/* The kernel parameters are stored a vtx buffer (ID=0), besides the explicit
* kernel parameters there are inplicit parameters that need to be stored
* in the vertex buffer as well. Here is how these parameters are organized in
* the buffer:
*
* DWORDS 0-2: Number of work groups in each dimension (x,y,z)
* DWORDS 3-5: Number of global work items in each dimension (x,y,z)
* DWORDS 6-8: Number of work items within each work group in each dimension
* (x,y,z)
* DWORDS 9+ : Kernel parameters
*/
void evergreen_compute_upload_input(
struct pipe_context *ctx_,
const uint *block_layout,
const uint *grid_layout,
const void *input)
{
struct r600_context *ctx = (struct r600_context *)ctx_;
struct r600_pipe_compute *shader = ctx->cs_shader_state.shader;
int i;
/* We need to reserve 9 dwords (36 bytes) for implicit kernel
* parameters.
*/
unsigned input_size = shader->input_size + 36;
uint32_t * num_work_groups_start;
uint32_t * global_size_start;
uint32_t * local_size_start;
uint32_t * kernel_parameters_start;
if (shader->input_size == 0) {
return;
}
if (!shader->kernel_param) {
/* Add space for the grid dimensions */
shader->kernel_param = r600_compute_buffer_alloc_vram(
ctx->screen, input_size);
}
num_work_groups_start = r600_buffer_mmap_sync_with_rings(ctx, shader->kernel_param, PIPE_TRANSFER_WRITE);
global_size_start = num_work_groups_start + (3 * (sizeof(uint) /4));
local_size_start = global_size_start + (3 * (sizeof(uint)) / 4);
kernel_parameters_start = local_size_start + (3 * (sizeof(uint)) / 4);
/* Copy the work group size */
memcpy(num_work_groups_start, grid_layout, 3 * sizeof(uint));
/* Copy the global size */
for (i = 0; i < 3; i++) {
global_size_start[i] = grid_layout[i] * block_layout[i];
}
/* Copy the local dimensions */
memcpy(local_size_start, block_layout, 3 * sizeof(uint));
/* Copy the kernel inputs */
memcpy(kernel_parameters_start, input, shader->input_size);
for (i = 0; i < (input_size / 4); i++) {
COMPUTE_DBG(ctx->screen, "input %i : %i\n", i,
((unsigned*)num_work_groups_start)[i]);
}
ctx->ws->buffer_unmap(shader->kernel_param->cs_buf);
/* ID=0 is reserved for the parameters */
evergreen_cs_set_constant_buffer(ctx, 0, 0, input_size,
(struct pipe_resource*)shader->kernel_param);
}
/* The kernel parameters are stored a vtx buffer (ID=0), besides the explicit
* kernel parameters there are implicit parameters that need to be stored
* in the vertex buffer as well. Here is how these parameters are organized in
* the buffer:
*
* DWORDS 0-2: Number of work groups in each dimension (x,y,z)
* DWORDS 3-5: Number of global work items in each dimension (x,y,z)
* DWORDS 6-8: Number of work items within each work group in each dimension
* (x,y,z)
* DWORDS 9+ : Kernel parameters
*/
void evergreen_compute_upload_input(
struct pipe_context *ctx_,
const uint *block_layout,
const uint *grid_layout,
const void *input)
{
struct r600_context *ctx = (struct r600_context *)ctx_;
struct r600_pipe_compute *shader = ctx->cs_shader_state.shader;
unsigned i;
/* We need to reserve 9 dwords (36 bytes) for implicit kernel
* parameters.
*/
unsigned input_size = shader->input_size + 36;
uint32_t * num_work_groups_start;
uint32_t * global_size_start;
uint32_t * local_size_start;
uint32_t * kernel_parameters_start;
struct pipe_box box;
struct pipe_transfer *transfer = NULL;
if (shader->input_size == 0) {
return;
}
if (!shader->kernel_param) {
/* Add space for the grid dimensions */
shader->kernel_param = (struct r600_resource *)
pipe_buffer_create(ctx_->screen, PIPE_BIND_CUSTOM,
PIPE_USAGE_IMMUTABLE, input_size);
}
u_box_1d(0, input_size, &box);
num_work_groups_start = ctx_->transfer_map(ctx_,
(struct pipe_resource*)shader->kernel_param,
0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD_RANGE,
&box, &transfer);
global_size_start = num_work_groups_start + (3 * (sizeof(uint) /4));
local_size_start = global_size_start + (3 * (sizeof(uint)) / 4);
kernel_parameters_start = local_size_start + (3 * (sizeof(uint)) / 4);
/* Copy the work group size */
memcpy(num_work_groups_start, grid_layout, 3 * sizeof(uint));
/* Copy the global size */
for (i = 0; i < 3; i++) {
global_size_start[i] = grid_layout[i] * block_layout[i];
}
/* Copy the local dimensions */
memcpy(local_size_start, block_layout, 3 * sizeof(uint));
/* Copy the kernel inputs */
memcpy(kernel_parameters_start, input, shader->input_size);
for (i = 0; i < (input_size / 4); i++) {
COMPUTE_DBG(ctx->screen, "input %i : %u\n", i,
((unsigned*)num_work_groups_start)[i]);
}
ctx_->transfer_unmap(ctx_, transfer);
/* ID=0 is reserved for the parameters */
evergreen_cs_set_constant_buffer(ctx, 0, 0, input_size,
(struct pipe_resource*)shader->kernel_param);
}
