Mesa (master): broadcom/vc5: Fix up the NIR types of FS outputs generated by NIR-to-TGSI.

[Eric Anholt]

Module: Mesa
Branch: master
Commit: baeb6a4b4a275bc418037b718c904180e9e5c690
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=baeb6a4b4a275bc418037b718c904180e9e5c690

Author: Eric Anholt <eric at anholt.net>
Date:   Wed Mar 21 12:05:54 2018 -0700

broadcom/vc5: Fix up the NIR types of FS outputs generated by NIR-to-TGSI.

Unfortunately TGSI doesn't record the type of the FS output like GLSL
does, but VC5's TLB writes depend on the output's base type.  Just record
the type in the key at variant compile time when we've got a TGSI input
and then fix it up.

Fixes KHR-GLES3.packed_pixels.pbo_rectangle.rgba32i/ui and apparently a
GPU hang that breaks most tests that come after it.

---

 src/broadcom/compiler/v3d_compiler.h  |  5 +++++
 src/broadcom/compiler/vir.c           | 33 +++++++++++++++++++++++++++++++++
 src/gallium/drivers/vc5/vc5_context.h |  7 +++++++
 src/gallium/drivers/vc5/vc5_program.c |  9 +++++++++
 4 files changed, 54 insertions(+)

diff --git a/src/broadcom/compiler/v3d_compiler.h b/src/broadcom/compiler/v3d_compiler.h
index df81f0757e..207e29733a 100644
--- a/src/broadcom/compiler/v3d_compiler.h
+++ b/src/broadcom/compiler/v3d_compiler.h
@@ -336,6 +336,11 @@ struct v3d_fs_key {
         uint8_t swap_color_rb;
         /* Mask of which render targets need to be written as 32-bit floats */
         uint8_t f32_color_rb;
+        /* Masks of which render targets need to be written as ints/uints.
+         * Used by gallium to work around lost information in TGSI.
+         */
+        uint8_t int_color_rb;
+        uint8_t uint_color_rb;
         uint8_t alpha_test_func;
         uint8_t logicop_func;
         uint32_t point_sprite_mask;
diff --git a/src/broadcom/compiler/vir.c b/src/broadcom/compiler/vir.c
index 05f557fbcd..ff9405e6c1 100644
--- a/src/broadcom/compiler/vir.c
+++ b/src/broadcom/compiler/vir.c
@@ -756,6 +756,36 @@ v3d_set_fs_prog_data_inputs(struct v3d_compile *c,
         }
 }
 
+static void
+v3d_fixup_fs_output_types(struct v3d_compile *c)
+{
+        nir_foreach_variable(var, &c->s->outputs) {
+                uint32_t mask = 0;
+
+                switch (var->data.location) {
+                case FRAG_RESULT_COLOR:
+                        mask = ~0;
+                        break;
+                case FRAG_RESULT_DATA0:
+                case FRAG_RESULT_DATA1:
+                case FRAG_RESULT_DATA2:
+                case FRAG_RESULT_DATA3:
+                        mask = 1 << (var->data.location - FRAG_RESULT_DATA0);
+                        break;
+                }
+
+                if (c->fs_key->int_color_rb & mask) {
+                        var->type =
+                                glsl_vector_type(GLSL_TYPE_INT,
+                                                 glsl_get_components(var->type));
+                } else if (c->fs_key->uint_color_rb & mask) {
+                        var->type =
+                                glsl_vector_type(GLSL_TYPE_UINT,
+                                                 glsl_get_components(var->type));
+                }
+        }
+}
+
 uint64_t *v3d_compile_fs(const struct v3d_compiler *compiler,
                          struct v3d_fs_key *key,
                          struct v3d_fs_prog_data *prog_data,
@@ -768,6 +798,9 @@ uint64_t *v3d_compile_fs(const struct v3d_compiler *compiler,
 
         c->fs_key = key;
 
+        if (key->int_color_rb || key->uint_color_rb)
+                v3d_fixup_fs_output_types(c);
+
         v3d_lower_nir(c);
 
         if (key->light_twoside)
diff --git a/src/gallium/drivers/vc5/vc5_context.h b/src/gallium/drivers/vc5/vc5_context.h
index 28b2e165a9..1ab5a6b153 100644
--- a/src/gallium/drivers/vc5/vc5_context.h
+++ b/src/gallium/drivers/vc5/vc5_context.h
@@ -132,6 +132,13 @@ struct vc5_uncompiled_shader {
         struct v3d_varying_slot *tf_outputs;
         uint16_t tf_specs[PIPE_MAX_SO_BUFFERS];
         uint32_t num_tf_specs;
+
+        /**
+         * Flag for if the NIR in this shader originally came from TGSI.  If
+         * so, we need to do some fixups at compile time, due to missing
+         * information in TGSI that exists in NIR.
+         */
+        bool was_tgsi;
 };
 
 struct vc5_compiled_shader {
diff --git a/src/gallium/drivers/vc5/vc5_program.c b/src/gallium/drivers/vc5/vc5_program.c
index ae3850a64b..87c21abe8b 100644
--- a/src/gallium/drivers/vc5/vc5_program.c
+++ b/src/gallium/drivers/vc5/vc5_program.c
@@ -170,6 +170,8 @@ vc5_shader_state_create(struct pipe_context *pctx,
                         fprintf(stderr, "\n");
                 }
                 s = tgsi_to_nir(cso->tokens, &v3d_nir_options);
+
+                so->was_tgsi = true;
         }
 
         NIR_PASS_V(s, nir_opt_global_to_local);
@@ -414,6 +416,13 @@ vc5_update_compiled_fs(struct vc5_context *vc5, uint8_t prim_mode)
                     desc->channel[0].size == 32) {
                         key->f32_color_rb |= 1 << i;
                 }
+
+                if (vc5->prog.bind_fs->was_tgsi) {
+                        if (util_format_is_pure_uint(cbuf->format))
+                                key->uint_color_rb |= 1 << i;
+                        else if (util_format_is_pure_sint(cbuf->format))
+                                key->int_color_rb |= 1 << i;
+                }
         }
 
         if (key->is_points) {