From: Nicolai Hähnle <nicolai.haeh...@amd.com>
This probably has only minor performance effects, but it simplifies some
subsequent code slightly.
Ideally, it could also be used to simplify the handling of slab buffers
in the same way, but unfortunately that's not possible as long as we need
indices for relocations.
---
src/gallium/winsys/amdgpu/drm/amdgpu_cs.c | 41 +++++++++++++++++++------------
src/gallium/winsys/amdgpu/drm/amdgpu_cs.h | 4 ++-
2 files changed, 28 insertions(+), 17 deletions(-)
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
index 2520aff..e2d3a45 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c
@@ -318,56 +318,42 @@ static int
amdgpu_do_add_real_buffer(struct amdgpu_cs_context *cs, struct
amdgpu_winsys_bo *bo)
{
struct amdgpu_cs_buffer *buffer;
int idx;
/* New buffer, check if the backing array is large enough. */
if (cs->num_real_buffers >= cs->max_real_buffers) {
unsigned new_max =
MAX2(cs->max_real_buffers + 16, (unsigned)(cs->max_real_buffers *
1.3));
struct amdgpu_cs_buffer *new_buffers;
- amdgpu_bo_handle *new_handles;
- uint8_t *new_flags;
new_buffers = MALLOC(new_max * sizeof(*new_buffers));
- new_handles = MALLOC(new_max * sizeof(*new_handles));
- new_flags = MALLOC(new_max * sizeof(*new_flags));
- if (!new_buffers || !new_handles || !new_flags) {
+ if (!new_buffers) {
fprintf(stderr, "amdgpu_do_add_buffer: allocation failed\n");
FREE(new_buffers);
- FREE(new_handles);
- FREE(new_flags);
return -1;
}
memcpy(new_buffers, cs->real_buffers, cs->num_real_buffers *
sizeof(*new_buffers));
- memcpy(new_handles, cs->handles, cs->num_real_buffers *
sizeof(*new_handles));
- memcpy(new_flags, cs->flags, cs->num_real_buffers * sizeof(*new_flags));
FREE(cs->real_buffers);
- FREE(cs->handles);
- FREE(cs->flags);
cs->max_real_buffers = new_max;
cs->real_buffers = new_buffers;
- cs->handles = new_handles;
- cs->flags = new_flags;
}
idx = cs->num_real_buffers;
buffer = &cs->real_buffers[idx];
memset(buffer, 0, sizeof(*buffer));
amdgpu_winsys_bo_reference(&buffer->bo, bo);
- cs->handles[idx] = bo->bo;
- cs->flags[idx] = 0;
p_atomic_inc(&bo->num_cs_references);
cs->num_real_buffers++;
return idx;
}
static int
amdgpu_lookup_or_add_real_buffer(struct amdgpu_cs *acs, struct
amdgpu_winsys_bo *bo)
{
struct amdgpu_cs_context *cs = acs->csc;
@@ -475,21 +461,20 @@ static unsigned amdgpu_cs_add_buffer(struct
radeon_winsys_cs *rcs,
index = buffer->u.slab.real_idx;
} else {
index = amdgpu_lookup_or_add_real_buffer(acs, bo);
if (index < 0)
return 0;
}
buffer = &cs->real_buffers[index];
buffer->u.real.priority_usage |= 1llu << priority;
buffer->usage |= usage;
- cs->flags[index] = MAX2(cs->flags[index], priority / 4);
cs->last_added_bo = bo;
cs->last_added_bo_index = index;
cs->last_added_bo_usage = buffer->usage;
cs->last_added_bo_priority_usage = buffer->u.real.priority_usage;
return index;
}
static bool amdgpu_ib_new_buffer(struct amdgpu_winsys *ws, struct amdgpu_ib
*ib)
{
@@ -1057,24 +1042,48 @@ void amdgpu_cs_submit_ib(void *job, int thread_index)
assert(num < ws->num_buffers);
handles[num++] = bo->bo;
}
r = amdgpu_bo_list_create(ws->dev, ws->num_buffers,
handles, NULL,
&cs->request.resources);
free(handles);
mtx_unlock(&ws->global_bo_list_lock);
} else {
+ if (cs->max_real_submit < cs->num_real_buffers) {
+ FREE(cs->handles);
+ FREE(cs->flags);
+
+ cs->handles = MALLOC(sizeof(*cs->handles) * cs->num_real_buffers);
+ cs->flags = MALLOC(sizeof(*cs->flags) * cs->num_real_buffers);
+
+ if (!cs->handles || !cs->flags) {
+ cs->max_real_submit = 0;
+ r = -ENOMEM;
+ goto bo_list_error;
+ }
+ }
+
+ for (i = 0; i < cs->num_real_buffers; ++i) {
+ struct amdgpu_cs_buffer *buffer = &cs->real_buffers[i];
+
+ assert(buffer->u.real.priority_usage != 0);
+
+ cs->handles[i] = buffer->bo->bo;
+ cs->flags[i] = (util_last_bit64(buffer->u.real.priority_usage) - 1) /
4;
+ }
+
r = amdgpu_bo_list_create(ws->dev, cs->num_real_buffers,
cs->handles, cs->flags,
&cs->request.resources);
}
+bo_list_error:
if (r) {
fprintf(stderr, "amdgpu: buffer list creation failed (%d)\n", r);
cs->request.resources = NULL;
amdgpu_fence_signalled(cs->fence);
cs->error_code = r;
goto cleanup;
}
if (acs->ctx->num_rejected_cs)
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.h
b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.h
index 495d55b..bdf7cb2 100644
--- a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.h
+++ b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.h
@@ -77,23 +77,25 @@ struct amdgpu_ib {
enum ib_type ib_type;
};
struct amdgpu_cs_context {
struct amdgpu_cs_request request;
struct amdgpu_cs_ib_info ib[IB_NUM];
/* Buffers. */
unsigned max_real_buffers;
unsigned num_real_buffers;
+ struct amdgpu_cs_buffer *real_buffers;
+
+ unsigned max_real_submit;
amdgpu_bo_handle *handles;
uint8_t *flags;
- struct amdgpu_cs_buffer *real_buffers;
unsigned num_slab_buffers;
unsigned max_slab_buffers;
struct amdgpu_cs_buffer *slab_buffers;
int buffer_indices_hashlist[4096];
struct amdgpu_winsys_bo *last_added_bo;
unsigned last_added_bo_index;
unsigned last_added_bo_usage;
--
2.9.3
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
