || yyextra->ARB_tessellation_shader_enable) { >>> return LAYOUT_TOK; >>> } else { >>> diff --git a/src/compiler/glsl/glsl_parser.yy >>> b/src/compiler/glsl/glsl_parser.yy >>> index 4f4a83c..cf2fb60 100644 >>> --- a/src/compiler/glsl/glsl_parser.yy >>> +++ b/src/compiler/glsl/glsl_parser.yy >>> @@ -1570,10 +1570,9 @@ layout_qualifier_id: >>> for (int i = 0; i < 3; i++) { >>> if (match_layout_qualifier(local_size_qualifiers[i], $1, >>> state) == 0) { >>> - if (!state->has_compute_shader()) { >>> + if (state->stage != MESA_SHADER_COMPUTE) { >>> _mesa_glsl_error(& @3, state, >>> - "%s qualifier requires GLSL 4.30 or " >>> - "GLSL ES 3.10 or ARB_compute_shader", >>> + "%s qualifier only valid in a compute >>> shader", >>> local_size_qualifiers[i]); >>> YYERROR; >>> } else { >>> diff --git a/src/compiler/glsl/glsl_parser_extras.cpp >>> b/src/compiler/glsl/glsl_parser_extras.cpp >>> index 14a5540..37aff38 100644 >>> --- a/src/compiler/glsl/glsl_parser_extras.cpp >>> +++ b/src/compiler/glsl/glsl_parser_extras.cpp >>> @@ -578,7 +578,6 @@ static const _mesa_glsl_extension >>> _mesa_glsl_supported_extensions[] = { >>> EXT(ARB_ES3_1_compatibility), >>> EXT(ARB_ES3_2_compatibility), >>> EXT(ARB_arrays_of_arrays), >>> - EXT(ARB_compute_shader), >> >> I think this will mean that having >> >> #extension GL_ARB_compute_shader: require >> >> will not compile. That seems bad. > > I see no indication that that should ever have worked in the first place.
There are a good number of conformance tests that do this. That's making me start to question the validity of this change at all. > -ilia > _______________________________________________ > mesa-stable mailing list > [email protected] > https://lists.freedesktop.org/mailman/listinfo/mesa-stable > _______________________________________________ mesa-dev mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/mesa-dev
