[Mesa-dev] [PATCH 1/2] radv: wrap cs_add_buffer in an inline.

Samuel Pitoiset samuel.pitoiset at gmail.com
Mon Nov 6 13:45:07 UTC 2017



On 11/06/2017 05:07 AM, Dave Airlie wrote:
> From: Dave Airlie <airlied at redhat.com>
> 
> The next patch will try and avoid calling the indirect function.
> 
> Signed-off-by: Dave Airlie <airlied at redhat.com>
> ---
>   src/amd/vulkan/radv_cmd_buffer.c     | 40 ++++++++++++++++++------------------
>   src/amd/vulkan/radv_descriptor_set.c |  6 +++---
>   src/amd/vulkan/radv_device.c         | 14 ++++++-------
>   src/amd/vulkan/radv_meta_buffer.c    |  8 ++++----
>   src/amd/vulkan/radv_query.c          | 10 ++++-----
>   src/amd/vulkan/radv_radeon_winsys.h  |  8 ++++++++
>   6 files changed, 47 insertions(+), 39 deletions(-)
> 
> diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c
> index 590c41a7bf2..5fb9b479324 100644
> --- a/src/amd/vulkan/radv_cmd_buffer.c
> +++ b/src/amd/vulkan/radv_cmd_buffer.c
> @@ -279,7 +279,7 @@ radv_reset_cmd_buffer(struct radv_cmd_buffer *cmd_buffer)
>   	cmd_buffer->sample_positions_needed = false;
>   
>   	if (cmd_buffer->upload.upload_bo)
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs,
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs,
>   						      cmd_buffer->upload.upload_bo, 8);

Indentation.

With this adjusted, patch is:

Reviewed-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>

>   	cmd_buffer->upload.offset = 0;
>   
> @@ -321,7 +321,7 @@ radv_cmd_buffer_resize_upload_buf(struct radv_cmd_buffer *cmd_buffer,
>   		return false;
>   	}
>   
> -	device->ws->cs_add_buffer(cmd_buffer->cs, bo, 8);
> +	radv_cs_add_buffer(device->ws, cmd_buffer->cs, bo, 8);
>   	if (cmd_buffer->upload.upload_bo) {
>   		upload = malloc(sizeof(*upload));
>   
> @@ -415,7 +415,7 @@ void radv_cmd_buffer_trace_emit(struct radv_cmd_buffer *cmd_buffer)
>   	MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 7);
>   
>   	++cmd_buffer->state.trace_id;
> -	device->ws->cs_add_buffer(cs, device->trace_bo, 8);
> +	radv_cs_add_buffer(device->ws, cs, device->trace_bo, 8);
>   	radv_emit_write_data_packet(cs, va, 1, &cmd_buffer->state.trace_id);
>   	radeon_emit(cs, PKT3(PKT3_NOP, 0, 0));
>   	radeon_emit(cs, AC_ENCODE_TRACE_POINT(cmd_buffer->state.trace_id));
> @@ -472,7 +472,7 @@ radv_save_pipeline(struct radv_cmd_buffer *cmd_buffer,
>   	data[0] = (uintptr_t)pipeline;
>   	data[1] = (uintptr_t)pipeline >> 32;
>   
> -	device->ws->cs_add_buffer(cs, device->trace_bo, 8);
> +	radv_cs_add_buffer(device->ws, cs, device->trace_bo, 8);
>   	radv_emit_write_data_packet(cs, va, 2, data);
>   }
>   
> @@ -508,7 +508,7 @@ radv_save_descriptors(struct radv_cmd_buffer *cmd_buffer)
>   		data[i * 2 + 1] = (uintptr_t)set >> 32;
>   	}
>   
> -	device->ws->cs_add_buffer(cs, device->trace_bo, 8);
> +	radv_cs_add_buffer(device->ws, cs, device->trace_bo, 8);
>   	radv_emit_write_data_packet(cs, va, MAX_SETS * 2, data);
>   }
>   
> @@ -1297,7 +1297,7 @@ radv_set_depth_clear_regs(struct radv_cmd_buffer *cmd_buffer,
>   	if (aspects & VK_IMAGE_ASPECT_DEPTH_BIT)
>   		++reg_count;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, image->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, image->bo, 8);
>   
>   	radeon_emit(cmd_buffer->cs, PKT3(PKT3_WRITE_DATA, 2 + reg_count, 0));
>   	radeon_emit(cmd_buffer->cs, S_370_DST_SEL(V_370_MEM_ASYNC) |
> @@ -1358,7 +1358,7 @@ radv_set_dcc_need_cmask_elim_pred(struct radv_cmd_buffer *cmd_buffer,
>   	if (!image->surface.dcc_size)
>   		return;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, image->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, image->bo, 8);
>   
>   	radeon_emit(cmd_buffer->cs, PKT3(PKT3_WRITE_DATA, 4, 0));
>   	radeon_emit(cmd_buffer->cs, S_370_DST_SEL(V_370_MEM_ASYNC) |
> @@ -1382,7 +1382,7 @@ radv_set_color_clear_regs(struct radv_cmd_buffer *cmd_buffer,
>   	if (!image->cmask.size && !image->surface.dcc_size)
>   		return;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, image->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, image->bo, 8);
>   
>   	radeon_emit(cmd_buffer->cs, PKT3(PKT3_WRITE_DATA, 4, 0));
>   	radeon_emit(cmd_buffer->cs, S_370_DST_SEL(V_370_MEM_ASYNC) |
> @@ -1445,7 +1445,7 @@ radv_emit_framebuffer_state(struct radv_cmd_buffer *cmd_buffer)
>   		int idx = subpass->color_attachments[i].attachment;
>   		struct radv_attachment_info *att = &framebuffer->attachments[idx];
>   
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, att->attachment->bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, att->attachment->bo, 8);
>   
>   		assert(att->attachment->aspect_mask & VK_IMAGE_ASPECT_COLOR_BIT);
>   		radv_emit_fb_color_state(cmd_buffer, i, &att->cb);
> @@ -1458,7 +1458,7 @@ radv_emit_framebuffer_state(struct radv_cmd_buffer *cmd_buffer)
>   		VkImageLayout layout = subpass->depth_stencil_attachment.layout;
>   		struct radv_attachment_info *att = &framebuffer->attachments[idx];
>   		struct radv_image *image = att->attachment->image;
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, att->attachment->bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, att->attachment->bo, 8);
>   		MAYBE_UNUSED uint32_t queue_mask = radv_image_queue_family_mask(image,
>   										cmd_buffer->queue_family_index,
>   										cmd_buffer->queue_family_index);
> @@ -1788,7 +1788,7 @@ radv_cmd_buffer_update_vertex_descriptors(struct radv_cmd_buffer *cmd_buffer, bo
>   			struct radv_buffer *buffer = cmd_buffer->vertex_bindings[vb].buffer;
>   			uint32_t stride = cmd_buffer->state.pipeline->binding_stride[vb];
>   
> -			device->ws->cs_add_buffer(cmd_buffer->cs, buffer->bo, 8);
> +			radv_cs_add_buffer(device->ws, cmd_buffer->cs, buffer->bo, 8);
>   			va = radv_buffer_get_va(buffer->bo);
>   
>   			offset = cmd_buffer->vertex_bindings[vb].offset + velems->offset[i];
> @@ -2185,7 +2185,7 @@ static void emit_gfx_buffer_state(struct radv_cmd_buffer *cmd_buffer)
>   	struct radv_device *device = cmd_buffer->device;
>   	if (device->gfx_init) {
>   		uint64_t va = radv_buffer_get_va(device->gfx_init);
> -		device->ws->cs_add_buffer(cmd_buffer->cs, device->gfx_init, 8);
> +		radv_cs_add_buffer(device->ws, cmd_buffer->cs, device->gfx_init, 8);
>   		radeon_emit(cmd_buffer->cs, PKT3(PKT3_INDIRECT_BUFFER_CIK, 2, 0));
>   		radeon_emit(cmd_buffer->cs, va);
>   		radeon_emit(cmd_buffer->cs, va >> 32);
> @@ -2304,7 +2304,7 @@ void radv_CmdBindIndexBuffer(
>   	int index_size_shift = cmd_buffer->state.index_type ? 2 : 1;
>   	cmd_buffer->state.max_index_count = (index_buffer->size - offset) >> index_size_shift;
>   	cmd_buffer->state.dirty |= RADV_CMD_DIRTY_INDEX_BUFFER;
> -	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, index_buffer->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, index_buffer->bo, 8);
>   }
>   
>   
> @@ -2322,10 +2322,10 @@ radv_bind_descriptor_set(struct radv_cmd_buffer *cmd_buffer,
>   
>   	for (unsigned j = 0; j < set->layout->buffer_count; ++j)
>   		if (set->descriptors[j])
> -			ws->cs_add_buffer(cmd_buffer->cs, set->descriptors[j], 7);
> +			radv_cs_add_buffer(ws, cmd_buffer->cs, set->descriptors[j], 7);
>   
>   	if(set->bo)
> -		ws->cs_add_buffer(cmd_buffer->cs, set->bo, 8);
> +		radv_cs_add_buffer(ws, cmd_buffer->cs, set->bo, 8);
>   }
>   
>   void radv_CmdBindDescriptorSets(
> @@ -3097,7 +3097,7 @@ radv_emit_draw_packets(struct radv_cmd_buffer *cmd_buffer,
>   
>   		va += info->indirect->offset + info->indirect_offset;
>   
> -		ws->cs_add_buffer(cs, info->indirect->bo, 8);
> +		radv_cs_add_buffer(ws, cs, info->indirect->bo, 8);
>   
>   		radeon_emit(cs, PKT3(PKT3_SET_BASE, 2, 0));
>   		radeon_emit(cs, 1);
> @@ -3109,7 +3109,7 @@ radv_emit_draw_packets(struct radv_cmd_buffer *cmd_buffer,
>   			count_va += info->count_buffer->offset +
>   				    info->count_buffer_offset;
>   
> -			ws->cs_add_buffer(cs, info->count_buffer->bo, 8);
> +			radv_cs_add_buffer(ws, cs, info->count_buffer->bo, 8);
>   		}
>   
>   		if (!state->subpass->view_mask) {
> @@ -3454,7 +3454,7 @@ radv_emit_dispatch_packets(struct radv_cmd_buffer *cmd_buffer,
>   
>   		va += info->indirect->offset + info->indirect_offset;
>   
> -		ws->cs_add_buffer(cs, info->indirect->bo, 8);
> +		radv_cs_add_buffer(ws, cs, info->indirect->bo, 8);
>   
>   		if (loc->sgpr_idx != -1) {
>   			for (unsigned i = 0; i < grid_used; ++i) {
> @@ -3921,7 +3921,7 @@ static void write_event(struct radv_cmd_buffer *cmd_buffer,
>   	struct radeon_winsys_cs *cs = cmd_buffer->cs;
>   	uint64_t va = radv_buffer_get_va(event->bo);
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cs, event->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cs, event->bo, 8);
>   
>   	MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 18);
>   
> @@ -3977,7 +3977,7 @@ void radv_CmdWaitEvents(VkCommandBuffer commandBuffer,
>   		RADV_FROM_HANDLE(radv_event, event, pEvents[i]);
>   		uint64_t va = radv_buffer_get_va(event->bo);
>   
> -		cmd_buffer->device->ws->cs_add_buffer(cs, event->bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cs, event->bo, 8);
>   
>   		MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 7);
>   
> diff --git a/src/amd/vulkan/radv_descriptor_set.c b/src/amd/vulkan/radv_descriptor_set.c
> index 424756c13f5..a98ff37ced6 100644
> --- a/src/amd/vulkan/radv_descriptor_set.c
> +++ b/src/amd/vulkan/radv_descriptor_set.c
> @@ -566,7 +566,7 @@ static void write_texel_buffer_descriptor(struct radv_device *device,
>   	memcpy(dst, buffer_view->state, 4 * 4);
>   
>   	if (cmd_buffer)
> -		device->ws->cs_add_buffer(cmd_buffer->cs, buffer_view->bo, 7);
> +		radv_cs_add_buffer(device->ws, cmd_buffer->cs, buffer_view->bo, 7);
>   	else
>   		*buffer_list = buffer_view->bo;
>   }
> @@ -596,7 +596,7 @@ static void write_buffer_descriptor(struct radv_device *device,
>   		S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
>   
>   	if (cmd_buffer)
> -		device->ws->cs_add_buffer(cmd_buffer->cs, buffer->bo, 7);
> +		radv_cs_add_buffer(device->ws, cmd_buffer->cs, buffer->bo, 7);
>   	else
>   		*buffer_list = buffer->bo;
>   }
> @@ -639,7 +639,7 @@ write_image_descriptor(struct radv_device *device,
>   	}
>   
>   	if (cmd_buffer)
> -		device->ws->cs_add_buffer(cmd_buffer->cs, iview->bo, 7);
> +		radv_cs_add_buffer(device->ws, cmd_buffer->cs, iview->bo, 7);
>   	else
>   		*buffer_list = iview->bo;
>   }
> diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
> index e34b19c648a..1ecf70d4a9d 100644
> --- a/src/amd/vulkan/radv_device.c
> +++ b/src/amd/vulkan/radv_device.c
> @@ -1564,22 +1564,22 @@ radv_get_preamble_cs(struct radv_queue *queue,
>   		dest_cs[i] = cs;
>   
>   		if (scratch_bo)
> -			queue->device->ws->cs_add_buffer(cs, scratch_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, scratch_bo, 8);
>   
>   		if (esgs_ring_bo)
> -			queue->device->ws->cs_add_buffer(cs, esgs_ring_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, esgs_ring_bo, 8);
>   
>   		if (gsvs_ring_bo)
> -			queue->device->ws->cs_add_buffer(cs, gsvs_ring_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, gsvs_ring_bo, 8);
>   
>   		if (tess_factor_ring_bo)
> -			queue->device->ws->cs_add_buffer(cs, tess_factor_ring_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, tess_factor_ring_bo, 8);
>   
>   		if (tess_offchip_ring_bo)
> -			queue->device->ws->cs_add_buffer(cs, tess_offchip_ring_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, tess_offchip_ring_bo, 8);
>   
>   		if (descriptor_bo)
> -			queue->device->ws->cs_add_buffer(cs, descriptor_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, descriptor_bo, 8);
>   
>   		if (descriptor_bo != queue->descriptor_bo) {
>   			uint32_t *map = (uint32_t*)queue->device->ws->buffer_map(descriptor_bo);
> @@ -1678,7 +1678,7 @@ radv_get_preamble_cs(struct radv_queue *queue,
>   			uint32_t rsrc1 = S_008F04_BASE_ADDRESS_HI(scratch_va >> 32) |
>   			                 S_008F04_SWIZZLE_ENABLE(1);
>   
> -			queue->device->ws->cs_add_buffer(cs, compute_scratch_bo, 8);
> +			radv_cs_add_buffer(queue->device->ws, cs, compute_scratch_bo, 8);
>   
>   			radeon_set_sh_reg_seq(cs, R_00B900_COMPUTE_USER_DATA_0, 2);
>   			radeon_emit(cs, scratch_va);
> diff --git a/src/amd/vulkan/radv_meta_buffer.c b/src/amd/vulkan/radv_meta_buffer.c
> index f7ffcbbc90b..41cdc76b95b 100644
> --- a/src/amd/vulkan/radv_meta_buffer.c
> +++ b/src/amd/vulkan/radv_meta_buffer.c
> @@ -421,7 +421,7 @@ uint32_t radv_fill_buffer(struct radv_cmd_buffer *cmd_buffer,
>   	} else if (size) {
>   		uint64_t va = radv_buffer_get_va(bo);
>   		va += offset;
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, bo, 8);
>   		si_cp_dma_clear_buffer(cmd_buffer, va, size, value);
>   	}
>   
> @@ -444,8 +444,8 @@ void radv_copy_buffer(struct radv_cmd_buffer *cmd_buffer,
>   		src_va += src_offset;
>   		dst_va += dst_offset;
>   
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, src_bo, 8);
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, dst_bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, src_bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, dst_bo, 8);
>   
>   		si_cp_dma_buffer_copy(cmd_buffer, src_va, dst_va, size);
>   	}
> @@ -512,7 +512,7 @@ void radv_CmdUpdateBuffer(
>   	if (dataSize < RADV_BUFFER_OPS_CS_THRESHOLD) {
>   		si_emit_cache_flush(cmd_buffer);
>   
> -		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, dst_buffer->bo, 8);
> +		radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, dst_buffer->bo, 8);
>   
>   		radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, words + 4);
>   
> diff --git a/src/amd/vulkan/radv_query.c b/src/amd/vulkan/radv_query.c
> index 16ffd846487..a3cae17ae11 100644
> --- a/src/amd/vulkan/radv_query.c
> +++ b/src/amd/vulkan/radv_query.c
> @@ -957,8 +957,8 @@ void radv_CmdCopyQueryPoolResults(
>   	uint64_t dest_va = radv_buffer_get_va(dst_buffer->bo);
>   	dest_va += dst_buffer->offset + dstOffset;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, pool->bo, 8);
> -	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, dst_buffer->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, pool->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, dst_buffer->bo, 8);
>   
>   	switch (pool->type) {
>   	case VK_QUERY_TYPE_OCCLUSION:
> @@ -1084,7 +1084,7 @@ void radv_CmdBeginQuery(
>   	uint64_t va = radv_buffer_get_va(pool->bo);
>   	va += pool->stride * query;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cs, pool->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cs, pool->bo, 8);
>   
>   	switch (pool->type) {
>   	case VK_QUERY_TYPE_OCCLUSION:
> @@ -1125,7 +1125,7 @@ void radv_CmdEndQuery(
>   	uint64_t avail_va = va + pool->availability_offset + 4 * query;
>   	va += pool->stride * query;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cs, pool->bo, 8);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cs, pool->bo, 8);
>   
>   	switch (pool->type) {
>   	case VK_QUERY_TYPE_OCCLUSION:
> @@ -1177,7 +1177,7 @@ void radv_CmdWriteTimestamp(
>   	uint64_t avail_va = va + pool->availability_offset + 4 * query;
>   	uint64_t query_va = va + pool->stride * query;
>   
> -	cmd_buffer->device->ws->cs_add_buffer(cs, pool->bo, 5);
> +	radv_cs_add_buffer(cmd_buffer->device->ws, cs, pool->bo, 5);
>   
>   	MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 28);
>   
> diff --git a/src/amd/vulkan/radv_radeon_winsys.h b/src/amd/vulkan/radv_radeon_winsys.h
> index 395c8499b3d..bab19a6233d 100644
> --- a/src/amd/vulkan/radv_radeon_winsys.h
> +++ b/src/amd/vulkan/radv_radeon_winsys.h
> @@ -279,4 +279,12 @@ static inline uint64_t radv_buffer_get_va(struct radeon_winsys_bo *bo)
>   	return bo->va;
>   }
>   
> +static inline void radv_cs_add_buffer(struct radeon_winsys *ws,
> +				      struct radeon_winsys_cs *cs,
> +				      struct radeon_winsys_bo *bo,
> +				      uint8_t priority)
> +{
> +	ws->cs_add_buffer(cs, bo, priority);
> +}
> +
>   #endif /* RADV_RADEON_WINSYS_H */
> 


More information about the mesa-dev mailing list