Commits

History for

Commits on Apr 16, 2024

Updated OMNI_SAMPLE to actually sample
Ram3sh
committed
bb64698
Copy full SHA for bb64698

Commits on Apr 13, 2024

Added OMNIPERF_* scripts. Besti runtime is 8.8ms total
Ram3sh
committed
c770846
Copy full SHA for c770846

Commits on Apr 12, 2024

Updated make.hip to MI300A. Added OMNI_* scripts
Ram3sh
committed
18fb95f
Copy full SHA for 18fb95f

Commits on Jul 18, 2023

remove extra files which are not needed
artv3
committed
d2324ae
Copy full SHA for d2324ae
add empty kernels to avoid measuring stream creation
artv3
committed
02c8e01
Copy full SHA for 02c8e01

Commits on Jul 5, 2023

update for latestest raja
artv3
committed
aea2564
Copy full SHA for aea2564

Commits on Mar 25, 2023

Added run_omniperf
Ram3sh
committed
ae4e19b
Copy full SHA for ae4e19b

Commits on Mar 24, 2023

Added Tioga sample output. Cleaned up make.hip
Ram3sh
committed
39399a9
Copy full SHA for 39399a9

Commits on Sep 20, 2022

Merge pull request #2 from artv3/artv3/hip-launch-bounds
Show description for d5dbb73
Ram3sh
authored
d5dbb73
Copy full SHA for d5dbb73

Commits on Sep 16, 2022

Added run_miper and compare_miperfs
Ram3sh
committed
edd133f
Copy full SHA for edd133f

Commits on Sep 15, 2022

Added Makefile.amdclang. Added best option to make.hip and make.andclang to access fastest compile option
Ram3sh
committed
12a6978
Copy full SHA for 12a6978

Commits on Aug 29, 2022

remove raja from config
artv3
committed
8aae204
Copy full SHA for 8aae204
optimizations for hip
artv3
committed
5042c79
Copy full SHA for 5042c79

Commits on Aug 24, 2022

use RAJA launch instead of kernel
artv3
committed
24f055d
Copy full SHA for 24f055d

Commits on Aug 16, 2022

Added kernel time output in microseconds
Ram3sh
committed
3e58fd1
Copy full SHA for 3e58fd1
Added run_roof. make.hip has options to turn on sync and register_count
Ram3sh
committed
54743c9
Copy full SHA for 54743c9

Commits on Aug 2, 2022

Added amd_unroll_fix option to make.CC. Slows code down from 41ms to 45ms
Ram3sh
committed
5d92f67
Copy full SHA for 5d92f67
Fixed CC build
Ram3sh
committed
59f3e26
Copy full SHA for 59f3e26

Commits on Aug 1, 2022

Revisiting CC. Fails at link time
Ram3sh
committed
66bee87
Copy full SHA for 66bee87

Commits on Jun 22, 2022

Added fixes from Alessandro Fanfarillo(AMD) that improve performance of K1 and K5 with rocm/5.1.1. make with amd_unroll_fix=yes to enable
Ram3sh
committed
ae383f1
Copy full SHA for ae383f1

Commits on Jun 20, 2022

Added Makefile.hip.isa to generate device ISA files for checking register usage
Ram3sh
committed
6104b8d
Copy full SHA for 6104b8d
Default HIP device is now MI250X. Added sample output for MI250X
Ram3sh
committed
5f37a97
Copy full SHA for 5f37a97

Commits on Jul 15, 2021

Updated build to use rocm4.2. New sample output for hipcc and CC
Ram3sh
committed
e2d0cfe
Copy full SHA for e2d0cfe

Commits on Jul 8, 2021

Makefile.CC/make.CC now uses rocm4.1
Ram3sh
committed
032e0dc
Copy full SHA for 032e0dc
Updated Makefile.CC/make.CC to use CCE12
Ram3sh
committed
e0c93a2
Copy full SHA for e0c93a2

Commits on Jun 1, 2021

hipcc(4.0.1 & 4.1.0) and CC times on Nevada, before and after adding __launch_bounds(256,2)__ to the offload kernel.
Ram3sh
committed
ed33944
Copy full SHA for ed33944

Commits on May 28, 2021

Added nevada.cmake. Modified Cmake files
Ram3sh
committed
9fd50be
Copy full SHA for 9fd50be

Commits on Apr 22, 2021

Added required notices Open Source release under GPV v2.0
Ram3sh
committed
e2e5815
Copy full SHA for e2e5815

Commits on Apr 13, 2021

Added error checking and output of HIP/Cuda/RAJA version info. Updated .cmake files to remove hardcoded paths and use loaded modules on Whamo. Sample outputs updated with runs using 4.1.0 and new v…
Show description for 99a87b5
Ram3sh
committed
99a87b5
Copy full SHA for 99a87b5

Commits on Mar 17, 2021

Updated make.hip to 4.0.1. Added .gitignorewq
Ram3sh
committed
8490a1e
Copy full SHA for 8490a1e
Added Makefile.hippat to build instrumented hipcc executable. Does not link.
Ram3sh
committed
4ea02a6
Copy full SHA for 4ea02a6

Commits on Mar 12, 2021

Added min&max of result vector. Updated samples
Ram3sh
committed
99f3b75
Copy full SHA for 99f3b75
Init changed to reduce magnitude of the L2 norm of the kernel results. Updated sample_output
Ram3sh
committed
3c02462
Copy full SHA for 3c02462

Commits on Mar 11, 2021

Added Makefile.CC for building with CC. Runs slower with incorrect results
Ram3sh
committed
fa0e173
Copy full SHA for fa0e173

Commits on Mar 9, 2021

Added -DENABLE_SPILL_WARNS=On flag to cmake to measure register usage and spills in Cuda & Hip. Updated INSTALL.txt
Ram3sh
committed
0895328
Copy full SHA for 0895328