panfrost: Implement dispatch helpers

Rather than open-coding workgroups_shift_* type fields, we include a
general routine for packing the vertex/tiler/compute descriptor based on
the provided dispatch parameters.

Signed-off-by: Alyssa Rosenzweig <[email protected]>

4 files changed, 165 insertions, 13 deletions

diff --git a/src/gallium/drivers/panfrost/meson.build b/src/gallium/drivers/panfrost/meson.build
index 4298242f6b9..80cfee794db 100644
--- a/src/gallium/drivers/panfrost/meson.build
+++ b/src/gallium/drivers/panfrost/meson.build
@@ -57,6 +57,7 @@ files_panfrost = files(
   'pan_blend_shaders.c',
   'pan_pretty_print.c',
   'pan_fragment.c',
+  'pan_invocation.c',
   'pan_scoreboard.c',
   'pan_sfbd.c',
   'pan_mfbd.c',
diff --git a/src/gallium/drivers/panfrost/pan_context.c b/src/gallium/drivers/panfrost/pan_context.c
index bf98d3853f1..871b168040c 100644
--- a/src/gallium/drivers/panfrost/pan_context.c
+++ b/src/gallium/drivers/panfrost/pan_context.c
@@ -283,11 +283,6 @@ static void
 panfrost_emit_vertex_payload(struct panfrost_context *ctx)
 {
         struct midgard_payload_vertex_tiler payload = {
-                .prefix = {
-                        .workgroups_z_shift = 32,
-                        .workgroups_x_shift_2 = 0x2,
-                        .workgroups_x_shift_3 = 0x5,
-                },
 		.gl_enables = 0x4 | (ctx->is_t6xx ? 0 : 0x2),
         };
 
@@ -299,10 +294,6 @@ panfrost_emit_tiler_payload(struct panfrost_context *ctx)
 {
         struct midgard_payload_vertex_tiler payload = {
                 .prefix = {
-                        .workgroups_z_shift = 32,
-                        .workgroups_x_shift_2 = 0x2,
-                        .workgroups_x_shift_3 = 0x6,
-
                         .zero1 = 0xffff, /* Why is this only seen on test-quad-textured? */
                 },
         };
@@ -1668,7 +1659,7 @@ panfrost_draw_vbo(
         ctx->vertex_count = info->count;
 
         /* For non-indexed draws, they're the same */
-        unsigned invocation_count = ctx->vertex_count;
+        unsigned vertex_count = ctx->vertex_count;
 
         unsigned draw_flags = 0;
 
@@ -1701,7 +1692,7 @@ panfrost_draw_vbo(
                 }
 
                 /* Use the corresponding values */
-                invocation_count = max_index - min_index + 1;
+                vertex_count = max_index - min_index + 1;
                 ctx->payload_vertex.draw_start = min_index;
                 ctx->payload_tiler.draw_start = min_index;
 
@@ -1724,8 +1715,15 @@ panfrost_draw_vbo(
                 ctx->payload_tiler.prefix.indices = (uintptr_t) NULL;
         }
 
-        ctx->payload_vertex.prefix.invocation_count = MALI_POSITIVE(invocation_count);
-        ctx->payload_tiler.prefix.invocation_count = MALI_POSITIVE(invocation_count);
+        /* Dispatch "compute jobs" for the vertex/tiler pair as (1,
+         * vertex_count, 1) */
+
+        panfrost_pack_work_groups_fused(
+                        &ctx->payload_vertex.prefix,
+                        &ctx->payload_tiler.prefix,
+                        1, vertex_count, 1,
+                        1, 1, 1);
+
         ctx->payload_tiler.prefix.unknown_draw = draw_flags;
 
         /* Fire off the draw itself */
diff --git a/src/gallium/drivers/panfrost/pan_context.h b/src/gallium/drivers/panfrost/pan_context.h
index ca6f8deef33..f83083be0fc 100644
--- a/src/gallium/drivers/panfrost/pan_context.h
+++ b/src/gallium/drivers/panfrost/pan_context.h
@@ -343,4 +343,27 @@ panfrost_fragment_job(struct panfrost_context *ctx, bool has_draws);
 void
 panfrost_shader_compile(struct panfrost_context *ctx, struct mali_shader_meta *meta, const char *src, int type, struct panfrost_shader_state *state);
 
+void
+panfrost_pack_work_groups_compute(
+                struct mali_vertex_tiler_prefix *out,
+                unsigned num_x,
+                unsigned num_y,
+                unsigned num_z,
+                unsigned size_x,
+                unsigned size_y,
+                unsigned size_z);
+
+void
+panfrost_pack_work_groups_fused(
+                struct mali_vertex_tiler_prefix *vertex,
+                struct mali_vertex_tiler_prefix *tiler,
+                unsigned num_x,
+                unsigned num_y,
+                unsigned num_z,
+                unsigned size_x,
+                unsigned size_y,
+                unsigned size_z);
+
+
+
 #endif
diff --git a/src/gallium/drivers/panfrost/pan_invocation.c b/src/gallium/drivers/panfrost/pan_invocation.c
new file mode 100644
index 00000000000..0d4945d05b1
--- /dev/null
+++ b/src/gallium/drivers/panfrost/pan_invocation.c
@@ -0,0 +1,130 @@
+/*
+ * Copyright (C) 2019 Collabora, Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Authors (Collabora):
+ *   Alyssa Rosenzweig <[email protected]>
+ *
+ */
+
+#include "pan_context.h"
+
+/* Compute shaders are invoked with a gl_NumWorkGroups X/Y/Z triplet. Vertex
+ * shaders, it turns out, are invoked with the same mechanism, with the triplet
+ * (1, vertex_count, instance_count).
+ *
+ * Alongside this triplet is the gl_WorkGroupSize X/Y/Z triplet.
+ *
+ * Unfortunately, the packing for these triplet into the
+ * mali_vertex_tiler_prefix is a little funky, using a dynamic bitfield. The
+ * routines here exist to pack this */
+
+void
+panfrost_pack_work_groups_compute(
+                struct mali_vertex_tiler_prefix *out,
+                unsigned num_x,
+                unsigned num_y,
+                unsigned num_z,
+                unsigned size_x,
+                unsigned size_y,
+                unsigned size_z)
+{
+        /* First of all, all 6 values are off-by-one (strictly positive).
+         * Account for that, first by ensuring all values are strictly positive
+         * and then by offsetting */
+
+        assert(num_x > 0);
+        assert(num_y > 0);
+        assert(num_z > 0);
+
+        assert(size_x > 0);
+        assert(size_y > 0);
+        assert(size_z > 0);
+
+        num_x = MALI_POSITIVE(num_x);
+        num_y = MALI_POSITIVE(num_y);
+        num_z = MALI_POSITIVE(num_z);
+
+        size_x = MALI_POSITIVE(size_x);
+        size_y = MALI_POSITIVE(size_y);
+        size_z = MALI_POSITIVE(size_z);
+
+        /* Next up is to pack in order */
+
+        uint32_t packed = 0;
+
+        /* The values needing packing, in order, and the corresponding shifts.
+         * Indicies into shift are off-by-one to make the logic easier */
+
+        unsigned shifts[7] = { 0 };
+        unsigned values[6] = { size_x, size_y, size_z, num_x, num_y, num_z };
+
+        for (unsigned i = 0; i < 6; ++i) {
+                /* OR it in, shifting as required */
+                packed |= (values[i] << shifts[i]);
+
+                /* How many bits did we use? */
+                unsigned bit_count = util_logbase2_ceil(values[i] + 1);
+
+                /* Set the next shift accordingly */
+                shifts[i + 1] = shifts[i] + bit_count;
+        }
+
+        /* We're packed, so upload everything */
+        out->invocation_count = packed;
+        out->size_y_shift = shifts[1];
+        out->size_z_shift = shifts[2];
+        out->workgroups_x_shift = shifts[3];
+        out->workgroups_y_shift = shifts[4];
+        out->workgroups_z_shift = shifts[5];
+
+        /* Special fields */
+        out->workgroups_x_shift_2 = MAX2(out->workgroups_x_shift, 2);
+        out->workgroups_x_shift_3 = out->workgroups_x_shift_2;
+}
+
+/* Packs vertex/tiler descriptors simultaneously */
+void
+panfrost_pack_work_groups_fused(
+                struct mali_vertex_tiler_prefix *vertex,
+                struct mali_vertex_tiler_prefix *tiler,
+                unsigned num_x,
+                unsigned num_y,
+                unsigned num_z,
+                unsigned size_x,
+                unsigned size_y,
+                unsigned size_z)
+{
+        panfrost_pack_work_groups_compute(vertex, num_x, num_y, num_z, size_x, size_y, size_z);
+
+        /* Copy results over */
+        tiler->invocation_count = vertex->invocation_count;
+        tiler->size_y_shift = vertex->size_y_shift;
+        tiler->size_z_shift = vertex->size_z_shift;
+        tiler->workgroups_x_shift = vertex->workgroups_x_shift;
+        tiler->workgroups_y_shift = vertex->workgroups_y_shift;
+        tiler->workgroups_z_shift = vertex->workgroups_z_shift;
+
+        /* Set special fields for each */
+        vertex->workgroups_x_shift_3 = 5;
+        tiler->workgroups_x_shift_3 = 6;
+}
+