#include "intel_bufmgr_ttm.h"
-#define BUFMGR_DEBUG 0
+#define DBG(...) do { \
+ if (bufmgr_ttm->bufmgr.debug) \
+ _mesa_printf(__VA_ARGS__); \
+} while (0)
-struct intel_reloc_info
-{
- GLuint type;
- GLuint reloc;
- GLuint delta;
- GLuint index;
- drm_handle_t handle;
-};
+/*
+ * These bits are always specified in each validation
+ * request. Other bits are not supported at this point
+ * as it would require a bit of investigation to figure
+ * out what mask value should be used.
+ */
+#define INTEL_BO_MASK (DRM_BO_MASK_MEM | \
+ DRM_BO_FLAG_READ | \
+ DRM_BO_FLAG_WRITE | \
+ DRM_BO_FLAG_EXE)
-struct intel_bo_node
-{
- drmMMListHead head;
- drmBO *buf;
+struct intel_validate_entry {
+ dri_bo *bo;
struct drm_i915_op_arg bo_arg;
- uint64_t flags;
- uint64_t mask;
- void (*destroy)(void *);
- void *priv;
-};
-
-struct intel_bo_reloc_list
-{
- drmMMListHead head;
- drmBO buf;
- uint32_t *relocs;
-};
-
-struct intel_bo_reloc_node
-{
- drmMMListHead head;
- drm_handle_t handle;
- uint32_t nr_reloc_types;
- struct intel_bo_reloc_list type_list;
-};
-
-struct intel_bo_list {
- unsigned numCurrent;
- drmMMListHead list;
- void (*destroy)(void *node);
};
typedef struct _dri_bufmgr_ttm {
dri_bufmgr bufmgr;
int fd;
- _glthread_Mutex mutex;
unsigned int fence_type;
unsigned int fence_type_flush;
uint32_t max_relocs;
- /** ttm relocation list */
- struct intel_bo_list list;
- struct intel_bo_list reloc_list;
+ struct intel_validate_entry *validate_array;
+ int validate_array_size;
+ int validate_count;
+
+ uint32_t *cached_reloc_buf_data;
} dri_bufmgr_ttm;
+/**
+ * Private information associated with a relocation that isn't already stored
+ * in the relocation buffer to be passed to the kernel.
+ */
+struct dri_ttm_reloc {
+ dri_bo *target_buf;
+ uint64_t validate_flags;
+};
+
typedef struct _dri_bo_ttm {
dri_bo bo;
- int refcount; /* Protected by bufmgr->mutex */
+ int refcount;
drmBO drm_bo;
const char *name;
+
+ uint64_t last_flags;
+
+ /**
+ * Index of the buffer within the validation list while preparing a
+ * batchbuffer execution.
+ */
+ int validate_index;
+
+ /** DRM buffer object containing relocation list */
+ uint32_t *reloc_buf_data;
+ struct dri_ttm_reloc *relocs;
+
+ /**
+ * Indicates that the buffer may be shared with other processes, so we
+ * can't hold maps beyond when the user does.
+ */
+ GLboolean shared;
+
+ GLboolean delayed_unmap;
+ /* Virtual address from the dri_bo_map whose unmap was delayed. */
+ void *saved_virtual;
} dri_bo_ttm;
typedef struct _dri_fence_ttm
{
dri_fence fence;
- int refcount; /* Protected by bufmgr->mutex */
+ int refcount;
const char *name;
drmFence drm_fence;
} dri_fence_ttm;
-
-static void
-intel_bo_free_list(struct intel_bo_list *list)
+static void dri_ttm_dump_validation_list(dri_bufmgr_ttm *bufmgr_ttm)
{
- struct intel_bo_node *node;
- drmMMListHead *l;
-
- l = list->list.next;
- while(l != &list->list) {
- DRMLISTDEL(l);
- node = DRMLISTENTRY(struct intel_bo_node, l, head);
- list->destroy(node);
- l = list->list.next;
- list->numCurrent--;
- }
-}
+ int i, j;
-static void
-generic_destroy(void *nodep)
-{
- free(nodep);
-}
+ for (i = 0; i < bufmgr_ttm->validate_count; i++) {
+ dri_bo *bo = bufmgr_ttm->validate_array[i].bo;
+ dri_bo_ttm *bo_ttm = (dri_bo_ttm *)bo;
-static int
-intel_create_bo_list(int numTarget, struct intel_bo_list *list,
- void (*destroy)(void *))
-{
- DRMINITLISTHEAD(&list->list);
- list->numCurrent = 0;
- if (destroy)
- list->destroy = destroy;
- else
- list->destroy = generic_destroy;
- return 0;
-}
+ if (bo_ttm->reloc_buf_data != NULL) {
+ for (j = 0; j < (bo_ttm->reloc_buf_data[0] & 0xffff); j++) {
+ uint32_t *reloc_entry = bo_ttm->reloc_buf_data +
+ I915_RELOC_HEADER +
+ j * I915_RELOC0_STRIDE;
+ dri_bo *target_bo =
+ bufmgr_ttm->validate_array[reloc_entry[2]].bo;
+ dri_bo_ttm *target_ttm = (dri_bo_ttm *)target_bo;
-
-static struct drm_i915_op_arg *
-intel_setup_validate_list(dri_bufmgr_ttm *bufmgr_ttm, GLuint *count_p)
-{
- struct intel_bo_list *list = &bufmgr_ttm->list;
- struct intel_bo_list *reloc_list = &bufmgr_ttm->reloc_list;
- struct intel_bo_node *node;
- struct intel_bo_reloc_node *rl_node;
- drmMMListHead *l, *rl;
- struct drm_i915_op_arg *arg, *first;
- struct drm_bo_op_req *req;
- uint64_t *prevNext = NULL;
- GLuint count = 0;
-
- first = NULL;
-
- for (l = list->list.next; l != &list->list; l = l->next) {
- node = DRMLISTENTRY(struct intel_bo_node, l, head);
-
- arg = &node->bo_arg;
- req = &arg->d.req;
-
- if (!first)
- first = arg;
-
- if (prevNext)
- *prevNext = (unsigned long) arg;
-
- memset(arg, 0, sizeof(*arg));
- prevNext = &arg->next;
- req->bo_req.handle = node->buf->handle;
- req->op = drm_bo_validate;
- req->bo_req.flags = node->flags;
- req->bo_req.hint = 0;
- req->bo_req.mask = node->mask;
- req->bo_req.fence_class = 0; /* Backwards compat. */
- arg->reloc_handle = 0;
-
- for (rl = reloc_list->list.next; rl != &reloc_list->list;
- rl = rl->next)
- {
- rl_node = DRMLISTENTRY(struct intel_bo_reloc_node, rl, head);
-
- if (rl_node->handle == node->buf->handle) {
- arg->reloc_handle = rl_node->type_list.buf.handle;
+ DBG("%2d: %s@0x%08x -> %s@0x%08x + 0x%08x\n",
+ i,
+ bo_ttm->name, reloc_entry[0],
+ target_ttm->name, target_bo->offset,
+ reloc_entry[1]);
}
+ } else {
+ DBG("%2d: %s\n", i, bo_ttm->name);
}
- count++;
}
-
- if (!first)
- return 0;
-
- *count_p = count;
- return first;
}
+/**
+ * Adds the given buffer to the list of buffers to be validated (moved into the
+ * appropriate memory type) with the next batch submission.
+ *
+ * If a buffer is validated multiple times in a batch submission, it ends up
+ * with the intersection of the memory type flags and the union of the
+ * access flags.
+ */
static void
-intel_free_validate_list(dri_bufmgr_ttm *bufmgr_ttm)
+intel_add_validate_buffer(dri_bo *buf,
+ uint64_t flags)
{
- struct intel_bo_list *list = &bufmgr_ttm->list;
- struct intel_bo_node *node;
- drmMMListHead *l;
-
- for (l = list->list.next; l != &list->list; l = l->next) {
- node = DRMLISTENTRY(struct intel_bo_node, l, head);
-
- if (node->destroy)
- (*node->destroy)(node->priv);
+ dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)buf->bufmgr;
+ dri_bo_ttm *ttm_buf = (dri_bo_ttm *)buf;
+ /* If we delayed doing an unmap to mitigate map/unmap syscall thrashing,
+ * do that now.
+ */
+ if (ttm_buf->delayed_unmap) {
+ drmBOUnmap(bufmgr_ttm->fd, &ttm_buf->drm_bo);
+ ttm_buf->delayed_unmap = GL_FALSE;
}
-}
-static void
-intel_free_reloc_list(dri_bufmgr_ttm *bufmgr_ttm)
-{
- struct intel_bo_list *reloc_list = &bufmgr_ttm->reloc_list;
- struct intel_bo_reloc_node *reloc_node;
- drmMMListHead *rl, *tmp;
+ if (ttm_buf->validate_index == -1) {
+ struct intel_validate_entry *entry;
+ struct drm_i915_op_arg *arg;
+ struct drm_bo_op_req *req;
+ int index;
- for (rl = reloc_list->list.next, tmp = rl->next; rl != &reloc_list->list;
- rl = tmp, tmp = rl->next)
- {
- reloc_node = DRMLISTENTRY(struct intel_bo_reloc_node, rl, head);
+ /* Extend the array of validation entries as necessary. */
+ if (bufmgr_ttm->validate_count == bufmgr_ttm->validate_array_size) {
+ int i, new_size = bufmgr_ttm->validate_array_size * 2;
- DRMLISTDEL(rl);
+ if (new_size == 0)
+ new_size = 5;
- if (reloc_node->nr_reloc_types > 1) {
- /* TODO */
+ bufmgr_ttm->validate_array =
+ realloc(bufmgr_ttm->validate_array,
+ sizeof(struct intel_validate_entry) * new_size);
+ bufmgr_ttm->validate_array_size = new_size;
+
+ /* Update pointers for realloced mem. */
+ for (i = 0; i < bufmgr_ttm->validate_count - 1; i++) {
+ bufmgr_ttm->validate_array[i].bo_arg.next = (unsigned long)
+ &bufmgr_ttm->validate_array[i + 1].bo_arg;
+ }
}
- drmBOUnmap(bufmgr_ttm->fd, &reloc_node->type_list.buf);
- drmBOUnreference(bufmgr_ttm->fd, &reloc_node->type_list.buf);
- free(reloc_node);
- }
-}
+ /* Pick out the new array entry for ourselves */
+ index = bufmgr_ttm->validate_count;
+ ttm_buf->validate_index = index;
+ entry = &bufmgr_ttm->validate_array[index];
+ bufmgr_ttm->validate_count++;
-static int
-intel_add_validate_buffer(struct intel_bo_list *list, dri_bo *buf,
- uint64_t flags, uint64_t mask,
- int *itemLoc, void (*destroy_cb)(void *))
-{
- struct intel_bo_list *list = &bufmgr_ttm->list;
- struct intel_bo_node *node, *cur;
- drmMMListHead *l;
- int count = 0;
- int ret = 0;
- drmBO *buf_bo = &((dri_bo_ttm *)buf)->drm_bo;
- cur = NULL;
-
- for (l = list->list.next; l != &list->list; l = l->next) {
- node = DRMLISTENTRY(struct intel_bo_node, l, head);
- if (node->buf->handle == buf_bo->handle) {
- cur = node;
- break;
- }
- count++;
- }
+ /* Fill in array entry */
+ entry->bo = buf;
+ dri_bo_reference(buf);
+
+ /* Fill in kernel arg */
+ arg = &entry->bo_arg;
+ req = &arg->d.req;
+
+ memset(arg, 0, sizeof(*arg));
+ req->bo_req.handle = ttm_buf->drm_bo.handle;
+ req->op = drm_bo_validate;
+ req->bo_req.flags = flags;
+ req->bo_req.hint = 0;
+#ifdef DRM_BO_HINT_PRESUMED_OFFSET
+ req->bo_req.hint |= DRM_BO_HINT_PRESUMED_OFFSET;
+ req->bo_req.presumed_offset = buf->offset;
+#endif
+ req->bo_req.mask = INTEL_BO_MASK;
+ req->bo_req.fence_class = 0; /* Backwards compat. */
+
+ if (ttm_buf->reloc_buf_data != NULL)
+ arg->reloc_ptr = (unsigned long)(void *)ttm_buf->reloc_buf_data;
+ else
+ arg->reloc_ptr = 0;
- if (!cur) {
- cur = drmMalloc(sizeof(*cur));
- if (!cur) {
- return -ENOMEM;
+ /* Hook up the linked list of args for the kernel */
+ arg->next = 0;
+ if (index != 0) {
+ bufmgr_ttm->validate_array[index - 1].bo_arg.next =
+ (unsigned long)arg;
}
- cur->buf = buf_bo;
- cur->priv = buf;
- cur->flags = flags;
- cur->mask = mask;
- cur->destroy = destroy_cb;
- ret = 1;
-
- DRMLISTADDTAIL(&cur->head, &list->list);
} else {
- uint64_t memMask = (cur->mask | mask) & DRM_BO_MASK_MEM;
- uint64_t memFlags = cur->flags & flags & memMask;
-
- if (!memFlags) {
- return -EINVAL;
+ struct intel_validate_entry *entry =
+ &bufmgr_ttm->validate_array[ttm_buf->validate_index];
+ struct drm_i915_op_arg *arg = &entry->bo_arg;
+ struct drm_bo_op_req *req = &arg->d.req;
+ uint64_t memFlags = req->bo_req.flags & flags & DRM_BO_MASK_MEM;
+ uint64_t modeFlags = (req->bo_req.flags | flags) & ~DRM_BO_MASK_MEM;
+
+ /* Buffer was already in the validate list. Extend its flags as
+ * necessary.
+ */
+
+ if (memFlags == 0) {
+ fprintf(stderr,
+ "%s: No shared memory types between "
+ "0x%16llx and 0x%16llx\n",
+ __FUNCTION__, req->bo_req.flags, flags);
+ abort();
}
- if (mask & cur->mask & ~DRM_BO_MASK_MEM & (cur->flags ^ flags)) {
- return -EINVAL;
+ if (flags & ~INTEL_BO_MASK) {
+ fprintf(stderr,
+ "%s: Flags bits 0x%16llx are not supposed to be used in a relocation\n",
+ __FUNCTION__, flags & ~INTEL_BO_MASK);
+ abort();
}
- cur->mask |= mask;
- cur->flags = memFlags | ((cur->flags | flags) &
- cur->mask & ~DRM_BO_MASK_MEM);
+ req->bo_req.flags = memFlags | modeFlags;
}
- *itemLoc = count;
- return ret;
}
sizeof(uint32_t))
static int
-intel_create_new_reloc_type_list(dri_bufmgr_ttm *bufmgr_ttm,
- struct intel_bo_reloc_list *cur_type)
+intel_setup_reloc_list(dri_bo *bo)
{
- int ret;
-
- /* should allocate a drmBO here */
- ret = drmBOCreate(bufmgr_ttm->fd, RELOC_BUF_SIZE(bufmgr_ttm->max_relocs), 0,
- NULL,
- DRM_BO_FLAG_MEM_LOCAL |
- DRM_BO_FLAG_READ |
- DRM_BO_FLAG_WRITE |
- DRM_BO_FLAG_MAPPABLE |
- DRM_BO_FLAG_CACHED,
- 0, &cur_type->buf);
- if (ret)
- return ret;
-
- ret = drmBOMap(bufmgr_ttm->fd, &cur_type->buf,
- DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
- 0, (void **)&cur_type->relocs);
- if (ret)
- return ret;
- return 0;
-}
-
-static int
-intel_add_validate_reloc(dri_bufmgr_ttm *bufmgr_ttm,
- struct intel_reloc_info *reloc_info)
-{
- struct intel_bo_list *reloc_list = &bufmgr_ttm->reloc_list;
- struct intel_bo_reloc_node *rl_node, *cur;
- drmMMListHead *rl, *l;
- int ret = 0;
- uint32_t *reloc_start;
- int num_relocs;
- struct intel_bo_reloc_list *cur_type;
-
- cur = NULL;
-
- for (rl = reloc_list->list.next; rl != &reloc_list->list; rl = rl->next) {
- rl_node = DRMLISTENTRY(struct intel_bo_reloc_node, rl, head);
- if (rl_node->handle == reloc_info->handle) {
- cur = rl_node;
- break;
- }
- }
-
- if (!cur) {
-
- cur = malloc(sizeof(*cur));
- if (!cur)
- return -ENOMEM;
+ dri_bo_ttm *bo_ttm = (dri_bo_ttm *)bo;
+ dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)bo->bufmgr;
- cur->nr_reloc_types = 1;
- cur->handle = reloc_info->handle;
- cur_type = &cur->type_list;
+ bo_ttm->relocs = malloc(sizeof(struct dri_ttm_reloc) *
+ bufmgr_ttm->max_relocs);
- DRMINITLISTHEAD(&cur->type_list.head);
- ret = intel_create_new_reloc_type_list(bufmgr_ttm, cur_type);
- if (ret) {
- return -1;
- }
- DRMLISTADDTAIL(&cur->head, &reloc_list->list);
-
- cur_type->relocs[0] = 0 | (reloc_info->type << 16);
- cur_type->relocs[1] = 0; // next reloc buffer handle is 0
+ if (bufmgr_ttm->cached_reloc_buf_data != NULL) {
+ bo_ttm->reloc_buf_data = bufmgr_ttm->cached_reloc_buf_data;
+ bufmgr_ttm->cached_reloc_buf_data = NULL;
} else {
- int found = 0;
- if ((cur->type_list.relocs[0] >> 16) == reloc_info->type) {
- cur_type = &cur->type_list;
- found = 1;
- } else {
- for (l = cur->type_list.head.next; l != &cur->type_list.head;
- l = l->next)
- {
- cur_type = DRMLISTENTRY(struct intel_bo_reloc_list, l, head);
- if (((cur_type->relocs[0] >> 16) & 0xffff) == reloc_info->type)
- found = 1;
- break;
- }
- }
-
- /* didn't find the relocation type */
- if (!found) {
- cur_type = malloc(sizeof(*cur_type));
- if (!cur_type) {
- return -ENOMEM;
- }
-
- ret = intel_create_new_reloc_type_list(bufmgr_ttm, cur_type);
- DRMLISTADDTAIL(&cur_type->head, &cur->type_list.head);
-
- cur_type->relocs[0] = (reloc_info->type << 16);
- cur_type->relocs[1] = 0;
-
- cur->nr_reloc_types++;
- }
+ bo_ttm->reloc_buf_data = calloc(1, RELOC_BUF_SIZE(bufmgr_ttm->max_relocs));
}
- reloc_start = cur_type->relocs;
-
- num_relocs = (reloc_start[0] & 0xffff);
+ /* Initialize the relocation list with the header:
+ * DWORD 0: relocation count
+ * DWORD 1: relocation type
+ * DWORD 2+3: handle to next relocation list (currently none) 64-bits
+ * DWORD 3: unused
+ */
+ bo_ttm->reloc_buf_data[0] = 0;
+ bo_ttm->reloc_buf_data[1] = I915_RELOC_TYPE_0;
+ bo_ttm->reloc_buf_data[2] = 0;
+ bo_ttm->reloc_buf_data[3] = 0;
- reloc_start[num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER] =
- reloc_info->reloc;
- reloc_start[num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER + 1] =
- reloc_info->delta;
- reloc_start[num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER + 2] =
- reloc_info->index;
- reloc_start[0]++;
- if (((reloc_start[0] & 0xffff)) > (bufmgr_ttm->max_relocs)) {
- return -ENOMEM;
- }
return 0;
}
-
#if 0
int
driFenceSignaled(DriFenceObject * fence, unsigned type)
if (fence == NULL)
return GL_TRUE;
- _glthread_LOCK_MUTEX(fence->mutex);
ret = drmFenceSignaled(bufmgr_ttm->fd, &fence->fence, type, &signaled);
- _glthread_UNLOCK_MUTEX(fence->mutex);
BM_CKFATAL(ret);
return signaled;
}
unsigned long size, unsigned int alignment,
uint64_t location_mask)
{
- dri_bufmgr_ttm *ttm_bufmgr;
+ dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)bufmgr;
dri_bo_ttm *ttm_buf;
unsigned int pageSize = getpagesize();
int ret;
unsigned int flags, hint;
- ttm_bufmgr = (dri_bufmgr_ttm *)bufmgr;
-
ttm_buf = malloc(sizeof(*ttm_buf));
if (!ttm_buf)
return NULL;
/* No hints we want to use. */
hint = 0;
- ret = drmBOCreate(ttm_bufmgr->fd, size, alignment / pageSize,
+ ret = drmBOCreate(bufmgr_ttm->fd, size, alignment / pageSize,
NULL, flags, hint, &ttm_buf->drm_bo);
if (ret != 0) {
free(ttm_buf);
ttm_buf->bo.bufmgr = bufmgr;
ttm_buf->name = name;
ttm_buf->refcount = 1;
+ ttm_buf->reloc_buf_data = NULL;
+ ttm_buf->relocs = NULL;
+ ttm_buf->last_flags = ttm_buf->drm_bo.flags;
+ ttm_buf->shared = GL_FALSE;
+ ttm_buf->delayed_unmap = GL_FALSE;
+ ttm_buf->validate_index = -1;
-#if BUFMGR_DEBUG
- fprintf(stderr, "bo_create: %p (%s)\n", &ttm_buf->bo, ttm_buf->name);
-#endif
+ DBG("bo_create: %p (%s) %db\n", &ttm_buf->bo, ttm_buf->name, size);
return &ttm_buf->bo;
}
intel_ttm_bo_create_from_handle(dri_bufmgr *bufmgr, const char *name,
unsigned int handle)
{
- dri_bufmgr_ttm *ttm_bufmgr;
+ dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)bufmgr;
dri_bo_ttm *ttm_buf;
int ret;
- ttm_bufmgr = (dri_bufmgr_ttm *)bufmgr;
-
ttm_buf = malloc(sizeof(*ttm_buf));
if (!ttm_buf)
return NULL;
- ret = drmBOReference(ttm_bufmgr->fd, handle, &ttm_buf->drm_bo);
+ ret = drmBOReference(bufmgr_ttm->fd, handle, &ttm_buf->drm_bo);
if (ret != 0) {
+ fprintf(stderr, "Couldn't reference %s handle 0x%08x: %s\n",
+ name, handle, strerror(-ret));
free(ttm_buf);
return NULL;
}
ttm_buf->bo.bufmgr = bufmgr;
ttm_buf->name = name;
ttm_buf->refcount = 1;
+ ttm_buf->reloc_buf_data = NULL;
+ ttm_buf->relocs = NULL;
+ ttm_buf->last_flags = ttm_buf->drm_bo.flags;
+ ttm_buf->shared = GL_TRUE;
+ ttm_buf->delayed_unmap = GL_FALSE;
+ ttm_buf->validate_index = -1;
-#if BUFMGR_DEBUG
- fprintf(stderr, "bo_create_from_handle: %p %08x (%s)\n",
- &ttm_buf->bo, handle, ttm_buf->name);
-#endif
+ DBG("bo_create_from_handle: %p %08x (%s)\n",
+ &ttm_buf->bo, handle, ttm_buf->name);
return &ttm_buf->bo;
}
static void
dri_ttm_bo_reference(dri_bo *buf)
{
- dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)buf->bufmgr;
dri_bo_ttm *ttm_buf = (dri_bo_ttm *)buf;
- _glthread_LOCK_MUTEX(bufmgr_ttm->mutex);
ttm_buf->refcount++;
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
}
static void
if (!buf)
return;
- _glthread_LOCK_MUTEX(bufmgr_ttm->mutex);
if (--ttm_buf->refcount == 0) {
int ret;
+ if (ttm_buf->reloc_buf_data) {
+ int i;
+
+ /* Unreference all the target buffers */
+ for (i = 0; i < (ttm_buf->reloc_buf_data[0] & 0xffff); i++)
+ dri_bo_unreference(ttm_buf->relocs[i].target_buf);
+ free(ttm_buf->relocs);
+
+ if (bufmgr_ttm->cached_reloc_buf_data == NULL) {
+ /* Cache a single relocation buffer allocation to avoid
+ * repeated create/map/unmap/destroy for batchbuffer
+ * relocations.
+ */
+ bufmgr_ttm->cached_reloc_buf_data = ttm_buf->reloc_buf_data;
+ } else {
+ /* Free the kernel BO containing relocation entries */
+ free(ttm_buf->reloc_buf_data);
+ ttm_buf->reloc_buf_data = NULL;
+ }
+ }
+
+ if (ttm_buf->delayed_unmap)
+ drmBOUnmap(bufmgr_ttm->fd, &ttm_buf->drm_bo);
+
ret = drmBOUnreference(bufmgr_ttm->fd, &ttm_buf->drm_bo);
if (ret != 0) {
fprintf(stderr, "drmBOUnreference failed (%s): %s\n",
ttm_buf->name, strerror(-ret));
}
-#if BUFMGR_DEBUG
- fprintf(stderr, "bo_unreference final: %p (%s)\n",
- &ttm_buf->bo, ttm_buf->name);
-#endif
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
+ DBG("bo_unreference final: %p (%s)\n", &ttm_buf->bo, ttm_buf->name);
+
free(buf);
return;
}
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
}
static int
assert(buf->virtual == NULL);
-#if BUFMGR_DEBUG
- fprintf(stderr, "bo_map: %p (%s)\n", &ttm_buf->bo, ttm_buf->name);
-#endif
+ DBG("bo_map: %p (%s)\n", &ttm_buf->bo, ttm_buf->name);
+
+ /* XXX: What about if we're upgrading from READ to WRITE? */
+ if (ttm_buf->delayed_unmap) {
+ buf->virtual = ttm_buf->saved_virtual;
+ return 0;
+ }
return drmBOMap(bufmgr_ttm->fd, &ttm_buf->drm_bo, flags, 0, &buf->virtual);
}
assert(buf->virtual != NULL);
- buf->virtual = NULL;
+ DBG("bo_unmap: %p (%s)\n", &ttm_buf->bo, ttm_buf->name);
-#if BUFMGR_DEBUG
- fprintf(stderr, "bo_unmap: %p (%s)\n", &ttm_buf->bo, ttm_buf->name);
-#endif
+ if (!ttm_buf->shared) {
+ ttm_buf->saved_virtual = buf->virtual;
+ ttm_buf->delayed_unmap = GL_TRUE;
+ buf->virtual = NULL;
+
+ return 0;
+ }
+
+ buf->virtual = NULL;
return drmBOUnmap(bufmgr_ttm->fd, &ttm_buf->drm_bo);
}
intel_ttm_fence_create_from_arg(dri_bufmgr *bufmgr, const char *name,
drm_fence_arg_t *arg)
{
- dri_bufmgr_ttm *ttm_bufmgr;
+ dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)bufmgr;
dri_fence_ttm *ttm_fence;
- ttm_bufmgr = (dri_bufmgr_ttm *)bufmgr;
-
ttm_fence = malloc(sizeof(*ttm_fence));
if (!ttm_fence)
return NULL;
ttm_fence->name = name;
ttm_fence->refcount = 1;
-#if BUFMGR_DEBUG
- fprintf(stderr, "fence_create_from_handle: %p (%s)\n", &ttm_fence->fence,
- ttm_fence->name);
-#endif
+ DBG("fence_create_from_handle: %p (%s)\n",
+ &ttm_fence->fence, ttm_fence->name);
return &ttm_fence->fence;
}
dri_fence_ttm *fence_ttm = (dri_fence_ttm *)fence;
dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)fence->bufmgr;
- _glthread_LOCK_MUTEX(bufmgr_ttm->mutex);
++fence_ttm->refcount;
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
-#if BUFMGR_DEBUG
- fprintf(stderr, "fence_reference: %p (%s)\n", &fence_ttm->fence,
- fence_ttm->name);
-#endif
+ DBG("fence_reference: %p (%s)\n", &fence_ttm->fence, fence_ttm->name);
}
static void
if (!fence)
return;
-#if BUFMGR_DEBUG
- fprintf(stderr, "fence_unreference: %p (%s)\n", &fence_ttm->fence,
- fence_ttm->name);
-#endif
- _glthread_LOCK_MUTEX(bufmgr_ttm->mutex);
+ DBG("fence_unreference: %p (%s)\n", &fence_ttm->fence, fence_ttm->name);
+
if (--fence_ttm->refcount == 0) {
int ret;
fence_ttm->name, strerror(-ret));
}
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
free(fence);
return;
}
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
}
static void
dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)fence->bufmgr;
int ret;
- _glthread_LOCK_MUTEX(bufmgr_ttm->mutex);
- ret = drmFenceWait(bufmgr_ttm->fd, 0, &fence_ttm->drm_fence, 0);
- _glthread_UNLOCK_MUTEX(bufmgr_ttm->mutex);
+ ret = drmFenceWait(bufmgr_ttm->fd, DRM_FENCE_FLAG_WAIT_LAZY, &fence_ttm->drm_fence, 0);
if (ret != 0) {
_mesa_printf("%s:%d: Error %d waiting for fence %s.\n",
__FILE__, __LINE__, ret, fence_ttm->name);
abort();
}
-#if BUFMGR_DEBUG
- fprintf(stderr, "fence_wait: %p (%s)\n", &fence_ttm->fence,
- fence_ttm->name);
-#endif
+ DBG("fence_wait: %p (%s)\n", &fence_ttm->fence, fence_ttm->name);
}
static void
{
dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)bufmgr;
- intel_bo_free_list(&bufmgr_ttm->list);
- intel_bo_free_list(&bufmgr_ttm->reloc_list);
+ if (bufmgr_ttm->cached_reloc_buf_data) {
+ free(bufmgr_ttm->cached_reloc_buf_data);
+ }
+
+ free(bufmgr_ttm->validate_array);
- _glthread_DESTROY_MUTEX(bufmgr_ttm->mutex);
free(bufmgr);
}
-
+/**
+ * Adds the target buffer to the validation list and adds the relocation
+ * to the reloc_buffer's relocation list.
+ *
+ * The relocation entry at the given offset must already contain the
+ * precomputed relocation value, because the kernel will optimize out
+ * the relocation entry write when the buffer hasn't moved from the
+ * last known offset in target_buf.
+ */
static void
-intel_dribo_destroy_callback(void *priv)
+dri_ttm_emit_reloc(dri_bo *reloc_buf, uint64_t flags, GLuint delta,
+ GLuint offset, dri_bo *target_buf)
{
- dri_bo *dribo = priv;
+ dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)reloc_buf->bufmgr;
+ dri_bo_ttm *reloc_buf_ttm = (dri_bo_ttm *)reloc_buf;
+ dri_bo_ttm *target_buf_ttm = (dri_bo_ttm *)target_buf;
+ int num_relocs;
+ uint32_t *this_reloc;
+
+ /* Create a new relocation list if needed */
+ if (reloc_buf_ttm->reloc_buf_data == NULL)
+ intel_setup_reloc_list(reloc_buf);
- if (dribo)
- dri_bo_unreference(dribo);
+ num_relocs = reloc_buf_ttm->reloc_buf_data[0];
+
+ /* Check overflow */
+ assert(num_relocs < bufmgr_ttm->max_relocs);
+
+ this_reloc = reloc_buf_ttm->reloc_buf_data + I915_RELOC_HEADER +
+ num_relocs * I915_RELOC0_STRIDE;
+
+ this_reloc[0] = offset;
+ this_reloc[1] = delta;
+ this_reloc[2] = target_buf_ttm->drm_bo.handle; /* To be filled in at exec time */
+ this_reloc[3] = 0;
+
+ reloc_buf_ttm->relocs[num_relocs].validate_flags = flags;
+ reloc_buf_ttm->relocs[num_relocs].target_buf = target_buf;
+ dri_bo_reference(target_buf);
+
+ reloc_buf_ttm->reloc_buf_data[0]++; /* Increment relocation count */
+ /* Check wraparound */
+ assert(reloc_buf_ttm->reloc_buf_data[0] != 0);
}
+/**
+ * Walk the tree of relocations rooted at BO and accumulate the list of
+ * validations to be performed and update the relocation buffers with
+ * index values into the validation list.
+ */
static void
-dri_ttm_emit_reloc(dri_bo *reloc_buf, uint64_t flags, GLuint delta,
- GLuint offset, dri_bo *target_buf)
+dri_ttm_bo_process_reloc(dri_bo *bo)
{
- dri_bo_ttm *ttm_buf = (dri_bo_ttm *)reloc_buf;
- dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)reloc_buf->bufmgr;
- int newItem;
- struct intel_reloc_info reloc;
- int mask;
- int ret;
+ dri_bo_ttm *bo_ttm = (dri_bo_ttm *)bo;
+ unsigned int nr_relocs;
+ int i;
- mask = DRM_BO_MASK_MEM;
- mask |= flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_EXE);
-
- ret = intel_add_validate_buffer(bufmgr_ttm, target_buf, flags, mask,
- &newItem, intel_dribo_destroy_callback);
- if (ret < 0)
+ if (bo_ttm->reloc_buf_data == NULL)
return;
- if (ret == 1)
- dri_bo_reference(target_buf);
+ nr_relocs = bo_ttm->reloc_buf_data[0] & 0xffff;
- reloc.type = I915_RELOC_TYPE_0;
- reloc.reloc = offset;
- reloc.delta = delta;
- reloc.index = newItem;
- reloc.handle = ttm_buf->drm_bo.handle;
+ for (i = 0; i < nr_relocs; i++) {
+ struct dri_ttm_reloc *r = &bo_ttm->relocs[i];
+ dri_bo_ttm *target_ttm = (dri_bo_ttm *)r->target_buf;
+ uint32_t *reloc_entry;
- intel_add_validate_reloc(bufmgr_ttm, &reloc);
-}
+ /* Continue walking the tree depth-first. */
+ dri_ttm_bo_process_reloc(r->target_buf);
+
+ /* Add the target to the validate list */
+ intel_add_validate_buffer(r->target_buf, r->validate_flags);
+ /* Update the index of the target in the relocation entry */
+ reloc_entry = bo_ttm->reloc_buf_data + I915_RELOC_HEADER +
+ i * I915_RELOC0_STRIDE;
+ }
+}
static void *
dri_ttm_process_reloc(dri_bo *batch_buf, GLuint *count)
{
dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)batch_buf->bufmgr;
- void *ptr;
- int itemLoc;
- dri_bo_unmap(batch_buf);
+ /* Update indices and set up the validate list. */
+ dri_ttm_bo_process_reloc(batch_buf);
+
+ /* Add the batch buffer to the validation list. There are no relocations
+ * pointing to it.
+ */
+ intel_add_validate_buffer(batch_buf,
+ DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE);
- intel_add_validate_buffer(bufmgr_ttm, batch_buf,
- DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE,
- DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE,
- &itemLoc, NULL);
+ *count = bufmgr_ttm->validate_count;
+ return &bufmgr_ttm->validate_array[0].bo_arg;
+}
- ptr = intel_setup_validate_list(bufmgr_ttm, count);
+static const char *
+intel_get_flags_mem_type_string(uint64_t flags)
+{
+ switch (flags & DRM_BO_MASK_MEM) {
+ case DRM_BO_FLAG_MEM_LOCAL: return "local";
+ case DRM_BO_FLAG_MEM_TT: return "ttm";
+ case DRM_BO_FLAG_MEM_VRAM: return "vram";
+ case DRM_BO_FLAG_MEM_PRIV0: return "priv0";
+ case DRM_BO_FLAG_MEM_PRIV1: return "priv1";
+ case DRM_BO_FLAG_MEM_PRIV2: return "priv2";
+ case DRM_BO_FLAG_MEM_PRIV3: return "priv3";
+ case DRM_BO_FLAG_MEM_PRIV4: return "priv4";
+ default: return NULL;
+ }
+}
- return ptr;
+static const char *
+intel_get_flags_caching_string(uint64_t flags)
+{
+ switch (flags & (DRM_BO_FLAG_CACHED | DRM_BO_FLAG_CACHED_MAPPED)) {
+ case 0: return "UU";
+ case DRM_BO_FLAG_CACHED: return "CU";
+ case DRM_BO_FLAG_CACHED_MAPPED: return "UC";
+ case DRM_BO_FLAG_CACHED | DRM_BO_FLAG_CACHED_MAPPED: return "CC";
+ default: return NULL;
+ }
+}
+
+static void
+intel_update_buffer_offsets (dri_bufmgr_ttm *bufmgr_ttm)
+{
+ int i;
+
+ for (i = 0; i < bufmgr_ttm->validate_count; i++) {
+ dri_bo *bo = bufmgr_ttm->validate_array[i].bo;
+ dri_bo_ttm *bo_ttm = (dri_bo_ttm *)bo;
+ struct drm_i915_op_arg *arg = &bufmgr_ttm->validate_array[i].bo_arg;
+ struct drm_bo_arg_rep *rep = &arg->d.rep;
+
+ /* Update the flags */
+ if (rep->bo_info.flags != bo_ttm->last_flags) {
+ DBG("BO %s migrated: %s/%s -> %s/%s\n",
+ bo_ttm->name,
+ intel_get_flags_mem_type_string(bo_ttm->last_flags),
+ intel_get_flags_caching_string(bo_ttm->last_flags),
+ intel_get_flags_mem_type_string(rep->bo_info.flags),
+ intel_get_flags_caching_string(rep->bo_info.flags));
+
+ bo_ttm->last_flags = rep->bo_info.flags;
+ }
+ /* Update the buffer offset */
+ if (rep->bo_info.offset != bo->offset) {
+ DBG("BO %s migrated: 0x%08x -> 0x%08x\n",
+ bo_ttm->name, bo->offset, rep->bo_info.offset);
+ bo->offset = rep->bo_info.offset;
+ }
+ }
}
static void
dri_ttm_post_submit(dri_bo *batch_buf, dri_fence **last_fence)
{
dri_bufmgr_ttm *bufmgr_ttm = (dri_bufmgr_ttm *)batch_buf->bufmgr;
+ int i;
- intel_free_validate_list(bufmgr_ttm);
- intel_free_reloc_list(bufmgr_ttm);
+ intel_update_buffer_offsets (bufmgr_ttm);
- intel_bo_free_list(&bufmgr_ttm->list);
+ if (bufmgr_ttm->bufmgr.debug)
+ dri_ttm_dump_validation_list(bufmgr_ttm);
+
+ for (i = 0; i < bufmgr_ttm->validate_count; i++) {
+ dri_bo *bo = bufmgr_ttm->validate_array[i].bo;
+ dri_bo_ttm *bo_ttm = (dri_bo_ttm *)bo;
+
+ /* Disconnect the buffer from the validate list */
+ bo_ttm->validate_index = -1;
+ dri_bo_unreference(bo);
+ bufmgr_ttm->validate_array[i].bo = NULL;
+ }
+ bufmgr_ttm->validate_count = 0;
}
/**
{
dri_bufmgr_ttm *bufmgr_ttm;
- bufmgr_ttm = malloc(sizeof(*bufmgr_ttm));
+ bufmgr_ttm = calloc(1, sizeof(*bufmgr_ttm));
bufmgr_ttm->fd = fd;
bufmgr_ttm->fence_type = fence_type;
bufmgr_ttm->fence_type_flush = fence_type_flush;
- _glthread_INIT_MUTEX(bufmgr_ttm->mutex);
+ bufmgr_ttm->cached_reloc_buf_data = NULL;
- /* lets go with one relocation per every four dwords - purely heuristic */
- bufmgr_ttm->max_relocs = batch_size / sizeof(uint32_t) / 4;
-
- intel_create_bo_list(10, &bufmgr_ttm->list, NULL);
- intel_create_bo_list(1, &bufmgr_ttm->reloc_list, NULL);
+ /* Let's go with one relocation per every 2 dwords (but round down a bit
+ * since a power of two will mean an extra page allocation for the reloc
+ * buffer).
+ *
+ * Every 4 was too few for the blender benchmark.
+ */
+ bufmgr_ttm->max_relocs = batch_size / sizeof(uint32_t) / 2 - 2;
bufmgr_ttm->bufmgr.bo_alloc = dri_ttm_alloc;
bufmgr_ttm->bufmgr.bo_alloc_static = dri_ttm_alloc_static;
bufmgr_ttm->bufmgr.emit_reloc = dri_ttm_emit_reloc;
bufmgr_ttm->bufmgr.process_relocs = dri_ttm_process_reloc;
bufmgr_ttm->bufmgr.post_submit = dri_ttm_post_submit;
+ bufmgr_ttm->bufmgr.debug = GL_FALSE;
return &bufmgr_ttm->bufmgr;
}
OSDN Git Service
