<div dir="ltr"> <div class="gmail_extra"> <div class="gmail_quote">On Mon, May 9, 2016 at 12:34 PM, Rob Clark <<a href="mailto:robdclark@gmail.com" target="_blank">robdclark@gmail.com</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">From: Rob Clark <<a href="mailto:robclark@freedesktop.org" target="_blank">robclark@freedesktop.org</a>> Signed-off-by: Rob Clark <<a href="mailto:robclark@freedesktop.org" target="_blank">robclark@freedesktop.org</a>> --- src/compiler/nir/nir.h | 3 +- src/compiler/nir/nir_lower_io_to_temporaries.c | 56 +++++++++++++++++++++----- src/mesa/drivers/dri/i965/brw_nir.c | 4 +- 3 files changed, 52 insertions(+), 11 deletions(-) diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h index 5410f0b..c96eaf9 100644 --- a/src/compiler/nir/nir.h +++ b/src/compiler/nir/nir.h @@ -2279,7 +2279,8 @@ bool nir_lower_indirect_derefs(nir_shader *shader, nir_variable_mode modes); bool nir_lower_locals_to_regs(nir_shader *shader); -void nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint); +void nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint, + bool outputs, bool inputs); void nir_shader_gather_info(nir_shader *shader, nir_function_impl *entrypoint); diff --git a/src/compiler/nir/nir_lower_io_to_temporaries.c b/src/compiler/nir/nir_lower_io_to_temporaries.c index 9df2ba0..34e7477 100644 --- a/src/compiler/nir/nir_lower_io_to_temporaries.c +++ b/src/compiler/nir/nir_lower_io_to_temporaries.c @@ -22,9 +22,12 @@ */ /* - * Implements a pass that lowers output variables to a temporary plus an - * output variable with a single copy at each exit point of the shader. - * This way the output variable is only ever written. + * Implements a pass that lowers output and/or input variables to a + * temporary plus an output variable with a single copy at each exit + * point of the shader and/or an input variable with a single copy + * at the entrance point of the shader. This way the output variable + * is only ever written once and/or input is only read once, and there + * are no indirect outut/input accesses. */ #include "nir.h" @@ -33,6 +36,7 @@ struct lower_io_state { nir_shader *shader; nir_function *entrypoint; struct exec_list old_outputs; + struct exec_list old_inputs; }; static void @@ -49,7 +53,6 @@ emit_copies(nir_cursor cursor, nir_shader *shader, struct exec_list *new_vars, nir_intrinsic_instr_create(shader, nir_intrinsic_copy_var); copy->variables[0] = nir_deref_var_create(copy, newv); copy->variables[1] = nir_deref_var_create(copy, temp); - </blockquote><div> </div><div>I don't think this was intended. </div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> nir_instr_insert(cursor, &copy->instr); } } @@ -90,6 +93,20 @@ emit_output_copies_impl(struct lower_io_state *state, nir_function_impl *impl) } } +static void +emit_input_copies(nir_cursor cursor, struct lower_io_state *state) +{ + emit_copies(cursor, state->shader, &state->old_inputs, &state->shader->inputs); +} + +static void +emit_input_copies_impl(struct lower_io_state *state, nir_function_impl *impl) +{ + if (impl->function == state->entrypoint) { + emit_input_copies(nir_before_block(nir_start_block(impl)), state); + } +} </blockquote><div> </div><div>I was questioning the need for the wrapper before, but this is a bit silly :-) </div><div>I think I'd like to see the extra layer of wrappers go if you're not too attached to them. and I had one other trivial change above. Other than that the lower_io_to_temporaries patches are </div><div>Reviewed-by: Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> </div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> + static nir_variable * create_shadow_temp(struct lower_io_state *state, nir_variable *var) { @@ -105,8 +122,8 @@ create_shadow_temp(struct lower_io_state *state, nir_variable *var) /* Reparent the constant initializer (if any) */ ralloc_steal(nvar, nvar->constant_initializer); - /* Give the output a new name with @out-temp appended */ - const char *mode = "out"; + /* Give the original a new name with @<mode>-temp appended */ + const char *mode = (temp->data.mode == nir_var_shader_in) ? "in" : "out"; temp->name = ralloc_asprintf(var, "%s@%s-temp", mode, nvar->name); temp->data.mode = nir_var_global; temp->constant_initializer = NULL; @@ -115,7 +132,8 @@ create_shadow_temp(struct lower_io_state *state, nir_variable *var) } void -nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint) +nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint, + bool outputs, bool inputs) { struct lower_io_state state; @@ -124,7 +142,16 @@ nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint) state.shader = shader; state.entrypoint = entrypoint; - exec_list_move_nodes_to(&shader->outputs, &state.old_outputs); + + if (inputs) + exec_list_move_nodes_to(&shader->inputs, &state.old_inputs); + else + exec_list_make_empty(&state.old_inputs); + + if (outputs) + exec_list_move_nodes_to(&shader->outputs, &state.old_outputs); + else + exec_list_make_empty(&state.old_outputs); /* Walk over all of the outputs turn each output into a temporary and * make a new variable for the actual output. @@ -134,15 +161,26 @@ nir_lower_io_to_temporaries(nir_shader *shader, nir_function *entrypoint) exec_list_push_tail(&shader->outputs, &output->node); } + /* and same for inputs: */ + nir_foreach_variable(var, &state.old_inputs) { + nir_variable *input = create_shadow_temp(&state, var); + exec_list_push_tail(&shader->inputs, &input->node); + } + nir_foreach_function(function, shader) { if (function->impl == NULL) continue; - emit_output_copies_impl(&state, function->impl); + if (inputs) + emit_input_copies_impl(&state, function->impl); + + if (outputs) + emit_output_copies_impl(&state, function->impl); nir_metadata_preserve(function->impl, nir_metadata_block_index | nir_metadata_dominance); } + exec_list_append(&shader->globals, &state.old_inputs); exec_list_append(&shader->globals, &state.old_outputs); } diff --git a/src/mesa/drivers/dri/i965/brw_nir.c b/src/mesa/drivers/dri/i965/brw_nir.c index 2ed46ca..c5cc07e 100644 --- a/src/mesa/drivers/dri/i965/brw_nir.c +++ b/src/mesa/drivers/dri/i965/brw_nir.c @@ -553,7 +553,9 @@ brw_create_nir(struct brw_context *brw, /* First, lower the GLSL IR or Mesa IR to NIR */ if (shader_prog) { nir = glsl_to_nir(shader_prog, stage, options); - OPT_V(nir_lower_io_to_temporaries, nir_shader_get_entrypoint(nir)); + OPT_V(nir_lower_io_to_temporaries, + nir_shader_get_entrypoint(nir), + true, false); } else { nir = prog_to_nir(prog, options); OPT_V(nir_convert_to_ssa); /* turn registers into SSA */ -- 2.5.5 </blockquote></div> </div></div>