static int
vc4_simulator_pin_bos(struct drm_device *dev, struct vc4_exec_info *exec)
{
struct drm_vc4_submit_cl *args = exec->args;
struct vc4_context *vc4 = dev->vc4;
struct vc4_bo **bos = vc4->bo_pointers.base;
exec->bo_count = args->bo_handle_count;
exec->bo = calloc(exec->bo_count, sizeof(void *));
for (int i = 0; i < exec->bo_count; i++) {
struct vc4_bo *bo = bos[i];
struct drm_gem_cma_object *obj = vc4_wrap_bo_with_cma(dev, bo);
struct drm_vc4_bo *drm_bo = to_vc4_bo(&obj->base);
#if 0
fprintf(stderr, "bo hindex %d: %s\n", i, bo->name);
#endif
vc4_bo_map(bo);
memcpy(obj->vaddr, bo->map, bo->size);
exec->bo[i] = obj;
/* The kernel does this validation at shader create ioctl
* time.
*/
if (strcmp(bo->name, "code") == 0) {
drm_bo->validated_shader = vc4_validate_shader(obj);
if (!drm_bo->validated_shader)
abort();
}
}
return 0;
}
static int
vc4_simulator_pin_bos(struct drm_device *dev, struct exec_info *exec)
{
struct drm_vc4_submit_cl *args = exec->args;
struct vc4_context *vc4 = dev->vc4;
struct vc4_bo **bos = vc4->bo_pointers.base;
exec->bo_count = args->bo_handle_count;
exec->bo = calloc(exec->bo_count, sizeof(struct vc4_bo_exec_state));
for (int i = 0; i < exec->bo_count; i++) {
struct vc4_bo *bo = bos[i];
struct drm_gem_cma_object *obj = vc4_wrap_bo_with_cma(dev, bo);
#if 0
fprintf(stderr, "bo hindex %d: %s\n", i, bo->name);
#endif
vc4_bo_map(bo);
memcpy(obj->vaddr, bo->map, bo->size);
exec->bo[i].bo = obj;
}
return 0;
}
static void *
vc4_resource_transfer_map(struct pipe_context *pctx,
struct pipe_resource *prsc,
unsigned level, unsigned usage,
const struct pipe_box *box,
struct pipe_transfer **pptrans)
{
struct vc4_context *vc4 = vc4_context(pctx);
struct vc4_resource *rsc = vc4_resource(prsc);
struct vc4_resource_slice *slice = &rsc->slices[level];
struct vc4_transfer *trans;
struct pipe_transfer *ptrans;
enum pipe_format format = prsc->format;
char *buf;
if (usage & PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE) {
vc4_resource_bo_alloc(rsc);
} else if (!(usage & PIPE_TRANSFER_UNSYNCHRONIZED)) {
if (vc4_cl_references_bo(pctx, rsc->bo)) {
if ((usage & PIPE_TRANSFER_DISCARD_RANGE) &&
prsc->last_level == 0 &&
prsc->width0 == box->width &&
prsc->height0 == box->height &&
prsc->depth0 == box->depth) {
vc4_resource_bo_alloc(rsc);
} else {
vc4_flush(pctx);
}
}
}
if (usage & PIPE_TRANSFER_WRITE)
rsc->writes++;
trans = util_slab_alloc(&vc4->transfer_pool);
if (!trans)
return NULL;
/* XXX: Handle DONTBLOCK, DISCARD_RANGE, PERSISTENT, COHERENT. */
/* util_slab_alloc() doesn't zero: */
memset(trans, 0, sizeof(*trans));
ptrans = &trans->base;
pipe_resource_reference(&ptrans->resource, prsc);
ptrans->level = level;
ptrans->usage = usage;
ptrans->box = *box;
/* Note that the current kernel implementation is synchronous, so no
* need to do syncing stuff here yet.
*/
buf = vc4_bo_map(rsc->bo);
if (!buf) {
fprintf(stderr, "Failed to map bo\n");
goto fail;
}
*pptrans = ptrans;
if (rsc->tiled) {
uint32_t utile_w = vc4_utile_width(rsc->cpp);
uint32_t utile_h = vc4_utile_height(rsc->cpp);
/* No direct mappings of tiled, since we need to manually
* tile/untile.
*/
if (usage & PIPE_TRANSFER_MAP_DIRECTLY)
return NULL;
/* We need to align the box to utile boundaries, since that's
* what load/store operate on.
*/
uint32_t box_start_x = ptrans->box.x & (utile_w - 1);
uint32_t box_start_y = ptrans->box.y & (utile_h - 1);
ptrans->box.width += box_start_x;
ptrans->box.x -= box_start_x;
ptrans->box.height += box_start_y;
ptrans->box.y -= box_start_y;
ptrans->box.width = align(ptrans->box.width, utile_w);
ptrans->box.height = align(ptrans->box.height, utile_h);
ptrans->stride = ptrans->box.width * rsc->cpp;
ptrans->layer_stride = ptrans->stride;
trans->map = malloc(ptrans->stride * ptrans->box.height);
if (usage & PIPE_TRANSFER_READ) {
vc4_load_tiled_image(trans->map, ptrans->stride,
buf + slice->offset +
box->z * rsc->cube_map_stride,
slice->stride,
slice->tiling, rsc->cpp,
&ptrans->box);
}
return (trans->map +
box_start_x * rsc->cpp +
box_start_y * ptrans->stride);
} else {
ptrans->stride = slice->stride;
ptrans->layer_stride = ptrans->stride;
return buf + slice->offset +
box->y / util_format_get_blockheight(format) * ptrans->stride +
box->x / util_format_get_blockwidth(format) * rsc->cpp +
box->z * rsc->cube_map_stride;
}
fail:
vc4_resource_transfer_unmap(pctx, ptrans);
return NULL;
}
