mirror of
https://gitlab.freedesktop.org/mesa/mesa.git
synced 2026-05-26 10:18:12 +02:00
This mega-commit primarily does two things. First, is to turn anv_batch into a better abstraction of a batch. Instead of actually having a BO, it now has a few pointers to some piece of memory that are used to add data to the "batch". If it gets to the end, there is a function pointer that it can call to attempt to grow the batch. The second change is to start using chained batch buffers. When the end of the current batch BO is reached, it automatically creates a new one and ineserts an MI_BATCH_BUFFER_START command to chain to it. In this way, our batch buffers are effectively infinite in length.
305 lines
9 KiB
C
305 lines
9 KiB
C
/*
|
|
* Copyright © 2015 Intel Corporation
|
|
*
|
|
* 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 <stdlib.h>
|
|
#include <stdio.h>
|
|
#include <stdint.h>
|
|
#include <string.h>
|
|
#include <unistd.h>
|
|
#include <assert.h>
|
|
#include <sys/types.h>
|
|
#include <sys/mman.h>
|
|
|
|
#include <drm.h>
|
|
#include <i915_drm.h>
|
|
|
|
#include "private.h"
|
|
#include "aub.h"
|
|
|
|
struct anv_aub_writer {
|
|
FILE *file;
|
|
uint32_t offset;
|
|
int gen;
|
|
};
|
|
|
|
static void
|
|
aub_out(struct anv_aub_writer *writer, uint32_t data)
|
|
{
|
|
fwrite(&data, 1, 4, writer->file);
|
|
}
|
|
|
|
static void
|
|
aub_out_data(struct anv_aub_writer *writer, const void *data, size_t size)
|
|
{
|
|
fwrite(data, 1, size, writer->file);
|
|
}
|
|
|
|
static struct anv_aub_writer *
|
|
get_anv_aub_writer(struct anv_device *device)
|
|
{
|
|
struct anv_aub_writer *writer = device->aub_writer;
|
|
int entry = 0x200003;
|
|
int i;
|
|
int gtt_size = 0x10000;
|
|
const char *filename;
|
|
|
|
if (geteuid() != getuid())
|
|
return NULL;
|
|
|
|
if (writer)
|
|
return writer;
|
|
|
|
writer = malloc(sizeof(*writer));
|
|
if (writer == NULL)
|
|
return NULL;
|
|
|
|
filename = "intel.aub";
|
|
writer->gen = device->info.gen;
|
|
writer->file = fopen(filename, "w+");
|
|
if (!writer->file) {
|
|
free(writer);
|
|
return NULL;
|
|
}
|
|
|
|
/* Start allocating objects from just after the GTT. */
|
|
writer->offset = gtt_size;
|
|
|
|
/* Start with a (required) version packet. */
|
|
aub_out(writer, CMD_AUB_HEADER | (13 - 2));
|
|
aub_out(writer,
|
|
(4 << AUB_HEADER_MAJOR_SHIFT) |
|
|
(0 << AUB_HEADER_MINOR_SHIFT));
|
|
for (i = 0; i < 8; i++) {
|
|
aub_out(writer, 0); /* app name */
|
|
}
|
|
aub_out(writer, 0); /* timestamp */
|
|
aub_out(writer, 0); /* timestamp */
|
|
aub_out(writer, 0); /* comment len */
|
|
|
|
/* Set up the GTT. The max we can handle is 256M */
|
|
aub_out(writer, CMD_AUB_TRACE_HEADER_BLOCK | ((writer->gen >= 8 ? 6 : 5) - 2));
|
|
aub_out(writer,
|
|
AUB_TRACE_MEMTYPE_GTT_ENTRY |
|
|
AUB_TRACE_TYPE_NOTYPE | AUB_TRACE_OP_DATA_WRITE);
|
|
aub_out(writer, 0); /* subtype */
|
|
aub_out(writer, 0); /* offset */
|
|
aub_out(writer, gtt_size); /* size */
|
|
if (writer->gen >= 8)
|
|
aub_out(writer, 0);
|
|
for (i = 0x000; i < gtt_size; i += 4, entry += 0x1000) {
|
|
aub_out(writer, entry);
|
|
}
|
|
|
|
return device->aub_writer = writer;
|
|
}
|
|
|
|
void
|
|
anv_aub_writer_destroy(struct anv_aub_writer *writer)
|
|
{
|
|
fclose(writer->file);
|
|
free(writer);
|
|
}
|
|
|
|
|
|
/**
|
|
* Break up large objects into multiple writes. Otherwise a 128kb VBO
|
|
* would overflow the 16 bits of size field in the packet header and
|
|
* everything goes badly after that.
|
|
*/
|
|
static void
|
|
aub_write_trace_block(struct anv_aub_writer *writer, uint32_t type,
|
|
void *virtual, uint32_t size, uint32_t gtt_offset)
|
|
{
|
|
uint32_t block_size;
|
|
uint32_t offset;
|
|
uint32_t subtype = 0;
|
|
static const char null_block[8 * 4096];
|
|
|
|
for (offset = 0; offset < size; offset += block_size) {
|
|
block_size = size - offset;
|
|
|
|
if (block_size > 8 * 4096)
|
|
block_size = 8 * 4096;
|
|
|
|
aub_out(writer,
|
|
CMD_AUB_TRACE_HEADER_BLOCK |
|
|
((writer->gen >= 8 ? 6 : 5) - 2));
|
|
aub_out(writer,
|
|
AUB_TRACE_MEMTYPE_GTT |
|
|
type | AUB_TRACE_OP_DATA_WRITE);
|
|
aub_out(writer, subtype);
|
|
aub_out(writer, gtt_offset + offset);
|
|
aub_out(writer, ALIGN_U32(block_size, 4));
|
|
if (writer->gen >= 8)
|
|
aub_out(writer, 0);
|
|
|
|
if (virtual)
|
|
aub_out_data(writer, (char *) virtual + offset, block_size);
|
|
else
|
|
aub_out_data(writer, null_block, block_size);
|
|
|
|
/* Pad to a multiple of 4 bytes. */
|
|
aub_out_data(writer, null_block, -block_size & 3);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Make a ringbuffer on fly and dump it
|
|
*/
|
|
static void
|
|
aub_build_dump_ringbuffer(struct anv_aub_writer *writer,
|
|
uint32_t batch_offset, uint32_t offset,
|
|
int ring_flag)
|
|
{
|
|
uint32_t ringbuffer[4096];
|
|
int ring = AUB_TRACE_TYPE_RING_PRB0; /* The default ring */
|
|
int ring_count = 0;
|
|
|
|
if (ring_flag == I915_EXEC_BSD)
|
|
ring = AUB_TRACE_TYPE_RING_PRB1;
|
|
else if (ring_flag == I915_EXEC_BLT)
|
|
ring = AUB_TRACE_TYPE_RING_PRB2;
|
|
|
|
/* Make a ring buffer to execute our batchbuffer. */
|
|
memset(ringbuffer, 0, sizeof(ringbuffer));
|
|
if (writer->gen >= 8) {
|
|
ringbuffer[ring_count++] = AUB_MI_BATCH_BUFFER_START | (3 - 2);
|
|
ringbuffer[ring_count++] = batch_offset;
|
|
ringbuffer[ring_count++] = 0;
|
|
} else {
|
|
ringbuffer[ring_count++] = AUB_MI_BATCH_BUFFER_START;
|
|
ringbuffer[ring_count++] = batch_offset;
|
|
}
|
|
|
|
/* Write out the ring. This appears to trigger execution of
|
|
* the ring in the simulator.
|
|
*/
|
|
aub_out(writer,
|
|
CMD_AUB_TRACE_HEADER_BLOCK |
|
|
((writer->gen >= 8 ? 6 : 5) - 2));
|
|
aub_out(writer,
|
|
AUB_TRACE_MEMTYPE_GTT | ring | AUB_TRACE_OP_COMMAND_WRITE);
|
|
aub_out(writer, 0); /* general/surface subtype */
|
|
aub_out(writer, offset);
|
|
aub_out(writer, ring_count * 4);
|
|
if (writer->gen >= 8)
|
|
aub_out(writer, 0);
|
|
|
|
/* FIXME: Need some flush operations here? */
|
|
aub_out_data(writer, ringbuffer, ring_count * 4);
|
|
}
|
|
|
|
struct aub_bo {
|
|
uint32_t offset;
|
|
void *map;
|
|
void *relocated;
|
|
};
|
|
|
|
static void
|
|
relocate_bo(struct anv_bo *bo, struct drm_i915_gem_relocation_entry *relocs,
|
|
size_t num_relocs, struct aub_bo *bos)
|
|
{
|
|
struct aub_bo *aub_bo = &bos[bo->index];
|
|
struct drm_i915_gem_relocation_entry *reloc;
|
|
uint32_t *dw;
|
|
|
|
aub_bo->relocated = malloc(bo->size);
|
|
memcpy(aub_bo->relocated, aub_bo->map, bo->size);
|
|
for (size_t i = 0; i < num_relocs; i++) {
|
|
reloc = &relocs[i];
|
|
assert(reloc->offset < bo->size);
|
|
dw = aub_bo->relocated + reloc->offset;
|
|
*dw = bos[reloc->target_handle].offset + reloc->delta;
|
|
}
|
|
}
|
|
|
|
void
|
|
anv_cmd_buffer_dump(struct anv_cmd_buffer *cmd_buffer)
|
|
{
|
|
struct anv_device *device = cmd_buffer->device;
|
|
struct anv_batch *batch = &cmd_buffer->batch;
|
|
struct anv_aub_writer *writer;
|
|
struct anv_bo *bo;
|
|
uint32_t ring_flag = 0;
|
|
uint32_t offset;
|
|
struct aub_bo *aub_bos;
|
|
|
|
writer = get_anv_aub_writer(device);
|
|
if (writer == NULL)
|
|
return;
|
|
|
|
aub_bos = malloc(cmd_buffer->bo_count * sizeof(aub_bos[0]));
|
|
offset = writer->offset;
|
|
for (uint32_t i = 0; i < cmd_buffer->bo_count; i++) {
|
|
bo = cmd_buffer->exec2_bos[i];
|
|
if (bo->map)
|
|
aub_bos[i].map = bo->map;
|
|
else
|
|
aub_bos[i].map = anv_gem_mmap(device, bo->gem_handle, 0, bo->size);
|
|
aub_bos[i].relocated = aub_bos[i].map;
|
|
aub_bos[i].offset = offset;
|
|
offset = ALIGN_U32(offset + bo->size + 4095, 4096);
|
|
}
|
|
|
|
struct anv_batch_bo *first_bbo;
|
|
for (struct anv_batch_bo *bbo = cmd_buffer->last_batch_bo;
|
|
bbo != NULL; bbo = bbo->prev_batch_bo) {
|
|
/* Keep stashing the current BO until we get to the beginning */
|
|
first_bbo = bbo;
|
|
|
|
/* Handle relocations for this batch BO */
|
|
relocate_bo(&bbo->bo, &batch->relocs.relocs[bbo->first_reloc],
|
|
bbo->num_relocs, aub_bos);
|
|
}
|
|
assert(first_bbo->prev_batch_bo == NULL);
|
|
|
|
relocate_bo(&cmd_buffer->surface_bo,
|
|
cmd_buffer->surface_relocs.relocs,
|
|
cmd_buffer->surface_relocs.num_relocs, aub_bos);
|
|
|
|
for (uint32_t i = 0; i < cmd_buffer->bo_count; i++) {
|
|
bo = cmd_buffer->exec2_bos[i];
|
|
if (i == cmd_buffer->bo_count - 1) {
|
|
assert(bo == &first_bbo->bo);
|
|
aub_write_trace_block(writer, AUB_TRACE_TYPE_BATCH,
|
|
aub_bos[i].relocated,
|
|
first_bbo->length, aub_bos[i].offset);
|
|
} else {
|
|
aub_write_trace_block(writer, AUB_TRACE_TYPE_NOTYPE,
|
|
aub_bos[i].relocated,
|
|
bo->size, aub_bos[i].offset);
|
|
}
|
|
if (aub_bos[i].relocated != aub_bos[i].map)
|
|
free(aub_bos[i].relocated);
|
|
if (aub_bos[i].map != bo->map)
|
|
anv_gem_munmap(aub_bos[i].map, bo->size);
|
|
}
|
|
|
|
/* Dump ring buffer */
|
|
aub_build_dump_ringbuffer(writer, aub_bos[first_bbo->bo.index].offset,
|
|
offset, ring_flag);
|
|
|
|
free(aub_bos);
|
|
|
|
fflush(writer->file);
|
|
}
|