freedreno/a6xx: split out const emit
authorRob Clark <robdclark@chromium.org>
Wed, 22 Apr 2020 18:20:25 +0000 (11:20 -0700)
committerMarge Bot <eric+marge@anholt.net>
Thu, 30 Apr 2020 20:03:17 +0000 (20:03 +0000)
In order to inline the const emit and drop the per-gen vfuncs to emit
the correct sort of packet, we should consolidate all of the entry-
points to const emit in one object file, otherwise we'll end up with
multiple copies per gen.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4813>

src/gallium/drivers/freedreno/Makefile.sources
src/gallium/drivers/freedreno/a6xx/fd6_compute.c
src/gallium/drivers/freedreno/a6xx/fd6_const.c [new file with mode: 0644]
src/gallium/drivers/freedreno/a6xx/fd6_const.h [new file with mode: 0644]
src/gallium/drivers/freedreno/a6xx/fd6_emit.c
src/gallium/drivers/freedreno/a6xx/fd6_program.c
src/gallium/drivers/freedreno/meson.build

index 0268bb11a553b4838f4513a9265398610a99d23e..7644da56e1d4f7a18443cdaa8ceb6e5e960bfa6b 100644 (file)
@@ -171,6 +171,8 @@ a6xx_SOURCES := \
        a6xx/fd6_blitter.h \
        a6xx/fd6_compute.c \
        a6xx/fd6_compute.h \
+       a6xx/fd6_const.c \
+       a6xx/fd6_const.h \
        a6xx/fd6_context.c \
        a6xx/fd6_context.h \
        a6xx/fd6_draw.c \
index e6d5b32564893e9a0f300b9936d477952f59493a..fb29e7d6061e46cdab4e2d2d6e118699dec54fd7 100644 (file)
@@ -31,6 +31,7 @@
 #include "freedreno_resource.h"
 
 #include "fd6_compute.h"
+#include "fd6_const.h"
 #include "fd6_context.h"
 #include "fd6_emit.h"
 
@@ -140,7 +141,7 @@ fd6_launch_grid(struct fd_context *ctx, const struct pipe_grid_info *info)
                cs_program_emit(ring, v);
 
        fd6_emit_cs_state(ctx, ring, v);
-       ir3_emit_cs_consts(v, ring, ctx, info);
+       fd6_emit_cs_consts(v, ring, ctx, info);
 
        foreach_bit(i, ctx->global_bindings.enabled_mask)
                nglobal++;
diff --git a/src/gallium/drivers/freedreno/a6xx/fd6_const.c b/src/gallium/drivers/freedreno/a6xx/fd6_const.c
new file mode 100644 (file)
index 0000000..e934266
--- /dev/null
@@ -0,0 +1,305 @@
+/*
+ * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org>
+ * Copyright © 2018 Google, Inc.
+ *
+ * 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.
+ */
+
+#include "fd6_const.h"
+
+/* regid:          base const register
+ * prsc or dwords: buffer containing constant values
+ * sizedwords:     size of const value buffer
+ */
+void
+fd6_emit_const(struct fd_ringbuffer *ring, gl_shader_stage type,
+               uint32_t regid, uint32_t offset, uint32_t sizedwords,
+               const uint32_t *dwords, struct pipe_resource *prsc)
+{
+       uint32_t i, sz, align_sz;
+       enum a6xx_state_src src;
+
+       debug_assert((regid % 4) == 0);
+
+       if (prsc) {
+               sz = 0;
+               src = SS6_INDIRECT;
+       } else {
+               sz = sizedwords;
+               src = SS6_DIRECT;
+       }
+
+       align_sz = align(sz, 4);
+
+       OUT_PKT7(ring, fd6_stage2opcode(type), 3 + align_sz);
+       OUT_RING(ring, CP_LOAD_STATE6_0_DST_OFF(regid/4) |
+                       CP_LOAD_STATE6_0_STATE_TYPE(ST6_CONSTANTS) |
+                       CP_LOAD_STATE6_0_STATE_SRC(src) |
+                       CP_LOAD_STATE6_0_STATE_BLOCK(fd6_stage2shadersb(type)) |
+                       CP_LOAD_STATE6_0_NUM_UNIT(DIV_ROUND_UP(sizedwords, 4)));
+       if (prsc) {
+               struct fd_bo *bo = fd_resource(prsc)->bo;
+               OUT_RELOC(ring, bo, offset, 0, 0);
+       } else {
+               OUT_RING(ring, CP_LOAD_STATE6_1_EXT_SRC_ADDR(0));
+               OUT_RING(ring, CP_LOAD_STATE6_2_EXT_SRC_ADDR_HI(0));
+               dwords = (uint32_t *)&((uint8_t *)dwords)[offset];
+       }
+
+       for (i = 0; i < sz; i++) {
+               OUT_RING(ring, dwords[i]);
+       }
+
+       /* Zero-pad to multiple of 4 dwords */
+       for (i = sz; i < align_sz; i++) {
+               OUT_RING(ring, 0);
+       }
+}
+
+void
+fd6_emit_const_bo(struct fd_ringbuffer *ring, gl_shader_stage type, boolean write,
+               uint32_t regid, uint32_t num, struct pipe_resource **prscs, uint32_t *offsets)
+{
+       uint32_t anum = align(num, 2);
+       uint32_t i;
+
+       debug_assert((regid % 4) == 0);
+
+       OUT_PKT7(ring, fd6_stage2opcode(type), 3 + (2 * anum));
+       OUT_RING(ring, CP_LOAD_STATE6_0_DST_OFF(regid/4) |
+                       CP_LOAD_STATE6_0_STATE_TYPE(ST6_CONSTANTS)|
+                       CP_LOAD_STATE6_0_STATE_SRC(SS6_DIRECT) |
+                       CP_LOAD_STATE6_0_STATE_BLOCK(fd6_stage2shadersb(type)) |
+                       CP_LOAD_STATE6_0_NUM_UNIT(anum/2));
+       OUT_RING(ring, CP_LOAD_STATE6_1_EXT_SRC_ADDR(0));
+       OUT_RING(ring, CP_LOAD_STATE6_2_EXT_SRC_ADDR_HI(0));
+
+       for (i = 0; i < num; i++) {
+               if (prscs[i]) {
+                       if (write) {
+                               OUT_RELOCW(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0);
+                       } else {
+                               OUT_RELOC(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0);
+                       }
+               } else {
+                       OUT_RING(ring, 0xbad00000 | (i << 16));
+                       OUT_RING(ring, 0xbad00000 | (i << 16));
+               }
+       }
+
+       for (; i < anum; i++) {
+               OUT_RING(ring, 0xffffffff);
+               OUT_RING(ring, 0xffffffff);
+       }
+}
+
+static void
+emit_tess_bos(struct fd_ringbuffer *ring, struct fd6_emit *emit, struct ir3_shader_variant *s)
+{
+       struct fd_context *ctx = emit->ctx;
+       const unsigned regid = s->shader->const_state.offsets.primitive_param * 4 + 4;
+       uint32_t dwords = 16;
+
+       OUT_PKT7(ring, fd6_stage2opcode(s->type), 3);
+       OUT_RING(ring, CP_LOAD_STATE6_0_DST_OFF(regid / 4) |
+                       CP_LOAD_STATE6_0_STATE_TYPE(ST6_CONSTANTS)|
+                       CP_LOAD_STATE6_0_STATE_SRC(SS6_INDIRECT) |
+                       CP_LOAD_STATE6_0_STATE_BLOCK(fd6_stage2shadersb(s->type)) |
+                       CP_LOAD_STATE6_0_NUM_UNIT(dwords / 4));
+       OUT_RB(ring, ctx->batch->tess_addrs_constobj);
+}
+
+static void
+emit_stage_tess_consts(struct fd_ringbuffer *ring, struct ir3_shader_variant *v,
+               uint32_t *params, int num_params)
+{
+       const unsigned regid = v->shader->const_state.offsets.primitive_param;
+       int size = MIN2(1 + regid, v->constlen) - regid;
+       if (size > 0)
+               fd6_emit_const(ring, v->type, regid * 4, 0, num_params, params, NULL);
+}
+
+static void
+emit_tess_consts(struct fd6_emit *emit)
+{
+       struct fd_context *ctx = emit->ctx;
+
+       struct fd_ringbuffer *constobj = fd_submit_new_ringbuffer(
+               ctx->batch->submit, 0x1000, FD_RINGBUFFER_STREAMING);
+
+       /* VS sizes are in bytes since that's what STLW/LDLW use, while the HS
+        * size is dwords, since that's what LDG/STG use.
+        */
+       unsigned num_vertices =
+               emit->hs ?
+               emit->info->vertices_per_patch :
+               emit->gs->shader->nir->info.gs.vertices_in;
+
+       uint32_t vs_params[4] = {
+               emit->vs->shader->output_size * num_vertices * 4,       /* vs primitive stride */
+               emit->vs->shader->output_size * 4,                                      /* vs vertex stride */
+               0,
+               0
+       };
+
+       emit_stage_tess_consts(constobj, emit->vs, vs_params, ARRAY_SIZE(vs_params));
+
+       if (emit->hs) {
+               uint32_t hs_params[4] = {
+                       emit->vs->shader->output_size * num_vertices * 4,       /* vs primitive stride */
+                       emit->vs->shader->output_size * 4,                                      /* vs vertex stride */
+                       emit->hs->shader->output_size,
+                       emit->info->vertices_per_patch
+               };
+
+               emit_stage_tess_consts(constobj, emit->hs, hs_params, ARRAY_SIZE(hs_params));
+               emit_tess_bos(constobj, emit, emit->hs);
+
+               if (emit->gs)
+                       num_vertices = emit->gs->shader->nir->info.gs.vertices_in;
+
+               uint32_t ds_params[4] = {
+                       emit->ds->shader->output_size * num_vertices * 4,       /* ds primitive stride */
+                       emit->ds->shader->output_size * 4,                                      /* ds vertex stride */
+                       emit->hs->shader->output_size,                      /* hs vertex stride (dwords) */
+                       emit->hs->shader->nir->info.tess.tcs_vertices_out
+               };
+
+               emit_stage_tess_consts(constobj, emit->ds, ds_params, ARRAY_SIZE(ds_params));
+               emit_tess_bos(constobj, emit, emit->ds);
+       }
+
+       if (emit->gs) {
+               struct ir3_shader_variant *prev;
+               if (emit->ds)
+                       prev = emit->ds;
+               else
+                       prev = emit->vs;
+
+               uint32_t gs_params[4] = {
+                       prev->shader->output_size * num_vertices * 4,   /* ds primitive stride */
+                       prev->shader->output_size * 4,                                  /* ds vertex stride */
+                       0,
+                       0,
+               };
+
+               num_vertices = emit->gs->shader->nir->info.gs.vertices_in;
+               emit_stage_tess_consts(constobj, emit->gs, gs_params, ARRAY_SIZE(gs_params));
+       }
+
+       fd6_emit_take_group(emit, constobj, FD6_GROUP_PRIMITIVE_PARAMS, ENABLE_ALL);
+}
+
+static void
+emit_user_consts(struct fd6_emit *emit)
+{
+       static const enum pipe_shader_type types[] = {
+                       PIPE_SHADER_VERTEX, PIPE_SHADER_TESS_CTRL, PIPE_SHADER_TESS_EVAL,
+                       PIPE_SHADER_GEOMETRY, PIPE_SHADER_FRAGMENT,
+       };
+       const struct ir3_shader_variant *variants[] = {
+                       emit->vs, emit->hs, emit->ds, emit->gs, emit->fs,
+       };
+       struct fd_context *ctx = emit->ctx;
+       unsigned sz = 0;
+
+       for (unsigned i = 0; i < ARRAY_SIZE(types); i++) {
+               if (!variants[i])
+                       continue;
+               sz += variants[i]->shader->ubo_state.cmdstream_size;
+       }
+
+       struct fd_ringbuffer *constobj = fd_submit_new_ringbuffer(
+                       ctx->batch->submit, sz, FD_RINGBUFFER_STREAMING);
+
+       for (unsigned i = 0; i < ARRAY_SIZE(types); i++) {
+               if (!variants[i])
+                       continue;
+               ir3_emit_user_consts(ctx->screen, variants[i], constobj, &ctx->constbuf[types[i]]);
+               ir3_emit_ubos(ctx->screen, variants[i], constobj, &ctx->constbuf[types[i]]);
+       }
+
+       fd6_emit_take_group(emit, constobj, FD6_GROUP_CONST, ENABLE_ALL);
+}
+
+void
+fd6_emit_consts(struct fd6_emit *emit)
+{
+       struct fd_context *ctx = emit->ctx;
+       struct fd6_context *fd6_ctx = fd6_context(ctx);
+
+       if (emit->dirty & (FD_DIRTY_CONST | FD_DIRTY_PROG))
+               emit_user_consts(emit);
+
+       if (emit->key.key.has_gs || emit->key.key.tessellation)
+               emit_tess_consts(emit);
+
+       /* if driver-params are needed, emit each time: */
+       const struct ir3_shader_variant *vs = emit->vs;
+       if (ir3_needs_vs_driver_params(vs)) {
+               struct fd_ringbuffer *dpconstobj = fd_submit_new_ringbuffer(
+                               ctx->batch->submit, IR3_DP_VS_COUNT * 4, FD_RINGBUFFER_STREAMING);
+               ir3_emit_vs_driver_params(vs, dpconstobj, ctx, emit->info);
+               fd6_emit_take_group(emit, dpconstobj, FD6_GROUP_VS_DRIVER_PARAMS, ENABLE_ALL);
+               fd6_ctx->has_dp_state = true;
+       } else if (fd6_ctx->has_dp_state) {
+               fd6_emit_take_group(emit, NULL, FD6_GROUP_VS_DRIVER_PARAMS, ENABLE_ALL);
+               fd6_ctx->has_dp_state = false;
+       }
+}
+
+void
+fd6_emit_ibo_consts(struct fd6_emit *emit, const struct ir3_shader_variant *v,
+               enum pipe_shader_type stage, struct fd_ringbuffer *ring)
+{
+       struct fd_context *ctx = emit->ctx;
+
+       ir3_emit_ssbo_sizes(ctx->screen, v, ring, &ctx->shaderbuf[stage]);
+       ir3_emit_image_dims(ctx->screen, v, ring, &ctx->shaderimg[stage]);
+}
+
+void
+fd6_emit_cs_consts(const struct ir3_shader_variant *v, struct fd_ringbuffer *ring,
+               struct fd_context *ctx, const struct pipe_grid_info *info)
+{
+       ir3_emit_cs_consts(v, ring, ctx, info);
+}
+
+void
+fd6_emit_immediates(struct fd_screen *screen, const struct ir3_shader_variant *v,
+               struct fd_ringbuffer *ring)
+{
+       ir3_emit_immediates(screen, v, ring);
+}
+
+void
+fd6_user_consts_size(struct ir3_ubo_analysis_state *state,
+               unsigned *packets, unsigned *size)
+{
+       ir3_user_consts_size(state, packets, size);
+}
+
+void
+fd6_emit_link_map(struct fd_screen *screen,
+               const struct ir3_shader_variant *producer,
+               const struct ir3_shader_variant *v, struct fd_ringbuffer *ring)
+{
+       ir3_emit_link_map(screen, producer, v, ring);
+}
diff --git a/src/gallium/drivers/freedreno/a6xx/fd6_const.h b/src/gallium/drivers/freedreno/a6xx/fd6_const.h
new file mode 100644 (file)
index 0000000..d53bf82
--- /dev/null
@@ -0,0 +1,50 @@
+/*
+ * Copyright (C) 2016 Rob Clark <robclark@freedesktop.org>
+ * Copyright © 2018 Google, Inc.
+ *
+ * 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.
+ */
+
+
+#ifndef FD6_CONST_H
+#define FD6_CONST_H
+
+#include "fd6_emit.h"
+
+void fd6_emit_const(struct fd_ringbuffer *ring, gl_shader_stage type,
+               uint32_t regid, uint32_t offset, uint32_t sizedwords,
+               const uint32_t *dwords, struct pipe_resource *prsc);
+void fd6_emit_const_bo(struct fd_ringbuffer *ring, gl_shader_stage type, boolean write,
+               uint32_t regid, uint32_t num, struct pipe_resource **prscs, uint32_t *offsets);
+
+void fd6_emit_consts(struct fd6_emit *emit);
+void fd6_emit_ibo_consts(struct fd6_emit *emit, const struct ir3_shader_variant *v,
+               enum pipe_shader_type stage, struct fd_ringbuffer *ring);
+void fd6_emit_cs_consts(const struct ir3_shader_variant *v, struct fd_ringbuffer *ring,
+               struct fd_context *ctx, const struct pipe_grid_info *info);
+void fd6_emit_immediates(struct fd_screen *screen, const struct ir3_shader_variant *v,
+               struct fd_ringbuffer *ring);
+void fd6_user_consts_size(struct ir3_ubo_analysis_state *state,
+               unsigned *packets, unsigned *size);
+void fd6_emit_link_map(struct fd_screen *screen,
+               const struct ir3_shader_variant *producer,
+               const struct ir3_shader_variant *v, struct fd_ringbuffer *ring);
+
+#endif /* FD6_CONST_H */
index 425efbbbe40326389246cf869ce43cc941503a62..3ec1c6c9ab611b2579bcb63a3b0136039e42068a 100644 (file)
@@ -38,6 +38,7 @@
 
 #include "fd6_emit.h"
 #include "fd6_blend.h"
+#include "fd6_const.h"
 #include "fd6_context.h"
 #include "fd6_image.h"
 #include "fd6_program.h"
 #include "fd6_format.h"
 #include "fd6_zsa.h"
 
-/* regid:          base const register
- * prsc or dwords: buffer containing constant values
- * sizedwords:     size of const value buffer
- */
-static void
-fd6_emit_const(struct fd_ringbuffer *ring, gl_shader_stage type,
-               uint32_t regid, uint32_t offset, uint32_t sizedwords,
-               const uint32_t *dwords, struct pipe_resource *prsc)
-{
-       uint32_t i, sz, align_sz;
-       enum a6xx_state_src src;
-
-       debug_assert((regid % 4) == 0);
-
-       if (prsc) {
-               sz = 0;
-               src = SS6_INDIRECT;
-       } else {
-               sz = sizedwords;
-               src = SS6_DIRECT;
-       }
-
-       align_sz = align(sz, 4);
-
-       OUT_PKT7(ring, fd6_stage2opcode(type), 3 + align_sz);
-       OUT_RING(ring, CP_LOAD_STATE6_0_DST_OFF(regid/4) |
-                       CP_LOAD_STATE6_0_STATE_TYPE(ST6_CONSTANTS) |
-                       CP_LOAD_STATE6_0_STATE_SRC(src) |
-                       CP_LOAD_STATE6_0_STATE_BLOCK(fd6_stage2shadersb(type)) |
-                       CP_LOAD_STATE6_0_NUM_UNIT(DIV_ROUND_UP(sizedwords, 4)));
-       if (prsc) {
-               struct fd_bo *bo = fd_resource(prsc)->bo;
-               OUT_RELOC(ring, bo, offset, 0, 0);
-       } else {
-               OUT_RING(ring, CP_LOAD_STATE6_1_EXT_SRC_ADDR(0));
-               OUT_RING(ring, CP_LOAD_STATE6_2_EXT_SRC_ADDR_HI(0));
-               dwords = (uint32_t *)&((uint8_t *)dwords)[offset];
-       }
-
-       for (i = 0; i < sz; i++) {
-               OUT_RING(ring, dwords[i]);
-       }
-
-       /* Zero-pad to multiple of 4 dwords */
-       for (i = sz; i < align_sz; i++) {
-               OUT_RING(ring, 0);
-       }
-}
-
-static void
-fd6_emit_const_bo(struct fd_ringbuffer *ring, gl_shader_stage type, boolean write,
-               uint32_t regid, uint32_t num, struct pipe_resource **prscs, uint32_t *offsets)
-{
-       uint32_t anum = align(num, 2);
-       uint32_t i;
-
-       debug_assert((regid % 4) == 0);
-
-       OUT_PKT7(ring, fd6_stage2opcode(type), 3 + (2 * anum));
-       OUT_RING(ring, CP_LOAD_STATE6_0_DST_OFF(regid/4) |
-                       CP_LOAD_STATE6_0_STATE_TYPE(ST6_CONSTANTS)|
-                       CP_LOAD_STATE6_0_STATE_SRC(SS6_DIRECT) |
-                       CP_LOAD_STATE6_0_STATE_BLOCK(fd6_stage2shadersb(type)) |
-                       CP_LOAD_STATE6_0_NUM_UNIT(anum/2));
-       OUT_RING(ring, CP_LOAD_STATE6_1_EXT_SRC_ADDR(0));
-       OUT_RING(ring, CP_LOAD_STATE6_2_EXT_SRC_ADDR_HI(0));
-
-       for (i = 0; i < num; i++) {
-               if (prscs[i]) {
-                       if (write) {
-                               OUT_RELOCW(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0);
-                       } else {
-                               OUT_RELOC(ring, fd_resource(prscs[i])->bo, offsets[i], 0, 0);
-                       }
-               } else {
-                       OUT_RING(ring, 0xbad00000 | (i << 16));
-                       OUT_RING(ring, 0xbad00000 | (i << 16));
-               }
-       }
-
-       for (; i < anum; i++) {
-               OUT_RING(ring, 0xffffffff);
-               OUT_RING(ring, 0xffffffff);
-       }
-}
-
 /* Border color layout is diff from a4xx/a5xx.. if it turns out to be
  * the same as a6xx then move this somewhere common ;-)
  *
@@ -807,140 +722,10 @@ fd6_emit_streamout(struct fd_ringbuffer *ring, struct fd6_emit *emit, struct ir3
        }
 }
 
-static void
-emit_tess_bos(struct fd_ringbuffer *ring, struct fd6_emit *emit, struct ir3_shader_variant *s)
-{
-       struct fd_context *ctx = emit->ctx;
-       const unsigned regid = s->shader->const_state.offsets.primitive_param * 4 + 4;
-       uint32_t dwords = 16;
-
-       OUT_PKT7(ring, fd6_stage2opcode(s->type), 3);
-       OUT_RING(ring, CP_LOAD_STATE6_0_DST_OFF(regid / 4) |
-                       CP_LOAD_STATE6_0_STATE_TYPE(ST6_CONSTANTS)|
-                       CP_LOAD_STATE6_0_STATE_SRC(SS6_INDIRECT) |
-                       CP_LOAD_STATE6_0_STATE_BLOCK(fd6_stage2shadersb(s->type)) |
-                       CP_LOAD_STATE6_0_NUM_UNIT(dwords / 4));
-       OUT_RB(ring, ctx->batch->tess_addrs_constobj);
-}
-
-static void
-emit_stage_tess_consts(struct fd_ringbuffer *ring, struct ir3_shader_variant *v,
-               uint32_t *params, int num_params)
-{
-       const unsigned regid = v->shader->const_state.offsets.primitive_param;
-       int size = MIN2(1 + regid, v->constlen) - regid;
-       if (size > 0)
-               fd6_emit_const(ring, v->type, regid * 4, 0, num_params, params, NULL);
-}
-
-static void
-fd6_emit_tess_const(struct fd6_emit *emit)
-{
-       struct fd_context *ctx = emit->ctx;
-
-       struct fd_ringbuffer *constobj = fd_submit_new_ringbuffer(
-               ctx->batch->submit, 0x1000, FD_RINGBUFFER_STREAMING);
-
-       /* VS sizes are in bytes since that's what STLW/LDLW use, while the HS
-        * size is dwords, since that's what LDG/STG use.
-        */
-       unsigned num_vertices =
-               emit->hs ?
-               emit->info->vertices_per_patch :
-               emit->gs->shader->nir->info.gs.vertices_in;
-
-       uint32_t vs_params[4] = {
-               emit->vs->shader->output_size * num_vertices * 4,       /* vs primitive stride */
-               emit->vs->shader->output_size * 4,                                      /* vs vertex stride */
-               0,
-               0
-       };
-
-       emit_stage_tess_consts(constobj, emit->vs, vs_params, ARRAY_SIZE(vs_params));
-
-       if (emit->hs) {
-               uint32_t hs_params[4] = {
-                       emit->vs->shader->output_size * num_vertices * 4,       /* vs primitive stride */
-                       emit->vs->shader->output_size * 4,                                      /* vs vertex stride */
-                       emit->hs->shader->output_size,
-                       emit->info->vertices_per_patch
-               };
-
-               emit_stage_tess_consts(constobj, emit->hs, hs_params, ARRAY_SIZE(hs_params));
-               emit_tess_bos(constobj, emit, emit->hs);
-
-               if (emit->gs)
-                       num_vertices = emit->gs->shader->nir->info.gs.vertices_in;
-
-               uint32_t ds_params[4] = {
-                       emit->ds->shader->output_size * num_vertices * 4,       /* ds primitive stride */
-                       emit->ds->shader->output_size * 4,                                      /* ds vertex stride */
-                       emit->hs->shader->output_size,                      /* hs vertex stride (dwords) */
-                       emit->hs->shader->nir->info.tess.tcs_vertices_out
-               };
-
-               emit_stage_tess_consts(constobj, emit->ds, ds_params, ARRAY_SIZE(ds_params));
-               emit_tess_bos(constobj, emit, emit->ds);
-       }
-
-       if (emit->gs) {
-               struct ir3_shader_variant *prev;
-               if (emit->ds)
-                       prev = emit->ds;
-               else
-                       prev = emit->vs;
-
-               uint32_t gs_params[4] = {
-                       prev->shader->output_size * num_vertices * 4,   /* ds primitive stride */
-                       prev->shader->output_size * 4,                                  /* ds vertex stride */
-                       0,
-                       0,
-               };
-
-               num_vertices = emit->gs->shader->nir->info.gs.vertices_in;
-               emit_stage_tess_consts(constobj, emit->gs, gs_params, ARRAY_SIZE(gs_params));
-       }
-
-       fd6_emit_take_group(emit, constobj, FD6_GROUP_PRIMITIVE_PARAMS, ENABLE_ALL);
-}
-
-static void
-fd6_emit_consts(struct fd6_emit *emit)
-{
-       static const enum pipe_shader_type types[] = {
-                       PIPE_SHADER_VERTEX, PIPE_SHADER_TESS_CTRL, PIPE_SHADER_TESS_EVAL,
-                       PIPE_SHADER_GEOMETRY, PIPE_SHADER_FRAGMENT,
-       };
-       const struct ir3_shader_variant *variants[] = {
-                       emit->vs, emit->hs, emit->ds, emit->gs, emit->fs,
-       };
-       struct fd_context *ctx = emit->ctx;
-       unsigned sz = 0;
-
-       for (unsigned i = 0; i < ARRAY_SIZE(types); i++) {
-               if (!variants[i])
-                       continue;
-               sz += variants[i]->shader->ubo_state.cmdstream_size;
-       }
-
-       struct fd_ringbuffer *constobj = fd_submit_new_ringbuffer(
-                       ctx->batch->submit, sz, FD_RINGBUFFER_STREAMING);
-
-       for (unsigned i = 0; i < ARRAY_SIZE(types); i++) {
-               if (!variants[i])
-                       continue;
-               ir3_emit_user_consts(ctx->screen, variants[i], constobj, &ctx->constbuf[types[i]]);
-               ir3_emit_ubos(ctx->screen, variants[i], constobj, &ctx->constbuf[types[i]]);
-       }
-
-       fd6_emit_take_group(emit, constobj, FD6_GROUP_CONST, ENABLE_ALL);
-}
-
 void
 fd6_emit_state(struct fd_ringbuffer *ring, struct fd6_emit *emit)
 {
        struct fd_context *ctx = emit->ctx;
-       struct fd6_context *fd6_ctx = fd6_context(ctx);
        struct pipe_framebuffer_state *pfb = &ctx->batch->framebuffer;
        const struct fd6_program_state *prog = fd6_emit_get_prog(emit);
        const struct ir3_shader_variant *vs = emit->vs;
@@ -1088,24 +873,7 @@ fd6_emit_state(struct fd_ringbuffer *ring, struct fd6_emit *emit)
                fd6_emit_take_group(emit, ring, FD6_GROUP_PROG_FB_RAST, ENABLE_DRAW);
        }
 
-       if (dirty & (FD_DIRTY_CONST | FD_DIRTY_PROG)) {
-               fd6_emit_consts(emit);
-       }
-
-       if (emit->key.key.has_gs || emit->key.key.tessellation)
-               fd6_emit_tess_const(emit);
-
-       /* if driver-params are needed, emit each time: */
-       if (ir3_needs_vs_driver_params(vs)) {
-               struct fd_ringbuffer *dpconstobj = fd_submit_new_ringbuffer(
-                               ctx->batch->submit, IR3_DP_VS_COUNT * 4, FD_RINGBUFFER_STREAMING);
-               ir3_emit_vs_driver_params(vs, dpconstobj, ctx, emit->info);
-               fd6_emit_take_group(emit, dpconstobj, FD6_GROUP_VS_DRIVER_PARAMS, ENABLE_ALL);
-               fd6_ctx->has_dp_state = true;
-       } else if (fd6_ctx->has_dp_state) {
-               fd6_emit_take_group(emit, NULL, FD6_GROUP_VS_DRIVER_PARAMS, ENABLE_ALL);
-               fd6_ctx->has_dp_state = false;
-       }
+       fd6_emit_consts(emit);
 
        struct ir3_stream_output_info *info = &fd6_last_shader(prog)->shader->stream_output;
        if (info->num_outputs)
@@ -1177,10 +945,7 @@ fd6_emit_state(struct fd_ringbuffer *ring, struct fd6_emit *emit)
                OUT_PKT4(obj, REG_A6XX_SP_IBO_COUNT, 1);
                OUT_RING(obj, ir3_shader_nibo(fs));
 
-               ir3_emit_ssbo_sizes(ctx->screen, fs, obj,
-                               &ctx->shaderbuf[PIPE_SHADER_FRAGMENT]);
-               ir3_emit_image_dims(ctx->screen, fs, obj,
-                               &ctx->shaderimg[PIPE_SHADER_FRAGMENT]);
+               fd6_emit_ibo_consts(emit, fs, PIPE_SHADER_FRAGMENT, ring);
 
                fd6_emit_take_group(emit, obj, FD6_GROUP_IBO, ENABLE_DRAW);
                fd_ringbuffer_del(state);
index fa98aae546fd33837b6a2850becb2b3bcaa0c350..80eb0c8b98868a47721f4a35dbb5d22c61cbccb4 100644 (file)
@@ -35,6 +35,7 @@
 #include "freedreno_program.h"
 
 #include "fd6_program.h"
+#include "fd6_const.h"
 #include "fd6_emit.h"
 #include "fd6_texture.h"
 #include "fd6_format.h"
@@ -425,7 +426,7 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
                        COND(vs->need_pixlod, A6XX_SP_VS_CTRL_REG0_PIXLODENABLE));
 
        fd6_emit_shader(ring, vs);
-       ir3_emit_immediates(screen, vs, ring);
+       fd6_emit_immediates(screen, vs, ring);
 
        struct ir3_shader_linkage l = {0};
        const struct ir3_shader_variant *last_shader = fd6_last_shader(state);
@@ -510,8 +511,8 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
                        COND(hs->need_pixlod, A6XX_SP_HS_CTRL_REG0_PIXLODENABLE));
 
                fd6_emit_shader(ring, hs);
-               ir3_emit_immediates(screen, hs, ring);
-               ir3_emit_link_map(screen, vs, hs, ring);
+               fd6_emit_immediates(screen, hs, ring);
+               fd6_emit_link_map(screen, vs, hs, ring);
 
                OUT_PKT4(ring, REG_A6XX_SP_DS_CTRL_REG0, 1);
                OUT_RING(ring, A6XX_SP_DS_CTRL_REG0_THREADSIZE(TWO_QUADS) |
@@ -520,8 +521,8 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
                        COND(ds->need_pixlod, A6XX_SP_DS_CTRL_REG0_PIXLODENABLE));
 
                fd6_emit_shader(ring, ds);
-               ir3_emit_immediates(screen, ds, ring);
-               ir3_emit_link_map(screen, hs, ds, ring);
+               fd6_emit_immediates(screen, ds, ring);
+               fd6_emit_link_map(screen, hs, ds, ring);
 
                shader_info *hs_info = &hs->shader->nir->info;
                OUT_PKT4(ring, REG_A6XX_PC_TESS_NUM_VERTEX, 1);
@@ -701,11 +702,11 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
                        COND(gs->need_pixlod, A6XX_SP_GS_CTRL_REG0_PIXLODENABLE));
 
                fd6_emit_shader(ring, gs);
-               ir3_emit_immediates(screen, gs, ring);
+               fd6_emit_immediates(screen, gs, ring);
                if (ds)
-                       ir3_emit_link_map(screen, ds, gs, ring);
+                       fd6_emit_link_map(screen, ds, gs, ring);
                else
-                       ir3_emit_link_map(screen, vs, gs, ring);
+                       fd6_emit_link_map(screen, vs, gs, ring);
 
                OUT_PKT4(ring, REG_A6XX_VPC_PACK_GS, 1);
                OUT_RING(ring, A6XX_VPC_PACK_GS_POSITIONLOC(pos_loc) |
@@ -818,7 +819,7 @@ setup_stateobj(struct fd_ringbuffer *ring, struct fd_screen *screen,
        OUT_RING(ring, COND(fragz, A6XX_GRAS_SU_DEPTH_PLANE_CNTL_FRAG_WRITES_Z));
 
        if (!binning_pass)
-               ir3_emit_immediates(screen, fs, ring);
+               fd6_emit_immediates(screen, fs, ring);
 }
 
 static struct fd_ringbuffer *
@@ -1026,7 +1027,7 @@ fd6_shader_state_create(struct pipe_context *pctx, const struct pipe_shader_stat
        unsigned packets, size;
 
        /* pre-calculate size required for userconst stateobj: */
-       ir3_user_consts_size(&shader->ubo_state, &packets, &size);
+       fd6_user_consts_size(&shader->ubo_state, &packets, &size);
 
        /* also account for UBO addresses: */
        packets += 1;
index 919baf070de10ad0e6255666d2880b305806fec1..86d45341e50f7bc5ff6431123a0c33979649aa23 100644 (file)
@@ -181,6 +181,8 @@ files_libfreedreno = files(
   'a6xx/fd6_blitter.h',
   'a6xx/fd6_compute.c',
   'a6xx/fd6_compute.h',
+  'a6xx/fd6_const.c',
+  'a6xx/fd6_const.h',
   'a6xx/fd6_context.c',
   'a6xx/fd6_context.h',
   'a6xx/fd6_draw.c',