Assertion failed on OpenCL rendering #53830

New Issue

Arto Kitula · 2018-01-18T19:44:30+01:00

Arto Kitula commented

2018-01-18 19:44:30 +01:00

macOS 10.12.6
AMD Radeon R9 M370X 2048 MB

Blender master, hash 42a562496b

GPU OpenCL compute render default cube scene ->

I0118 20:43:47.349573 229617664 util_system.cpp:77] Detected 1 CPU groups.
I0118 20:43:47.349611 229617664 util_system.cpp:80] Group 0 has 8 threads.
I0118 20:43:47.349617 229617664 util_task.cpp:203] Creating pool of 8 threads.
I0118 20:43:47.349787 229617664 device_opencl.cpp:41] Using mega kernel.
I0118 20:43:47.349866 229617664 opencl_base.cpp:99] Creating new Cycles device for OpenCL platform Apple, device AMD Radeon R9 M370X Compute Engine.
Assertion failed: (alignment < 1024), function MEM_lockfree_mallocN_aligned, file /Users/azo/Code/blender/blender-build/blender/intern/guardedalloc/intern/mallocn_lockfree_impl.c, line 366.

macOS 10.12.6 AMD Radeon R9 M370X 2048 MB Blender master, hash 42a562496b5 GPU OpenCL compute render default cube scene -> ``` I0118 20:43:47.349573 229617664 util_system.cpp:77] Detected 1 CPU groups. I0118 20:43:47.349611 229617664 util_system.cpp:80] Group 0 has 8 threads. I0118 20:43:47.349617 229617664 util_task.cpp:203] Creating pool of 8 threads. I0118 20:43:47.349787 229617664 device_opencl.cpp:41] Using mega kernel. I0118 20:43:47.349866 229617664 opencl_base.cpp:99] Creating new Cycles device for OpenCL platform Apple, device AMD Radeon R9 M370X Compute Engine. Assertion failed: (alignment < 1024), function MEM_lockfree_mallocN_aligned, file /Users/azo/Code/blender/blender-build/blender/intern/guardedalloc/intern/mallocn_lockfree_impl.c, line 366. ```

Arto Kitula commented

2018-01-18 19:44:30 +01:00

Added subscriber: @ArtoKitula

Brecht Van Lommel was assigned by Arto Kitula

2018-01-18 19:45:06 +01:00

blender-admin commented

2018-01-18 21:06:35 +01:00

This issue was referenced by blender/cycles@eff8353a44

This issue was referenced by blender/cycles@eff8353a442f37737e4f8c916ef84b27a5c3ac2d

blender-admin commented

2018-01-18 21:06:35 +01:00

This issue was referenced by 0fe41009f0

This issue was referenced by 0fe41009f095835bda961a67697b36cbc8cade49

Brecht Van Lommel commented

2018-01-18 21:17:30 +01:00

Can you test if this helps? I don't have a Mac with AMD card.

diff --git a/intern/cycles/device/opencl/opencl_util.cpp b/intern/cycles/device/opencl/opencl_util.cpp
index 459d512172f..47d099a1f94 100644
--- a/intern/cycles/device/opencl/opencl_util.cpp
+++ b/intern/cycles/device/opencl/opencl_util.cpp
@@ -1137,14 +1137,14 @@ bool OpenCLInfo::get_driver_version(cl_device_id device_id,

 int OpenCLInfo::mem_address_alignment(cl_device_id device_id)
 {
-       int base_align_bits;
+       int base_align_bytes;
        if(clGetDeviceInfo(device_id,
-                          CL_DEVICE_MEM_BASE_ADDR_ALIGN,
+                          CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE,
                           sizeof(int),
-                          &base_align_bits,
+                          &base_align_bytes,
                           NULL) == CL_SUCCESS)
        {
-               return base_align_bits/8;
+               return base_align_bytes;
        }
        return 1;
 }

Aligning to a very high value is likely harmless, but no good reason to do it. Alternatively we could just bump that 1024 in the assert a bit higher, but would need to know the value you were getting then.

Can you test if this helps? I don't have a Mac with AMD card. ``` diff --git a/intern/cycles/device/opencl/opencl_util.cpp b/intern/cycles/device/opencl/opencl_util.cpp index 459d512172f..47d099a1f94 100644 --- a/intern/cycles/device/opencl/opencl_util.cpp +++ b/intern/cycles/device/opencl/opencl_util.cpp @@ -1137,14 +1137,14 @@ bool OpenCLInfo::get_driver_version(cl_device_id device_id, int OpenCLInfo::mem_address_alignment(cl_device_id device_id) { - int base_align_bits; + int base_align_bytes; if(clGetDeviceInfo(device_id, - CL_DEVICE_MEM_BASE_ADDR_ALIGN, + CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE, sizeof(int), - &base_align_bits, + &base_align_bytes, NULL) == CL_SUCCESS) { - return base_align_bits/8; + return base_align_bytes; } return 1; } ``` Aligning to a very high value is likely harmless, but no good reason to do it. Alternatively we could just bump that 1024 in the assert a bit higher, but would need to know the value you were getting then.

Arto Kitula commented

2018-01-18 22:11:21 +01:00

Didn't work, now giving:

Assertion failed: (alignment == 16), function aligned_malloc, file /Users/azo/Code/blender/blender-build/blender/intern/guardedalloc/intern/mallocn.c, line 77.

Didn't work, now giving: ``` Assertion failed: (alignment == 16), function aligned_malloc, file /Users/azo/Code/blender/blender-build/blender/intern/guardedalloc/intern/mallocn.c, line 77. ```

Brecht Van Lommel commented

2018-01-19 11:35:15 +01:00

Changed status from 'Open' to: 'Resolved'

Brecht Van Lommel closed this issue

2018-01-19 11:35:15 +01:00

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Assertion failed on OpenCL rendering #53830