mesa/state_tracker/st_shader_cache.c

/*
 * Copyright © 2017 Timothy Arceri
 *
 * 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 <stdio.h>
#include "st_debug.h"
#include "st_program.h"
#include "st_shader_cache.h"
#include "st_util.h"
#include "compiler/glsl/program.h"
#include "compiler/nir/nir.h"
#include "compiler/nir/nir_serialize.h"
#include "pipe/p_shader_tokens.h"
#include "program/ir_to_mesa.h"
#include "tgsi/tgsi_parse.h"
#include "util/u_memory.h"

void
st_get_program_binary_driver_sha1(struct gl_context *ctx, uint8_t *sha1)
{
   disk_cache_compute_key(ctx->Cache, NULL, 0, sha1);
}

static void
write_stream_out_to_cache(struct blob *blob,
                          struct pipe_shader_state *state)
{
   blob_write_uint32(blob, state->stream_output.num_outputs);
   if (state->stream_output.num_outputs) {
      blob_write_bytes(blob, &state->stream_output.stride,
                       sizeof(state->stream_output.stride));
      blob_write_bytes(blob, &state->stream_output.output,
                       sizeof(state->stream_output.output));
   }
}

static void
copy_blob_to_driver_cache_blob(struct blob *blob, struct gl_program *prog)
{
   prog->driver_cache_blob = ralloc_size(NULL, blob->size);
   memcpy(prog->driver_cache_blob, blob->data, blob->size);
   prog->driver_cache_blob_size = blob->size;
}

static void
write_tgsi_to_cache(struct blob *blob, const struct tgsi_token *tokens,
                    struct gl_program *prog)
{
   unsigned num_tokens = tgsi_num_tokens(tokens);

   blob_write_uint32(blob, num_tokens);
   blob_write_bytes(blob, tokens, num_tokens * sizeof(struct tgsi_token));
   copy_blob_to_driver_cache_blob(blob, prog);
}

static void
write_nir_to_cache(struct blob *blob, struct gl_program *prog)
{
   struct st_program *stp = (struct st_program *)prog;

   st_serialize_nir(stp);

   blob_write_intptr(blob, stp->serialized_nir_size);
   blob_write_bytes(blob, stp->serialized_nir, stp->serialized_nir_size);

   copy_blob_to_driver_cache_blob(blob, prog);
}

static void
st_serialise_ir_program(struct gl_context *ctx, struct gl_program *prog,
                        bool nir)
{
   if (prog->driver_cache_blob)
      return;

   struct st_program *stp = (struct st_program *)prog;
   struct blob blob;
   blob_init(&blob);

   if (prog->info.stage == MESA_SHADER_VERTEX) {
      struct st_vertex_program *stvp = (struct st_vertex_program *)stp;

      blob_write_uint32(&blob, stvp->num_inputs);
      blob_write_uint32(&blob, stvp->vert_attrib_mask);
      blob_write_bytes(&blob, stvp->result_to_output,
                       sizeof(stvp->result_to_output));
   }

   if (prog->info.stage == MESA_SHADER_VERTEX ||
       prog->info.stage == MESA_SHADER_TESS_EVAL ||
       prog->info.stage == MESA_SHADER_GEOMETRY)
      write_stream_out_to_cache(&blob, &stp->state);

   if (nir)
      write_nir_to_cache(&blob, prog);
   else
      write_tgsi_to_cache(&blob, stp->state.tokens, prog);

   blob_finish(&blob);
}

/**
 * Store TGSI or NIR and any other required state in on-disk shader cache.
 */
void
st_store_ir_in_disk_cache(struct st_context *st, struct gl_program *prog,
                          bool nir)
{
   if (!st->ctx->Cache)
      return;

   /* Exit early when we are dealing with a ff shader with no source file to
    * generate a source from.
    */
   static const char zero[sizeof(prog->sh.data->sha1)] = {0};
   if (memcmp(prog->sh.data->sha1, zero, sizeof(prog->sh.data->sha1)) == 0)
      return;

   st_serialise_ir_program(st->ctx, prog, nir);

   if (st->ctx->_Shader->Flags & GLSL_CACHE_INFO) {
      fprintf(stderr, "putting %s state tracker IR in cache\n",
              _mesa_shader_stage_to_string(prog->info.stage));
   }
}

static void
read_stream_out_from_cache(struct blob_reader *blob_reader,
                           struct pipe_shader_state *state)
{
   memset(&state->stream_output, 0, sizeof(state->stream_output));
   state->stream_output.num_outputs = blob_read_uint32(blob_reader);
   if (state->stream_output.num_outputs) {
      blob_copy_bytes(blob_reader, &state->stream_output.stride,
                      sizeof(state->stream_output.stride));
      blob_copy_bytes(blob_reader, &state->stream_output.output,
                      sizeof(state->stream_output.output));
   }
}

static void
read_tgsi_from_cache(struct blob_reader *blob_reader,
                     const struct tgsi_token **tokens)
{
   unsigned num_tokens  = blob_read_uint32(blob_reader);
   unsigned tokens_size = num_tokens * sizeof(struct tgsi_token);
   *tokens = (const struct tgsi_token*) MALLOC(tokens_size);
   blob_copy_bytes(blob_reader, (uint8_t *) *tokens, tokens_size);
}

static void
st_deserialise_ir_program(struct gl_context *ctx,
                          struct gl_shader_program *shProg,
                          struct gl_program *prog, bool nir)
{
   struct st_context *st = st_context(ctx);
   size_t size = prog->driver_cache_blob_size;
   uint8_t *buffer = (uint8_t *) prog->driver_cache_blob;

   st_set_prog_affected_state_flags(prog);

   /* Avoid reallocation of the program parameter list, because the uniform
    * storage is only associated with the original parameter list.
    * This should be enough for Bitmap and DrawPixels constants.
    */
   _mesa_ensure_and_associate_uniform_storage(ctx, shProg, prog, 16);

   assert(prog->driver_cache_blob && prog->driver_cache_blob_size > 0);

   struct st_program *stp = st_program(prog);
   struct blob_reader blob_reader;
   blob_reader_init(&blob_reader, buffer, size);

   st_release_variants(st, stp);

   if (prog->info.stage == MESA_SHADER_VERTEX) {
      struct st_vertex_program *stvp = (struct st_vertex_program *)stp;
      stvp->num_inputs = blob_read_uint32(&blob_reader);
      stvp->vert_attrib_mask = blob_read_uint32(&blob_reader);
      blob_copy_bytes(&blob_reader, (uint8_t *) stvp->result_to_output,
                      sizeof(stvp->result_to_output));
   }

   if (prog->info.stage == MESA_SHADER_VERTEX ||
       prog->info.stage == MESA_SHADER_TESS_EVAL ||
       prog->info.stage == MESA_SHADER_GEOMETRY)
      read_stream_out_from_cache(&blob_reader, &stp->state);

   if (nir) {
      assert(prog->nir == NULL);
      assert(stp->serialized_nir == NULL);

      stp->state.type = PIPE_SHADER_IR_NIR;
      stp->serialized_nir_size = blob_read_intptr(&blob_reader);
      stp->serialized_nir = malloc(stp->serialized_nir_size);
      blob_copy_bytes(&blob_reader, stp->serialized_nir, stp->serialized_nir_size);
      stp->shader_program = shProg;
   } else {
      read_tgsi_from_cache(&blob_reader, &stp->state.tokens);
   }

   /* Make sure we don't try to read more data than we wrote. This should
    * never happen in release builds but its useful to have this check to
    * catch development bugs.
    */
   if (blob_reader.current != blob_reader.end || blob_reader.overrun) {
      assert(!"Invalid TGSI shader disk cache item!");

      if (ctx->_Shader->Flags & GLSL_CACHE_INFO) {
         fprintf(stderr, "Error reading program from cache (invalid "
                 "TGSI cache item)\n");
      }
   }

   st_finalize_program(st, prog);
}

bool
st_load_ir_from_disk_cache(struct gl_context *ctx,
                           struct gl_shader_program *prog,
                           bool nir)
{
   if (!ctx->Cache)
      return false;

   /* If we didn't load the GLSL metadata from cache then we could not have
    * loaded TGSI or NIR either.
    */
   if (prog->data->LinkStatus != LINKING_SKIPPED)
      return false;

   for (unsigned i = 0; i < MESA_SHADER_STAGES; i++) {
      if (prog->_LinkedShaders[i] == NULL)
         continue;

      struct gl_program *glprog = prog->_LinkedShaders[i]->Program;
      st_deserialise_ir_program(ctx, prog, glprog, nir);

      /* We don't need the cached blob anymore so free it */
      ralloc_free(glprog->driver_cache_blob);
      glprog->driver_cache_blob = NULL;
      glprog->driver_cache_blob_size = 0;

      if (ctx->_Shader->Flags & GLSL_CACHE_INFO) {
         fprintf(stderr, "%s state tracker IR retrieved from cache\n",
                 _mesa_shader_stage_to_string(i));
      }
   }

   return true;
}

void
st_serialise_tgsi_program(struct gl_context *ctx, struct gl_program *prog)
{
   st_serialise_ir_program(ctx, prog, false);
}

void
st_serialise_tgsi_program_binary(struct gl_context *ctx,
                                 struct gl_shader_program *shProg,
                                 struct gl_program *prog)
{
   st_serialise_ir_program(ctx, prog, false);
}

void
st_deserialise_tgsi_program(struct gl_context *ctx,
                            struct gl_shader_program *shProg,
                            struct gl_program *prog)
{
   st_deserialise_ir_program(ctx, shProg, prog, false);
}

void
st_serialise_nir_program(struct gl_context *ctx, struct gl_program *prog)
{
   st_serialise_ir_program(ctx, prog, true);
}

void
st_serialise_nir_program_binary(struct gl_context *ctx,
                                struct gl_shader_program *shProg,
                                struct gl_program *prog)
{
   st_serialise_ir_program(ctx, prog, true);
}

void
st_deserialise_nir_program(struct gl_context *ctx,
                           struct gl_shader_program *shProg,
                           struct gl_program *prog)
{
   st_deserialise_ir_program(ctx, shProg, prog, true);
}