diff options
author | Cedric Nugteren <web@cedricnugteren.nl> | 2017-09-14 21:27:33 +0200 |
---|---|---|
committer | Cedric Nugteren <web@cedricnugteren.nl> | 2017-09-14 21:27:33 +0200 |
commit | 0d13d814c21f1fdb3b0781b84bf12914aec5225a (patch) | |
tree | f4ed5bfe6b13a5e865b1973ca571df87f1af4bc2 /src/database/kernels/xgemm/xgemm_16.hpp | |
parent | 14a61d24258498811ac0dfd11cad6087a231fa2a (diff) |
Added architecture layer in the tuning database for better performance on unseen devices
Diffstat (limited to 'src/database/kernels/xgemm/xgemm_16.hpp')
-rw-r--r-- | src/database/kernels/xgemm/xgemm_16.hpp | 19 |
1 files changed, 14 insertions, 5 deletions
diff --git a/src/database/kernels/xgemm/xgemm_16.hpp b/src/database/kernels/xgemm/xgemm_16.hpp index 53a8ccd9..c9954a9e 100644 --- a/src/database/kernels/xgemm/xgemm_16.hpp +++ b/src/database/kernels/xgemm/xgemm_16.hpp @@ -14,19 +14,28 @@ const DatabaseEntry XgemmHalf = { "Xgemm", Precision::kHalf, {"KWG", "KWI", "MDIMA", "MDIMC", "MWG", "NDIMB", "NDIMC", "NWG", "SA", "SB", "STRM", "STRN", "VWM", "VWN"}, { { // AMD GPUs kDeviceTypeGPU, "AMD", { - { "Ellesmere", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, - { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + { "Ellesmere", { + { "AMD Radeon RX 480", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + } }, + { "default", { + { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + } }, } }, { // Intel GPUs kDeviceTypeGPU, "Intel", { - { "Intel(R) HD Graphics Skylake ULT GT2", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, - { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + { "default", { + { "Intel(R) HD Graphics Skylake ULT GT2", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + } }, } }, { // Default kDeviceTypeAll, "default", { - { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + { "default", { + { "default", { 32, 2, 8, 8, 64, 16, 16, 64, 1, 1, 0, 0, 4, 4 } }, + } }, } }, } |