Nouveau - Aug 2013 - [PATCH 1/4] nouveau: fix number of surfaces in video buffer, use defines

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
---

The pipe_surface miscount caused issues for my (failed) attempt at getting
vdpau to work with pmpeg.

 src/gallium/drivers/nouveau/nouveau_video.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/gallium/drivers/nouveau/nouveau_video.h
b/src/gallium/drivers/nouveau/nouveau_video.h
index 1d6ced0..be53758 100644
--- a/src/gallium/drivers/nouveau/nouveau_video.h
+++ b/src/gallium/drivers/nouveau/nouveau_video.h
@@ -10,10 +10,10 @@
 struct nouveau_video_buffer {
    struct pipe_video_buffer base;
    unsigned num_planes;
-   struct pipe_resource     *resources[3];
-   struct pipe_sampler_view *sampler_view_planes[3];
-   struct pipe_sampler_view *sampler_view_components[3];
-   struct pipe_surface      *surfaces[3];
+   struct pipe_resource     *resources[VL_NUM_COMPONENTS];
+   struct pipe_sampler_view *sampler_view_planes[VL_NUM_COMPONENTS];
+   struct pipe_sampler_view *sampler_view_components[VL_NUM_COMPONENTS];
+   struct pipe_surface      *surfaces[VL_NUM_COMPONENTS * 2];
 };
 
 struct nouveau_decoder {
-- 
1.8.1.5

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
---

I don't think anything looks at this right now, but seems good to store.
(Also
ran into it with my vdpau attempt.)

 src/gallium/drivers/nouveau/nouveau_video.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/gallium/drivers/nouveau/nouveau_video.c
b/src/gallium/drivers/nouveau/nouveau_video.c
index 9357508..2e7f9b1 100644
--- a/src/gallium/drivers/nouveau/nouveau_video.c
+++ b/src/gallium/drivers/nouveau/nouveau_video.c
@@ -810,6 +810,7 @@ nouveau_video_buffer_create(struct pipe_context *pipe,
    buffer->base.get_sampler_view_components =
nouveau_video_buffer_sampler_view_components;
    buffer->base.get_surfaces = nouveau_video_buffer_surfaces;
    buffer->base.chroma_format = templat->chroma_format;
+   buffer->base.buffer_format = templat->buffer_format;
    buffer->base.width = width;
    buffer->base.height = height;
    buffer->num_planes = 2;
-- 
1.8.1.5

Force the format to be the reasonable format that doesn't require an
inverse z-scan.

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
---

Tested this on both nv44 and nv96, and it works, all at the same time. Without
the 0x72 command, they default to different scan orders. Putting the 0x72
command only at the beginning (in nouveau_vpe_init) seemed to cause pmpeg to
hang.

 src/gallium/drivers/nouveau/nouveau_video.c | 27 ++++++++++++---------------
 src/gallium/drivers/nv30/nv30_context.c     |  2 ++
 src/gallium/drivers/nv30/nv30_screen.c      |  1 +
 3 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/src/gallium/drivers/nouveau/nouveau_video.c
b/src/gallium/drivers/nouveau/nouveau_video.c
index 2e7f9b1..cbe2477 100644
--- a/src/gallium/drivers/nouveau/nouveau_video.c
+++ b/src/gallium/drivers/nouveau/nouveau_video.c
@@ -108,21 +108,10 @@ nouveau_vpe_mb_dct_blocks(struct nouveau_decoder *dec,
const struct pipe_mpeg12_
    short *db = mb->blocks;
    for (cbb = 0x20; cbb > 0; cbb >>= 1) {
       if (cbb & cbp) {
-         static const int lookup[64] = {
-             0, 1, 8,16, 9, 2, 3,10,
-            17,24,32,25,18,11, 4, 5,
-            12,19,26,33,40,48,41,34,
-            27,20,13, 6, 7,14,21,28,
-            35,42,49,56,57,50,43,36,
-            29,22,15,23,30,37,44,51,
-            58,59,52,45,38,31,39,46,
-            53,60,61,54,47,55,62,63
-         };
-         int i, j = 0, found = 0;
+         int i, found = 0;
          for (i = 0; i < 64; ++i) {
-            if (!db[lookup[i]]) { j += 2; continue; }
-            dec->data[dec->data_pos++] = (db[lookup[i]] << 16) | j;
-            j = 0;
+            if (!db[i]) continue;
+            dec->data[dec->data_pos++] = (db[i] << 16) | (i * 2);
             found = 1;
          }
          if (found)
@@ -443,6 +432,11 @@ nouveau_decoder_decode_macroblock(struct pipe_video_decoder
*decoder,
       dec->past = nouveau_decoder_surface_index(dec, desc->ref[0]);
 
    if (nouveau_vpe_init(dec)) return;
+
+   /* initialize scan order */
+   nouveau_vpe_write(dec, 0x720000c0);
+   nouveau_vpe_write(dec, dec->data_pos);
+
    mb = (const struct pipe_mpeg12_macroblock *)pipe_mb;
    for (i = 0; i < num_macroblocks; ++i, mb++) {
       if (mb->macroblock_type & PIPE_MPEG12_MB_TYPE_INTRA) {
@@ -528,6 +522,8 @@ nouveau_create_decoder(struct pipe_context *context,
       goto vl;
    if (screen->device->chipset >= 0x98 &&
screen->device->chipset != 0xa0)
       goto vl;
+   if (screen->device->chipset < 0x31 || screen->device->chipset
== 0x35)
+      goto vl;
 
    dec = CALLOC_STRUCT(nouveau_decoder);
    if (!dec)
@@ -793,7 +789,8 @@ nouveau_video_buffer_create(struct pipe_context *pipe,
     * and it only supports the NV12 format
     */
    if (templat->buffer_format != PIPE_FORMAT_NV12 ||
getenv("XVMC_VL") ||
-       (screen->device->chipset >= 0x98 &&
screen->device->chipset != 0xa0))
+       (screen->device->chipset >= 0x98 &&
screen->device->chipset != 0xa0) ||
+       screen->device->chipset < 0x31 || screen->device->chipset
== 0x35)
       return vl_video_buffer_create(pipe, templat);
 
    assert(templat->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420);
diff --git a/src/gallium/drivers/nv30/nv30_context.c
b/src/gallium/drivers/nv30/nv30_context.c
index bd05042..e872c02 100644
--- a/src/gallium/drivers/nv30/nv30_context.c
+++ b/src/gallium/drivers/nv30/nv30_context.c
@@ -257,5 +257,7 @@ nv30_context_create(struct pipe_screen *pscreen, void *priv)
       return NULL;
    }
 
+   nouveau_context_init_vdec(&nv30->base);
+
    return pipe;
 }
diff --git a/src/gallium/drivers/nv30/nv30_screen.c
b/src/gallium/drivers/nv30/nv30_screen.c
index 3d55d6f..40e8b5f 100644
--- a/src/gallium/drivers/nv30/nv30_screen.c
+++ b/src/gallium/drivers/nv30/nv30_screen.c
@@ -377,6 +377,7 @@ nv30_screen_create(struct nouveau_device *dev)
    pscreen->context_create = nv30_context_create;
    pscreen->is_format_supported = nv30_screen_is_format_supported;
    nv30_resource_screen_init(pscreen);
+   nouveau_screen_init_vdec(&screen->base);
 
    screen->base.fence.emit = nv30_screen_fence_emit;
    screen->base.fence.update = nv30_screen_fence_update;
-- 
1.8.1.5

This also allows people who don't want to install the binary blobs
required for VP2 to still get MPEG decoding.

Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
---
 src/gallium/drivers/nv50/nv50_context.c | 3 ++-
 src/gallium/drivers/nv50/nv50_screen.c  | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/src/gallium/drivers/nv50/nv50_context.c
b/src/gallium/drivers/nv50/nv50_context.c
index 79a0473..185d241 100644
--- a/src/gallium/drivers/nv50/nv50_context.c
+++ b/src/gallium/drivers/nv50/nv50_context.c
@@ -258,7 +258,8 @@ nv50_create(struct pipe_screen *pscreen, void *priv)
    draw_set_rasterize_stage(nv50->draw, nv50_draw_render_stage(nv50));
 #endif
 
-   if (screen->base.device->chipset < 0x84) {
+   if (screen->base.device->chipset < 0x84 ||
+       debug_get_bool_option("NOUVEAU_PMPEG", FALSE)) {
       /* PMPEG */
       nouveau_context_init_vdec(&nv50->base);
    } else if (screen->base.device->chipset < 0x98 ||
diff --git a/src/gallium/drivers/nv50/nv50_screen.c
b/src/gallium/drivers/nv50/nv50_screen.c
index 2951eb4..0cbee5d 100644
--- a/src/gallium/drivers/nv50/nv50_screen.c
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -647,7 +647,8 @@ nv50_screen_create(struct nouveau_device *dev)
 
    nv50_screen_init_resource_functions(pscreen);
 
-   if (screen->base.device->chipset < 0x84) {
+   if (screen->base.device->chipset < 0x84 ||
+       debug_get_bool_option("NOUVEAU_PMPEG", FALSE)) {
       /* PMPEG */
       nouveau_screen_init_vdec(&screen->base);
    } else if (screen->base.device->chipset < 0x98 ||
-- 
1.8.1.5