<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Thu, Nov 2, 2017 at 11:17 AM, Chema Casanova <<a href="mailto:jmcasanova@igalia.com" target="_blank">jmcasanova@igalia.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> El 01/11/17 a las 22:07, Jason Ekstrand escribió: > On Tue, Oct 17, 2017 at 10:05 AM, Jose Maria Casanova Crespo <div><div class="h5">> <<a href="mailto:jmcasanova@igalia.com">jmcasanova@igalia.com</a> <mailto:<a href="mailto:jmcasanova@igalia.com">jmcasanova@igalia.com</a>>> wrote: > > New shader attribute to mark when a location has 16-bit > value. This patch includes support on mesa glsl and nir. > > v2: Remove use of is_half_slot as is a duplicate of is_16bit > (Topi Pohjolainen) > --- > src/compiler/glsl_types.h | 15 +++++++++++++++ > src/compiler/nir/nir_gather_info.c | 21 ++++++++++++++------- > src/compiler/shader_info.h | 2 ++ > 3 files changed, 31 insertions(+), 7 deletions(-) > > diff --git a/src/compiler/glsl_types.h b/src/compiler/glsl_types.h > index 32399df351..e35e8d8f88 100644 > --- a/src/compiler/glsl_types.h > +++ b/src/compiler/glsl_types.h > @@ -93,6 +93,13 @@ static inline bool > glsl_base_type_is_integer(enum glsl_base_type type) > type == GLSL_TYPE_IMAGE; > } > > +static inline bool glsl_base_type_is_16bit(enum glsl_base_type type) > +{ > + return type == GLSL_TYPE_FLOAT16 || > + type == GLSL_TYPE_UINT16 || > + type == GLSL_TYPE_INT16; > +} > + > enum glsl_sampler_dim { > GLSL_SAMPLER_DIM_1D = 0, > GLSL_SAMPLER_DIM_2D, > @@ -555,6 +562,14 @@ struct glsl_type { > } > > /** > + * Query whether or not a type is 16-bit > + */ > + bool is_16bit() const > + { > + return glsl_base_type_is_16bit(base_type); > + } > + > + /** > * Query whether or not a type is a non-array boolean type > */ > bool is_boolean() const > diff --git a/src/compiler/nir/nir_gather_info.c > b/src/compiler/nir/nir_gather_info.c > index ac87bec46c..cce64f9c84 100644 > --- a/src/compiler/nir/nir_gather_info.c > +++ b/src/compiler/nir/nir_gather_info.c > @@ -212,14 +212,20 @@ gather_intrinsic_info(nir_intrinsic_instr > *instr, nir_shader *shader) > if (!try_mask_partial_io(shader, instr->variables[0])) > mark_whole_variable(shader, var); > > - /* We need to track which input_reads bits correspond to a > - * dvec3/dvec4 input attribute */ > + /* We need to track which input_reads bits correspond to > + * dvec3/dvec4 or 16-bit input attributes */ > if (shader->stage == MESA_SHADER_VERTEX && > - var->data.mode == nir_var_shader_in && > - glsl_type_is_dual_slot(glsl_without_array(var->type))) { > - for (uint i = 0; i < > glsl_count_attribute_slots(var->type, false); i++) { > - int idx = var->data.location + i; > - shader->info.double_inputs_read |= > BITFIELD64_BIT(idx); > + var->data.mode == nir_var_shader_in) { > + if > (glsl_type_is_dual_slot(glsl_without_array(var->type))) { > + for (uint i = 0; i < > glsl_count_attribute_slots(var->type, false); i++) { > + int idx = var->data.location + i; > + shader->info.double_inputs_read |= > BITFIELD64_BIT(idx); > + } > + } else if > (glsl_get_bit_size(glsl_without_array(var->type)) == 16) { > + for (uint i = 0; i < > glsl_count_attribute_slots(var->type, false); i++) { > + int idx = var->data.location + i; > + shader->info.half_inputs_read |= > BITFIELD64_BIT(idx); > + } > } > } > } > @@ -312,6 +318,7 @@ nir_shader_gather_info(nir_shader *shader, > nir_function_impl *entrypoint) > shader->info.outputs_written = 0; > shader->info.outputs_read = 0; > shader->info.double_inputs_read = 0; > + shader->info.half_inputs_read = 0; > shader->info.patch_inputs_read = 0; > shader->info.patch_outputs_written = 0; > shader->info.system_values_read = 0; > diff --git a/src/compiler/shader_info.h b/src/compiler/shader_info.h > index 38413940d6..98111fa1e0 100644 > --- a/src/compiler/shader_info.h > +++ b/src/compiler/shader_info.h > @@ -55,6 +55,8 @@ typedef struct shader_info { > uint64_t inputs_read; > /* Which inputs are actually read and are double */ > uint64_t double_inputs_read; > + /* Which inputs are actually read and are half */ > + uint64_t half_inputs_read; > > > Given that we're flagging this for 16-bit integers, I don't think > "half" is really appropriate. How about just 16bit_inputs_read? </div></div>I thought about that, but we can not do that. As the C restriction of variable names starting with alphabet or underscore. As the logic was the same as for double I didn't want to go for a inputs_read_16bits. I didn't come up with a better name. :( </blockquote><div> </div><div>Ugh. :-( You could always do _16bit_inputs_read. Or inputs_read_16bit. Really the double one will probably have to get renamed too since we have int64. </div></div></div></div>