radeonsi: Allocate chunks of CE ram.

v2: Use 32 byte alignment.

v3: Don't allocate CE space for vertex buffer descriptors.

Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index 7fc1461ec9b6d3ea78194cc87702741ea45d4a40..c41923dabc82cc10cf7e9cac9bafd71cd97e10fd 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -98,7 +98,8 @@ static void si_init_descriptors(struct si_descriptors *desc,
                                unsigned shader_userdata_index,
                                unsigned element_dw_size,
                                unsigned num_elements,
-                               const uint32_t *null_descriptor)
+                               const uint32_t *null_descriptor,
+                               unsigned *ce_offset)
 {
        int i;
 
@@ -110,6 +111,13 @@ static void si_init_descriptors(struct si_descriptors *desc,
        desc->list_dirty = true; /* upload the list before the next draw */
        desc->shader_userdata_offset = shader_userdata_index * 4;
 
+       if (ce_offset) {
+               desc->ce_offset = *ce_offset;
+
+               /* make sure that ce_offset stays 32 byte aligned */
+               *ce_offset += align(element_dw_size * num_elements * 4, 32);
+       }
+
        /* Initialize the array to NULL descriptors if the element size is 8. */
        if (null_descriptor) {
                assert(element_dw_size % 8 == 0);
@@ -511,14 +519,15 @@ static void si_init_buffer_resources(struct si_buffer_resources *buffers,
                                     unsigned num_buffers,
                                     unsigned shader_userdata_index,
                                     enum radeon_bo_usage shader_usage,
-                                    enum radeon_bo_priority priority)
+                                    enum radeon_bo_priority priority,
+                                    unsigned *ce_offset)
 {
        buffers->shader_usage = shader_usage;
        buffers->priority = priority;
        buffers->buffers = CALLOC(num_buffers, sizeof(struct pipe_resource*));
 
        si_init_descriptors(&buffers->desc, shader_userdata_index, 4,
-                           num_buffers, NULL);
+                           num_buffers, NULL, ce_offset);
 }
 
 static void si_release_buffer_resources(struct si_buffer_resources *buffers)
@@ -1326,29 +1335,35 @@ void si_emit_shader_userdata(struct si_context *sctx, struct r600_atom *atom)
 void si_init_all_descriptors(struct si_context *sctx)
 {
        int i;
+       unsigned ce_offset = 0;
 
        for (i = 0; i < SI_NUM_SHADERS; i++) {
                si_init_buffer_resources(&sctx->const_buffers[i],
                                         SI_NUM_CONST_BUFFERS, SI_SGPR_CONST_BUFFERS,
-                                        RADEON_USAGE_READ, RADEON_PRIO_CONST_BUFFER);
+                                        RADEON_USAGE_READ, RADEON_PRIO_CONST_BUFFER,
+                                        &ce_offset);
                si_init_buffer_resources(&sctx->rw_buffers[i],
                                         SI_NUM_RW_BUFFERS, SI_SGPR_RW_BUFFERS,
-                                        RADEON_USAGE_READWRITE, RADEON_PRIO_RINGS_STREAMOUT);
+                                        RADEON_USAGE_READWRITE, RADEON_PRIO_RINGS_STREAMOUT,
+                                        &ce_offset);
                si_init_buffer_resources(&sctx->shader_buffers[i],
                                         SI_NUM_SHADER_BUFFERS, SI_SGPR_SHADER_BUFFERS,
-                                        RADEON_USAGE_READWRITE, RADEON_PRIO_SHADER_RW_BUFFER);
+                                        RADEON_USAGE_READWRITE, RADEON_PRIO_SHADER_RW_BUFFER,
+                                        &ce_offset);
 
                si_init_descriptors(&sctx->samplers[i].views.desc,
                                    SI_SGPR_SAMPLERS, 16, SI_NUM_SAMPLERS,
-                                   null_texture_descriptor);
+                                   null_texture_descriptor, &ce_offset);
 
                si_init_descriptors(&sctx->images[i].desc,
                                    SI_SGPR_IMAGES, 8, SI_NUM_IMAGES,
-                                   null_image_descriptor);
+                                   null_image_descriptor, &ce_offset);
        }
 
        si_init_descriptors(&sctx->vertex_buffers, SI_SGPR_VERTEX_BUFFERS,
-                           4, SI_NUM_VERTEX_BUFFERS, NULL);
+                           4, SI_NUM_VERTEX_BUFFERS, NULL, NULL);
+
+       assert(ce_offset <= 32768);
 
        /* Set pipe_context functions. */
        sctx->b.b.bind_sampler_states = si_bind_sampler_states;

diff --git a/src/gallium/drivers/radeonsi/si_state.h b/src/gallium/drivers/radeonsi/si_state.h
index 6748f802c7d10e4aab83dc8d6faf18d7ef928031..fbdc8ee64bc3732b4aec881e6b3f0e9fac8ed098 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_state.h
+++ b/src/gallium/drivers/radeonsi/si_state.h
@@ -198,6 +198,9 @@ struct si_descriptors {
        struct r600_resource *buffer;
        unsigned buffer_offset;
 
+       /* Offset in CE RAM */
+       unsigned ce_offset;
+
        /* The i-th bit is set if that element is enabled (non-NULL resource). */
        uint64_t enabled_mask;