Any way we could get this fixed ?
On a fresh install with 13.9 AMD drivers, thus having to recompile the OpenCL kernels :
cgminer --benchmark --verbose --debug 2>logfile.txt
[2013-10-12 11:28:20] Global quota greatest common denominator set to 1
[2013-10-12 11:28:20] Started cgminer 3.5.1
[2013-10-12 11:28:20] RES: thread starting
[2013-10-12 11:28:20] CL Platform 0 vendor: Advanced Micro Devices, Inc.
[2013-10-12 11:28:20] CL Platform 0 name: AMD Accelerated Parallel Processing
[2013-10-12 11:28:20] CL Platform 0 version: OpenCL 1.2 AMD-APP (1268.1)
[2013-10-12 11:28:20] Platform 0 devices: 1
[2013-10-12 11:28:20] 0 Barts
[2013-10-12 11:28:20] GPU 0 iAdapterIndex 0 strUDID PCI_VEN_1002&DEV_6738&SUBSYS_23051787&REV_00_4&B77C4C1&0&0008A iBusNumber 1 iDeviceNumber 0 iFunctionNumber 0 iVendorID 1002 strAdapterName AMD Radeon HD 6800 Series
[2013-10-12 11:28:20] GPU 0 AMD Radeon HD 6800 Series hardware monitoring enabled
[2013-10-12 11:28:20] USB scan devices: checking for BFL devices
...
[2013-10-12 11:28:21] Init GPU thread 0 GPU 0 virtual GPU 0
[2013-10-12 11:28:21] CL Platform vendor: Advanced Micro Devices, Inc.
[2013-10-12 11:28:21] CL Platform name: AMD Accelerated Parallel Processing
[2013-10-12 11:28:21] CL Platform version: OpenCL 1.2 AMD-APP (1268.1)
[2013-10-12 11:28:21] List of devices:
[2013-10-12 11:28:21] 0 Barts
[2013-10-12 11:28:21] Selected 0: Barts
[2013-10-12 11:28:21] Popping work to stage thread
[2013-10-12 11:28:21] Preferred vector width reported 4
[2013-10-12 11:28:21] Max work group size reported 256
[2013-10-12 11:28:21] Max mem alloc size is 536870912
[2013-10-12 11:28:21] Selecting phatk kernel
[2013-10-12 11:28:21] No binary found, generating from source
[2013-10-12 11:28:21] Setting worksize to 128
[2013-10-12 11:28:21] Patched source to suit 2 vectors
[2013-10-12 11:28:21] cl_amd_media_ops found, setting BITALIGN
[2013-10-12 11:28:21] BFI_INT patch requiring device found, patched source with BFI_INT
[2013-10-12 11:28:21] CompilerOptions: -D WORKSIZE=128 -D VECTORS2 -D WORKVEC=256 -D BITALIGN -D BFI_INT
[2013-10-12 11:28:22] Binary size for gpu 0 found in binary slot 0: 556192
[2013-10-12 11:28:22] At 07e37494 (4290041900 rem. bytes), to begin patching
and then it exits (older versions used to crash there).
[2013-10-12 11:28:22] At 07e37494 (
4290041900 rem. bytes), to begin patching
Looks like an integer overflow as a negative number (-4925396) got passed to applog(LOG_DEBUG, "At %p (%u rem. bytes), to begin patching", w, remaining); Why this happens I don't know - did the kernel fail to compile and cgminer doesn't check for that or is it something else ? In any case this is very annoying not to have any more diagnostics and the only fallback is to blindly try different driver versions until it hopefully works.
Thanks !