i965/nir: Add OPT() and OPT_V() macros for invoking NIR passes.

OPT() is the normal macro for passes that return booleans, while OPT_V()
is a variant that works for passes that don't properly report progress.
(Such passes should be fixed to return a boolean, eventually.)

These macros take care of calling nir_validate_shader() and setting
progress appropriately.  In the future, it would be easy to add shader
dumping similar to INTEL_DEBUG=optimizer by extending the macro.

v2 (Jason Ekstrand):
 - Fix an unused variable warning

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
This commit is contained in:
Kenneth Graunke 2015-11-03 00:31:15 -08:00 committed by Jason Ekstrand
parent d27ae2cf8c
commit 7bc0978999

View file

@ -56,8 +56,9 @@ remap_vs_attrs(nir_block *block, void *closure)
}
static void
brw_nir_lower_inputs(const struct brw_device_info *devinfo,
nir_shader *nir, bool is_scalar)
brw_nir_lower_inputs(nir_shader *nir,
const struct brw_device_info *devinfo,
bool is_scalar)
{
switch (nir->stage) {
case MESA_SHADER_VERTEX:
@ -170,46 +171,49 @@ brw_nir_lower_outputs(nir_shader *nir, bool is_scalar)
}
}
#define _OPT(do_pass) (({ \
bool this_progress = true; \
do_pass \
nir_validate_shader(nir); \
this_progress; \
}))
#define OPT(pass, ...) _OPT( \
this_progress = pass(nir ,##__VA_ARGS__); \
progress = progress || this_progress; \
)
#define OPT_V(pass, ...) _OPT( \
pass(nir, ##__VA_ARGS__); \
)
static void
nir_optimize(nir_shader *nir, bool is_scalar)
{
bool progress;
do {
progress = false;
nir_lower_vars_to_ssa(nir);
nir_validate_shader(nir);
OPT_V(nir_lower_vars_to_ssa);
if (is_scalar) {
nir_lower_alu_to_scalar(nir);
nir_validate_shader(nir);
OPT_V(nir_lower_alu_to_scalar);
}
progress |= nir_copy_prop(nir);
nir_validate_shader(nir);
OPT(nir_copy_prop);
if (is_scalar) {
nir_lower_phis_to_scalar(nir);
nir_validate_shader(nir);
OPT_V(nir_lower_phis_to_scalar);
}
progress |= nir_copy_prop(nir);
nir_validate_shader(nir);
progress |= nir_opt_dce(nir);
nir_validate_shader(nir);
progress |= nir_opt_cse(nir);
nir_validate_shader(nir);
progress |= nir_opt_peephole_select(nir);
nir_validate_shader(nir);
progress |= nir_opt_algebraic(nir);
nir_validate_shader(nir);
progress |= nir_opt_constant_folding(nir);
nir_validate_shader(nir);
progress |= nir_opt_dead_cf(nir);
nir_validate_shader(nir);
progress |= nir_opt_remove_phis(nir);
nir_validate_shader(nir);
progress |= nir_opt_undef(nir);
nir_validate_shader(nir);
OPT(nir_copy_prop);
OPT(nir_opt_dce);
OPT(nir_opt_cse);
OPT(nir_opt_peephole_select);
OPT(nir_opt_algebraic);
OPT(nir_opt_constant_folding);
OPT(nir_opt_dead_cf);
OPT(nir_opt_remove_phis);
OPT(nir_opt_undef);
} while (progress);
}
@ -228,6 +232,7 @@ brw_create_nir(struct brw_context *brw,
.lower_txp = ~0,
};
bool debug_enabled = INTEL_DEBUG & intel_debug_flag_for_shader_stage(stage);
bool progress = false;
nir_shader *nir;
/* First, lower the GLSL IR or Mesa IR to NIR */
@ -235,80 +240,63 @@ brw_create_nir(struct brw_context *brw,
nir = glsl_to_nir(shader_prog, stage, options);
} else {
nir = prog_to_nir(prog, options);
nir_convert_to_ssa(nir); /* turn registers into SSA */
OPT_V(nir_convert_to_ssa); /* turn registers into SSA */
}
nir_validate_shader(nir);
if (stage == MESA_SHADER_GEOMETRY) {
nir_lower_gs_intrinsics(nir);
nir_validate_shader(nir);
OPT(nir_lower_gs_intrinsics);
}
nir_lower_global_vars_to_local(nir);
nir_validate_shader(nir);
OPT(nir_lower_global_vars_to_local);
nir_lower_tex(nir, &tex_options);
nir_validate_shader(nir);
OPT_V(nir_lower_tex, &tex_options);
nir_normalize_cubemap_coords(nir);
nir_validate_shader(nir);
OPT(nir_normalize_cubemap_coords);
nir_split_var_copies(nir);
nir_validate_shader(nir);
OPT(nir_split_var_copies);
nir_optimize(nir, is_scalar);
/* Lower a bunch of stuff */
nir_lower_var_copies(nir);
nir_validate_shader(nir);
OPT_V(nir_lower_var_copies);
/* Get rid of split copies */
nir_optimize(nir, is_scalar);
brw_nir_lower_inputs(devinfo, nir, is_scalar);
brw_nir_lower_outputs(nir, is_scalar);
OPT_V(brw_nir_lower_inputs, devinfo, is_scalar);
OPT_V(brw_nir_lower_outputs, is_scalar);
nir_assign_var_locations(&nir->uniforms,
&nir->num_uniforms,
is_scalar ? type_size_scalar : type_size_vec4);
nir_lower_io(nir, -1, is_scalar ? type_size_scalar : type_size_vec4);
nir_validate_shader(nir);
OPT_V(nir_lower_io, -1, is_scalar ? type_size_scalar : type_size_vec4);
nir_remove_dead_variables(nir);
nir_validate_shader(nir);
OPT(nir_remove_dead_variables);
if (shader_prog) {
nir_lower_samplers(nir, shader_prog);
nir_validate_shader(nir);
OPT_V(nir_lower_samplers, shader_prog);
}
nir_lower_system_values(nir);
nir_validate_shader(nir);
OPT(nir_lower_system_values);
if (shader_prog) {
nir_lower_atomics(nir, shader_prog);
nir_validate_shader(nir);
OPT_V(nir_lower_atomics, shader_prog);
}
nir_optimize(nir, is_scalar);
if (brw->gen >= 6) {
/* Try and fuse multiply-adds */
brw_nir_opt_peephole_ffma(nir);
nir_validate_shader(nir);
OPT(brw_nir_opt_peephole_ffma);
}
nir_opt_algebraic_late(nir);
nir_validate_shader(nir);
OPT(nir_opt_algebraic_late);
nir_lower_locals_to_regs(nir);
nir_validate_shader(nir);
OPT(nir_lower_locals_to_regs);
nir_lower_to_source_mods(nir);
nir_validate_shader(nir);
nir_copy_prop(nir);
nir_validate_shader(nir);
nir_opt_dce(nir);
nir_validate_shader(nir);
OPT_V(nir_lower_to_source_mods);
OPT(nir_copy_prop);
OPT(nir_opt_dce);
if (unlikely(debug_enabled)) {
/* Re-index SSA defs so we print more sensible numbers. */
@ -322,17 +310,16 @@ brw_create_nir(struct brw_context *brw,
nir_print_shader(nir, stderr);
}
nir_convert_from_ssa(nir, true);
nir_validate_shader(nir);
OPT_V(nir_convert_from_ssa, true);
if (!is_scalar) {
nir_move_vec_src_uses_to_dest(nir);
nir_validate_shader(nir);
nir_lower_vec_to_movs(nir);
nir_validate_shader(nir);
OPT_V(nir_move_vec_src_uses_to_dest);
OPT(nir_lower_vec_to_movs);
}
/* Needed only so that OPT and OPT_V can set it */
(void)progress;
/* This is the last pass we run before we start emitting stuff. It
* determines when we need to insert boolean resolves on Gen <= 5. We
* run it last because it stashes data in instr->pass_flags and we don't