v3d: rename vertex shader key (num)_fs_inputs fields

Until now this made sense because we always paired vertex shaders
with fragment shaders, but as soon as we implement geometry and
tessellation shaders that will no longer be the case, so rename
this to (num_)used_outputs.

v2: Use 'used_outputs' instead of ns_outputs, which is more explicit (Eric).

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

diff --git a/src/broadcom/common/v3d_limits.h b/src/broadcom/common/v3d_limits.h
index 7768476..d65eddd 100644 (file)
--- a/src/broadcom/common/v3d_limits.h
+++ b/src/broadcom/common/v3d_limits.h
@@ -31,6 +31,7 @@
 
 #define V3D_MAX_FS_INPUTS 64
 #define V3D_MAX_VS_INPUTS 64
+#define V3D_MAX_ANY_STAGE_INPUTS MAX2(V3D_MAX_VS_INPUTS, V3D_MAX_FS_INPUTS)
 
 /* Not specifically a hardware limit, just coordination between compiler and
  * driver.

diff --git a/src/broadcom/compiler/v3d_compiler.h b/src/broadcom/compiler/v3d_compiler.h
index fbb4b64..e2a4665 100644 (file)
--- a/src/broadcom/compiler/v3d_compiler.h
+++ b/src/broadcom/compiler/v3d_compiler.h
@@ -374,8 +374,8 @@ struct v3d_fs_key {
 struct v3d_vs_key {
         struct v3d_key base;
 
-        struct v3d_varying_slot fs_inputs[V3D_MAX_FS_INPUTS];
-        uint8_t num_fs_inputs;
+        struct v3d_varying_slot used_outputs[V3D_MAX_ANY_STAGE_INPUTS];
+        uint8_t num_used_outputs;
 
         bool is_coord;
         bool per_vertex_point_size;

diff --git a/src/broadcom/compiler/v3d_nir_lower_io.c b/src/broadcom/compiler/v3d_nir_lower_io.c
index 2a68efb..ee9b29d 100644 (file)
--- a/src/broadcom/compiler/v3d_nir_lower_io.c
+++ b/src/broadcom/compiler/v3d_nir_lower_io.c
@@ -45,7 +45,7 @@ struct v3d_nir_lower_io_state {
         int psiz_vpm_offset;
         int varyings_vpm_offset;
 
-        BITSET_WORD varyings_stored[BITSET_WORDS(V3D_MAX_FS_INPUTS)];
+        BITSET_WORD varyings_stored[BITSET_WORDS(V3D_MAX_ANY_STAGE_INPUTS)];
 
         nir_ssa_def *pos[4];
 };
@@ -91,8 +91,8 @@ v3d_varying_slot_vpm_offset(struct v3d_compile *c, nir_variable *var, int chan)
 {
         int component = var->data.location_frac + chan;
 
-        for (int i = 0; i < c->vs_key->num_fs_inputs; i++) {
-                struct v3d_varying_slot slot = c->vs_key->fs_inputs[i];
+        for (int i = 0; i < c->vs_key->num_used_outputs; i++) {
+                struct v3d_varying_slot slot = c->vs_key->used_outputs[i];
 
                 if (v3d_slot_get_slot(slot) == var->data.location &&
                     v3d_slot_get_component(slot) == component) {
@@ -255,7 +255,7 @@ v3d_nir_setup_vpm_layout(struct v3d_compile *c,
 
         state->varyings_vpm_offset = vpm_offset;
 
-        c->vpm_output_size = vpm_offset + c->vs_key->num_fs_inputs;
+        c->vpm_output_size = vpm_offset + c->vs_key->num_used_outputs;
 }
 
 static void
@@ -306,7 +306,7 @@ v3d_nir_emit_ff_vpm_outputs(struct v3d_compile *c, nir_builder *b,
          * This should be undefined behavior, but glsl-routing seems to rely
          * on it.
          */
-        for (int i = 0; i < c->vs_key->num_fs_inputs; i++) {
+        for (int i = 0; i < c->vs_key->num_used_outputs; i++) {
                 if (!BITSET_TEST(state->varyings_stored, i)) {
                         v3d_nir_store_output(b, state->varyings_vpm_offset + i,
                                              nir_imm_int(b, 0));

diff --git a/src/broadcom/compiler/vir.c b/src/broadcom/compiler/vir.c
index da42340..1200900 100644 (file)
--- a/src/broadcom/compiler/vir.c
+++ b/src/broadcom/compiler/vir.c
@@ -749,9 +749,9 @@ v3d_nir_lower_vs_early(struct v3d_compile *c)
         NIR_PASS_V(c->s, nir_lower_io_to_scalar_early,
                    nir_var_shader_in | nir_var_shader_out);
         uint64_t used_outputs[4] = {0};
-        for (int i = 0; i < c->vs_key->num_fs_inputs; i++) {
-                int slot = v3d_slot_get_slot(c->vs_key->fs_inputs[i]);
-                int comp = v3d_slot_get_component(c->vs_key->fs_inputs[i]);
+        for (int i = 0; i < c->vs_key->num_used_outputs; i++) {
+                int slot = v3d_slot_get_slot(c->vs_key->used_outputs[i]);
+                int comp = v3d_slot_get_component(c->vs_key->used_outputs[i]);
                 used_outputs[comp] |= 1ull << slot;
         }
         NIR_PASS_V(c->s, nir_remove_unused_io_vars,

diff --git a/src/gallium/drivers/v3d/v3d_program.c b/src/gallium/drivers/v3d/v3d_program.c
index ff91b33..75022d8 100644 (file)
--- a/src/gallium/drivers/v3d/v3d_program.c
+++ b/src/gallium/drivers/v3d/v3d_program.c
@@ -220,7 +220,7 @@ v3d_shader_precompile(struct v3d_context *v3d,
                         int slot = var->data.location;
                         for (int i = 0; i < glsl_get_components(var->type); i++) {
                                 int swiz = var->data.location_frac + i;
-                                key.fs_inputs[key.num_fs_inputs++] =
+                                key.used_outputs[key.num_used_outputs++] =
                                         v3d_slot_from_slot_and_component(slot,
                                                                          swiz);
                         }
@@ -230,9 +230,9 @@ v3d_shader_precompile(struct v3d_context *v3d,
 
                 /* Compile VS bin shader: only position (XXX: include TF) */
                 key.is_coord = true;
-                key.num_fs_inputs = 0;
+                key.num_used_outputs = 0;
                 for (int i = 0; i < 4; i++) {
-                        key.fs_inputs[key.num_fs_inputs++] =
+                        key.used_outputs[key.num_used_outputs++] =
                                 v3d_slot_from_slot_and_component(VARYING_SLOT_POS,
                                                                  i);
                 }
@@ -627,11 +627,11 @@ v3d_update_compiled_vs(struct v3d_context *v3d, uint8_t prim_mode)
         v3d_setup_shared_key(v3d, &key->base, &v3d->tex[PIPE_SHADER_VERTEX]);
         key->base.shader_state = v3d->prog.bind_vs;
         key->base.ucp_enables = v3d->rasterizer->base.clip_plane_enable;
-        key->num_fs_inputs = v3d->prog.fs->prog_data.fs->num_inputs;
-        STATIC_ASSERT(sizeof(key->fs_inputs) ==
+        key->num_used_outputs = v3d->prog.fs->prog_data.fs->num_inputs;
+        STATIC_ASSERT(sizeof(key->used_outputs) ==
                       sizeof(v3d->prog.fs->prog_data.fs->input_slots));
-        memcpy(key->fs_inputs, v3d->prog.fs->prog_data.fs->input_slots,
-               sizeof(key->fs_inputs));
+        memcpy(key->used_outputs, v3d->prog.fs->prog_data.fs->input_slots,
+               sizeof(key->used_outputs));
         key->clamp_color = v3d->rasterizer->base.clamp_vertex_color;
 
         key->per_vertex_point_size =
@@ -648,15 +648,15 @@ v3d_update_compiled_vs(struct v3d_context *v3d, uint8_t prim_mode)
         key->is_coord = true;
         /* Coord shaders only output varyings used by transform feedback. */
         struct v3d_uncompiled_shader *shader_state = key->base.shader_state;
-        memcpy(key->fs_inputs, shader_state->tf_outputs,
-               sizeof(*key->fs_inputs) * shader_state->num_tf_outputs);
-        if (shader_state->num_tf_outputs < key->num_fs_inputs) {
-                memset(&key->fs_inputs[shader_state->num_tf_outputs],
+        memcpy(key->used_outputs, shader_state->tf_outputs,
+               sizeof(*key->used_outputs) * shader_state->num_tf_outputs);
+        if (shader_state->num_tf_outputs < key->num_used_outputs) {
+                memset(&key->used_outputs[shader_state->num_tf_outputs],
                        0,
-                       sizeof(*key->fs_inputs) * (key->num_fs_inputs -
+                       sizeof(*key->used_outputs) * (key->num_used_outputs -
                                                   shader_state->num_tf_outputs));
         }
-        key->num_fs_inputs = shader_state->num_tf_outputs;
+        key->num_used_outputs = shader_state->num_tf_outputs;
 
         struct v3d_compiled_shader *cs =
                 v3d_get_compiled_shader(v3d, &key->base, sizeof(*key));