gallium/radeon: remove buffer_unmap calls that can potentially decrease perf

buffer_unmap is currently a no-op on radeon and done correctly on amdgpu.
I plan to fix it for radeon, but before that, all occurences of buffer_unmap
that can negatively affect performance in the future must be removed.

There are 2 reasons for removing buffer_unmap calls:
- There is a likelihood that buffer_map will be called again, so we don't
  want to unmap yet.
- The buffer is being released, which automatically unmaps it.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>

src/gallium/drivers/r300/r300_query.c

@@ -168,8 +168,6 @@ static boolean r300_get_query_result(struct pipe_context* pipe,
         map++;
     }
 
-    r300->rws->buffer_unmap(q->cs_buf);
-
     if (q->type == PIPE_QUERY_OCCLUSION_PREDICATE) {
         vresult->b = temp != 0;
     } else {

src/gallium/drivers/r300/r300_transfer.c

@@ -251,16 +251,12 @@ void r300_texture_transfer_unmap(struct pipe_context *ctx,
     struct r300_resource *tex = r300_resource(transfer->resource);
 
     if (trans->linear_texture) {
-        rws->buffer_unmap(trans->linear_texture->cs_buf);
-
         if (transfer->usage & PIPE_TRANSFER_WRITE) {
             r300_copy_into_tiled_texture(ctx, trans);
         }
 
         pipe_resource_reference(
             (struct pipe_resource**)&trans->linear_texture, NULL);
-    } else {
-        rws->buffer_unmap(tex->cs_buf);
     }
     FREE(transfer);
 }

src/gallium/drivers/r600/r600_state_common.c

@@ -1409,7 +1409,6 @@ static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info
 					data += info.indirect_offset / sizeof(unsigned);
 					start = data[2] * ib.index_size;
 					count = data[0];
-					rctx->b.ws->buffer_unmap(indirect_resource->cs_buf);
 				}
 				else {
 					start = 0;

src/gallium/drivers/radeon/r600_pipe_common.c

@@ -939,10 +939,8 @@ void r600_destroy_common_screen(struct r600_common_screen *rscreen)
 	pipe_mutex_destroy(rscreen->aux_context_lock);
 	rscreen->aux_context->destroy(rscreen->aux_context);
 
-	if (rscreen->trace_bo) {
-		rscreen->ws->buffer_unmap(rscreen->trace_bo->cs_buf);
+	if (rscreen->trace_bo)
 		pipe_resource_reference((struct pipe_resource**)&rscreen->trace_bo, NULL);
-	}
 
 	rscreen->ws->destroy(rscreen->ws);
 	FREE(rscreen);

src/gallium/drivers/radeon/r600_query.c

@@ -118,7 +118,6 @@ static struct r600_resource *r600_new_query_buffer(struct r600_common_context *c
 			}
 			results += 4 * ctx->max_db;
 		}
-		ctx->ws->buffer_unmap(buf->cs_buf);
 		break;
 	case PIPE_QUERY_TIME_ELAPSED:
 	case PIPE_QUERY_TIMESTAMP:
@@ -130,7 +129,6 @@ static struct r600_resource *r600_new_query_buffer(struct r600_common_context *c
 	case PIPE_QUERY_PIPELINE_STATISTICS:
 		results = r600_buffer_map_sync_with_rings(ctx, buf, PIPE_TRANSFER_WRITE);
 		memset(results, 0, buf_size);
-		ctx->ws->buffer_unmap(buf->cs_buf);
 		break;
 	default:
 		assert(0);
@@ -751,7 +749,6 @@ static boolean r600_get_query_buffer_result(struct r600_common_context *ctx,
 		assert(0);
 	}
 
-	ctx->ws->buffer_unmap(qbuf->buf->cs_buf);
 	return TRUE;
 }
 
@@ -919,7 +916,6 @@ void r600_query_init_backend_mask(struct r600_common_context *ctx)
 	results = r600_buffer_map_sync_with_rings(ctx, buffer, PIPE_TRANSFER_WRITE);
 	if (results) {
 		memset(results, 0, ctx->max_db * 4 * 4);
-		ctx->ws->buffer_unmap(buffer->cs_buf);
 
 		/* emit EVENT_WRITE for ZPASS_DONE */
 		radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
@@ -937,7 +933,6 @@ void r600_query_init_backend_mask(struct r600_common_context *ctx)
 				if (results[i*4 + 1])
 					mask |= (1<<i);
 			}
-			ctx->ws->buffer_unmap(buffer->cs_buf);
 		}
 	}
 

src/gallium/drivers/radeon/r600_texture.c

@@ -1059,18 +1059,9 @@ static void r600_texture_transfer_unmap(struct pipe_context *ctx,
 					struct pipe_transfer* transfer)
 {
 	struct r600_transfer *rtransfer = (struct r600_transfer*)transfer;
-	struct r600_common_context *rctx = (struct r600_common_context*)ctx;
-	struct radeon_winsys_cs_handle *buf;
 	struct pipe_resource *texture = transfer->resource;
 	struct r600_texture *rtex = (struct r600_texture*)texture;
 
-	if (rtransfer->staging) {
-		buf = rtransfer->staging->cs_buf;
-	} else {
-		buf = r600_resource(transfer->resource)->cs_buf;
-	}
-	rctx->ws->buffer_unmap(buf);
-
 	if ((transfer->usage & PIPE_TRANSFER_WRITE) && rtransfer->staging) {
 		if (rtex->is_depth && rtex->resource.b.b.nr_samples <= 1) {
 			ctx->resource_copy_region(ctx, texture, transfer->level,

src/gallium/drivers/radeonsi/si_compute.c

@@ -309,8 +309,6 @@ static void si_launch_grid(
 			kernel_args[i]);
 	}
 
-	sctx->b.ws->buffer_unmap(input_buffer->cs_buf);
-
 	kernel_args_va = input_buffer->gpu_address;
 	kernel_args_va += kernel_args_offset;