Skip to content

Commit 4558beb

Browse files
committed
GPU: Tune some kernel parameters for AMD Vega
1 parent 8fbf7d0 commit 4558beb

File tree

1 file changed

+8
-5
lines changed

1 file changed

+8
-5
lines changed

GPU/Common/GPUDefGPUParameters.h

Lines changed: 8 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -34,15 +34,15 @@
3434
#define GPUCA_LB_GPUTPCCreateSliceData 128
3535
#define GPUCA_LB_GPUTPCStartHitsSorter 1024, 2
3636
#define GPUCA_LB_GPUTPCStartHitsFinder 1024
37-
#define GPUCA_LB_GPUTPCTrackletConstructor 512, 1
37+
#define GPUCA_LB_GPUTPCTrackletConstructor 256, 2
3838
#define GPUCA_LB_GPUTPCTrackletSelector 256, 8
3939
#define GPUCA_LB_GPUTPCNeighboursFinder 1024, 1
4040
#define GPUCA_LB_GPUTPCNeighboursCleaner 896
4141
#define GPUCA_LB_GPUTPCGlobalTracking 256
4242
#define GPUCA_LB_GPUTPCCFDecodeZS 64, 4
4343
#define GPUCA_LB_GPUTPCCFGather 1024, 1
4444
#define GPUCA_LB_GPUTPCGMMergerTrackFit 64, 1
45-
#define GPUCA_LB_GPUTPCGMMergerFollowLoopers 256, 1, 200
45+
#define GPUCA_LB_GPUTPCGMMergerFollowLoopers 256, 4, 200
4646
#define GPUCA_LB_GPUTPCGMMergerSliceRefit 256
4747
#define GPUCA_LB_GPUTPCGMMergerUnpackResetIds 256
4848
#define GPUCA_LB_GPUTPCGMMergerUnpackGlobal 256
@@ -68,8 +68,8 @@
6868
#define GPUCA_LB_GPUTPCGMMergerFinalize_0 256
6969
#define GPUCA_LB_GPUTPCGMMergerFinalize_1 256
7070
#define GPUCA_LB_GPUTPCGMMergerFinalize_2 256
71-
#define GPUCA_LB_GPUTPCCompressionKernels_step0attached 256
72-
#define GPUCA_LB_GPUTPCCompressionKernels_step1unattached 512
71+
#define GPUCA_LB_GPUTPCCompressionKernels_step0attached 192, 2
72+
#define GPUCA_LB_GPUTPCCompressionKernels_step1unattached 512, 2
7373
#define GPUCA_LB_COMPRESSION_GATHER 1024
7474
#define GPUCA_LB_CLUSTER_FINDER 512
7575
#define GPUCA_NEIGHBOURS_FINDER_MAX_NNEIGHUP 5
@@ -81,7 +81,10 @@
8181
#define GPUCA_MERGER_SPLIT_LOOP_INTERPOLATION 1
8282
#define GPUCA_TRACKLET_SELECTOR_SLICE_COUNT 1
8383
#define GPUCA_NO_ATOMIC_PRECHECK 1
84-
#define GPUCA_COMP_GATHER_KERNEL 3
84+
#define GPUCA_DEDX_STORAGE_TYPE unsigned short
85+
#define GPUCA_MERGER_INTERPOLATION_ERROR_TYPE half
86+
#define GPUCA_COMP_GATHER_KERNEL 4
87+
#define GPUCA_COMP_GATHER_MODE 3
8588
#elif defined(GPUCA_GPUTYPE_AMPERE)
8689
#define GPUCA_WARP_SIZE 32
8790
#define GPUCA_THREAD_COUNT 512

0 commit comments

Comments
 (0)