On 05/12/14 14:18, Ilia Mirkin wrote:
On Fri, Dec 5, 2014 at 9:16 AM, Jose Fonseca <jfons...@vmware.com> wrote:
From: José Fonseca <jfons...@vmware.com>
Matches what u_vbuf_get_minmax_index() does.
Hm, nouveau nv50 (and probably nvc0) does:
if (ib->buffer) {
nv50->idxbuf.offset = ib->offset;
BCTX_REFN(nv50->bufctx_3d, INDEX, nv04_resource(ib->buffer), RD);
} else {
nv50->idxbuf.user_buffer = ib->user_buffer;
}
Is the offset really supposed to be applied to the user buffer?
Nouveau is the odd ball here, as softpipe/llvmpipe apply it:
src/gallium/drivers/llvmpipe/lp_draw_arrays.c
src/gallium/drivers/softpipe/sp_draw_arrays.c
> I
> figured the point of the offset was that you couldn't provide an
> offset into a pipe_resource otherwise, while with a user ptr, you can
> just add it in yourself...
True. Still it's better to apply these things uniformly, because the
user vs buffer pointer is a distinction which rapidly disappears inside
a pipe driver. (E.g, the driver can promote the user pointer to a buffer
internally, and then the information of whether offset should or not be
taken in account gets lost)
Jose
---
src/gallium/auxiliary/indices/u_primconvert.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/gallium/auxiliary/indices/u_primconvert.c
b/src/gallium/auxiliary/indices/u_primconvert.c
index 4632781..eba1f9e 100644
--- a/src/gallium/auxiliary/indices/u_primconvert.c
+++ b/src/gallium/auxiliary/indices/u_primconvert.c
@@ -137,8 +137,9 @@ util_primconvert_draw_vbo(struct primconvert_context *pc,
src = ib->user_buffer;
if (!src) {
src = pipe_buffer_map(pc->pipe, ib->buffer,
- PIPE_TRANSFER_READ, &src_transfer) + ib->offset;
+ PIPE_TRANSFER_READ, &src_transfer);
}
+ src = (const uint8_t *)src + ib->offset;
}
else {
u_index_generator(pc->primtypes_mask,
--
2.1.0
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
