ShuriZma
/
xemu
mirror of https://github.com/xemu-project/xemu.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fix and refactor NV2A code 

This patch does the following:
- Fixes up things for Qemu 2.x compat
- Factors out the high-level NV2A blocks into separate files
- Updates g-lru-cache for latest glib compat (github.com/chergert/glrucache@c10af24)
- Changes texture hashing algorithm from FNV to xxH v0.6.5
This commit is contained in:
Matt Borgerson 2018-06-26 14:40:01 -07:00
parent cff0d97e35
commit 584dbda1d6
33 changed files with 6247 additions and 4397 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

338
hw/xbox/g-lru-cache.c
View File

@ -1,338 +0,0 @@
/* g-lru-cache.c
*
* Copyright (C) 2009 - Christian Hergert
*
* This is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
/*
* Ideally, you want to use fast_get. This is because we are using a
* GStaticRWLock which is indeed slower than a mutex if you have lots of writer
* acquisitions. This doesn't make it a true LRU, though, as the oldest
* retrieval from strorage is the first item evicted.
*/
#include "g-lru-cache.h"
// #define DEBUG
#define LRU_CACHE_PRIVATE(object) \
(G_TYPE_INSTANCE_GET_PRIVATE((object), \
G_TYPE_LRU_CACHE, \
GLruCachePrivate))
struct _GLruCachePrivate
{
GStaticRWLock rw_lock;
guint max_size;
gboolean fast_get;
GHashTable *hash_table;
GEqualFunc key_equal_func;
GCopyFunc key_copy_func;
GList *newest;
GList *oldest;
GLookupFunc retrieve_func;
gpointer user_data;
GDestroyNotify user_destroy_func;
};
G_DEFINE_TYPE (GLruCache, g_lru_cache, G_TYPE_OBJECT);
static void
g_lru_cache_finalize (GObject *object)
{
GLruCachePrivate *priv = LRU_CACHE_PRIVATE (object);
if (priv->user_data && priv->user_destroy_func)
priv->user_destroy_func (priv->user_data);
priv->user_data = NULL;
priv->user_destroy_func = NULL;
g_hash_table_destroy (priv->hash_table);
priv->hash_table = NULL;
g_list_free (priv->newest);
priv->newest = NULL;
priv->oldest = NULL;
G_OBJECT_CLASS (g_lru_cache_parent_class)->finalize (object);
}
static void
g_lru_cache_class_init (GLruCacheClass *klass)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
object_class->finalize = g_lru_cache_finalize;
g_type_class_add_private (object_class, sizeof (GLruCachePrivate));
}
static void
g_lru_cache_init (GLruCache *self)
{
self->priv = LRU_CACHE_PRIVATE (self);
self->priv->max_size = 1024;
self->priv->fast_get = FALSE;
g_static_rw_lock_init (&self->priv->rw_lock);
}
static void
g_lru_cache_evict_n_oldest_locked (GLruCache *self, gint n)
{
GList *victim;
gint i;
for (i = 0; i < n; i++)
{
victim = self->priv->oldest;
if (victim == NULL)
return;
if (victim->prev)
victim->prev->next = NULL;
self->priv->oldest = victim->prev;
g_hash_table_remove (self->priv->hash_table, victim->data);
if (self->priv->newest == victim)
self->priv->newest = NULL;
g_list_free1 (victim); /* victim->data is owned by hashtable */
}
#ifdef DEBUG
g_assert (g_hash_table_size (self->priv->hash_table) == g_list_length (self->priv->newest));
#endif
}
GLruCache*
g_lru_cache_new (GHashFunc hash_func,
GEqualFunc key_equal_func,
GCopyFunc key_copy_func,
GLookupFunc retrieve_func,
GDestroyNotify key_destroy_func,
GDestroyNotify value_destroy_func,
gpointer user_data,
GDestroyNotify user_destroy_func)
{
GLruCache *self = g_object_new (G_TYPE_LRU_CACHE, NULL);
self->priv->hash_table = g_hash_table_new_full (hash_func,
key_equal_func,
key_destroy_func,
value_destroy_func);
self->priv->key_equal_func = key_equal_func;
self->priv->key_copy_func = key_copy_func;
self->priv->retrieve_func = retrieve_func;
self->priv->user_data = user_data;
self->priv->user_destroy_func = user_destroy_func;
return self;
}
void
g_lru_cache_set_max_size (GLruCache *self, guint max_size)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
guint old_max_size = self->priv->max_size;
g_static_rw_lock_writer_lock (&(self->priv->rw_lock));
self->priv->max_size = max_size;
if (old_max_size > max_size)
g_lru_cache_evict_n_oldest_locked (self, old_max_size - max_size);
g_static_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
guint
g_lru_cache_get_max_size (GLruCache *self)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), -1);
return self->priv->max_size;
}
guint
g_lru_cache_get_size (GLruCache *self)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), -1);
return g_hash_table_size (self->priv->hash_table);
}
gpointer
g_lru_cache_get (GLruCache *self, gpointer key)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), NULL);
gpointer value;
g_static_rw_lock_reader_lock (&(self->priv->rw_lock));
value = g_hash_table_lookup (self->priv->hash_table, key);
#ifdef DEBUG
if (value)
g_debug ("Cache Hit!");
else
g_debug ("Cache miss");
#endif
g_static_rw_lock_reader_unlock (&(self->priv->rw_lock));
if (!value)
{
g_static_rw_lock_writer_lock (&(self->priv->rw_lock));
if (!g_hash_table_lookup (self->priv->hash_table, key))
{
if (g_hash_table_size (self->priv->hash_table) >= self->priv->max_size)
#ifdef DEBUG
{
g_debug ("We are at capacity, must evict oldest");
#endif
g_lru_cache_evict_n_oldest_locked (self, 1);
#ifdef DEBUG
}
g_debug ("Retrieving value from external resource");
#endif
value = self->priv->retrieve_func (key, self->priv->user_data);
if (self->priv->key_copy_func)
g_hash_table_insert (self->priv->hash_table,
self->priv->key_copy_func (key, self->priv->user_data),
value);
else
g_hash_table_insert (self->priv->hash_table, key, value);
self->priv->newest = g_list_prepend (self->priv->newest, key);
if (self->priv->oldest == NULL)
self->priv->oldest = self->priv->newest;
}
#ifdef DEBUG
else g_debug ("Lost storage race with another thread");
#endif
g_static_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
/* fast_get means that we do not reposition the item to the head
* of the list. it essentially makes the lru, a lru from storage,
* not lru to user.
*/
else if (!self->priv->fast_get &&
!self->priv->key_equal_func (key, self->priv->newest->data))
{
#ifdef DEBUG
g_debug ("Making item most recent");
#endif
g_static_rw_lock_writer_lock (&(self->priv->rw_lock));
GList *list = self->priv->newest;
GList *tmp;
GEqualFunc equal = self->priv->key_equal_func;
for (tmp = list; tmp; tmp = tmp->next)
{
if (equal (key, tmp->data))
{
GList *tmp1 = g_list_remove_link (list, tmp);
self->priv->newest = g_list_prepend (tmp1, tmp);
break;
}
}
g_static_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
return value;
}
void
g_lru_cache_evict (GLruCache *self, gpointer key)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
GEqualFunc equal = self->priv->key_equal_func;
GList *list = NULL;
g_static_rw_lock_writer_lock (&(self->priv->rw_lock));
if (equal (key, self->priv->oldest))
{
g_lru_cache_evict_n_oldest_locked (self, 1);
}
else
{
for (list = self->priv->newest; list; list = list->next)
{
/* key, list->data is owned by hashtable */
if (equal (key, list->data))
{
self->priv->newest = g_list_remove_link (self->priv->newest, list);
g_list_free (list);
break;
}
}
g_hash_table_remove (self->priv->hash_table, key);
}
g_static_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
void
g_lru_cache_clear (GLruCache *self)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
g_static_rw_lock_writer_lock (&(self->priv->rw_lock));
g_hash_table_remove_all (self->priv->hash_table);
g_list_free (self->priv->newest);
self->priv->oldest = NULL;
self->priv->newest = NULL;
g_static_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
void
g_lru_cache_set_fast_get (GLruCache *self, gboolean fast_get)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
self->priv->fast_get = fast_get;
}
gboolean
g_lru_cache_get_fast_get (GLruCache *self)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), FALSE);
return self->priv->fast_get;
}

32
hw/xbox/nv2a/Makefile.objs Normal file
View File

@ -0,0 +1,32 @@
obj-y += g-lru-cache.o
obj-y += swizzle.o
obj-y += nv2a.o
obj-y += nv2a_debug.o
obj-y += nv2a_shaders.o
###
# These are just #included into nv2a.c for build time savings
#
# obj-y += nv2a_pbus.o
# obj-y += nv2a_pcrtc.o
# obj-y += nv2a_pfb.o
# obj-y += nv2a_pfifo.o
# obj-y += nv2a_pgraph.o
# obj-y += nv2a_pmc.o
# obj-y += nv2a_pramdac.o
# obj-y += nv2a_prmcio.o
# obj-y += nv2a_prmvio.o
# obj-y += nv2a_ptimer.o
# obj-y += nv2a_pvideo.o
# obj-y += nv2a_user.o
# obj-y += nv2a_stubs.o
###
obj-y += nv2a_psh.o
obj-y += nv2a_vsh.o
obj-y += gl/
obj-y += xxhash.o
xxhash.o-cflags := -O3 -DXXH_FORCE_MEMORY_ACCESS=2

372
hw/xbox/nv2a/g-lru-cache.c Normal file
View File

@ -0,0 +1,372 @@
/* g-lru-cache.c
*
* Copyright (C) 2009 - Christian Hergert
*
* This is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
/*
* Ideally, you want to use fast_get. This is because we are using a
* GStaticRWLock which is indeed slower than a mutex if you have lots of writer
* acquisitions. This doesn't make it a true LRU, though, as the oldest
* retrieval from strorage is the first item evicted.
*/
#include "g-lru-cache.h"
#ifndef DEBUG
#define DEBUG 0
#endif
#define LRU_CACHE_PRIVATE(object) \
(G_TYPE_INSTANCE_GET_PRIVATE((object), \
G_TYPE_LRU_CACHE, \
GLruCachePrivate))
struct _GLruCachePrivate
{
GRWLock rw_lock;
guint max_size;
gboolean fast_get;
GHashTable *hash_table;
GEqualFunc key_equal_func;
GCopyFunc key_copy_func;
GList *newest;
GList *oldest;
GLookupFunc retrieve_func;
gpointer user_data;
GDestroyNotify user_destroy_func;
};
G_DEFINE_TYPE (GLruCache, g_lru_cache, G_TYPE_OBJECT);
static void
g_lru_cache_finalize (GObject *object)
{
GLruCachePrivate *priv = LRU_CACHE_PRIVATE (object);
if (priv->user_data && priv->user_destroy_func)
priv->user_destroy_func (priv->user_data);
priv->user_data = NULL;
priv->user_destroy_func = NULL;
g_hash_table_destroy (priv->hash_table);
priv->hash_table = NULL;
g_list_free (priv->newest);
priv->newest = NULL;
priv->oldest = NULL;
G_OBJECT_CLASS (g_lru_cache_parent_class)->finalize (object);
}
static void
g_lru_cache_class_init (GLruCacheClass *klass)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
object_class->finalize = g_lru_cache_finalize;
g_type_class_add_private (object_class, sizeof (GLruCachePrivate));
}
static void
g_lru_cache_init (GLruCache *self)
{
self->priv = LRU_CACHE_PRIVATE (self);
self->priv->max_size = 1024;
self->priv->fast_get = FALSE;
g_rw_lock_init (&self->priv->rw_lock);
}
static void
g_lru_cache_evict_n_oldest_locked (GLruCache *self, gint n)
{
GList *victim;
gint i;
for (i = 0; i < n; i++)
{
victim = self->priv->oldest;
if (victim == NULL)
return;
if (victim->prev)
victim->prev->next = NULL;
self->priv->oldest = victim->prev;
g_hash_table_remove (self->priv->hash_table, victim->data);
if (self->priv->newest == victim)
self->priv->newest = NULL;
g_list_free1 (victim); /* victim->data is owned by hashtable */
}
#if DEBUG
g_assert (g_hash_table_size (self->priv->hash_table) == g_list_length (self->priv->newest));
#endif
}
GLruCache*
g_lru_cache_new (GHashFunc key_hash_func,
GEqualFunc key_equal_func,
GLookupFunc retrieve_func,
gpointer user_data,
GDestroyNotify user_destroy_func)
{
return g_lru_cache_new_full (0,
NULL,
NULL,
0,
NULL,
NULL,
key_hash_func,
key_equal_func,
retrieve_func,
user_data,
user_destroy_func);
}
GLruCache*
g_lru_cache_new_full (GType key_type,
GCopyFunc key_copy_func,
GDestroyNotify key_destroy_func,
GType value_type,
GCopyFunc value_copy_func,
GDestroyNotify value_destroy_func,
GHashFunc key_hash_func,
GEqualFunc key_equal_func,
GLookupFunc retrieve_func,
gpointer user_data,
GDestroyNotify user_destroy_func)
{
GLruCache *self = g_object_new (G_TYPE_LRU_CACHE, NULL);
self->priv->hash_table = g_hash_table_new_full (key_hash_func,
key_equal_func,
key_destroy_func,
value_destroy_func);
self->priv->key_equal_func = key_equal_func;
self->priv->key_copy_func = key_copy_func;
self->priv->retrieve_func = retrieve_func;
self->priv->user_data = user_data;
self->priv->user_destroy_func = user_destroy_func;
return self;
}
void
g_lru_cache_set_max_size (GLruCache *self, guint max_size)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
guint old_max_size = self->priv->max_size;
g_rw_lock_writer_lock (&(self->priv->rw_lock));
self->priv->max_size = max_size;
if (old_max_size > max_size)
g_lru_cache_evict_n_oldest_locked (self, old_max_size - max_size);
g_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
guint
g_lru_cache_get_max_size (GLruCache *self)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), -1);
return self->priv->max_size;
}
guint
g_lru_cache_get_size (GLruCache *self)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), -1);
return g_hash_table_size (self->priv->hash_table);
}
gpointer
g_lru_cache_get (GLruCache *self, gpointer key, GError **error)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), NULL);
gpointer value;
GError *retrieve_error = NULL;
g_rw_lock_reader_lock (&(self->priv->rw_lock));
value = g_hash_table_lookup (self->priv->hash_table, key);
#if DEBUG
if (value)
g_debug ("Cache Hit!");
else
g_debug ("Cache miss");
#endif
g_rw_lock_reader_unlock (&(self->priv->rw_lock));
if (!value)
{
g_rw_lock_writer_lock (&(self->priv->rw_lock));
if (!g_hash_table_lookup (self->priv->hash_table, key))
{
if (g_hash_table_size (self->priv->hash_table) >= self->priv->max_size)
#if DEBUG
{
g_debug ("We are at capacity, must evict oldest");
#endif
g_lru_cache_evict_n_oldest_locked (self, 1);
#if DEBUG
}
g_debug ("Retrieving value from external resource");
#endif
value = self->priv->retrieve_func (key,
self->priv->user_data,
&retrieve_error);
if (G_UNLIKELY (retrieve_error != NULL))
{
g_propagate_error (error, retrieve_error);
return value; /* likely 'NULL', but we should be transparent */
}
if (self->priv->key_copy_func)
g_hash_table_insert (self->priv->hash_table,
self->priv->key_copy_func (key, self->priv->user_data),
value);
else
g_hash_table_insert (self->priv->hash_table, key, value);
self->priv->newest = g_list_prepend (self->priv->newest, key);
if (self->priv->oldest == NULL)
self->priv->oldest = self->priv->newest;
}
#if DEBUG
else g_debug ("Lost storage race with another thread");
#endif
g_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
/* fast_get means that we do not reposition the item to the head
* of the list. it essentially makes the lru, a lru from storage,
* not lru to user.
*/
else if (!self->priv->fast_get &&
!self->priv->key_equal_func (key, self->priv->newest->data))
{
#if DEBUG
g_debug ("Making item most recent");
#endif
g_rw_lock_writer_lock (&(self->priv->rw_lock));
GList *list = self->priv->newest;
GList *tmp;
GEqualFunc equal = self->priv->key_equal_func;
for (tmp = list; tmp; tmp = tmp->next)
{
if (equal (key, tmp->data))
{
GList *tmp1 = g_list_remove_link (list, tmp);
self->priv->newest = g_list_prepend (tmp1, tmp);
break;
}
}
g_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
return value;
}
void
g_lru_cache_evict (GLruCache *self, gpointer key)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
GEqualFunc equal = self->priv->key_equal_func;
GList *list = NULL;
g_rw_lock_writer_lock (&(self->priv->rw_lock));
if (equal (key, self->priv->oldest))
{
g_lru_cache_evict_n_oldest_locked (self, 1);
}
else
{
g_hash_table_remove (self->priv->hash_table, key);
for (list = self->priv->newest; list; list = list->next)
{
if (equal (key, list->data))
{
self->priv->newest = g_list_remove_link (self->priv->newest, list);
g_list_free (list);
break;
}
}
}
g_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
void
g_lru_cache_clear (GLruCache *self)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
g_rw_lock_writer_lock (&(self->priv->rw_lock));
g_hash_table_remove_all (self->priv->hash_table);
g_list_free (self->priv->newest);
self->priv->oldest = NULL;
self->priv->newest = NULL;
g_rw_lock_writer_unlock (&(self->priv->rw_lock));
}
void
g_lru_cache_set_fast_get (GLruCache *self, gboolean fast_get)
{
g_return_if_fail (G_IS_LRU_CACHE (self));
self->priv->fast_get = fast_get;
}
gboolean
g_lru_cache_get_fast_get (GLruCache *self)
{
g_return_val_if_fail (G_IS_LRU_CACHE (self), FALSE);
return self->priv->fast_get;
}

53
hw/xbox/g-lru-cache.h → hw/xbox/nv2a/g-lru-cache.h
View File

@ -20,46 +20,59 @@
#ifndef __G_LRU_CACHE_H__ #ifndef __G_LRU_CACHE_H__
#define __G_LRU_CACHE_H__ #define __G_LRU_CACHE_H__
#ifdef __cplusplus
extern "C" {
#endif
#include <glib.h> #include <glib.h>
#include <glib-object.h> #include <glib-object.h>
G_BEGIN_DECLS G_BEGIN_DECLS
#define G_TYPE_LRU_CACHE (g_lru_cache_get_type ()) #define G_TYPE_LRU_CACHE (g_lru_cache_get_type ())
#define G_LRU_CACHE(obj) (G_TYPE_CHECK_INSTANCE_CAST ((obj), G_TYPE_LRU_CACHE, GLruCache)) #define G_LRU_CACHE(obj) (G_TYPE_CHECK_INSTANCE_CAST ((obj), G_TYPE_LRU_CACHE, GLruCache))
#define G_LRU_CACHE_CONST(obj) (G_TYPE_CHECK_INSTANCE_CAST ((obj), G_TYPE_LRU_CACHE, GLruCache const)) #define G_LRU_CACHE_CONST(obj) (G_TYPE_CHECK_INSTANCE_CAST ((obj), G_TYPE_LRU_CACHE, GLruCache const))
#define G_LRU_CACHE_CLASS(klass) (G_TYPE_CHECK_CLASS_CAST ((klass), G_TYPE_LRU_CACHE, GLruCacheClass)) #define G_LRU_CACHE_CLASS(klass) (G_TYPE_CHECK_CLASS_CAST ((klass), G_TYPE_LRU_CACHE, GLruCacheClass))
#define G_IS_LRU_CACHE(obj) (G_TYPE_CHECK_INSTANCE_TYPE ((obj), G_TYPE_LRU_CACHE)) #define G_IS_LRU_CACHE(obj) (G_TYPE_CHECK_INSTANCE_TYPE ((obj), G_TYPE_LRU_CACHE))
#define G_IS_LRU_CACHE_CLASS(klass) (G_TYPE_CHECK_CLASS_TYPE ((klass), G_TYPE_LRU_CACHE)) #define G_IS_LRU_CACHE_CLASS(klass) (G_TYPE_CHECK_CLASS_TYPE ((klass), G_TYPE_LRU_CACHE))
#define G_LRU_CACHE_GET_CLASS(obj) (G_TYPE_INSTANCE_GET_CLASS ((obj), G_TYPE_LRU_CACHE, GLruCacheClass)) #define G_LRU_CACHE_GET_CLASS(obj) (G_TYPE_INSTANCE_GET_CLASS ((obj), G_TYPE_LRU_CACHE, GLruCacheClass))
#define G_LOOKUP_FUNC(func) ((GLookupFunc)func) #define G_LOOKUP_FUNC(func) ((GLookupFunc)func)
typedef struct _GLruCache GLruCache; typedef struct _GLruCache GLruCache;
typedef struct _GLruCacheClass GLruCacheClass; typedef struct _GLruCacheClass GLruCacheClass;
typedef struct _GLruCachePrivate GLruCachePrivate; typedef struct _GLruCachePrivate GLruCachePrivate;
typedef gpointer (*GLookupFunc) (gpointer key, gpointer user_data); typedef gpointer (*GLookupFunc) (gpointer key, gpointer user_data, GError **error);
struct _GLruCache struct _GLruCache
{ {
GObject parent; GObject parent;
GLruCachePrivate *priv; GLruCachePrivate *priv;
}; };
struct _GLruCacheClass struct _GLruCacheClass
{ {
GObjectClass parent_class; GObjectClass parent_class;
}; };
GType g_lru_cache_get_type (void) G_GNUC_CONST; GType g_lru_cache_get_type (void) G_GNUC_CONST;
GLruCache* g_lru_cache_new (GHashFunc hash_func, GLruCache* g_lru_cache_new (GHashFunc key_hash_func,
GEqualFunc key_equal_func, GEqualFunc key_equal_func,
GCopyFunc key_copy_func,
GLookupFunc retrieve_func, GLookupFunc retrieve_func,
gpointer user_data,
GDestroyNotify user_destroy_func);
GLruCache* g_lru_cache_new_full (GType key_type,
GCopyFunc key_copy_func,
GDestroyNotify key_destroy_func, GDestroyNotify key_destroy_func,
GType value_type,
GCopyFunc value_copy_func,
GDestroyNotify value_destroy_func, GDestroyNotify value_destroy_func,
GHashFunc key_hash_func,
GEqualFunc key_equal_func,
GLookupFunc retrieve_func,
gpointer user_data, gpointer user_data,
GDestroyNotify user_destroy_func); GDestroyNotify user_destroy_func);
@ -68,7 +81,7 @@ guint g_lru_cache_get_max_size (GLruCache *self);
guint g_lru_cache_get_size (GLruCache *self); guint g_lru_cache_get_size (GLruCache *self);
gpointer g_lru_cache_get (GLruCache *self, gpointer key); gpointer g_lru_cache_get (GLruCache *self, gpointer key, GError **error);
void g_lru_cache_evict (GLruCache *self, gpointer key); void g_lru_cache_evict (GLruCache *self, gpointer key);
void g_lru_cache_clear (GLruCache *self); void g_lru_cache_clear (GLruCache *self);
@ -77,4 +90,8 @@ void g_lru_cache_set_fast_get (GLruCache *self, gboolean fast_get);
G_END_DECLS G_END_DECLS
#ifdef __cplusplus
}
#endif
#endif /* __G_LRU_CACHE_H__ */ #endif /* __G_LRU_CACHE_H__ */

568
hw/xbox/nv2a/nv2a.c Normal file
View File

@ -0,0 +1,568 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
#include "qapi/error.h"
#include "qemu/error-report.h"
#include "qemu/error-report.h"
#include <assert.h>
#include "nv2a.h"
#include "hw/display/vga_regs.h"
#ifdef __WINNT__
// HACK: mingw-w64 doesn't provide ffs, for now we just shove it here
// TODO: decide on a better location
int ffs(register int valu)
{
register int bit;
if (valu == 0)
return 0;
for (bit = 1; !(valu & 1); bit++)
valu >>= 1;
return bit;
}
#endif
DMAObject nv_dma_load(NV2AState *d, hwaddr dma_obj_address);
void *nv_dma_map(NV2AState *d, hwaddr dma_obj_address, hwaddr *len);
void nv2a_init(PCIBus *bus, int devfn, MemoryRegion *ram);
void update_irq(NV2AState *d)
{
/* PFIFO */
if (d->pfifo.pending_interrupts & d->pfifo.enabled_interrupts) {
d->pmc.pending_interrupts |= NV_PMC_INTR_0_PFIFO;
} else {
d->pmc.pending_interrupts &= ~NV_PMC_INTR_0_PFIFO;
}
/* PCRTC */
if (d->pcrtc.pending_interrupts & d->pcrtc.enabled_interrupts) {
d->pmc.pending_interrupts |= NV_PMC_INTR_0_PCRTC;
} else {
d->pmc.pending_interrupts &= ~NV_PMC_INTR_0_PCRTC;
}
/* PGRAPH */
if (d->pgraph.pending_interrupts & d->pgraph.enabled_interrupts) {
d->pmc.pending_interrupts |= NV_PMC_INTR_0_PGRAPH;
} else {
d->pmc.pending_interrupts &= ~NV_PMC_INTR_0_PGRAPH;
}
if (d->pmc.pending_interrupts && d->pmc.enabled_interrupts) {
NV2A_DPRINTF("raise irq\n");
pci_irq_assert(&d->dev);
} else {
pci_irq_deassert(&d->dev);
}
}
DMAObject nv_dma_load(NV2AState *d, hwaddr dma_obj_address)
{
assert(dma_obj_address < memory_region_size(&d->ramin));
uint32_t *dma_obj = (uint32_t*)(d->ramin_ptr + dma_obj_address);
uint32_t flags = ldl_le_p(dma_obj);
uint32_t limit = ldl_le_p(dma_obj + 1);
uint32_t frame = ldl_le_p(dma_obj + 2);
return (DMAObject){
.dma_class = GET_MASK(flags, NV_DMA_CLASS),
.dma_target = GET_MASK(flags, NV_DMA_TARGET),
.address = (frame & NV_DMA_ADDRESS) | GET_MASK(flags, NV_DMA_ADJUST),
.limit = limit,
};
}
void *nv_dma_map(NV2AState *d, hwaddr dma_obj_address, hwaddr *len)
{
assert(dma_obj_address < memory_region_size(&d->ramin));
DMAObject dma = nv_dma_load(d, dma_obj_address);
/* TODO: Handle targets and classes properly */
NV2A_DPRINTF("dma_map %x, %x, %" HWADDR_PRIx " %" HWADDR_PRIx "\n",
dma.dma_class, dma.dma_target, dma.address, dma.limit);
dma.address &= 0x07FFFFFF;
// assert(dma.address + dma.limit < memory_region_size(d->vram));
*len = dma.limit;
return d->vram_ptr + dma.address;
}
#define STUB 0
#if STUB
void *pfifo_puller_thread(void *opaque) { return NULL; }
void pgraph_init(NV2AState *d){}
static void pfifo_run_pusher(NV2AState *d){}
void pgraph_destroy(PGRAPHState *pg){}
static uint8_t cliptobyte(int x)
{
return (uint8_t)((x < 0) ? 0 : ((x > 255) ? 255 : x));
}
static void convert_yuy2_to_rgb(const uint8_t *line, unsigned int ix,
uint8_t *r, uint8_t *g, uint8_t* b) {
int c, d, e;
c = (int)line[ix * 2] - 16;
if (ix % 2) {
d = (int)line[ix * 2 - 1] - 128;
e = (int)line[ix * 2 + 1] - 128;
} else {
d = (int)line[ix * 2 + 1] - 128;
e = (int)line[ix * 2 + 3] - 128;
}
*r = cliptobyte((298 * c + 409 * e + 128) >> 8);
*g = cliptobyte((298 * c - 100 * d - 208 * e + 128) >> 8);
*b = cliptobyte((298 * c + 516 * d + 128) >> 8);
}
#endif
#define DEFINE_PROTO(prefix) \
uint64_t prefix ## _read(void *opaque, hwaddr addr, unsigned int size); \
void prefix ## _write(void *opaque, hwaddr addr, uint64_t val, unsigned int size);
DEFINE_PROTO(pmc)
DEFINE_PROTO(pbus)
DEFINE_PROTO(pfifo)
DEFINE_PROTO(prma)
DEFINE_PROTO(pvideo)
DEFINE_PROTO(ptimer)
DEFINE_PROTO(pcounter)
DEFINE_PROTO(pvpe)
DEFINE_PROTO(ptv)
DEFINE_PROTO(prmfb)
DEFINE_PROTO(prmvio)
DEFINE_PROTO(pfb)
DEFINE_PROTO(pstraps)
DEFINE_PROTO(pgraph)
DEFINE_PROTO(pcrtc)
DEFINE_PROTO(prmcio)
DEFINE_PROTO(pramdac)
DEFINE_PROTO(prmdio)
DEFINE_PROTO(pramin)
DEFINE_PROTO(user)
#undef DEFINE_PROTO
#include "nv2a_pbus.c"
#include "nv2a_pcrtc.c"
#include "nv2a_pfb.c"
#if !STUB
#include "nv2a_pgraph.c"
#include "nv2a_pfifo.c"
#endif
#include "nv2a_pmc.c"
#include "nv2a_pramdac.c"
#include "nv2a_prmcio.c"
#include "nv2a_prmvio.c"
#include "nv2a_ptimer.c"
#include "nv2a_pvideo.c"
#include "nv2a_stubs.c"
#include "nv2a_user.c"
#if STUB
void pgraph_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
reg_log_write(NV_PGRAPH, addr, val);
}
uint64_t pgraph_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PGRAPH, addr, 0);
return 0;
}
void pfifo_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
reg_log_write(NV_PFIFO, addr, val);
}
uint64_t pfifo_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PFIFO, addr, 0);
return 0;
}
#endif
const struct NV2ABlockInfo blocktable[] = {
#define ENTRY(NAME, OFFSET, SIZE, RDFUNC, WRFUNC) \
[ NV_ ## NAME ] = { \
.name = #NAME, .offset = OFFSET, .size = SIZE, \
.ops = { .read = RDFUNC, .write = WRFUNC }, \
}, \
ENTRY(PMC, 0x000000, 0x001000, pmc_read, pmc_write)
ENTRY(PBUS, 0x001000, 0x001000, pbus_read, pbus_write)
ENTRY(PFIFO, 0x002000, 0x002000, pfifo_read, pfifo_write)
ENTRY(PRMA, 0x007000, 0x001000, prma_read, prma_write)
ENTRY(PVIDEO, 0x008000, 0x001000, pvideo_read, pvideo_write)
ENTRY(PTIMER, 0x009000, 0x001000, ptimer_read, ptimer_write)
ENTRY(PCOUNTER, 0x00a000, 0x001000, pcounter_read, pcounter_write)
ENTRY(PVPE, 0x00b000, 0x001000, pvpe_read, pvpe_write)
ENTRY(PTV, 0x00d000, 0x001000, ptv_read, ptv_write)
ENTRY(PRMFB, 0x0a0000, 0x020000, prmfb_read, prmfb_write)
ENTRY(PRMVIO, 0x0c0000, 0x001000, prmvio_read, prmvio_write)
ENTRY(PFB, 0x100000, 0x001000, pfb_read, pfb_write)
ENTRY(PSTRAPS, 0x101000, 0x001000, pstraps_read, pstraps_write)
ENTRY(PGRAPH, 0x400000, 0x002000, pgraph_read, pgraph_write)
ENTRY(PCRTC, 0x600000, 0x001000, pcrtc_read, pcrtc_write)
ENTRY(PRMCIO, 0x601000, 0x001000, prmcio_read, prmcio_write)
ENTRY(PRAMDAC, 0x680000, 0x001000, pramdac_read, pramdac_write)
ENTRY(PRMDIO, 0x681000, 0x001000, prmdio_read, prmdio_write)
// ENTRY(PRAMIN, 0x700000, 0x100000, pramin_read, pramin_write)
ENTRY(USER, 0x800000, 0x800000, user_read, user_write)
#undef ENTRY
};
const int blocktable_len = ARRAY_SIZE(blocktable);
// static const char* nv2a_reg_names[] = {};
void reg_log_read(int block, hwaddr addr, uint64_t val) {
if (blocktable[block].name) {
// hwaddr naddr = blocktable[block].offset + addr;
// if (naddr < ARRAY_SIZE(nv2a_reg_names) && nv2a_reg_names[naddr]) {
// NV2A_DPRINTF("%s: read [%s] -> 0x%" PRIx64 "\n",
// blocktable[block].name, nv2a_reg_names[naddr], val);
// } else {
NV2A_DPRINTF("%s: read [%" HWADDR_PRIx "] -> 0x%" PRIx64 "\n",
blocktable[block].name, addr, val);
// }
} else {
NV2A_DPRINTF("(%d?): read [%" HWADDR_PRIx "] -> 0x%" PRIx64 "\n",
block, addr, val);
}
}
void reg_log_write(int block, hwaddr addr, uint64_t val) {
if (blocktable[block].name) {
// hwaddr naddr = blocktable[block].offset + addr;
// if (naddr < ARRAY_SIZE(nv2a_reg_names) && nv2a_reg_names[naddr]) {
// NV2A_DPRINTF("%s: [%s] = 0x%" PRIx64 "\n",
// blocktable[block].name, nv2a_reg_names[naddr], val);
// } else {
NV2A_DPRINTF("%s: [%" HWADDR_PRIx "] = 0x%" PRIx64 "\n",
blocktable[block].name, addr, val);
// }
} else {
NV2A_DPRINTF("(%d?): [%" HWADDR_PRIx "] = 0x%" PRIx64 "\n",
block, addr, val);
}
}
#if 0
/* FIXME: Probably totally wrong */
static inline unsigned int rgb_to_pixel8(unsigned int r, unsigned int g,
unsigned int b)
{
return ((r >> 5) << 5) | ((g >> 5) << 2) | (b >> 6);
}
static inline unsigned int rgb_to_pixel16(unsigned int r, unsigned int g,
unsigned int b)
{
return ((r >> 3) << 11) | ((g >> 2) << 5) | (b >> 3);
}
static inline unsigned int rgb_to_pixel32(unsigned int r, unsigned int g,
unsigned int b)
{
return (r << 16) | (g << 8) | b;
}
static void nv2a_overlay_draw_line(VGACommonState *vga, uint8_t *line, int y)
{
NV2A_DPRINTF("nv2a_overlay_draw_line\n");
NV2AState *d = container_of(vga, NV2AState, vga);
DisplaySurface *surface = qemu_console_surface(d->vga.con);
int surf_bpp = surface_bytes_per_pixel(surface);
int surf_width = surface_width(surface);
if (!(d->pvideo.regs[NV_PVIDEO_BUFFER] & NV_PVIDEO_BUFFER_0_USE)) return;
hwaddr base = d->pvideo.regs[NV_PVIDEO_BASE];
hwaddr limit = d->pvideo.regs[NV_PVIDEO_LIMIT];
hwaddr offset = d->pvideo.regs[NV_PVIDEO_OFFSET];
int in_width = GET_MASK(d->pvideo.regs[NV_PVIDEO_SIZE_IN],
NV_PVIDEO_SIZE_IN_WIDTH);
int in_height = GET_MASK(d->pvideo.regs[NV_PVIDEO_SIZE_IN],
NV_PVIDEO_SIZE_IN_HEIGHT);
int in_s = GET_MASK(d->pvideo.regs[NV_PVIDEO_POINT_IN],
NV_PVIDEO_POINT_IN_S);
int in_t = GET_MASK(d->pvideo.regs[NV_PVIDEO_POINT_IN],
NV_PVIDEO_POINT_IN_T);
int in_pitch = GET_MASK(d->pvideo.regs[NV_PVIDEO_FORMAT],
NV_PVIDEO_FORMAT_PITCH);
int in_color = GET_MASK(d->pvideo.regs[NV_PVIDEO_FORMAT],
NV_PVIDEO_FORMAT_COLOR);
// TODO: support other color formats
assert(in_color == NV_PVIDEO_FORMAT_COLOR_LE_CR8YB8CB8YA8);
int out_width = GET_MASK(d->pvideo.regs[NV_PVIDEO_SIZE_OUT],
NV_PVIDEO_SIZE_OUT_WIDTH);
int out_height = GET_MASK(d->pvideo.regs[NV_PVIDEO_SIZE_OUT],
NV_PVIDEO_SIZE_OUT_HEIGHT);
int out_x = GET_MASK(d->pvideo.regs[NV_PVIDEO_POINT_OUT],
NV_PVIDEO_POINT_OUT_X);
int out_y = GET_MASK(d->pvideo.regs[NV_PVIDEO_POINT_OUT],
NV_PVIDEO_POINT_OUT_Y);
if (y < out_y || y >= out_y + out_height) return;
// TODO: scaling, color keys
int in_y = y - out_y;
if (in_y >= in_height) return;
assert(offset + in_pitch * (in_y + 1) <= limit);
uint8_t *in_line = d->vram_ptr + base + offset + in_pitch * in_y;
int x;
for (x=0; x<out_width; x++) {
int ox = out_x + x;
if (ox >= surf_width) break;
int ix = in_s + x;
if (ix >= in_width) break;
uint8_t r,g,b;
convert_yuy2_to_rgb(in_line, ix, &r, &g, &b);
// unsigned int pixel = vga->rgb_to_pixel(r, g, b);
switch (surf_bpp) {
case 1:
((uint8_t*)line)[ox] = (uint8_t)rgb_to_pixel8(r,g,b);
break;
case 2:
((uint16_t*)line)[ox] = (uint16_t)rgb_to_pixel16(r,g,b);
break;
case 4:
((uint32_t*)line)[ox] = (uint32_t)rgb_to_pixel32(r,g,b);
break;
default:
assert(false);
break;
}
}
}
#endif
static int nv2a_get_bpp(VGACommonState *s)
{
if ((s->cr[0x28] & 3) == 3) {
return 32;
}
return (s->cr[0x28] & 3) * 8;
}
static void nv2a_get_offsets(VGACommonState *s,
uint32_t *pline_offset,
uint32_t *pstart_addr,
uint32_t *pline_compare)
{
NV2AState *d = container_of(s, NV2AState, vga);
uint32_t start_addr, line_offset, line_compare;
line_offset = s->cr[0x13]
| ((s->cr[0x19] & 0xe0) << 3)
| ((s->cr[0x25] & 0x20) << 6);
line_offset <<= 3;
*pline_offset = line_offset;
start_addr = d->pcrtc.start / 4;
*pstart_addr = start_addr;
line_compare = s->cr[VGA_CRTC_LINE_COMPARE] |
((s->cr[VGA_CRTC_OVERFLOW] & 0x10) << 4) |
((s->cr[VGA_CRTC_MAX_SCAN] & 0x40) << 3);
*pline_compare = line_compare;
}
static void nv2a_vga_gfx_update(void *opaque)
{
VGACommonState *vga = opaque;
vga->hw_ops->gfx_update(vga);
NV2AState *d = container_of(vga, NV2AState, vga);
d->pcrtc.pending_interrupts |= NV_PCRTC_INTR_0_VBLANK;
update_irq(d);
}
static void nv2a_init_memory(NV2AState *d, MemoryRegion *ram)
{
/* xbox is UMA - vram *is* ram */
d->vram = ram;
/* PCI exposed vram */
memory_region_init_alias(&d->vram_pci, OBJECT(d), "nv2a-vram-pci", d->vram,
0, memory_region_size(d->vram));
pci_register_bar(&d->dev, 1, PCI_BASE_ADDRESS_MEM_PREFETCH, &d->vram_pci);
/* RAMIN - should be in vram somewhere, but not quite sure where atm */
memory_region_init_ram(&d->ramin, OBJECT(d), "nv2a-ramin", 0x100000, &error_fatal);
/* memory_region_init_alias(&d->ramin, "nv2a-ramin", &d->vram,
memory_region_size(d->vram) - 0x100000,
0x100000); */
memory_region_add_subregion(&d->mmio, 0x700000, &d->ramin);
d->vram_ptr = memory_region_get_ram_ptr(d->vram);
d->ramin_ptr = memory_region_get_ram_ptr(&d->ramin);
memory_region_set_log(d->vram, true, DIRTY_MEMORY_NV2A);
memory_region_set_dirty(d->vram, 0, memory_region_size(d->vram));
/* hacky. swap out vga's vram */
memory_region_destroy(&d->vga.vram);
// memory_region_unref(&d->vga.vram); // FIXME: Is ths right?
memory_region_init_alias(&d->vga.vram, OBJECT(d), "vga.vram",
d->vram, 0, memory_region_size(d->vram));
d->vga.vram_ptr = memory_region_get_ram_ptr(&d->vga.vram);
vga_dirty_log_start(&d->vga);
pgraph_init(d);
/* fire up puller */
qemu_thread_create(&d->pfifo.puller_thread, "nv2a.puller_thread",
pfifo_puller_thread,
d, QEMU_THREAD_JOINABLE);
}
static void nv2a_realize(PCIDevice *dev, Error **errp)
{
int i;
NV2AState *d;
d = NV2A_DEVICE(dev);
dev->config[PCI_INTERRUPT_PIN] = 0x01;
d->pcrtc.start = 0;
d->pramdac.core_clock_coeff = 0x00011c01; /* 189MHz...? */
d->pramdac.core_clock_freq = 189000000;
d->pramdac.memory_clock_coeff = 0;
d->pramdac.video_clock_coeff = 0x0003C20D; /* 25182Khz...? */
/* legacy VGA shit */
VGACommonState *vga = &d->vga;
vga_common_reset(vga);
vga->vram_size_mb = 64;
/* seems to start in color mode */
vga->msr = VGA_MIS_COLOR;
vga_common_init(vga, OBJECT(dev), false); // FIXME: true or false? idk
vga->get_bpp = nv2a_get_bpp;
vga->get_offsets = nv2a_get_offsets;
// vga->overlay_draw_line = nv2a_overlay_draw_line;
d->hw_ops = *vga->hw_ops;
d->hw_ops.gfx_update = nv2a_vga_gfx_update;
vga->con = graphic_console_init(DEVICE(dev), 0, &d->hw_ops, vga);
/* mmio */
memory_region_init(&d->mmio, OBJECT(dev), "nv2a-mmio", 0x1000000);
pci_register_bar(&d->dev, 0, PCI_BASE_ADDRESS_SPACE_MEMORY, &d->mmio);
for (i=0; i<ARRAY_SIZE(blocktable); i++) {
if (!blocktable[i].name) continue;
memory_region_init_io(&d->block_mmio[i], OBJECT(dev),
&blocktable[i].ops, d,
blocktable[i].name, blocktable[i].size);
memory_region_add_subregion(&d->mmio, blocktable[i].offset,
&d->block_mmio[i]);
}
/* init fifo cache1 */
qemu_mutex_init(&d->pfifo.cache1.cache_lock);
qemu_cond_init(&d->pfifo.cache1.cache_cond);
QSIMPLEQ_INIT(&d->pfifo.cache1.cache);
QSIMPLEQ_INIT(&d->pfifo.cache1.working_cache);
}
static void nv2a_exitfn(PCIDevice *dev)
{
NV2AState *d;
d = NV2A_DEVICE(dev);
d->exiting = true;
qemu_cond_signal(&d->pfifo.cache1.cache_cond);
qemu_thread_join(&d->pfifo.puller_thread);
qemu_mutex_destroy(&d->pfifo.cache1.cache_lock);
qemu_cond_destroy(&d->pfifo.cache1.cache_cond);
pgraph_destroy(&d->pgraph);
}
static void nv2a_class_init(ObjectClass *klass, void *data)
{
DeviceClass *dc = DEVICE_CLASS(klass);
PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
k->vendor_id = PCI_VENDOR_ID_NVIDIA;
k->device_id = PCI_DEVICE_ID_NVIDIA_GEFORCE_NV2A;
k->revision = 161;
k->class_id = PCI_CLASS_DISPLAY_3D;
k->realize = nv2a_realize;
k->exit = nv2a_exitfn;
dc->desc = "GeForce NV2A Integrated Graphics";
}
static const TypeInfo nv2a_info = {
.name = "nv2a",
.parent = TYPE_PCI_DEVICE,
.instance_size = sizeof(NV2AState),
.class_init = nv2a_class_init,
.interfaces = (InterfaceInfo[]) {
{ INTERFACE_CONVENTIONAL_PCI_DEVICE },
{ },
},
};
static void nv2a_register(void)
{
type_register_static(&nv2a_info);
}
type_init(nv2a_register);
void nv2a_init(PCIBus *bus, int devfn, MemoryRegion *ram)
{
PCIDevice *dev = pci_create_simple(bus, devfn, "nv2a");
NV2AState *d = NV2A_DEVICE(dev);
nv2a_init_memory(d, ram);
}

446
hw/xbox/nv2a/nv2a.h Normal file
View File

@ -0,0 +1,446 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#ifndef HW_NV2A_H
#define HW_NV2A_H
#include "hw/hw.h"
#include "hw/i386/pc.h"
#include "ui/console.h"
#include "hw/pci/pci.h"
#include "ui/console.h"
#include "hw/display/vga.h"
#include "hw/display/vga_int.h"
#include "qemu/thread.h"
#include "qapi/qmp/qstring.h"
#include "cpu.h"
#include "g-lru-cache.h"
#include "swizzle.h"
#include "nv2a_shaders.h"
#include "nv2a_debug.h"
#include "nv2a_int.h"
#include "gl/gloffscreen.h"
#include "gl/glextensions.h"
#define USE_TEXTURE_CACHE
#define GET_MASK(v, mask) (((v) & (mask)) >> (ffs(mask)-1))
#define SET_MASK(v, mask, val) ({ \
const unsigned int __val = (val); \
const unsigned int __mask = (mask); \
(v) &= ~(__mask); \
(v) |= ((__val) << (ffs(__mask)-1)) & (__mask); \
})
#define CASE_4(v, step) \
case (v): \
case (v)+(step): \
case (v)+(step)*2: \
case (v)+(step)*3
#define NV2A_DEVICE(obj) \
OBJECT_CHECK(NV2AState, (obj), "nv2a")
void reg_log_read(int block, hwaddr addr, uint64_t val);
void reg_log_write(int block, hwaddr addr, uint64_t val);
enum FifoMode {
FIFO_PIO = 0,
FIFO_DMA = 1,
};
enum FIFOEngine {
ENGINE_SOFTWARE = 0,
ENGINE_GRAPHICS = 1,
ENGINE_DVD = 2,
};
typedef struct DMAObject {
unsigned int dma_class;
unsigned int dma_target;
hwaddr address;
hwaddr limit;
} DMAObject;
typedef struct VertexAttribute {
bool dma_select;
hwaddr offset;
/* inline arrays are packed in order?
* Need to pass the offset to converted attributes */
unsigned int inline_array_offset;
float inline_value[4];
unsigned int format;
unsigned int size; /* size of the data type */
unsigned int count; /* number of components */
uint32_t stride;
bool needs_conversion;
uint8_t *converted_buffer;
unsigned int converted_elements;
unsigned int converted_size;
unsigned int converted_count;
float *inline_buffer;
GLint gl_count;
GLenum gl_type;
GLboolean gl_normalize;
GLuint gl_converted_buffer;
GLuint gl_inline_buffer;
} VertexAttribute;
typedef struct Surface {
bool draw_dirty;
bool buffer_dirty;
bool write_enabled_cache;
unsigned int pitch;
hwaddr offset;
} Surface;
typedef struct SurfaceShape {
unsigned int z_format;
unsigned int color_format;
unsigned int zeta_format;
unsigned int log_width, log_height;
unsigned int clip_x, clip_y;
unsigned int clip_width, clip_height;
unsigned int anti_aliasing;
} SurfaceShape;
typedef struct TextureShape {
bool cubemap;
unsigned int dimensionality;
unsigned int color_format;
unsigned int levels;
unsigned int width, height, depth;
unsigned int min_mipmap_level, max_mipmap_level;
unsigned int pitch;
} TextureShape;
typedef struct TextureKey {
TextureShape state;
uint64_t data_hash;
uint8_t* texture_data;
uint8_t* palette_data;
} TextureKey;
typedef struct TextureBinding {
GLenum gl_target;
GLuint gl_texture;
unsigned int refcnt;
} TextureBinding;
typedef struct KelvinState {
hwaddr dma_notifies;
hwaddr dma_state;
hwaddr dma_semaphore;
unsigned int semaphore_offset;
} KelvinState;
typedef struct ContextSurfaces2DState {
hwaddr dma_image_source;
hwaddr dma_image_dest;
unsigned int color_format;
unsigned int source_pitch, dest_pitch;
hwaddr source_offset, dest_offset;
} ContextSurfaces2DState;
typedef struct ImageBlitState {
hwaddr context_surfaces;
unsigned int operation;
unsigned int in_x, in_y;
unsigned int out_x, out_y;
unsigned int width, height;
} ImageBlitState;
typedef struct GraphicsObject {
uint8_t graphics_class;
union {
ContextSurfaces2DState context_surfaces_2d;
ImageBlitState image_blit;
KelvinState kelvin;
} data;
} GraphicsObject;
typedef struct GraphicsSubchannel {
hwaddr object_instance;
GraphicsObject object;
uint32_t object_cache[5];
} GraphicsSubchannel;
typedef struct GraphicsContext {
bool channel_3d;
unsigned int subchannel;
} GraphicsContext;
typedef struct PGRAPHState {
QemuMutex lock;
uint32_t pending_interrupts;
uint32_t enabled_interrupts;
QemuCond interrupt_cond;
hwaddr context_table;
hwaddr context_address;
unsigned int trapped_method;
unsigned int trapped_subchannel;
unsigned int trapped_channel_id;
uint32_t trapped_data[2];
uint32_t notify_source;
bool fifo_access;
QemuCond fifo_access_cond;
QemuCond flip_3d;
unsigned int channel_id;
bool channel_valid;
GraphicsContext context[NV2A_NUM_CHANNELS];
hwaddr dma_color, dma_zeta;
Surface surface_color, surface_zeta;
unsigned int surface_type;
SurfaceShape surface_shape;
SurfaceShape last_surface_shape;
hwaddr dma_a, dma_b;
GLruCache *texture_cache;
bool texture_dirty[NV2A_MAX_TEXTURES];
TextureBinding *texture_binding[NV2A_MAX_TEXTURES];
GHashTable *shader_cache;
ShaderBinding *shader_binding;
bool texture_matrix_enable[NV2A_MAX_TEXTURES];
/* FIXME: Move to NV_PGRAPH_BUMPMAT... */
float bump_env_matrix[NV2A_MAX_TEXTURES-1][4]; /* 3 allowed stages with 2x2 matrix each */
GloContext *gl_context;
GLuint gl_framebuffer;
GLuint gl_color_buffer, gl_zeta_buffer;
GraphicsSubchannel subchannel_data[NV2A_NUM_SUBCHANNELS];
hwaddr dma_report;
hwaddr report_offset;
bool zpass_pixel_count_enable;
unsigned int zpass_pixel_count_result;
unsigned int gl_zpass_pixel_count_query_count;
GLuint* gl_zpass_pixel_count_queries;
hwaddr dma_vertex_a, dma_vertex_b;
unsigned int primitive_mode;
bool enable_vertex_program_write;
uint32_t program_data[NV2A_MAX_TRANSFORM_PROGRAM_LENGTH][VSH_TOKEN_SIZE];
uint32_t vsh_constants[NV2A_VERTEXSHADER_CONSTANTS][4];
bool vsh_constants_dirty[NV2A_VERTEXSHADER_CONSTANTS];
/* lighting constant arrays */
uint32_t ltctxa[NV2A_LTCTXA_COUNT][4];
bool ltctxa_dirty[NV2A_LTCTXA_COUNT];
uint32_t ltctxb[NV2A_LTCTXB_COUNT][4];
bool ltctxb_dirty[NV2A_LTCTXB_COUNT];
uint32_t ltc1[NV2A_LTC1_COUNT][4];
bool ltc1_dirty[NV2A_LTC1_COUNT];
// should figure out where these are in lighting context
float light_infinite_half_vector[NV2A_MAX_LIGHTS][3];
float light_infinite_direction[NV2A_MAX_LIGHTS][3];
float light_local_position[NV2A_MAX_LIGHTS][3];
float light_local_attenuation[NV2A_MAX_LIGHTS][3];
VertexAttribute vertex_attributes[NV2A_VERTEXSHADER_ATTRIBUTES];
unsigned int inline_array_length;
uint32_t inline_array[NV2A_MAX_BATCH_LENGTH];
GLuint gl_inline_array_buffer;
unsigned int inline_elements_length;
uint32_t inline_elements[NV2A_MAX_BATCH_LENGTH];
unsigned int inline_buffer_length;
unsigned int draw_arrays_length;
unsigned int draw_arrays_max_count;
/* FIXME: Unknown size, possibly endless, 1000 will do for now */
GLint gl_draw_arrays_start[1000];
GLsizei gl_draw_arrays_count[1000];
GLuint gl_element_buffer;
GLuint gl_memory_buffer;
GLuint gl_vertex_array;
uint32_t regs[0x2000];
} PGRAPHState;
typedef struct CacheEntry {
QSIMPLEQ_ENTRY(CacheEntry) entry;
unsigned int method : 14;
unsigned int subchannel : 3;
bool nonincreasing;
uint32_t parameter;
} CacheEntry;
typedef struct Cache1State {
unsigned int channel_id;
enum FifoMode mode;
/* Pusher state */
bool push_enabled;
bool dma_push_enabled;
bool dma_push_suspended;
hwaddr dma_instance;
bool method_nonincreasing;
unsigned int method : 14;
unsigned int subchannel : 3;
unsigned int method_count : 24;
uint32_t dcount;
bool subroutine_active;
hwaddr subroutine_return;
hwaddr get_jmp_shadow;
uint32_t rsvd_shadow;
uint32_t data_shadow;
uint32_t error;
bool pull_enabled;
enum FIFOEngine bound_engines[NV2A_NUM_SUBCHANNELS];
enum FIFOEngine last_engine;
/* The actual command queue */
QemuMutex cache_lock;
QemuCond cache_cond;
QSIMPLEQ_HEAD(, CacheEntry) cache;
QSIMPLEQ_HEAD(, CacheEntry) working_cache;
} Cache1State;
typedef struct ChannelControl {
hwaddr dma_put;
hwaddr dma_get;
uint32_t ref;
} ChannelControl;
typedef struct NV2AState {
PCIDevice dev;
qemu_irq irq;
bool exiting;
VGACommonState vga;
GraphicHwOps hw_ops;
QEMUTimer *vblank_timer;
MemoryRegion *vram;
MemoryRegion vram_pci;
uint8_t *vram_ptr;
MemoryRegion ramin;
uint8_t *ramin_ptr;
MemoryRegion mmio;
MemoryRegion block_mmio[NV_NUM_BLOCKS];
struct {
uint32_t pending_interrupts;
uint32_t enabled_interrupts;
} pmc;
struct {
QemuThread puller_thread;
uint32_t pending_interrupts;
uint32_t enabled_interrupts;
Cache1State cache1;
uint32_t regs[0x2000];
} pfifo;
struct {
uint32_t regs[0x1000];
} pvideo;
struct {
uint32_t pending_interrupts;
uint32_t enabled_interrupts;
uint32_t numerator;
uint32_t denominator;
uint32_t alarm_time;
} ptimer;
struct {
uint32_t regs[0x1000];
} pfb;
struct PGRAPHState pgraph;
struct {
uint32_t pending_interrupts;
uint32_t enabled_interrupts;
hwaddr start;
} pcrtc;
struct {
uint32_t core_clock_coeff;
uint64_t core_clock_freq;
uint32_t memory_clock_coeff;
uint32_t video_clock_coeff;
} pramdac;
struct {
ChannelControl channel_control[NV2A_NUM_CHANNELS];
} user;
} NV2AState;
typedef struct NV2ABlockInfo {
const char* name;
hwaddr offset;
uint64_t size;
MemoryRegionOps ops;
} NV2ABlockInfo;
extern const struct NV2ABlockInfo blocktable[];
extern const int blocktable_len;
void pgraph_init(NV2AState *d);
void *pfifo_puller_thread(void *opaque);
void pgraph_destroy(PGRAPHState *pg);
void update_irq(NV2AState *d);
#endif

5
hw/xbox/nv2a_debug.c → hw/xbox/nv2a/nv2a_debug.c
View File

@ -18,14 +18,15 @@
* along with this program; if not, see <http://www.gnu.org/licenses/>. * along with this program; if not, see <http://www.gnu.org/licenses/>.
*/ */
#include "hw/xbox/nv2a_debug.h"
#ifdef DEBUG_NV2A_GL #ifdef DEBUG_NV2A_GL
#include "qemu/osdep.h"
#include <stdio.h> #include <stdio.h>
#include <stdarg.h> #include <stdarg.h>
#include <assert.h> #include <assert.h>
#include "nv2a_debug.h"
#include "gl/glextensions.h" #include "gl/glextensions.h"
void gl_debug_message(bool cc, const char *fmt, ...) void gl_debug_message(bool cc, const char *fmt, ...)

0
hw/xbox/nv2a_debug.h → hw/xbox/nv2a/nv2a_debug.h
View File

0
hw/xbox/nv2a_int.h → hw/xbox/nv2a/nv2a_int.h
View File

59
hw/xbox/nv2a/nv2a_pbus.c Normal file
View File

@ -0,0 +1,59 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/* PBUS - bus control */
uint64_t pbus_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = opaque;
uint64_t r = 0;
switch (addr) {
case NV_PBUS_PCI_NV_0:
r = pci_get_long(d->dev.config + PCI_VENDOR_ID);
break;
case NV_PBUS_PCI_NV_1:
r = pci_get_long(d->dev.config + PCI_COMMAND);
break;
case NV_PBUS_PCI_NV_2:
r = pci_get_long(d->dev.config + PCI_CLASS_REVISION);
break;
default:
break;
}
reg_log_read(NV_PBUS, addr, r);
return r;
}
void pbus_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = opaque;
reg_log_write(NV_PBUS, addr, val);
switch (addr) {
case NV_PBUS_PCI_NV_1:
pci_set_long(d->dev.config + PCI_COMMAND, val);
break;
default:
break;
}
}

72
hw/xbox/nv2a/nv2a_pcrtc.c Normal file
View File

@ -0,0 +1,72 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
uint64_t pcrtc_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
uint64_t r = 0;
switch (addr) {
case NV_PCRTC_INTR_0:
r = d->pcrtc.pending_interrupts;
break;
case NV_PCRTC_INTR_EN_0:
r = d->pcrtc.enabled_interrupts;
break;
case NV_PCRTC_START:
r = d->pcrtc.start;
break;
default:
break;
}
reg_log_read(NV_PCRTC, addr, r);
return r;
}
void pcrtc_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
reg_log_write(NV_PCRTC, addr, val);
switch (addr) {
case NV_PCRTC_INTR_0:
d->pcrtc.pending_interrupts &= ~val;
update_irq(d);
break;
case NV_PCRTC_INTR_EN_0:
d->pcrtc.enabled_interrupts = val;
update_irq(d);
break;
case NV_PCRTC_START:
val &= 0x07FFFFFF;
// assert(val < memory_region_size(d->vram));
d->pcrtc.start = val;
NV2A_DPRINTF("PCRTC_START - %x %x %x %x\n",
d->vram_ptr[val+64], d->vram_ptr[val+64+1],
d->vram_ptr[val+64+2], d->vram_ptr[val+64+3]);
break;
default:
break;
}
}

58
hw/xbox/nv2a/nv2a_pfb.c Normal file
View File

@ -0,0 +1,58 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
uint64_t pfb_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
uint64_t r = 0;
switch (addr) {
case NV_PFB_CFG0:
/* 3-4 memory partitions. The debug bios checks this. */
r = 3;
break;
case NV_PFB_CSTATUS:
r = memory_region_size(d->vram);
break;
case NV_PFB_WBC:
r = 0; /* Flush not pending. */
break;
default:
r = d->pfb.regs[addr];
break;
}
reg_log_read(NV_PFB, addr, r);
return r;
}
void pfb_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
reg_log_write(NV_PFB, addr, val);
switch (addr) {
default:
d->pfb.regs[addr] = val;
break;
}
}

513
hw/xbox/nv2a/nv2a_pfifo.c Normal file
View File

@ -0,0 +1,513 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
typedef struct RAMHTEntry {
uint32_t handle;
hwaddr instance;
enum FIFOEngine engine;
unsigned int channel_id : 5;
bool valid;
} RAMHTEntry;
static void pfifo_run_pusher(NV2AState *d);
void *pfifo_puller_thread(void *opaque);
static uint32_t ramht_hash(NV2AState *d, uint32_t handle);
static RAMHTEntry ramht_lookup(NV2AState *d, uint32_t handle);
/* PFIFO - MMIO and DMA FIFO submission to PGRAPH and VPE */
uint64_t pfifo_read(void *opaque, hwaddr addr, unsigned int size)
{
int i;
NV2AState *d = (NV2AState *)opaque;
uint64_t r = 0;
switch (addr) {
case NV_PFIFO_INTR_0:
r = d->pfifo.pending_interrupts;
break;
case NV_PFIFO_INTR_EN_0:
r = d->pfifo.enabled_interrupts;
break;
case NV_PFIFO_RUNOUT_STATUS:
r = NV_PFIFO_RUNOUT_STATUS_LOW_MARK; /* low mark empty */
break;
case NV_PFIFO_CACHE1_PUSH0:
r = d->pfifo.cache1.push_enabled;
break;
case NV_PFIFO_CACHE1_PUSH1:
SET_MASK(r, NV_PFIFO_CACHE1_PUSH1_CHID, d->pfifo.cache1.channel_id);
SET_MASK(r, NV_PFIFO_CACHE1_PUSH1_MODE, d->pfifo.cache1.mode);
break;
case NV_PFIFO_CACHE1_STATUS:
qemu_mutex_lock(&d->pfifo.cache1.cache_lock);
if (QSIMPLEQ_EMPTY(&d->pfifo.cache1.cache)) {
r |= NV_PFIFO_CACHE1_STATUS_LOW_MARK; /* low mark empty */
}
qemu_mutex_unlock(&d->pfifo.cache1.cache_lock);
break;
case NV_PFIFO_CACHE1_DMA_PUSH:
SET_MASK(r, NV_PFIFO_CACHE1_DMA_PUSH_ACCESS,
d->pfifo.cache1.dma_push_enabled);
SET_MASK(r, NV_PFIFO_CACHE1_DMA_PUSH_STATUS,
d->pfifo.cache1.dma_push_suspended);
SET_MASK(r, NV_PFIFO_CACHE1_DMA_PUSH_BUFFER, 1); /* buffer emoty */
break;
case NV_PFIFO_CACHE1_DMA_STATE:
SET_MASK(r, NV_PFIFO_CACHE1_DMA_STATE_METHOD_TYPE,
d->pfifo.cache1.method_nonincreasing);
SET_MASK(r, NV_PFIFO_CACHE1_DMA_STATE_METHOD,
d->pfifo.cache1.method >> 2);
SET_MASK(r, NV_PFIFO_CACHE1_DMA_STATE_SUBCHANNEL,
d->pfifo.cache1.subchannel);
SET_MASK(r, NV_PFIFO_CACHE1_DMA_STATE_METHOD_COUNT,
d->pfifo.cache1.method_count);
SET_MASK(r, NV_PFIFO_CACHE1_DMA_STATE_ERROR,
d->pfifo.cache1.error);
break;
case NV_PFIFO_CACHE1_DMA_INSTANCE:
SET_MASK(r, NV_PFIFO_CACHE1_DMA_INSTANCE_ADDRESS,
d->pfifo.cache1.dma_instance >> 4);
break;
case NV_PFIFO_CACHE1_DMA_PUT:
r = d->user.channel_control[d->pfifo.cache1.channel_id].dma_put;
break;
case NV_PFIFO_CACHE1_DMA_GET:
r = d->user.channel_control[d->pfifo.cache1.channel_id].dma_get;
break;
case NV_PFIFO_CACHE1_DMA_SUBROUTINE:
r = d->pfifo.cache1.subroutine_return
| d->pfifo.cache1.subroutine_active;
break;
case NV_PFIFO_CACHE1_PULL0:
qemu_mutex_lock(&d->pfifo.cache1.cache_lock);
r = d->pfifo.cache1.pull_enabled;
qemu_mutex_unlock(&d->pfifo.cache1.cache_lock);
break;
case NV_PFIFO_CACHE1_ENGINE:
qemu_mutex_lock(&d->pfifo.cache1.cache_lock);
for (i=0; i<NV2A_NUM_SUBCHANNELS; i++) {
r |= d->pfifo.cache1.bound_engines[i] << (i*2);
}
qemu_mutex_unlock(&d->pfifo.cache1.cache_lock);
break;
case NV_PFIFO_CACHE1_DMA_DCOUNT:
r = d->pfifo.cache1.dcount;
break;
case NV_PFIFO_CACHE1_DMA_GET_JMP_SHADOW:
r = d->pfifo.cache1.get_jmp_shadow;
break;
case NV_PFIFO_CACHE1_DMA_RSVD_SHADOW:
r = d->pfifo.cache1.rsvd_shadow;
break;
case NV_PFIFO_CACHE1_DMA_DATA_SHADOW:
r = d->pfifo.cache1.data_shadow;
break;
default:
r = d->pfifo.regs[addr];
break;
}
reg_log_read(NV_PFIFO, addr, r);
return r;
}
void pfifo_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
int i;
NV2AState *d = (NV2AState *)opaque;
reg_log_write(NV_PFIFO, addr, val);
switch (addr) {
case NV_PFIFO_INTR_0:
d->pfifo.pending_interrupts &= ~val;
update_irq(d);
break;
case NV_PFIFO_INTR_EN_0:
d->pfifo.enabled_interrupts = val;
update_irq(d);
break;
case NV_PFIFO_CACHE1_PUSH0:
d->pfifo.cache1.push_enabled = val & NV_PFIFO_CACHE1_PUSH0_ACCESS;
break;
case NV_PFIFO_CACHE1_PUSH1:
d->pfifo.cache1.channel_id = GET_MASK(val, NV_PFIFO_CACHE1_PUSH1_CHID);
d->pfifo.cache1.mode = (enum FifoMode)GET_MASK(val, NV_PFIFO_CACHE1_PUSH1_MODE);
assert(d->pfifo.cache1.channel_id < NV2A_NUM_CHANNELS);
break;
case NV_PFIFO_CACHE1_DMA_PUSH:
d->pfifo.cache1.dma_push_enabled =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_PUSH_ACCESS);
if (d->pfifo.cache1.dma_push_suspended
&& !GET_MASK(val, NV_PFIFO_CACHE1_DMA_PUSH_STATUS)) {
d->pfifo.cache1.dma_push_suspended = false;
pfifo_run_pusher(d);
}
d->pfifo.cache1.dma_push_suspended =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_PUSH_STATUS);
break;
case NV_PFIFO_CACHE1_DMA_STATE:
d->pfifo.cache1.method_nonincreasing =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_STATE_METHOD_TYPE);
d->pfifo.cache1.method =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_STATE_METHOD) << 2;
d->pfifo.cache1.subchannel =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_STATE_SUBCHANNEL);
d->pfifo.cache1.method_count =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_STATE_METHOD_COUNT);
d->pfifo.cache1.error =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_STATE_ERROR);
break;
case NV_PFIFO_CACHE1_DMA_INSTANCE:
d->pfifo.cache1.dma_instance =
GET_MASK(val, NV_PFIFO_CACHE1_DMA_INSTANCE_ADDRESS) << 4;
break;
case NV_PFIFO_CACHE1_DMA_PUT:
d->user.channel_control[d->pfifo.cache1.channel_id].dma_put = val;
break;
case NV_PFIFO_CACHE1_DMA_GET:
d->user.channel_control[d->pfifo.cache1.channel_id].dma_get = val;
break;
case NV_PFIFO_CACHE1_DMA_SUBROUTINE:
d->pfifo.cache1.subroutine_return =
(val & NV_PFIFO_CACHE1_DMA_SUBROUTINE_RETURN_OFFSET);
d->pfifo.cache1.subroutine_active =
(val & NV_PFIFO_CACHE1_DMA_SUBROUTINE_STATE);
break;
case NV_PFIFO_CACHE1_PULL0:
qemu_mutex_lock(&d->pfifo.cache1.cache_lock);
if ((val & NV_PFIFO_CACHE1_PULL0_ACCESS)
&& !d->pfifo.cache1.pull_enabled) {
d->pfifo.cache1.pull_enabled = true;
/* the puller thread should wake up */
qemu_cond_signal(&d->pfifo.cache1.cache_cond);
} else if (!(val & NV_PFIFO_CACHE1_PULL0_ACCESS)
&& d->pfifo.cache1.pull_enabled) {
d->pfifo.cache1.pull_enabled = false;
}
qemu_mutex_unlock(&d->pfifo.cache1.cache_lock);
break;
case NV_PFIFO_CACHE1_ENGINE:
qemu_mutex_lock(&d->pfifo.cache1.cache_lock);
for (i=0; i<NV2A_NUM_SUBCHANNELS; i++) {
d->pfifo.cache1.bound_engines[i] = (enum FIFOEngine)((val >> (i*2)) & 3);
}
qemu_mutex_unlock(&d->pfifo.cache1.cache_lock);
break;
case NV_PFIFO_CACHE1_DMA_DCOUNT:
d->pfifo.cache1.dcount =
(val & NV_PFIFO_CACHE1_DMA_DCOUNT_VALUE);
break;
case NV_PFIFO_CACHE1_DMA_GET_JMP_SHADOW:
d->pfifo.cache1.get_jmp_shadow =
(val & NV_PFIFO_CACHE1_DMA_GET_JMP_SHADOW_OFFSET);
break;
case NV_PFIFO_CACHE1_DMA_RSVD_SHADOW:
d->pfifo.cache1.rsvd_shadow = val;
break;
case NV_PFIFO_CACHE1_DMA_DATA_SHADOW:
d->pfifo.cache1.data_shadow = val;
break;
default:
d->pfifo.regs[addr] = val;
break;
}
}
/* pusher should be fine to run from a mimo handler
* whenever's it's convenient */
static void pfifo_run_pusher(NV2AState *d) {
uint8_t channel_id;
ChannelControl *control;
Cache1State *state;
CacheEntry *command;
uint8_t *dma;
hwaddr dma_len;
uint32_t word;
/* TODO: How is cache1 selected? */
state = &d->pfifo.cache1;
channel_id = state->channel_id;
control = &d->user.channel_control[channel_id];
if (!state->push_enabled) return;
/* only handling DMA for now... */
/* Channel running DMA */
uint32_t channel_modes = d->pfifo.regs[NV_PFIFO_MODE];
assert(channel_modes & (1 << channel_id));
assert(state->mode == FIFO_DMA);
if (!state->dma_push_enabled) return;
if (state->dma_push_suspended) return;
/* We're running so there should be no pending errors... */
assert(state->error == NV_PFIFO_CACHE1_DMA_STATE_ERROR_NONE);
dma = (uint8_t*)nv_dma_map(d, state->dma_instance, &dma_len);
NV2A_DPRINTF("DMA pusher: max 0x%" HWADDR_PRIx ", 0x%" HWADDR_PRIx " - 0x%" HWADDR_PRIx "\n",
dma_len, control->dma_get, control->dma_put);
/* based on the convenient pseudocode in envytools */
while (control->dma_get != control->dma_put) {
if (control->dma_get >= dma_len) {
state->error = NV_PFIFO_CACHE1_DMA_STATE_ERROR_PROTECTION;
break;
}
word = ldl_le_p((uint32_t*)(dma + control->dma_get));
control->dma_get += 4;
if (state->method_count) {
/* data word of methods command */
state->data_shadow = word;
command = (CacheEntry*)g_malloc0(sizeof(CacheEntry));
command->method = state->method;
command->subchannel = state->subchannel;
command->nonincreasing = state->method_nonincreasing;
command->parameter = word;
qemu_mutex_lock(&state->cache_lock);
QSIMPLEQ_INSERT_TAIL(&state->cache, command, entry);
qemu_cond_signal(&state->cache_cond);
qemu_mutex_unlock(&state->cache_lock);
if (!state->method_nonincreasing) {
state->method += 4;
}
state->method_count--;
state->dcount++;
} else {
/* no command active - this is the first word of a new one */
state->rsvd_shadow = word;
/* match all forms */
if ((word & 0xe0000003) == 0x20000000) {
/* old jump */
state->get_jmp_shadow = control->dma_get;
control->dma_get = word & 0x1fffffff;
NV2A_DPRINTF("pb OLD_JMP 0x%" HWADDR_PRIx "\n", control->dma_get);
} else if ((word & 3) == 1) {
/* jump */
state->get_jmp_shadow = control->dma_get;
control->dma_get = word & 0xfffffffc;
NV2A_DPRINTF("pb JMP 0x%" HWADDR_PRIx "\n", control->dma_get);
} else if ((word & 3) == 2) {
/* call */
if (state->subroutine_active) {
state->error = NV_PFIFO_CACHE1_DMA_STATE_ERROR_CALL;
break;
}
state->subroutine_return = control->dma_get;
state->subroutine_active = true;
control->dma_get = word & 0xfffffffc;
NV2A_DPRINTF("pb CALL 0x%" HWADDR_PRIx "\n", control->dma_get);
} else if (word == 0x00020000) {
/* return */
if (!state->subroutine_active) {
state->error = NV_PFIFO_CACHE1_DMA_STATE_ERROR_RETURN;
break;
}
control->dma_get = state->subroutine_return;
state->subroutine_active = false;
NV2A_DPRINTF("pb RET 0x%" HWADDR_PRIx "\n", control->dma_get);
} else if ((word & 0xe0030003) == 0) {
/* increasing methods */
state->method = word & 0x1fff;
state->subchannel = (word >> 13) & 7;
state->method_count = (word >> 18) & 0x7ff;
state->method_nonincreasing = false;
state->dcount = 0;
} else if ((word & 0xe0030003) == 0x40000000) {
/* non-increasing methods */
state->method = word & 0x1fff;
state->subchannel = (word >> 13) & 7;
state->method_count = (word >> 18) & 0x7ff;
state->method_nonincreasing = true;
state->dcount = 0;
} else {
NV2A_DPRINTF("pb reserved cmd 0x%" HWADDR_PRIx " - 0x%x\n",
control->dma_get, word);
state->error = NV_PFIFO_CACHE1_DMA_STATE_ERROR_RESERVED_CMD;
break;
}
}
}
NV2A_DPRINTF("DMA pusher done: max 0x%" HWADDR_PRIx ", 0x%" HWADDR_PRIx " - 0x%" HWADDR_PRIx "\n",
dma_len, control->dma_get, control->dma_put);
if (state->error) {
NV2A_DPRINTF("pb error: %d\n", state->error);
assert(false);
state->dma_push_suspended = true;
d->pfifo.pending_interrupts |= NV_PFIFO_INTR_0_DMA_PUSHER;
update_irq(d);
}
}
void *pfifo_puller_thread(void *opaque)
{
NV2AState *d = (NV2AState*)opaque;
Cache1State *state = &d->pfifo.cache1;
glo_set_current(d->pgraph.gl_context);
while (true) {
qemu_mutex_lock(&state->cache_lock);
while (QSIMPLEQ_EMPTY(&state->cache) || !state->pull_enabled) {
qemu_cond_wait(&state->cache_cond, &state->cache_lock);
if (d->exiting) {
qemu_mutex_unlock(&state->cache_lock);
glo_set_current(NULL);
return 0;
}
}
QSIMPLEQ_CONCAT(&state->working_cache, &state->cache);
qemu_mutex_unlock(&state->cache_lock);
qemu_mutex_lock(&d->pgraph.lock);
while (!QSIMPLEQ_EMPTY(&state->working_cache)) {
CacheEntry * command = QSIMPLEQ_FIRST(&state->working_cache);
QSIMPLEQ_REMOVE_HEAD(&state->working_cache, entry);
if (command->method == 0) {
// qemu_mutex_lock_iothread();
RAMHTEntry entry = ramht_lookup(d, command->parameter);
assert(entry.valid);
assert(entry.channel_id == state->channel_id);
// qemu_mutex_unlock_iothread();
switch (entry.engine) {
case ENGINE_GRAPHICS:
pgraph_context_switch(d, entry.channel_id);
pgraph_wait_fifo_access(d);
pgraph_method(d, command->subchannel, 0, entry.instance);
break;
default:
assert(false);
break;
}
/* the engine is bound to the subchannel */
qemu_mutex_lock(&state->cache_lock);
state->bound_engines[command->subchannel] = entry.engine;
state->last_engine = entry.engine;
qemu_mutex_unlock(&state->cache_lock);
} else if (command->method >= 0x100) {
/* method passed to engine */
uint32_t parameter = command->parameter;
/* methods that take objects.
* TODO: Check this range is correct for the nv2a */
if (command->method >= 0x180 && command->method < 0x200) {
//qemu_mutex_lock_iothread();
RAMHTEntry entry = ramht_lookup(d, parameter);
assert(entry.valid);
assert(entry.channel_id == state->channel_id);
parameter = entry.instance;
//qemu_mutex_unlock_iothread();
}
// qemu_mutex_lock(&state->cache_lock);
enum FIFOEngine engine = state->bound_engines[command->subchannel];
// qemu_mutex_unlock(&state->cache_lock);
switch (engine) {
case ENGINE_GRAPHICS:
pgraph_wait_fifo_access(d);
pgraph_method(d, command->subchannel,
command->method, parameter);
break;
default:
assert(false);
break;
}
// qemu_mutex_lock(&state->cache_lock);
state->last_engine = state->bound_engines[command->subchannel];
// qemu_mutex_unlock(&state->cache_lock);
}
g_free(command);
}
qemu_mutex_unlock(&d->pgraph.lock);
}
return 0;
}
static uint32_t ramht_hash(NV2AState *d, uint32_t handle)
{
unsigned int ramht_size =
1 << (GET_MASK(d->pfifo.regs[NV_PFIFO_RAMHT], NV_PFIFO_RAMHT_SIZE)+12);
/* XXX: Think this is different to what nouveau calculates... */
unsigned int bits = ffs(ramht_size)-2;
uint32_t hash = 0;
while (handle) {
hash ^= (handle & ((1 << bits) - 1));
handle >>= bits;
}
hash ^= d->pfifo.cache1.channel_id << (bits - 4);
return hash;
}
static RAMHTEntry ramht_lookup(NV2AState *d, uint32_t handle)
{
unsigned int ramht_size =
1 << (GET_MASK(d->pfifo.regs[NV_PFIFO_RAMHT], NV_PFIFO_RAMHT_SIZE)+12);
uint32_t hash = ramht_hash(d, handle);
assert(hash * 8 < ramht_size);
uint32_t ramht_address =
GET_MASK(d->pfifo.regs[NV_PFIFO_RAMHT],
NV_PFIFO_RAMHT_BASE_ADDRESS) << 12;
uint8_t *entry_ptr = d->ramin_ptr + ramht_address + hash * 8;
uint32_t entry_handle = ldl_le_p((uint32_t*)entry_ptr);
uint32_t entry_context = ldl_le_p((uint32_t*)(entry_ptr + 4));
return (RAMHTEntry){
.handle = entry_handle,
.instance = (entry_context & NV_RAMHT_INSTANCE) << 4,
.engine = (enum FIFOEngine)((entry_context & NV_RAMHT_ENGINE) >> 16),
.channel_id = (entry_context & NV_RAMHT_CHID) >> 24,
.valid = entry_context & NV_RAMHT_STATUS,
};
}

6089
hw/xbox/nv2a.c → hw/xbox/nv2a/nv2a_pgraph.c
View File

File diff suppressed because it is too large Load Diff

71
hw/xbox/nv2a/nv2a_pmc.c Normal file
View File

@ -0,0 +1,71 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/* PMC - card master control */
uint64_t pmc_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
uint64_t r = 0;
switch (addr) {
case NV_PMC_BOOT_0:
/* chipset and stepping:
* NV2A, A02, Rev 0 */
r = 0x02A000A2;
break;
case NV_PMC_INTR_0:
/* Shows which functional units have pending IRQ */
r = d->pmc.pending_interrupts;
break;
case NV_PMC_INTR_EN_0:
/* Selects which functional units can cause IRQs */
r = d->pmc.enabled_interrupts;
break;
default:
break;
}
reg_log_read(NV_PMC, addr, r);
return r;
}
void pmc_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
reg_log_write(NV_PMC, addr, val);
switch (addr) {
case NV_PMC_INTR_0:
/* the bits of the interrupts to clear are wrtten */
d->pmc.pending_interrupts &= ~val;
update_irq(d);
break;
case NV_PMC_INTR_EN_0:
d->pmc.enabled_interrupts = val;
update_irq(d);
break;
default:
break;
}
}

87
hw/xbox/nv2a/nv2a_pramdac.c Normal file
View File

@ -0,0 +1,87 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
uint64_t pramdac_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
uint64_t r = 0;
switch (addr & ~3) {
case NV_PRAMDAC_NVPLL_COEFF:
r = d->pramdac.core_clock_coeff;
break;
case NV_PRAMDAC_MPLL_COEFF:
r = d->pramdac.memory_clock_coeff;
break;
case NV_PRAMDAC_VPLL_COEFF:
r = d->pramdac.video_clock_coeff;
break;
case NV_PRAMDAC_PLL_TEST_COUNTER:
/* emulated PLLs locked instantly? */
r = NV_PRAMDAC_PLL_TEST_COUNTER_VPLL2_LOCK
| NV_PRAMDAC_PLL_TEST_COUNTER_NVPLL_LOCK
| NV_PRAMDAC_PLL_TEST_COUNTER_MPLL_LOCK
| NV_PRAMDAC_PLL_TEST_COUNTER_VPLL_LOCK;
break;
default:
break;
}
/* Surprisingly, QEMU doesn't handle unaligned access for you properly */
r >>= 32 - 8 * size - 8 * (addr & 3);
NV2A_DPRINTF("PRAMDAC: read %d [0x%" HWADDR_PRIx "] -> %llx\n", size, addr, r);
return r;
}
void pramdac_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
uint32_t m, n, p;
reg_log_write(NV_PRAMDAC, addr, val);
switch (addr) {
case NV_PRAMDAC_NVPLL_COEFF:
d->pramdac.core_clock_coeff = val;
m = val & NV_PRAMDAC_NVPLL_COEFF_MDIV;
n = (val & NV_PRAMDAC_NVPLL_COEFF_NDIV) >> 8;
p = (val & NV_PRAMDAC_NVPLL_COEFF_PDIV) >> 16;
if (m == 0) {
d->pramdac.core_clock_freq = 0;
} else {
d->pramdac.core_clock_freq = (NV2A_CRYSTAL_FREQ * n)
/ (1 << p) / m;
}
break;
case NV_PRAMDAC_MPLL_COEFF:
d->pramdac.memory_clock_coeff = val;
break;
case NV_PRAMDAC_VPLL_COEFF:
d->pramdac.video_clock_coeff = val;
break;
default:
break;
}
}

55
hw/xbox/nv2a/nv2a_prmcio.c Normal file
View File

@ -0,0 +1,55 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/* PRMCIO - aliases VGA CRTC and attribute controller registers */
uint64_t prmcio_read(void *opaque,
hwaddr addr, unsigned int size)
{
NV2AState *d = opaque;
uint64_t r = vga_ioport_read(&d->vga, addr);
reg_log_read(NV_PRMCIO, addr, r);
return r;
}
void prmcio_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
NV2AState *d = opaque;
reg_log_write(NV_PRMCIO, addr, val);
switch (addr) {
case VGA_ATT_W:
/* Cromwell sets attrs without enabling VGA_AR_ENABLE_DISPLAY
* (which should result in a blank screen).
* Either nvidia's hardware is lenient or it is set through
* something else. The former seems more likely.
*/
if (d->vga.ar_flip_flop == 0) {
val |= VGA_AR_ENABLE_DISPLAY;
}
break;
default:
break;
}
vga_ioport_write(&d->vga, addr, val);
}

23
hw/xbox/nv2a.h → hw/xbox/nv2a/nv2a_prmvio.c
View File

@ -2,6 +2,8 @@
* QEMU Geforce NV2A implementation * QEMU Geforce NV2A implementation
* *
* Copyright (c) 2012 espes * Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
* *
* This program is free software; you can redistribute it and/or * This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as * modify it under the terms of the GNU General Public License as
@ -17,9 +19,22 @@
* along with this program; if not, see <http://www.gnu.org/licenses/>. * along with this program; if not, see <http://www.gnu.org/licenses/>.
*/ */
#ifndef HW_NV2A_H /* PRMVIO - aliases VGA sequencer and graphics controller registers */
#define HW_NV2A_H uint64_t prmvio_read(void *opaque,
hwaddr addr, unsigned int size)
{
NV2AState *d = opaque;
uint64_t r = vga_ioport_read(&d->vga, addr);
void nv2a_init(PCIBus *bus, int devfn, MemoryRegion *ram); reg_log_read(NV_PRMVIO, addr, r);
return r;
}
void prmvio_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
NV2AState *d = opaque;
#endif reg_log_write(NV_PRMVIO, addr, val);
vga_ioport_write(&d->vga, addr, val);
}

6
hw/xbox/nv2a_psh.c → hw/xbox/nv2a/nv2a_psh.c
View File

@ -25,6 +25,8 @@
* along with this program; if not, see <http://www.gnu.org/licenses/>. * along with this program; if not, see <http://www.gnu.org/licenses/>.
*/ */
#include "qemu/osdep.h"
#include <stdio.h> #include <stdio.h>
#include <stdlib.h> #include <stdlib.h>
#include <string.h> #include <string.h>
@ -33,8 +35,8 @@
#include "qapi/qmp/qstring.h" #include "qapi/qmp/qstring.h"
#include "hw/xbox/nv2a_shaders_common.h" #include "nv2a_shaders_common.h"
#include "hw/xbox/nv2a_psh.h" #include "nv2a_psh.h"
/* /*
* This implements translation of register combiners into glsl * This implements translation of register combiners into glsl

0
hw/xbox/nv2a_psh.h → hw/xbox/nv2a/nv2a_psh.h
View File

89
hw/xbox/nv2a/nv2a_ptimer.c Normal file
View File

@ -0,0 +1,89 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/* PIMTER - time measurement and time-based alarms */
static uint64_t ptimer_get_clock(NV2AState *d)
{
return muldiv64(qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL),
d->pramdac.core_clock_freq * d->ptimer.numerator,
NANOSECONDS_PER_SECOND * d->ptimer.denominator);
}
uint64_t ptimer_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = opaque;
uint64_t r = 0;
switch (addr) {
case NV_PTIMER_INTR_0:
r = d->ptimer.pending_interrupts;
break;
case NV_PTIMER_INTR_EN_0:
r = d->ptimer.enabled_interrupts;
break;
case NV_PTIMER_NUMERATOR:
r = d->ptimer.numerator;
break;
case NV_PTIMER_DENOMINATOR:
r = d->ptimer.denominator;
break;
case NV_PTIMER_TIME_0:
r = (ptimer_get_clock(d) & 0x7ffffff) << 5;
break;
case NV_PTIMER_TIME_1:
r = (ptimer_get_clock(d) >> 27) & 0x1fffffff;
break;
default:
break;
}
reg_log_read(NV_PTIMER, addr, r);
return r;
}
void ptimer_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = opaque;
reg_log_write(NV_PTIMER, addr, val);
switch (addr) {
case NV_PTIMER_INTR_0:
d->ptimer.pending_interrupts &= ~val;
update_irq(d);
break;
case NV_PTIMER_INTR_EN_0:
d->ptimer.enabled_interrupts = val;
update_irq(d);
break;
case NV_PTIMER_DENOMINATOR:
d->ptimer.denominator = val;
break;
case NV_PTIMER_NUMERATOR:
d->ptimer.numerator = val;
break;
case NV_PTIMER_ALARM_0:
d->ptimer.alarm_time = val;
break;
default:
break;
}
}

73
hw/xbox/nv2a/nv2a_pvideo.c Normal file
View File

@ -0,0 +1,73 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
static void pvideo_vga_invalidate(NV2AState *d)
{
int y1 = GET_MASK(d->pvideo.regs[NV_PVIDEO_POINT_OUT],
NV_PVIDEO_POINT_OUT_Y);
int y2 = y1 + GET_MASK(d->pvideo.regs[NV_PVIDEO_SIZE_OUT],
NV_PVIDEO_SIZE_OUT_HEIGHT);
NV2A_DPRINTF("pvideo_vga_invalidate %d %d\n", y1, y2);
vga_invalidate_scanlines(&d->vga, y1, y2);
}
uint64_t pvideo_read(void *opaque,
hwaddr addr, unsigned int size)
{
NV2AState *d = opaque;
uint64_t r = 0;
switch (addr) {
case NV_PVIDEO_STOP:
r = 0;
break;
default:
r = d->pvideo.regs[addr];
break;
}
reg_log_read(NV_PVIDEO, addr, r);
return r;
}
void pvideo_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
NV2AState *d = opaque;
reg_log_write(NV_PVIDEO, addr, val);
switch (addr) {
case NV_PVIDEO_BUFFER:
d->pvideo.regs[addr] = val;
// d->vga.enable_overlay = true;
pvideo_vga_invalidate(d);
break;
case NV_PVIDEO_STOP:
d->pvideo.regs[NV_PVIDEO_BUFFER] = 0;
// d->vga.enable_overlay = false;
pvideo_vga_invalidate(d);
break;
default:
d->pvideo.regs[addr] = val;
break;
}
}

52
hw/xbox/nv2a_shaders.c → hw/xbox/nv2a/nv2a_shaders.c
View File

@ -18,10 +18,56 @@
* along with this program; if not, see <http://www.gnu.org/licenses/>. * along with this program; if not, see <http://www.gnu.org/licenses/>.
*/ */
#include "qemu/osdep.h"
#include "qemu-common.h" #include "qemu-common.h"
#include "hw/xbox/nv2a_debug.h" #include "nv2a_debug.h"
#include "hw/xbox/nv2a_shaders_common.h" #include "nv2a_shaders_common.h"
#include "hw/xbox/nv2a_shaders.h" #include "nv2a_shaders.h"
void qstring_append_fmt(QString *qstring, const char *fmt, ...)
{
va_list ap;
va_start(ap, fmt);
qstring_append_va(qstring, fmt, ap);
va_end(ap);
}
QString *qstring_from_fmt(const char *fmt, ...)
{
QString *ret = qstring_new();
va_list ap;
va_start(ap, fmt);
qstring_append_va(ret, fmt, ap);
va_end(ap);
return ret;
}
void qstring_append_va(QString *qstring, const char *fmt, va_list va)
{
char scratch[256];
va_list ap;
va_copy(ap, va);
const int len = vsnprintf(scratch, sizeof(scratch), fmt, ap);
va_end(ap);
if (len == 0) {
return;
} else if (len < sizeof(scratch)) {
qstring_append(qstring, scratch);
return;
}
/* overflowed out scratch buffer, alloc and try again */
char *buf = g_malloc(len + 1);
va_copy(ap, va);
vsnprintf(buf, len + 1, fmt, ap);
va_end(ap);
qstring_append(qstring, buf);
g_free(buf);
}
static QString* generate_geometry_shader( static QString* generate_geometry_shader(
enum ShaderPolygonMode polygon_front_mode, enum ShaderPolygonMode polygon_front_mode,

0
hw/xbox/nv2a_shaders.h → hw/xbox/nv2a/nv2a_shaders.h
View File

5
hw/xbox/nv2a_shaders_common.h → hw/xbox/nv2a/nv2a_shaders_common.h
View File

@ -34,4 +34,9 @@
" vec4 T3;\n" \ " vec4 T3;\n" \
"};\n" "};\n"
void qstring_append_fmt(QString *qstring, const char *fmt, ...);
QString *qstring_from_fmt(const char *fmt, ...);
void qstring_append_va(QString *qstring, const char *fmt, va_list va);
#endif #endif

118
hw/xbox/nv2a/nv2a_stubs.c Normal file
View File

@ -0,0 +1,118 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
uint64_t prma_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PRMA, addr, 0);
return 0;
}
void prma_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PRMA, addr, val);
}
uint64_t pcounter_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PCOUNTER, addr, 0);
return 0;
}
void pcounter_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PCOUNTER, addr, val);
}
uint64_t pvpe_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PVPE, addr, 0);
return 0;
}
void pvpe_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PVPE, addr, val);
}
uint64_t ptv_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PTV, addr, 0);
return 0;
}
void ptv_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PTV, addr, val);
}
uint64_t prmfb_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PRMFB, addr, 0);
return 0;
}
void prmfb_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PRMFB, addr, val);
}
uint64_t prmdio_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PRMDIO, addr, 0);
return 0;
}
void prmdio_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PRMDIO, addr, val);
}
uint64_t pstraps_read(void *opaque,
hwaddr addr, unsigned int size)
{
reg_log_read(NV_PSTRAPS, addr, 0);
return 0;
}
void pstraps_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
reg_log_write(NV_PSTRAPS, addr, val);
}
/* PRAMIN - RAMIN access */
/*
uint64_t pramin_read(void *opaque,
hwaddr addr, unsigned int size)
{
NV2A_DPRINTF("nv2a PRAMIN: read [0x%" HWADDR_PRIx "] -> 0x%" HWADDR_PRIx "\n", addr, r);
return 0;
}
void pramin_write(void *opaque, hwaddr addr,
uint64_t val, unsigned int size)
{
NV2A_DPRINTF("nv2a PRAMIN: [0x%" HWADDR_PRIx "] = 0x%02llx\n", addr, val);
}*/

95
hw/xbox/nv2a/nv2a_user.c Normal file
View File

@ -0,0 +1,95 @@
/*
* QEMU Geforce NV2A implementation
*
* Copyright (c) 2012 espes
* Copyright (c) 2015 Jannik Vogel
* Copyright (c) 2018 Matt Borgerson
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 or
* (at your option) version 3 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/* USER - PFIFO MMIO and DMA submission area */
uint64_t user_read(void *opaque, hwaddr addr, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
unsigned int channel_id = addr >> 16;
assert(channel_id < NV2A_NUM_CHANNELS);
ChannelControl *control = &d->user.channel_control[channel_id];
uint32_t channel_modes = d->pfifo.regs[NV_PFIFO_MODE];
uint64_t r = 0;
if (channel_modes & (1 << channel_id)) {
/* DMA Mode */
switch (addr & 0xFFFF) {
case NV_USER_DMA_PUT:
r = control->dma_put;
break;
case NV_USER_DMA_GET:
r = control->dma_get;
break;
case NV_USER_REF:
r = control->ref;
break;
default:
break;
}
} else {
/* PIO Mode */
assert(false);
}
reg_log_read(NV_USER, addr, r);
return r;
}
void user_write(void *opaque, hwaddr addr, uint64_t val, unsigned int size)
{
NV2AState *d = (NV2AState *)opaque;
reg_log_write(NV_USER, addr, val);
unsigned int channel_id = addr >> 16;
assert(channel_id < NV2A_NUM_CHANNELS);
ChannelControl *control = &d->user.channel_control[channel_id];
uint32_t channel_modes = d->pfifo.regs[NV_PFIFO_MODE];
if (channel_modes & (1 << channel_id)) {
/* DMA Mode */
switch (addr & 0xFFFF) {
case NV_USER_DMA_PUT:
control->dma_put = val;
if (d->pfifo.cache1.push_enabled) {
pfifo_run_pusher(d);
}
break;
case NV_USER_DMA_GET:
control->dma_get = val;
break;
case NV_USER_REF:
control->ref = val;
break;
default:
break;
}
} else {
/* PIO Mode */
assert(false);
}
}

6
hw/xbox/nv2a_vsh.c → hw/xbox/nv2a/nv2a_vsh.c
View File

@ -25,13 +25,15 @@
* along with this program; if not, see <http://www.gnu.org/licenses/>. * along with this program; if not, see <http://www.gnu.org/licenses/>.
*/ */
#include "qemu/osdep.h"
#include <stdio.h> #include <stdio.h>
#include <string.h> #include <string.h>
#include <stdbool.h> #include <stdbool.h>
#include <assert.h> #include <assert.h>
#include "hw/xbox/nv2a_shaders_common.h" #include "nv2a_shaders_common.h"
#include "hw/xbox/nv2a_vsh.h" #include "nv2a_vsh.h"
#define VSH_D3DSCM_CORRECTION 96 #define VSH_D3DSCM_CORRECTION 96

0
hw/xbox/nv2a_vsh.h → hw/xbox/nv2a/nv2a_vsh.h
View File

2
hw/xbox/swizzle.c → hw/xbox/nv2a/swizzle.c
View File

@ -26,7 +26,7 @@
#include <assert.h> #include <assert.h>
#include "qemu/osdep.h" #include "qemu/osdep.h"
#include "hw/xbox/swizzle.h" #include "swizzle.h"
/* This should be pretty straightforward. /* This should be pretty straightforward.
* It creates a bit pattern like ..zyxzyxzyx from ..xxx, ..yyy and ..zzz * It creates a bit pattern like ..zyxzyxzyx from ..xxx, ..yyy and ..zzz

0
hw/xbox/swizzle.h → hw/xbox/nv2a/swizzle.h
View File

1029
hw/xbox/nv2a/xxhash.c Normal file
View File

File diff suppressed because it is too large Load Diff

328
hw/xbox/nv2a/xxhash.h Normal file
View File

@ -0,0 +1,328 @@
/*
xxHash - Extremely Fast Hash algorithm
Header File
Copyright (C) 2012-2016, Yann Collet.
BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
You can contact the author at :
- xxHash source repository : https://github.com/Cyan4973/xxHash
*/
/* Notice extracted from xxHash homepage :
xxHash is an extremely fast Hash algorithm, running at RAM speed limits.
It also successfully passes all tests from the SMHasher suite.
Comparison (single thread, Windows Seven 32 bits, using SMHasher on a Core 2 Duo @3GHz)
Name Speed Q.Score Author
xxHash 5.4 GB/s 10
CrapWow 3.2 GB/s 2 Andrew
MumurHash 3a 2.7 GB/s 10 Austin Appleby
SpookyHash 2.0 GB/s 10 Bob Jenkins
SBox 1.4 GB/s 9 Bret Mulvey
Lookup3 1.2 GB/s 9 Bob Jenkins
SuperFastHash 1.2 GB/s 1 Paul Hsieh
CityHash64 1.05 GB/s 10 Pike & Alakuijala
FNV 0.55 GB/s 5 Fowler, Noll, Vo
CRC32 0.43 GB/s 9
MD5-32 0.33 GB/s 10 Ronald L. Rivest
SHA1-32 0.28 GB/s 10
Q.Score is a measure of quality of the hash function.
It depends on successfully passing SMHasher test set.
10 is a perfect score.
A 64-bit version, named XXH64, is available since r35.
It offers much better speed, but for 64-bit applications only.
Name Speed on 64 bits Speed on 32 bits
XXH64 13.8 GB/s 1.9 GB/s
XXH32 6.8 GB/s 6.0 GB/s
*/
#ifndef XXHASH_H_5627135585666179
#define XXHASH_H_5627135585666179 1
#if defined (__cplusplus)
extern "C" {
#endif
/* ****************************
* Definitions
******************************/
#include <stddef.h> /* size_t */
typedef enum { XXH_OK=0, XXH_ERROR } XXH_errorcode;
/* ****************************
* API modifier
******************************/
/** XXH_INLINE_ALL (and XXH_PRIVATE_API)
* This is useful to include xxhash functions in `static` mode
* in order to inline them, and remove their symbol from the public list.
* Inlining can offer dramatic performance improvement on small keys.
* Methodology :
* #define XXH_INLINE_ALL
* #include "xxhash.h"
* `xxhash.c` is automatically included.
* It's not useful to compile and link it as a separate module.
*/
#if defined(XXH_INLINE_ALL) || defined(XXH_PRIVATE_API)
# ifndef XXH_STATIC_LINKING_ONLY
# define XXH_STATIC_LINKING_ONLY
# endif
# if defined(__GNUC__)
# define XXH_PUBLIC_API static __inline __attribute__((unused))
# elif defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
# define XXH_PUBLIC_API static inline
# elif defined(_MSC_VER)
# define XXH_PUBLIC_API static __inline
# else
/* this version may generate warnings for unused static functions */
# define XXH_PUBLIC_API static
# endif
#else
# define XXH_PUBLIC_API /* do nothing */
#endif /* XXH_INLINE_ALL || XXH_PRIVATE_API */
/*! XXH_NAMESPACE, aka Namespace Emulation :
*
* If you want to include _and expose_ xxHash functions from within your own library,
* but also want to avoid symbol collisions with other libraries which may also include xxHash,
*
* you can use XXH_NAMESPACE, to automatically prefix any public symbol from xxhash library
* with the value of XXH_NAMESPACE (therefore, avoid NULL and numeric values).
*
* Note that no change is required within the calling program as long as it includes `xxhash.h` :
* regular symbol name will be automatically translated by this header.
*/
#ifdef XXH_NAMESPACE
# define XXH_CAT(A,B) A##B
# define XXH_NAME2(A,B) XXH_CAT(A,B)
# define XXH_versionNumber XXH_NAME2(XXH_NAMESPACE, XXH_versionNumber)
# define XXH32 XXH_NAME2(XXH_NAMESPACE, XXH32)
# define XXH32_createState XXH_NAME2(XXH_NAMESPACE, XXH32_createState)
# define XXH32_freeState XXH_NAME2(XXH_NAMESPACE, XXH32_freeState)
# define XXH32_reset XXH_NAME2(XXH_NAMESPACE, XXH32_reset)
# define XXH32_update XXH_NAME2(XXH_NAMESPACE, XXH32_update)
# define XXH32_digest XXH_NAME2(XXH_NAMESPACE, XXH32_digest)
# define XXH32_copyState XXH_NAME2(XXH_NAMESPACE, XXH32_copyState)
# define XXH32_canonicalFromHash XXH_NAME2(XXH_NAMESPACE, XXH32_canonicalFromHash)
# define XXH32_hashFromCanonical XXH_NAME2(XXH_NAMESPACE, XXH32_hashFromCanonical)
# define XXH64 XXH_NAME2(XXH_NAMESPACE, XXH64)
# define XXH64_createState XXH_NAME2(XXH_NAMESPACE, XXH64_createState)
# define XXH64_freeState XXH_NAME2(XXH_NAMESPACE, XXH64_freeState)
# define XXH64_reset XXH_NAME2(XXH_NAMESPACE, XXH64_reset)
# define XXH64_update XXH_NAME2(XXH_NAMESPACE, XXH64_update)
# define XXH64_digest XXH_NAME2(XXH_NAMESPACE, XXH64_digest)
# define XXH64_copyState XXH_NAME2(XXH_NAMESPACE, XXH64_copyState)
# define XXH64_canonicalFromHash XXH_NAME2(XXH_NAMESPACE, XXH64_canonicalFromHash)
# define XXH64_hashFromCanonical XXH_NAME2(XXH_NAMESPACE, XXH64_hashFromCanonical)
#endif
/* *************************************
* Version
***************************************/
#define XXH_VERSION_MAJOR 0
#define XXH_VERSION_MINOR 6
#define XXH_VERSION_RELEASE 5
#define XXH_VERSION_NUMBER (XXH_VERSION_MAJOR *100*100 + XXH_VERSION_MINOR *100 + XXH_VERSION_RELEASE)
XXH_PUBLIC_API unsigned XXH_versionNumber (void);
/*-**********************************************************************
* 32-bit hash
************************************************************************/
typedef unsigned int XXH32_hash_t;
/*! XXH32() :
Calculate the 32-bit hash of sequence "length" bytes stored at memory address "input".
The memory between input & input+length must be valid (allocated and read-accessible).
"seed" can be used to alter the result predictably.
Speed on Core 2 Duo @ 3 GHz (single thread, SMHasher benchmark) : 5.4 GB/s */
XXH_PUBLIC_API XXH32_hash_t XXH32 (const void* input, size_t length, unsigned int seed);
/*====== Streaming ======*/
typedef struct XXH32_state_s XXH32_state_t; /* incomplete type */
XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void);
XXH_PUBLIC_API XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr);
XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* dst_state, const XXH32_state_t* src_state);
XXH_PUBLIC_API XXH_errorcode XXH32_reset (XXH32_state_t* statePtr, unsigned int seed);
XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* statePtr, const void* input, size_t length);
XXH_PUBLIC_API XXH32_hash_t XXH32_digest (const XXH32_state_t* statePtr);
/*
* Streaming functions generate the xxHash of an input provided in multiple segments.
* Note that, for small input, they are slower than single-call functions, due to state management.
* For small inputs, prefer `XXH32()` and `XXH64()`, which are better optimized.
*
* XXH state must first be allocated, using XXH*_createState() .
*
* Start a new hash by initializing state with a seed, using XXH*_reset().
*
* Then, feed the hash state by calling XXH*_update() as many times as necessary.
* The function returns an error code, with 0 meaning OK, and any other value meaning there is an error.
*
* Finally, a hash value can be produced anytime, by using XXH*_digest().
* This function returns the nn-bits hash as an int or long long.
*
* It's still possible to continue inserting input into the hash state after a digest,
* and generate some new hashes later on, by calling again XXH*_digest().
*
* When done, free XXH state space if it was allocated dynamically.
*/
/*====== Canonical representation ======*/
typedef struct { unsigned char digest[4]; } XXH32_canonical_t;
XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash);
XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src);
/* Default result type for XXH functions are primitive unsigned 32 and 64 bits.
* The canonical representation uses human-readable write convention, aka big-endian (large digits first).
* These functions allow transformation of hash result into and from its canonical format.
* This way, hash values can be written into a file / memory, and remain comparable on different systems and programs.
*/
#ifndef XXH_NO_LONG_LONG
/*-**********************************************************************
* 64-bit hash
************************************************************************/
typedef unsigned long long XXH64_hash_t;
/*! XXH64() :
Calculate the 64-bit hash of sequence of length "len" stored at memory address "input".
"seed" can be used to alter the result predictably.
This function runs faster on 64-bit systems, but slower on 32-bit systems (see benchmark).
*/
XXH_PUBLIC_API XXH64_hash_t XXH64 (const void* input, size_t length, unsigned long long seed);
/*====== Streaming ======*/
typedef struct XXH64_state_s XXH64_state_t; /* incomplete type */
XXH_PUBLIC_API XXH64_state_t* XXH64_createState(void);
XXH_PUBLIC_API XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr);
XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* dst_state, const XXH64_state_t* src_state);
XXH_PUBLIC_API XXH_errorcode XXH64_reset (XXH64_state_t* statePtr, unsigned long long seed);
XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* statePtr, const void* input, size_t length);
XXH_PUBLIC_API XXH64_hash_t XXH64_digest (const XXH64_state_t* statePtr);
/*====== Canonical representation ======*/
typedef struct { unsigned char digest[8]; } XXH64_canonical_t;
XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t hash);
XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src);
#endif /* XXH_NO_LONG_LONG */
#ifdef XXH_STATIC_LINKING_ONLY
/* ================================================================================================
This section contains declarations which are not guaranteed to remain stable.
They may change in future versions, becoming incompatible with a different version of the library.
These declarations should only be used with static linking.
Never use them in association with dynamic linking !
=================================================================================================== */
/* These definitions are only present to allow
* static allocation of XXH state, on stack or in a struct for example.
* Never **ever** use members directly. */
#if !defined (__VMS) \
&& (defined (__cplusplus) \
|| (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
# include <stdint.h>
struct XXH32_state_s {
uint32_t total_len_32;
uint32_t large_len;
uint32_t v1;
uint32_t v2;
uint32_t v3;
uint32_t v4;
uint32_t mem32[4];
uint32_t memsize;
uint32_t reserved; /* never read nor write, might be removed in a future version */
}; /* typedef'd to XXH32_state_t */
struct XXH64_state_s {
uint64_t total_len;
uint64_t v1;
uint64_t v2;
uint64_t v3;
uint64_t v4;
uint64_t mem64[4];
uint32_t memsize;
uint32_t reserved[2]; /* never read nor write, might be removed in a future version */
}; /* typedef'd to XXH64_state_t */
# else
struct XXH32_state_s {
unsigned total_len_32;
unsigned large_len;
unsigned v1;
unsigned v2;
unsigned v3;
unsigned v4;
unsigned mem32[4];
unsigned memsize;
unsigned reserved; /* never read nor write, might be removed in a future version */
}; /* typedef'd to XXH32_state_t */
# ifndef XXH_NO_LONG_LONG /* remove 64-bit support */
struct XXH64_state_s {
unsigned long long total_len;
unsigned long long v1;
unsigned long long v2;
unsigned long long v3;
unsigned long long v4;
unsigned long long mem64[4];
unsigned memsize;
unsigned reserved[2]; /* never read nor write, might be removed in a future version */
}; /* typedef'd to XXH64_state_t */
# endif
# endif
#if defined(XXH_INLINE_ALL) || defined(XXH_PRIVATE_API)
# include "xxhash.c" /* include xxhash function bodies as `static`, for inlining */
#endif
#endif /* XXH_STATIC_LINKING_ONLY */
#if defined (__cplusplus)
}
#endif
#endif /* XXHASH_H_5627135585666179 */