Skip to content

L
linux
Commit 054b93e4 authored by Ben Skeggs's avatar Ben Skeggs
Browse files
Options

drm/nv40: implement ctxprog/state generation 

The context programs are *very* simple compared to the ones used by
the binary driver.  There's notes in nv40_grctx.c explaining most of
the things we don't implement.  If we discover if/why any of it is
required further down the track, we'll handle it then.

The PGRAPH state generated for each chipset should match what NVIDIA
do almost exactly (there's a couple of exceptions).  If someone has
a lot of time on their hands, they could figure out the mapping of
object/method to PGRAPH register and demagic the initial state a little,
it's not terribly important however.

At time of commit, confirmed to be working at least well enough for
accelerated X (and where tested, for 3D apps) on NV40, NV43, NV44, NV46,
NV49, NV4A, NV4B and NV4E.

A module option has been added to force the use of external firmware
blobs if it becomes required.
Signed-off-by: default avatarBen Skeggs <bskeggs@redhat.com>
parent 15bee69e
Show whitespace changes
Inline Side-by-side
Showing with 1031 additions and 182 deletions
drivers/gpu/drm/nouveau/Makefile
View file @ 054b93e4
......@@ -9,13 +9,14 @@ nouveau-y := nouveau_drv.o nouveau_state.o nouveau_channel.o nouveau_mem.o \
nouveau_bo.o nouveau_fence.o nouveau_gem.o nouveau_ttm.o \
nouveau_hw.o nouveau_calc.o nouveau_bios.o nouveau_i2c.o \
nouveau_display.o nouveau_connector.o nouveau_fbcon.o \
nouveau_dp.o \
nouveau_dp.o nouveau_grctx.o \
nv04_timer.o \
nv04_mc.o nv40_mc.o nv50_mc.o \
nv04_fb.o nv10_fb.o nv40_fb.o \
nv04_fifo.o nv10_fifo.o nv40_fifo.o nv50_fifo.o \
nv04_graph.o nv10_graph.o nv20_graph.o \
nv40_graph.o nv50_graph.o \
nv40_grctx.o \
nv04_instmem.o nv50_instmem.o \
nv50_crtc.o nv50_dac.o nv50_sor.o \
nv50_cursor.o nv50_display.o nv50_fbcon.o \
......
drivers/gpu/drm/nouveau/nouveau_drv.c
View file @ 054b93e4
......@@ -35,6 +35,10 @@
#include "drm_pciids.h"
MODULE_PARM_DESC(ctxfw, "Use external firmware blob for grctx init (NV40)");
int nouveau_ctxfw = 0;
module_param_named(ctxfw, nouveau_ctxfw, int, 0400);
MODULE_PARM_DESC(noagp, "Disable AGP");
int nouveau_noagp;
module_param_named(noagp, nouveau_noagp, int, 0400);
......
drivers/gpu/drm/nouveau/nouveau_drv.h
View file @ 054b93e4
......@@ -54,6 +54,7 @@ struct nouveau_fpriv {
#include "nouveau_drm.h"
#include "nouveau_reg.h"
#include "nouveau_bios.h"
struct nouveau_grctx;
#define MAX_NUM_DCB_ENTRIES 16
......@@ -317,6 +318,7 @@ struct nouveau_pgraph_engine {
bool accel_blocked;
void *ctxprog;
void *ctxvals;
int grctx_size;
int (*init)(struct drm_device *);
void (*takedown)(struct drm_device *);
......@@ -647,6 +649,7 @@ extern int nouveau_fbpercrtc;
extern char *nouveau_tv_norm;
extern int nouveau_reg_debug;
extern char *nouveau_vbios;
extern int nouveau_ctxfw;
/* nouveau_state.c */
extern void nouveau_preclose(struct drm_device *dev, struct drm_file *);
......@@ -959,9 +962,7 @@ extern int nv40_graph_create_context(struct nouveau_channel *);
extern void nv40_graph_destroy_context(struct nouveau_channel *);
extern int nv40_graph_load_context(struct nouveau_channel *);
extern int nv40_graph_unload_context(struct drm_device *);
extern int nv40_grctx_init(struct drm_device *);
extern void nv40_grctx_fini(struct drm_device *);
extern void nv40_grctx_vals_load(struct drm_device *, struct nouveau_gpuobj *);
extern void nv40_grctx_init(struct nouveau_grctx *);
/* nv50_graph.c */
extern struct nouveau_pgraph_object_class nv50_graph_grclass[];
......@@ -975,6 +976,12 @@ extern int nv50_graph_load_context(struct nouveau_channel *);
extern int nv50_graph_unload_context(struct drm_device *);
extern void nv50_graph_context_switch(struct drm_device *);
/* nouveau_grctx.c */
extern int nouveau_grctx_prog_load(struct drm_device *);
extern void nouveau_grctx_vals_load(struct drm_device *,
struct nouveau_gpuobj *);
extern void nouveau_grctx_fini(struct drm_device *);
/* nv04_instmem.c */
extern int nv04_instmem_init(struct drm_device *);
extern void nv04_instmem_takedown(struct drm_device *);
......
drivers/gpu/drm/nouveau/nouveau_grctx.c 0 → 100644
View file @ 054b93e4
/*
* Copyright 2009 Red Hat Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
* Authors: Ben Skeggs
*/
#include <linux/firmware.h>
#include "drmP.h"
#include "nouveau_drv.h"
struct nouveau_ctxprog {
uint32_t signature;
uint8_t version;
uint16_t length;
uint32_t data[];
} __attribute__ ((packed));
struct nouveau_ctxvals {
uint32_t signature;
uint8_t version;
uint32_t length;
struct {
uint32_t offset;
uint32_t value;
} data[];
} __attribute__ ((packed));
int
nouveau_grctx_prog_load(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
const int chipset = dev_priv->chipset;
const struct firmware *fw;
const struct nouveau_ctxprog *cp;
const struct nouveau_ctxvals *cv;
char name[32];
int ret, i;
if (!pgraph->ctxprog) {
sprintf(name, "nouveau/nv%02x.ctxprog", chipset);
ret = request_firmware(&fw, name, &dev->pdev->dev);
if (ret) {
NV_ERROR(dev, "No ctxprog for NV%02x\n", chipset);
return ret;
}
pgraph->ctxprog = kmalloc(fw->size, GFP_KERNEL);
if (!pgraph->ctxprog) {
NV_ERROR(dev, "OOM copying ctxprog\n");
release_firmware(fw);
return -ENOMEM;
}
memcpy(pgraph->ctxprog, fw->data, fw->size);
cp = pgraph->ctxprog;
if (le32_to_cpu(cp->signature) != 0x5043564e ||
cp->version != 0 ||
le16_to_cpu(cp->length) != ((fw->size - 7) / 4)) {
NV_ERROR(dev, "ctxprog invalid\n");
release_firmware(fw);
nouveau_grctx_fini(dev);
return -EINVAL;
}
release_firmware(fw);
}
if (!pgraph->ctxvals) {
sprintf(name, "nouveau/nv%02x.ctxvals", chipset);
ret = request_firmware(&fw, name, &dev->pdev->dev);
if (ret) {
NV_ERROR(dev, "No ctxvals for NV%02x\n", chipset);
nouveau_grctx_fini(dev);
return ret;
}
pgraph->ctxvals = kmalloc(fw->size, GFP_KERNEL);
if (!pgraph->ctxprog) {
NV_ERROR(dev, "OOM copying ctxprog\n");
release_firmware(fw);
nouveau_grctx_fini(dev);
return -ENOMEM;
}
memcpy(pgraph->ctxvals, fw->data, fw->size);
cv = (void *)pgraph->ctxvals;
if (le32_to_cpu(cv->signature) != 0x5643564e ||
cv->version != 0 ||
le32_to_cpu(cv->length) != ((fw->size - 9) / 8)) {
NV_ERROR(dev, "ctxvals invalid\n");
release_firmware(fw);
nouveau_grctx_fini(dev);
return -EINVAL;
}
release_firmware(fw);
}
cp = pgraph->ctxprog;
nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
for (i = 0; i < le16_to_cpu(cp->length); i++)
nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA,
le32_to_cpu(cp->data[i]));
return 0;
}
void
nouveau_grctx_fini(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
if (pgraph->ctxprog) {
kfree(pgraph->ctxprog);
pgraph->ctxprog = NULL;
}
if (pgraph->ctxvals) {
kfree(pgraph->ctxprog);
pgraph->ctxvals = NULL;
}
}
void
nouveau_grctx_vals_load(struct drm_device *dev, struct nouveau_gpuobj *ctx)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
struct nouveau_ctxvals *cv = pgraph->ctxvals;
int i;
if (!cv)
return;
for (i = 0; i < le32_to_cpu(cv->length); i++)
nv_wo32(dev, ctx, le32_to_cpu(cv->data[i].offset),
le32_to_cpu(cv->data[i].value));
}
drivers/gpu/drm/nouveau/nouveau_grctx.h 0 → 100644
View file @ 054b93e4
#ifndef __NOUVEAU_GRCTX_H__
#define __NOUVEAU_GRCTX_H__
struct nouveau_grctx {
struct drm_device *dev;
enum {
NOUVEAU_GRCTX_PROG,
NOUVEAU_GRCTX_VALS
} mode;
void *data;
uint32_t ctxprog_max;
uint32_t ctxprog_len;
uint32_t ctxprog_reg;
int ctxprog_label[32];
uint32_t ctxvals_pos;
uint32_t ctxvals_base;
};
#ifdef CP_CTX
static inline void
cp_out(struct nouveau_grctx *ctx, uint32_t inst)
{
uint32_t *ctxprog = ctx->data;
if (ctx->mode != NOUVEAU_GRCTX_PROG)
return;
BUG_ON(ctx->ctxprog_len == ctx->ctxprog_max);
ctxprog[ctx->ctxprog_len++] = inst;
}
static inline void
cp_lsr(struct nouveau_grctx *ctx, uint32_t val)
{
cp_out(ctx, CP_LOAD_SR | val);
}
static inline void
cp_ctx(struct nouveau_grctx *ctx, uint32_t reg, uint32_t length)
{
ctx->ctxprog_reg = (reg - 0x00400000) >> 2;
ctx->ctxvals_base = ctx->ctxvals_pos;
ctx->ctxvals_pos = ctx->ctxvals_base + length;
if (length > (CP_CTX_COUNT >> CP_CTX_COUNT_SHIFT)) {
cp_lsr(ctx, length);
length = 0;
}
cp_out(ctx, CP_CTX | (length << CP_CTX_COUNT_SHIFT) | ctx->ctxprog_reg);
}
static inline void
cp_name(struct nouveau_grctx *ctx, int name)
{
uint32_t *ctxprog = ctx->data;
int i;
if (ctx->mode != NOUVEAU_GRCTX_PROG)
return;
ctx->ctxprog_label[name] = ctx->ctxprog_len;
for (i = 0; i < ctx->ctxprog_len; i++) {
if ((ctxprog[i] & 0xfff00000) != 0xff400000)
continue;
if ((ctxprog[i] & CP_BRA_IP) != ((name) << CP_BRA_IP_SHIFT))
continue;
ctxprog[i] = (ctxprog[i] & 0x00ff00ff) |
(ctx->ctxprog_len << CP_BRA_IP_SHIFT);
}
}
static inline void
_cp_bra(struct nouveau_grctx *ctx, u32 mod, int flag, int state, int name)
{
int ip = 0;
if (mod != 2) {
ip = ctx->ctxprog_label[name] << CP_BRA_IP_SHIFT;
if (ip == 0)
ip = 0xff000000 | (name << CP_BRA_IP_SHIFT);
}
cp_out(ctx, CP_BRA | (mod << 18) | ip | flag |
(state ? 0 : CP_BRA_IF_CLEAR));
}
#define cp_bra(c,f,s,n) _cp_bra((c), 0, CP_FLAG_##f, CP_FLAG_##f##_##s, n)
#ifdef CP_BRA_MOD
#define cp_cal(c,f,s,n) _cp_bra((c), 1, CP_FLAG_##f, CP_FLAG_##f##_##s, n)
#define cp_ret(c,f,s) _cp_bra((c), 2, CP_FLAG_##f, CP_FLAG_##f##_##s, 0)
#endif
static inline void
_cp_wait(struct nouveau_grctx *ctx, int flag, int state)
{
cp_out(ctx, CP_WAIT | flag | (state ? CP_WAIT_SET : 0));
}
#define cp_wait(c,f,s) _cp_wait((c), CP_FLAG_##f, CP_FLAG_##f##_##s)
static inline void
_cp_set(struct nouveau_grctx *ctx, int flag, int state)
{
cp_out(ctx, CP_SET | flag | (state ? CP_SET_1 : 0));
}
#define cp_set(c,f,s) _cp_set((c), CP_FLAG_##f, CP_FLAG_##f##_##s)
static inline void
cp_pos(struct nouveau_grctx *ctx, int offset)
{
ctx->ctxvals_pos = offset;
ctx->ctxvals_base = ctx->ctxvals_pos;
cp_lsr(ctx, ctx->ctxvals_pos);
cp_out(ctx, CP_SET_CONTEXT_POINTER);
}
static inline void
gr_def(struct nouveau_grctx *ctx, uint32_t reg, uint32_t val)
{
if (ctx->mode != NOUVEAU_GRCTX_VALS)
return;
reg = (reg - 0x00400000) / 4;
reg = (reg - ctx->ctxprog_reg) + ctx->ctxvals_base;
nv_wo32(ctx->dev, ctx->data, reg, val);
}
#endif
#endif
drivers/gpu/drm/nouveau/nv40_graph.c
View file @ 054b93e4
......@@ -24,36 +24,10 @@
*
*/
#include <linux/firmware.h>
#include "drmP.h"
#include "drm.h"
#include "nouveau_drv.h"
MODULE_FIRMWARE("nouveau/nv40.ctxprog");
MODULE_FIRMWARE("nouveau/nv40.ctxvals");
MODULE_FIRMWARE("nouveau/nv41.ctxprog");
MODULE_FIRMWARE("nouveau/nv41.ctxvals");
MODULE_FIRMWARE("nouveau/nv42.ctxprog");
MODULE_FIRMWARE("nouveau/nv42.ctxvals");
MODULE_FIRMWARE("nouveau/nv43.ctxprog");
MODULE_FIRMWARE("nouveau/nv43.ctxvals");
MODULE_FIRMWARE("nouveau/nv44.ctxprog");
MODULE_FIRMWARE("nouveau/nv44.ctxvals");
MODULE_FIRMWARE("nouveau/nv46.ctxprog");
MODULE_FIRMWARE("nouveau/nv46.ctxvals");
MODULE_FIRMWARE("nouveau/nv47.ctxprog");
MODULE_FIRMWARE("nouveau/nv47.ctxvals");
MODULE_FIRMWARE("nouveau/nv49.ctxprog");
MODULE_FIRMWARE("nouveau/nv49.ctxvals");
MODULE_FIRMWARE("nouveau/nv4a.ctxprog");
MODULE_FIRMWARE("nouveau/nv4a.ctxvals");
MODULE_FIRMWARE("nouveau/nv4b.ctxprog");
MODULE_FIRMWARE("nouveau/nv4b.ctxvals");
MODULE_FIRMWARE("nouveau/nv4c.ctxprog");
MODULE_FIRMWARE("nouveau/nv4c.ctxvals");
MODULE_FIRMWARE("nouveau/nv4e.ctxprog");
MODULE_FIRMWARE("nouveau/nv4e.ctxvals");
#include "nouveau_grctx.h"
struct nouveau_channel *
nv40_graph_channel(struct drm_device *dev)
......@@ -83,27 +57,30 @@ nv40_graph_create_context(struct nouveau_channel *chan)
{
struct drm_device *dev = chan->dev;
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_gpuobj *ctx;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
int ret;
/* Allocate a 175KiB block of PRAMIN to store the context. This
* is massive overkill for a lot of chipsets, but it should be safe
* until we're able to implement this properly (will happen at more
* or less the same time we're able to write our own context programs.
*/
ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, 175*1024, 16,
NVOBJ_FLAG_ZERO_ALLOC,
ret = nouveau_gpuobj_new_ref(dev, chan, NULL, 0, pgraph->grctx_size,
16, NVOBJ_FLAG_ZERO_ALLOC,
&chan->ramin_grctx);
if (ret)
return ret;
ctx = chan->ramin_grctx->gpuobj;
/* Initialise default context values */
dev_priv->engine.instmem.prepare_access(dev, true);
nv40_grctx_vals_load(dev, ctx);
nv_wo32(dev, ctx, 0, ctx->im_pramin->start);
dev_priv->engine.instmem.finish_access(dev);
if (!pgraph->ctxprog) {
struct nouveau_grctx ctx = {};
ctx.dev = chan->dev;
ctx.mode = NOUVEAU_GRCTX_VALS;
ctx.data = chan->ramin_grctx->gpuobj;
nv40_grctx_init(&ctx);
} else {
nouveau_grctx_vals_load(dev, chan->ramin_grctx->gpuobj);
}
nv_wo32(dev, chan->ramin_grctx->gpuobj, 0,
chan->ramin_grctx->gpuobj->im_pramin->start);
dev_priv->engine.instmem.finish_access(dev);
return 0;
}
......@@ -204,139 +181,6 @@ nv40_graph_unload_context(struct drm_device *dev)
return ret;
}
struct nouveau_ctxprog {
uint32_t signature;
uint8_t version;
uint16_t length;
uint32_t data[];
} __attribute__ ((packed));
struct nouveau_ctxvals {
uint32_t signature;
uint8_t version;
uint32_t length;
struct {
uint32_t offset;
uint32_t value;
} data[];
} __attribute__ ((packed));
int
nv40_grctx_init(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
const int chipset = dev_priv->chipset;
const struct firmware *fw;
const struct nouveau_ctxprog *cp;
const struct nouveau_ctxvals *cv;
char name[32];
int ret, i;
pgraph->accel_blocked = true;
if (!pgraph->ctxprog) {
sprintf(name, "nouveau/nv%02x.ctxprog", chipset);
ret = request_firmware(&fw, name, &dev->pdev->dev);
if (ret) {
NV_ERROR(dev, "No ctxprog for NV%02x\n", chipset);
return ret;
}
pgraph->ctxprog = kmalloc(fw->size, GFP_KERNEL);
if (!pgraph->ctxprog) {
NV_ERROR(dev, "OOM copying ctxprog\n");
release_firmware(fw);
return -ENOMEM;
}
memcpy(pgraph->ctxprog, fw->data, fw->size);
cp = pgraph->ctxprog;
if (le32_to_cpu(cp->signature) != 0x5043564e ||
cp->version != 0 ||
le16_to_cpu(cp->length) != ((fw->size - 7) / 4)) {
NV_ERROR(dev, "ctxprog invalid\n");
release_firmware(fw);
nv40_grctx_fini(dev);
return -EINVAL;
}
release_firmware(fw);
}
if (!pgraph->ctxvals) {
sprintf(name, "nouveau/nv%02x.ctxvals", chipset);
ret = request_firmware(&fw, name, &dev->pdev->dev);
if (ret) {
NV_ERROR(dev, "No ctxvals for NV%02x\n", chipset);
nv40_grctx_fini(dev);
return ret;
}
pgraph->ctxvals = kmalloc(fw->size, GFP_KERNEL);
if (!pgraph->ctxprog) {
NV_ERROR(dev, "OOM copying ctxprog\n");
release_firmware(fw);
nv40_grctx_fini(dev);
return -ENOMEM;
}
memcpy(pgraph->ctxvals, fw->data, fw->size);
cv = (void *)pgraph->ctxvals;
if (le32_to_cpu(cv->signature) != 0x5643564e ||
cv->version != 0 ||
le32_to_cpu(cv->length) != ((fw->size - 9) / 8)) {
NV_ERROR(dev, "ctxvals invalid\n");
release_firmware(fw);
nv40_grctx_fini(dev);
return -EINVAL;
}
release_firmware(fw);
}
cp = pgraph->ctxprog;
nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
for (i = 0; i < le16_to_cpu(cp->length); i++)
nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA,
le32_to_cpu(cp->data[i]));
pgraph->accel_blocked = false;
return 0;
}
void
nv40_grctx_fini(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
if (pgraph->ctxprog) {
kfree(pgraph->ctxprog);
pgraph->ctxprog = NULL;
}
if (pgraph->ctxvals) {
kfree(pgraph->ctxprog);
pgraph->ctxvals = NULL;
}
}
void
nv40_grctx_vals_load(struct drm_device *dev, struct nouveau_gpuobj *ctx)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
struct nouveau_ctxvals *cv = pgraph->ctxvals;
int i;
if (!cv)
return;
for (i = 0; i < le32_to_cpu(cv->length); i++)
nv_wo32(dev, ctx, le32_to_cpu(cv->data[i].offset),
le32_to_cpu(cv->data[i].value));
}
/*
* G70 0x47
* G71 0x49
......@@ -359,7 +203,26 @@ nv40_graph_init(struct drm_device *dev)
nv_wr32(dev, NV03_PMC_ENABLE, nv_rd32(dev, NV03_PMC_ENABLE) |
NV_PMC_ENABLE_PGRAPH);
nv40_grctx_init(dev);
if (nouveau_ctxfw) {
nouveau_grctx_prog_load(dev);
dev_priv->engine.graph.grctx_size = 175 * 1024;
}
if (!dev_priv->engine.graph.ctxprog) {
struct nouveau_grctx ctx = {};
uint32_t cp[256];
ctx.dev = dev;
ctx.mode = NOUVEAU_GRCTX_PROG;
ctx.data = cp;
ctx.ctxprog_max = 256;
nv40_grctx_init(&ctx);
dev_priv->engine.graph.grctx_size = ctx.ctxvals_pos * 4;
nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_INDEX, 0);
for (i = 0; i < ctx.ctxprog_len; i++)
nv_wr32(dev, NV40_PGRAPH_CTXCTL_UCODE_DATA, cp[i]);
}
/* No context present currently */
nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0x00000000);
......@@ -539,6 +402,7 @@ nv40_graph_init(struct drm_device *dev)
void nv40_graph_takedown(struct drm_device *dev)
{
nouveau_grctx_fini(dev);
}
struct nouveau_pgraph_object_class nv40_graph_grclass[] = {
......
drivers/gpu/drm/nouveau/nv40_grctx.c 0 → 100644
View file @ 054b93e4
/*
* Copyright 2009 Red Hat Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
* Authors: Ben Skeggs
*/
/* NVIDIA context programs handle a number of other conditions which are
* not implemented in our versions. It's not clear why NVIDIA context
* programs have this code, nor whether it's strictly necessary for
* correct operation. We'll implement additional handling if/when we
* discover it's necessary.
*
* - On context save, NVIDIA set 0x400314 bit 0 to 1 if the "3D state"
* flag is set, this gets saved into the context.
* - On context save, the context program for all cards load nsource
* into a flag register and check for ILLEGAL_MTHD. If it's set,
* opcode 0x60000d is called before resuming normal operation.
* - Some context programs check more conditions than the above. NV44
* checks: ((nsource & 0x0857) || (0x400718 & 0x0100) || (intr & 0x0001))
* and calls 0x60000d before resuming normal operation.
* - At the very beginning of NVIDIA's context programs, flag 9 is checked
* and if true 0x800001 is called with count=0, pos=0, the flag is cleared
* and then the ctxprog is aborted. It looks like a complicated NOP,
* its purpose is unknown.
* - In the section of code that loads the per-vs state, NVIDIA check
* flag 10. If it's set, they only transfer the small 0x300 byte block
* of state + the state for a single vs as opposed to the state for
* all vs units. It doesn't seem likely that it'll occur in normal
* operation, especially seeing as it appears NVIDIA may have screwed
* up the ctxprogs for some cards and have an invalid instruction
* rather than a cp_lsr(ctx, dwords_for_1_vs_unit) instruction.
* - There's a number of places where context offset 0 (where we place
* the PRAMIN offset of the context) is loaded into either 0x408000,
* 0x408004 or 0x408008. Not sure what's up there either.
* - The ctxprogs for some cards save 0x400a00 again during the cleanup
* path for auto-loadctx.
*/
#define CP_FLAG_CLEAR 0
#define CP_FLAG_SET 1
#define CP_FLAG_SWAP_DIRECTION ((0 * 32) + 0)
#define CP_FLAG_SWAP_DIRECTION_LOAD 0
#define CP_FLAG_SWAP_DIRECTION_SAVE 1
#define CP_FLAG_USER_SAVE ((0 * 32) + 5)
#define CP_FLAG_USER_SAVE_NOT_PENDING 0
#define CP_FLAG_USER_SAVE_PENDING 1
#define CP_FLAG_USER_LOAD ((0 * 32) + 6)
#define CP_FLAG_USER_LOAD_NOT_PENDING 0
#define CP_FLAG_USER_LOAD_PENDING 1
#define CP_FLAG_STATUS ((3 * 32) + 0)
#define CP_FLAG_STATUS_IDLE 0
#define CP_FLAG_STATUS_BUSY 1
#define CP_FLAG_AUTO_SAVE ((3 * 32) + 4)
#define CP_FLAG_AUTO_SAVE_NOT_PENDING 0
#define CP_FLAG_AUTO_SAVE_PENDING 1
#define CP_FLAG_AUTO_LOAD ((3 * 32) + 5)
#define CP_FLAG_AUTO_LOAD_NOT_PENDING 0
#define CP_FLAG_AUTO_LOAD_PENDING 1
#define CP_FLAG_UNK54 ((3 * 32) + 6)
#define CP_FLAG_UNK54_CLEAR 0
#define CP_FLAG_UNK54_SET 1
#define CP_FLAG_ALWAYS ((3 * 32) + 8)
#define CP_FLAG_ALWAYS_FALSE 0
#define CP_FLAG_ALWAYS_TRUE 1
#define CP_FLAG_UNK57 ((3 * 32) + 9)
#define CP_FLAG_UNK57_CLEAR 0
#define CP_FLAG_UNK57_SET 1
#define CP_CTX 0x00100000
#define CP_CTX_COUNT 0x000fc000
#define CP_CTX_COUNT_SHIFT 14
#define CP_CTX_REG 0x00003fff
#define CP_LOAD_SR 0x00200000
#define CP_LOAD_SR_VALUE 0x000fffff
#define CP_BRA 0x00400000
#define CP_BRA_IP 0x0000ff00
#define CP_BRA_IP_SHIFT 8
#define CP_BRA_IF_CLEAR 0x00000080
#define CP_BRA_FLAG 0x0000007f
#define CP_WAIT 0x00500000
#define CP_WAIT_SET 0x00000080
#define CP_WAIT_FLAG 0x0000007f
#define CP_SET 0x00700000
#define CP_SET_1 0x00000080
#define CP_SET_FLAG 0x0000007f
#define CP_NEXT_TO_SWAP 0x00600007
#define CP_NEXT_TO_CURRENT 0x00600009
#define CP_SET_CONTEXT_POINTER 0x0060000a
#define CP_END 0x0060000e
#define CP_LOAD_MAGIC_UNK01 0x00800001 /* unknown */
#define CP_LOAD_MAGIC_NV44TCL 0x00800029 /* per-vs state (0x4497) */
#define CP_LOAD_MAGIC_NV40TCL 0x00800041 /* per-vs state (0x4097) */
#include "drmP.h"
#include "nouveau_drv.h"
#include "nouveau_grctx.h"
/* TODO:
* - get vs count from 0x1540
* - document unimplemented bits compared to nvidia
* - nsource handling
* - R0 & 0x0200 handling
* - single-vs handling
* - 400314 bit 0
*/
static int
nv40_graph_4097(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
if ((dev_priv->chipset & 0xf0) == 0x60)
return 0;
return !!(0x0baf & (1 << dev_priv->chipset));
}
static int
nv40_graph_vs_count(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
switch (dev_priv->chipset) {
case 0x47:
case 0x49:
case 0x4b:
return 8;
case 0x40:
return 6;
case 0x41:
case 0x42:
return 5;
case 0x43:
case 0x44:
case 0x46:
case 0x4a:
return 3;
case 0x4c:
case 0x4e:
case 0x67:
default:
return 1;
}
}
enum cp_label {
cp_check_load = 1,
cp_setup_auto_load,
cp_setup_load,
cp_setup_save,
cp_swap_state,
cp_swap_state3d_3_is_save,
cp_prepare_exit,
cp_exit,
};
static void
nv40_graph_construct_general(struct nouveau_grctx *ctx)
{
struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
int i;
cp_ctx(ctx, 0x4000a4, 1);
gr_def(ctx, 0x4000a4, 0x00000008);
cp_ctx(ctx, 0x400144, 58);
gr_def(ctx, 0x400144, 0x00000001);
cp_ctx(ctx, 0x400314, 1);
gr_def(ctx, 0x400314, 0x00000000);
cp_ctx(ctx, 0x400400, 10);
cp_ctx(ctx, 0x400480, 10);
cp_ctx(ctx, 0x400500, 19);
gr_def(ctx, 0x400514, 0x00040000);
gr_def(ctx, 0x400524, 0x55555555);
gr_def(ctx, 0x400528, 0x55555555);
gr_def(ctx, 0x40052c, 0x55555555);
gr_def(ctx, 0x400530, 0x55555555);
cp_ctx(ctx, 0x400560, 6);
gr_def(ctx, 0x400568, 0x0000ffff);
gr_def(ctx, 0x40056c, 0x0000ffff);
cp_ctx(ctx, 0x40057c, 5);
cp_ctx(ctx, 0x400710, 3);
gr_def(ctx, 0x400710, 0x20010001);
gr_def(ctx, 0x400714, 0x0f73ef00);
cp_ctx(ctx, 0x400724, 1);
gr_def(ctx, 0x400724, 0x02008821);
cp_ctx(ctx, 0x400770, 3);
if (dev_priv->chipset == 0x40) {
cp_ctx(ctx, 0x400814, 4);
cp_ctx(ctx, 0x400828, 5);
cp_ctx(ctx, 0x400840, 5);
gr_def(ctx, 0x400850, 0x00000040);
cp_ctx(ctx, 0x400858, 4);
gr_def(ctx, 0x400858, 0x00000040);
gr_def(ctx, 0x40085c, 0x00000040);
gr_def(ctx, 0x400864, 0x80000000);
cp_ctx(ctx, 0x40086c, 9);
gr_def(ctx, 0x40086c, 0x80000000);
gr_def(ctx, 0x400870, 0x80000000);
gr_def(ctx, 0x400874, 0x80000000);
gr_def(ctx, 0x400878, 0x80000000);
gr_def(ctx, 0x400888, 0x00000040);
gr_def(ctx, 0x40088c, 0x80000000);
cp_ctx(ctx, 0x4009c0, 8);
gr_def(ctx, 0x4009cc, 0x80000000);
gr_def(ctx, 0x4009dc, 0x80000000);
} else {
cp_ctx(ctx, 0x400840, 20);
if (!nv40_graph_4097(ctx->dev)) {
for (i = 0; i < 8; i++)
gr_def(ctx, 0x400860 + (i * 4), 0x00000001);
}
gr_def(ctx, 0x400880, 0x00000040);
gr_def(ctx, 0x400884, 0x00000040);
gr_def(ctx, 0x400888, 0x00000040);
cp_ctx(ctx, 0x400894, 11);
gr_def(ctx, 0x400894, 0x00000040);
if (nv40_graph_4097(ctx->dev)) {
for (i = 0; i < 8; i++)
gr_def(ctx, 0x4008a0 + (i * 4), 0x80000000);
}
cp_ctx(ctx, 0x4008e0, 2);
cp_ctx(ctx, 0x4008f8, 2);
if (dev_priv->chipset == 0x4c ||
(dev_priv->chipset & 0xf0) == 0x60)
cp_ctx(ctx, 0x4009f8, 1);
}
cp_ctx(ctx, 0x400a00, 73);
gr_def(ctx, 0x400b0c, 0x0b0b0b0c);
cp_ctx(ctx, 0x401000, 4);
cp_ctx(ctx, 0x405004, 1);
switch (dev_priv->chipset) {
case 0x47:
case 0x49:
case 0x4b:
cp_ctx(ctx, 0x403448, 1);
gr_def(ctx, 0x403448, 0x00001010);
break;
default:
cp_ctx(ctx, 0x403440, 1);
switch (dev_priv->chipset) {
case 0x40:
gr_def(ctx, 0x403440, 0x00000010);
break;
case 0x44:
case 0x46:
case 0x4a:
gr_def(ctx, 0x403440, 0x00003010);
break;
case 0x41:
case 0x42:
case 0x43:
case 0x4c:
case 0x4e:
case 0x67:
default:
gr_def(ctx, 0x403440, 0x00001010);
break;
}
break;
}
}
static void
nv40_graph_construct_state3d(struct nouveau_grctx *ctx)
{
struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
int i;
if (dev_priv->chipset == 0x40) {
cp_ctx(ctx, 0x401880, 51);
gr_def(ctx, 0x401940, 0x00000100);
} else
if (dev_priv->chipset == 0x46 || dev_priv->chipset == 0x47 ||
dev_priv->chipset == 0x49 || dev_priv->chipset == 0x4b) {
cp_ctx(ctx, 0x401880, 32);
for (i = 0; i < 16; i++)
gr_def(ctx, 0x401880 + (i * 4), 0x00000111);
if (dev_priv->chipset == 0x46)
cp_ctx(ctx, 0x401900, 16);
cp_ctx(ctx, 0x401940, 3);
}
cp_ctx(ctx, 0x40194c, 18);
gr_def(ctx, 0x401954, 0x00000111);
gr_def(ctx, 0x401958, 0x00080060);
gr_def(ctx, 0x401974, 0x00000080);
gr_def(ctx, 0x401978, 0xffff0000);
gr_def(ctx, 0x40197c, 0x00000001);
gr_def(ctx, 0x401990, 0x46400000);
if (dev_priv->chipset == 0x40) {
cp_ctx(ctx, 0x4019a0, 2);
cp_ctx(ctx, 0x4019ac, 5);
} else {
cp_ctx(ctx, 0x4019a0, 1);
cp_ctx(ctx, 0x4019b4, 3);
}
gr_def(ctx, 0x4019bc, 0xffff0000);
switch (dev_priv->chipset) {
case 0x46:
case 0x47:
case 0x49:
case 0x4b:
cp_ctx(ctx, 0x4019c0, 18);
for (i = 0; i < 16; i++)
gr_def(ctx, 0x4019c0 + (i * 4), 0x88888888);
break;
}
cp_ctx(ctx, 0x401a08, 8);
gr_def(ctx, 0x401a10, 0x0fff0000);
gr_def(ctx, 0x401a14, 0x0fff0000);
gr_def(ctx, 0x401a1c, 0x00011100);
cp_ctx(ctx, 0x401a2c, 4);
cp_ctx(ctx, 0x401a44, 26);
for (i = 0; i < 16; i++)
gr_def(ctx, 0x401a44 + (i * 4), 0x07ff0000);
gr_def(ctx, 0x401a8c, 0x4b7fffff);
if (dev_priv->chipset == 0x40) {
cp_ctx(ctx, 0x401ab8, 3);
} else {
cp_ctx(ctx, 0x401ab8, 1);
cp_ctx(ctx, 0x401ac0, 1);
}
cp_ctx(ctx, 0x401ad0, 8);
gr_def(ctx, 0x401ad0, 0x30201000);
gr_def(ctx, 0x401ad4, 0x70605040);
gr_def(ctx, 0x401ad8, 0xb8a89888);
gr_def(ctx, 0x401adc, 0xf8e8d8c8);
cp_ctx(ctx, 0x401b10, dev_priv->chipset == 0x40 ? 2 : 1);
gr_def(ctx, 0x401b10, 0x40100000);
cp_ctx(ctx, 0x401b18, dev_priv->chipset == 0x40 ? 6 : 5);
gr_def(ctx, 0x401b28, dev_priv->chipset == 0x40 ?
0x00000004 : 0x00000000);
cp_ctx(ctx, 0x401b30, 25);
gr_def(ctx, 0x401b34, 0x0000ffff);
gr_def(ctx, 0x401b68, 0x435185d6);
gr_def(ctx, 0x401b6c, 0x2155b699);
gr_def(ctx, 0x401b70, 0xfedcba98);
gr_def(ctx, 0x401b74, 0x00000098);
gr_def(ctx, 0x401b84, 0xffffffff);
gr_def(ctx, 0x401b88, 0x00ff7000);
gr_def(ctx, 0x401b8c, 0x0000ffff);
if (dev_priv->chipset != 0x44 && dev_priv->chipset != 0x4a &&
dev_priv->chipset != 0x4e)
cp_ctx(ctx, 0x401b94, 1);
cp_ctx(ctx, 0x401b98, 8);
gr_def(ctx, 0x401b9c, 0x00ff0000);
cp_ctx(ctx, 0x401bc0, 9);
gr_def(ctx, 0x401be0, 0x00ffff00);
cp_ctx(ctx, 0x401c00, 192);
for (i = 0; i < 16; i++) { /* fragment texture units */
gr_def(ctx, 0x401c40 + (i * 4), 0x00018488);
gr_def(ctx, 0x401c80 + (i * 4), 0x00028202);
gr_def(ctx, 0x401d00 + (i * 4), 0x0000aae4);
gr_def(ctx, 0x401d40 + (i * 4), 0x01012000);
gr_def(ctx, 0x401d80 + (i * 4), 0x00080008);
gr_def(ctx, 0x401e00 + (i * 4), 0x00100008);
}
for (i = 0; i < 4; i++) { /* vertex texture units */
gr_def(ctx, 0x401e90 + (i * 4), 0x0001bc80);
gr_def(ctx, 0x401ea0 + (i * 4), 0x00000202);
gr_def(ctx, 0x401ec0 + (i * 4), 0x00000008);
gr_def(ctx, 0x401ee0 + (i * 4), 0x00080008);
}
cp_ctx(ctx, 0x400f5c, 3);
gr_def(ctx, 0x400f5c, 0x00000002);
cp_ctx(ctx, 0x400f84, 1);
}
static void
nv40_graph_construct_state3d_2(struct nouveau_grctx *ctx)
{
struct drm_nouveau_private *dev_priv = ctx->dev->dev_private;
int i;
cp_ctx(ctx, 0x402000, 1);
cp_ctx(ctx, 0x402404, dev_priv->chipset == 0x40 ? 1 : 2);
switch (dev_priv->chipset) {
case 0x40:
gr_def(ctx, 0x402404, 0x00000001);
break;
case 0x4c:
case 0x4e:
case 0x67:
gr_def(ctx, 0x402404, 0x00000020);
break;
case 0x46:
case 0x49:
case 0x4b:
gr_def(ctx, 0x402404, 0x00000421);
break;
default:
gr_def(ctx, 0x402404, 0x00000021);
}
if (dev_priv->chipset != 0x40)
gr_def(ctx, 0x402408, 0x030c30c3);
switch (dev_priv->chipset) {
case 0x44:
case 0x46:
case 0x4a:
case 0x4c:
case 0x4e:
case 0x67:
cp_ctx(ctx, 0x402440, 1);
gr_def(ctx, 0x402440, 0x00011001);
break;
default:
break;
}
cp_ctx(ctx, 0x402480, dev_priv->chipset == 0x40 ? 8 : 9);
gr_def(ctx, 0x402488, 0x3e020200);
gr_def(ctx, 0x40248c, 0x00ffffff);
switch (dev_priv->chipset) {
case 0x40:
gr_def(ctx, 0x402490, 0x60103f00);
break;
case 0x47:
gr_def(ctx, 0x402490, 0x40103f00);
break;
case 0x41:
case 0x42:
case 0x49:
case 0x4b:
gr_def(ctx, 0x402490, 0x20103f00);
break;
default:
gr_def(ctx, 0x402490, 0x0c103f00);
break;
}
gr_def(ctx, 0x40249c, dev_priv->chipset <= 0x43 ?
0x00020000 : 0x00040000);
cp_ctx(ctx, 0x402500, 31);
gr_def(ctx, 0x402530, 0x00008100);
if (dev_priv->chipset == 0x40)
cp_ctx(ctx, 0x40257c, 6);
cp_ctx(ctx, 0x402594, 16);
cp_ctx(ctx, 0x402800, 17);
gr_def(ctx, 0x402800, 0x00000001);
switch (dev_priv->chipset) {
case 0x47:
case 0x49:
case 0x4b:
cp_ctx(ctx, 0x402864, 1);
gr_def(ctx, 0x402864, 0x00001001);
cp_ctx(ctx, 0x402870, 3);
gr_def(ctx, 0x402878, 0x00000003);
if (dev_priv->chipset != 0x47) { /* belong at end!! */
cp_ctx(ctx, 0x402900, 1);
cp_ctx(ctx, 0x402940, 1);
cp_ctx(ctx, 0x402980, 1);
cp_ctx(ctx, 0x4029c0, 1);
cp_ctx(ctx, 0x402a00, 1);
cp_ctx(ctx, 0x402a40, 1);
cp_ctx(ctx, 0x402a80, 1);
cp_ctx(ctx, 0x402ac0, 1);
}
break;
case 0x40:
cp_ctx(ctx, 0x402844, 1);
gr_def(ctx, 0x402844, 0x00000001);
cp_ctx(ctx, 0x402850, 1);
break;
default:
cp_ctx(ctx, 0x402844, 1);
gr_def(ctx, 0x402844, 0x00001001);
cp_ctx(ctx, 0x402850, 2);
gr_def(ctx, 0x402854, 0x00000003);
break;
}
cp_ctx(ctx, 0x402c00, 4);
gr_def(ctx, 0x402c00, dev_priv->chipset == 0x40 ?
0x80800001 : 0x00888001);
switch (dev_priv->chipset) {
case 0x47:
case 0x49:
case 0x4b:
cp_ctx(ctx, 0x402c20, 40);
for (i = 0; i < 32; i++)
gr_def(ctx, 0x402c40 + (i * 4), 0xffffffff);
cp_ctx(ctx, 0x4030b8, 13);
gr_def(ctx, 0x4030dc, 0x00000005);
gr_def(ctx, 0x4030e8, 0x0000ffff);
break;
default:
cp_ctx(ctx, 0x402c10, 4);
if (dev_priv->chipset == 0x40)
cp_ctx(ctx, 0x402c20, 36);
else
if (dev_priv->chipset <= 0x42)
cp_ctx(ctx, 0x402c20, 24);
else
if (dev_priv->chipset <= 0x4a)
cp_ctx(ctx, 0x402c20, 16);
else
cp_ctx(ctx, 0x402c20, 8);
cp_ctx(ctx, 0x402cb0, dev_priv->chipset == 0x40 ? 12 : 13);
gr_def(ctx, 0x402cd4, 0x00000005);
if (dev_priv->chipset != 0x40)
gr_def(ctx, 0x402ce0, 0x0000ffff);
break;
}
cp_ctx(ctx, 0x403400, dev_priv->chipset == 0x40 ? 4 : 3);
cp_ctx(ctx, 0x403410, dev_priv->chipset == 0x40 ? 4 : 3);
cp_ctx(ctx, 0x403420, nv40_graph_vs_count(ctx->dev));
for (i = 0; i < nv40_graph_vs_count(ctx->dev); i++)
gr_def(ctx, 0x403420 + (i * 4), 0x00005555);
if (dev_priv->chipset != 0x40) {
cp_ctx(ctx, 0x403600, 1);
gr_def(ctx, 0x403600, 0x00000001);
}
cp_ctx(ctx, 0x403800, 1);
cp_ctx(ctx, 0x403c18, 1);
gr_def(ctx, 0x403c18, 0x00000001);
switch (dev_priv->chipset) {
case 0x46:
case 0x47:
case 0x49:
case 0x4b:
cp_ctx(ctx, 0x405018, 1);
gr_def(ctx, 0x405018, 0x08e00001);
cp_ctx(ctx, 0x405c24, 1);
gr_def(ctx, 0x405c24, 0x000e3000);
break;
}
if (dev_priv->chipset != 0x4e)
cp_ctx(ctx, 0x405800, 11);
cp_ctx(ctx, 0x407000, 1);
}
static void
nv40_graph_construct_state3d_3(struct nouveau_grctx *ctx)
{
int len = nv40_graph_4097(ctx->dev) ? 0x0684 : 0x0084;
cp_out (ctx, 0x300000);
cp_lsr (ctx, len - 4);
cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_swap_state3d_3_is_save);
cp_lsr (ctx, len);
cp_name(ctx, cp_swap_state3d_3_is_save);
cp_out (ctx, 0x800001);
ctx->ctxvals_pos += len;
}
static void
nv40_graph_construct_shader(struct nouveau_grctx *ctx)
{
struct drm_device *dev = ctx->dev;
struct drm_nouveau_private *dev_priv = dev->dev_private;
struct nouveau_gpuobj *obj = ctx->data;
int vs, vs_nr, vs_len, vs_nr_b0, vs_nr_b1, b0_offset, b1_offset;
int offset, i;
vs_nr = nv40_graph_vs_count(ctx->dev);
vs_nr_b0 = 363;
vs_nr_b1 = dev_priv->chipset == 0x40 ? 128 : 64;
if (dev_priv->chipset == 0x40) {
b0_offset = 0x2200/4; /* 33a0 */
b1_offset = 0x55a0/4; /* 1500 */
vs_len = 0x6aa0/4;
} else
if (dev_priv->chipset == 0x41 || dev_priv->chipset == 0x42) {
b0_offset = 0x2200/4; /* 2200 */
b1_offset = 0x4400/4; /* 0b00 */
vs_len = 0x4f00/4;
} else {
b0_offset = 0x1d40/4; /* 2200 */
b1_offset = 0x3f40/4; /* 0b00 : 0a40 */
vs_len = nv40_graph_4097(dev) ? 0x4a40/4 : 0x4980/4;
}
cp_lsr(ctx, vs_len * vs_nr + 0x300/4);
cp_out(ctx, nv40_graph_4097(dev) ? 0x800041 : 0x800029);
offset = ctx->ctxvals_pos;
ctx->ctxvals_pos += (0x0300/4 + (vs_nr * vs_len));
if (ctx->mode != NOUVEAU_GRCTX_VALS)
return;
offset += 0x0280/4;
for (i = 0; i < 16; i++, offset += 2)
nv_wo32(dev, obj, offset, 0x3f800000);
for (vs = 0; vs < vs_nr; vs++, offset += vs_len) {
for (i = 0; i < vs_nr_b0 * 6; i += 6)
nv_wo32(dev, obj, offset + b0_offset + i, 0x00000001);
for (i = 0; i < vs_nr_b1 * 4; i += 4)
nv_wo32(dev, obj, offset + b1_offset + i, 0x3f800000);
}
}
void
nv40_grctx_init(struct nouveau_grctx *ctx)
{
/* decide whether we're loading/unloading the context */
cp_bra (ctx, AUTO_SAVE, PENDING, cp_setup_save);
cp_bra (ctx, USER_SAVE, PENDING, cp_setup_save);
cp_name(ctx, cp_check_load);
cp_bra (ctx, AUTO_LOAD, PENDING, cp_setup_auto_load);
cp_bra (ctx, USER_LOAD, PENDING, cp_setup_load);
cp_bra (ctx, ALWAYS, TRUE, cp_exit);
/* setup for context load */
cp_name(ctx, cp_setup_auto_load);
cp_wait(ctx, STATUS, IDLE);
cp_out (ctx, CP_NEXT_TO_SWAP);
cp_name(ctx, cp_setup_load);
cp_wait(ctx, STATUS, IDLE);
cp_set (ctx, SWAP_DIRECTION, LOAD);
cp_out (ctx, 0x00910880); /* ?? */
cp_out (ctx, 0x00901ffe); /* ?? */
cp_out (ctx, 0x01940000); /* ?? */
cp_lsr (ctx, 0x20);
cp_out (ctx, 0x0060000b); /* ?? */
cp_wait(ctx, UNK57, CLEAR);
cp_out (ctx, 0x0060000c); /* ?? */
cp_bra (ctx, ALWAYS, TRUE, cp_swap_state);
/* setup for context save */
cp_name(ctx, cp_setup_save);
cp_set (ctx, SWAP_DIRECTION, SAVE);
/* general PGRAPH state */
cp_name(ctx, cp_swap_state);
cp_pos (ctx, 0x00020/4);
nv40_graph_construct_general(ctx);
cp_wait(ctx, STATUS, IDLE);
/* 3D state, block 1 */
cp_bra (ctx, UNK54, CLEAR, cp_prepare_exit);
nv40_graph_construct_state3d(ctx);
cp_wait(ctx, STATUS, IDLE);
/* 3D state, block 2 */
nv40_graph_construct_state3d_2(ctx);
/* Some other block of "random" state */
nv40_graph_construct_state3d_3(ctx);
/* Per-vertex shader state */
cp_pos (ctx, ctx->ctxvals_pos);
nv40_graph_construct_shader(ctx);
/* pre-exit state updates */
cp_name(ctx, cp_prepare_exit);
cp_bra (ctx, SWAP_DIRECTION, SAVE, cp_check_load);
cp_bra (ctx, USER_SAVE, PENDING, cp_exit);
cp_out (ctx, CP_NEXT_TO_CURRENT);
cp_name(ctx, cp_exit);
cp_set (ctx, USER_SAVE, NOT_PENDING);
cp_set (ctx, USER_LOAD, NOT_PENDING);
cp_out (ctx, CP_END);
}
drivers/gpu/drm/nouveau/nv50_graph.c
View file @ 054b93e4
......@@ -107,9 +107,13 @@ nv50_graph_init_regs(struct drm_device *dev)
static int
nv50_graph_init_ctxctl(struct drm_device *dev)
{
struct drm_nouveau_private *dev_priv = dev->dev_private;
NV_DEBUG(dev, "\n");
nv40_grctx_init(dev);
nouveau_grctx_prog_load(dev);
if (!dev_priv->engine.graph.ctxprog)
dev_priv->engine.graph.accel_blocked = true;
nv_wr32(dev, 0x400320, 4);
nv_wr32(dev, NV40_PGRAPH_CTXCTL_CUR, 0);
......@@ -140,7 +144,7 @@ void
nv50_graph_takedown(struct drm_device *dev)
{
NV_DEBUG(dev, "\n");
nv40_grctx_fini(dev);
nouveau_grctx_fini(dev);
}
void
......@@ -207,7 +211,7 @@ nv50_graph_create_context(struct nouveau_channel *chan)
dev_priv->engine.instmem.finish_access(dev);
dev_priv->engine.instmem.prepare_access(dev, true);
nv40_grctx_vals_load(dev, ctx);
nouveau_grctx_vals_load(dev, ctx);
nv_wo32(dev, ctx, 0x00000/4, chan->ramin->instance >> 12);
if ((dev_priv->chipset & 0xf0) == 0xa0)
nv_wo32(dev, ctx, 0x00004/4, 0x00000000);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7