/*
 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
 *
 * 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
 * on the rights to use, copy, modify, merge, publish, distribute, sub
 * license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS 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:
 *      Jerome Glisse
 */
#include "r600.h"
#include "r600d.h"

/* state creation functions prototype */
int r600_framebuffer_emit(struct r600_winsys *rdev, struct r600_atom *atom, void *data, struct radeon_ib *ib);
int r600_framebuffer_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_cb_cntl_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_rasterizer_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_viewport_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_scissor_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_blend_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_constants_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_dsa_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_vs_shader_emit(struct r600_winsys *rdev, struct r600_atom *atom, void *data, struct radeon_ib *ib);
int r600_vs_shader_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_fs_shader_emit(struct r600_winsys *rdev, struct r600_atom *atom, void *data, struct radeon_ib *ib);
int r600_fs_shader_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);
int r600_shader_resource_emit(struct r600_winsys *rdev, struct r600_atom *atom, void *data, struct radeon_ib *ib);
int r600_shader_resource_create(struct r600_winsys *rdev, struct r600_atom *atom, void *data);

/*
 * helpers functions
 */
static void r600_emit_flush(struct r600_winsys *rdev,
				struct radeon_ib *ib,
				struct radeon_bo *bo,
				u32 flags)
{
	ib->ptr[ib->cpkts++] = PKT3(PKT3_SURFACE_SYNC, 3);
	ib->ptr[ib->cpkts++] = flags;
	ib->ptr[ib->cpkts++] = bo->size >> 8;
	ib->ptr[ib->cpkts++] = 0x00000000;
	ib->ptr[ib->cpkts++] = 0x0000000A;
	ib->ptr[ib->cpkts++] = PKT3(PKT3_NOP, 0);
	ib->ptr[ib->cpkts++] = radeon_ib_reloc(ib, bo, RADEON_GEM_DOMAIN_VRAM |
						RADEON_GEM_DOMAIN_GTT);
}

void r600_atom_flush_add(struct r600_work *batch, struct radeon_bo *bo, u32 flags)
{
	int i;

	for (i = 0; i < batch->nflush; i++) {
		if (batch->flush[i].bo == bo) {
			batch->flush[i].flags |= flags;
			return;
		}
	}
	batch->flush[batch->nflush].bo = bo;
	batch->flush[batch->nflush].flags = flags;
	batch->nflush++;
}

/*
 * r600_scheduler
 */
static void r600_scheduler_clear_locked(struct r600_winsys *rw, struct r600_scheduler *scheduler)
{
	struct r600_work *batch, *n;
	int i, j;

	for (i = 0; i < scheduler->nwork; i++) {
		for (j = 0; j < R600_BATCH_NATOMS; j++) {
			r600_atom_destroy(scheduler->work[i].atoms[j]);
		}
	}
	scheduler->ib->cpkts = 0;
	scheduler->nwork = 0;
	r700_scheduler_states_default(rw, scheduler);
	scheduler->npkts = scheduler->ib->cpkts;
}

int r600_scheduler_flush(struct r600_winsys *rw)
{
	struct r600_scheduler *scheduler =  &rw->scheduler;
	int r, i, j;

	for (i = 0; i < scheduler->nwork; i++) {
		for (j = 0; j < scheduler->work[i].nflush; j++) {
			r600_emit_flush(rw, scheduler->ib,
					scheduler->work[i].flush[j].bo,
					scheduler->work[i].flush[j].flags);
		}
		for (j = 0; j < scheduler->work[i].nemit_atoms; j++) {
			r = scheduler->work[i].emit_atoms[j]->emit(rw,
							scheduler->work[i].emit_atoms[j],
							&scheduler->work[i],
							scheduler->ib);
			if (r)
				goto out_err;
		}
		r = r600_draw_cmd_emit(scheduler->ib, &scheduler->work[i].drm);
		/* flush + wait until */
		scheduler->ib->ptr[scheduler->ib->cpkts++] = PKT3(PKT3_EVENT_WRITE, 0);
		scheduler->ib->ptr[scheduler->ib->cpkts++] = 0x00000016;
		scheduler->ib->ptr[scheduler->ib->cpkts++] = PKT3(PKT3_SET_CONFIG_REG, 1);
		scheduler->ib->ptr[scheduler->ib->cpkts++] = 0x00000010;
		scheduler->ib->ptr[scheduler->ib->cpkts++] = 0x00028000;
	}
	printf("ib %d dw\n", scheduler->ib->cpkts);
	r = radeon_ib_schedule(rw, scheduler->ib);
out_err:
	/* FIXME helper function */
	scheduler->ib->cpkts = 0;
	scheduler->ib->nrelocs = 0;
	r600_scheduler_clear_locked(rw, scheduler);
	return r;
}

int r600_scheduler_queue(struct r600_winsys *rw, struct r600_request *rq)
{
	struct r600_batch *batch = rq->data;
	struct r600_work *rbatch;
	struct r600_scheduler *scheduler =  &rw->scheduler;
	int r, i, j;

	if (scheduler->nwork >= R600_MAX_WORK) {
		r = r600_scheduler_flush(rw);
		if (r)
			return r;
	}
	rbatch = &scheduler->work[scheduler->nwork];
	memset(rbatch, 0, sizeof(struct r600_work));
	i = 0;
	if (batch->blend == NULL || batch->cb_cntl == NULL ||
		batch->rasterizer == NULL || batch->viewport == NULL ||
		batch->framebuffer == NULL || batch->dsa == NULL ||
		batch->vs_shader == NULL || batch->fs_shader == NULL ||
		batch->scissor == NULL || batch->vs_resource == NULL) {
		free(rbatch);
		fprintf(stderr, "invalid batch\n");
		return -EINVAL;
	}
	rbatch->atoms[i++] = batch->blend;
	rbatch->atoms[i++] = batch->cb_cntl;
	rbatch->atoms[i++] = batch->rasterizer;
	rbatch->atoms[i++] = batch->viewport;
	rbatch->atoms[i++] = batch->scissor;
	rbatch->atoms[i++] = batch->vs_shader;
	rbatch->atoms[i++] = batch->fs_shader;
	rbatch->atoms[i++] = batch->dsa;
	rbatch->atoms[i++] = batch->vs_constants;
	rbatch->atoms[i++] = batch->ps_constants;
	rbatch->atoms[i++] = batch->vs_resource;
	rbatch->atoms[i++] = batch->framebuffer;
	for (i = 0; i < R600_BATCH_NATOMS; i++) {
		r600_atom_ref(rbatch->atoms[i]);
	}
	memcpy(&rbatch->drm, batch, sizeof(struct r600_batch));
reprocess:
	rbatch->nflush = 0;
	rbatch->npkts = 0;
	/* flush + wait until = 5dw */
	rbatch->npkts += 5;
	rbatch->npkts += r600_draw_cmd_size(batch);
	for (i = 0; i < R600_BATCH_NATOMS; i++) {
		if (rbatch->atoms[i]) {
			for (j = 0; j < rbatch->atoms[i]->nbo; j++) {
				r600_atom_flush_add(rbatch, rbatch->atoms[i]->bo[j],  rbatch->atoms[i]->flags[j]);
			}
			rbatch->emit_atoms[rbatch->nemit_atoms++] = rbatch->atoms[i];
			rbatch->npkts += rbatch->atoms[i]->npkts;
		}
	}
	/* add flush */
	rbatch->npkts += rbatch->nflush * 7;
	/* if batch is bigger than ib size it's an invalid one, this should
	 * not happen
	 */
	if (rbatch->npkts > scheduler->ib->length_dw) {
		fprintf(stderr, "single batch to big (%d) to fit into ib (%d)\n",
			rbatch->npkts, scheduler->ib->length_dw);
		goto out_err;
	}
	/* flush or not ? */
	if (scheduler->npkts + rbatch->npkts > scheduler->ib->length_dw) {
		r = r600_scheduler_flush(rw);
		if (r)
			goto out_err;
		goto reprocess;
	}
	/* batch is queued */
	for (i = 0; i < R600_BATCH_NATOMS; i++) {
		if (rbatch->atoms[i]) {
			scheduler->last_id[i] = rbatch->atoms[i]->id;
		}
	}
	printf("batch %d dw scheduler with %d dw\n", rbatch->npkts, scheduler->npkts);
	scheduler->npkts += rbatch->npkts;
	scheduler->nwork++;
	return 0;
out_err:
	for (i = 0; i < R600_BATCH_NATOMS; i++) {
		if (rbatch->atoms[i])
			r600_atom_destroy(rbatch->atoms[i]);
	}
	free(rbatch);
	return r;
}

static int r600_scheduler_init(struct r600_winsys *rw, struct r600_scheduler *scheduler)
{
	int r;

	memset(scheduler, 0 , sizeof(struct r600_scheduler));
	r = radeon_ib_get(rw, &scheduler->ib);
	if (r)
		return r;
	r600_scheduler_clear_locked(rw, scheduler);
	return 0;
}

static void r600_scheduler_cleanup_locked(struct r600_winsys *rw, struct r600_scheduler *scheduler)
{
	r600_scheduler_clear_locked(rw, scheduler);
	radeon_ib_free(scheduler->ib);
	scheduler->ib = NULL;
}

int r600_atoms_init(struct r600_winsys *rdev)
{
	rdev->npipes = 2;
	rdev->nbanks = 4;
	rdev->group_bytes = 256;
	return r600_scheduler_init(rdev, &rdev->scheduler);
}

void r600_atoms_release(struct r600_winsys *rdev)
{
	r600_scheduler_cleanup_locked(rdev, &rdev->scheduler);
}