[Beignet] [PATCH] Runtime: fix a userptr bug.

[Guo, Yejun]

LGTM, thanks.

-----Original Message-----
From: Beignet [mailto:beignet-bounces at lists.freedesktop.org] On Behalf Of Yang Rong
Sent: Tuesday, July 26, 2016 4:50 PM
To: beignet at lists.freedesktop.org
Cc: Yang, Rong R
Subject: [Beignet] [PATCH] Runtime: fix a userptr bug.

Userptr also require size cache alignment, otherwise, the remained
memory may be allocated in CPU side, when gpu flush the last cacheline
to memory, will override the value changed by CPU.

Signed-off-by: Yang Rong <rong.r.yang at intel.com>
---
 src/cl_mem.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/cl_mem.c b/src/cl_mem.c
index 229bc0a..9e796ef 100644
--- a/src/cl_mem.c
+++ b/src/cl_mem.c
@@ -295,7 +295,8 @@ cl_mem_allocate(enum cl_mem_type type,
           assert(host_ptr != NULL);
           /* userptr not support tiling */
           if (!is_tiled) {
-            if (ALIGN((unsigned long)host_ptr, cacheline_size) == (unsigned long)host_ptr) {
+            if ((ALIGN((unsigned long)host_ptr, cacheline_size) == (unsigned long)host_ptr) &&
+                (ALIGN((unsigned long)sz, cacheline_size) == (unsigned long)sz)) {
               void* aligned_host_ptr = (void*)(((unsigned long)host_ptr) & (~(page_size - 1)));
               mem->offset = host_ptr - aligned_host_ptr;
               mem->is_userptr = 1;
@@ -851,6 +852,7 @@ _cl_mem_new_image(cl_context ctx,
     cl_get_device_info(ctx->device, CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE, sizeof(cacheline_size), &cacheline_size, NULL);
     if (ALIGN((unsigned long)data, cacheline_size) == (unsigned long)data &&
         ALIGN(h, cl_buffer_get_tiling_align(ctx, CL_NO_TILE, 1)) == h &&
+        ALIGN(h * pitch * depth, cacheline_size) == h * pitch * depth && //h and pitch should same as aligned_h and aligned_pitch if enable userptr
         ((image_type != CL_MEM_OBJECT_IMAGE3D && image_type != CL_MEM_OBJECT_IMAGE1D_ARRAY && image_type != CL_MEM_OBJECT_IMAGE2D_ARRAY) || pitch * h == slice_pitch)) {
       tiling = CL_NO_TILE;
       enableUserptr = 1;
-- 
2.1.4

_______________________________________________
Beignet mailing list
Beignet at lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/beignet