From dc93523204ebe8562145997673f25f8e59f9d2f5 Mon Sep 17 00:00:00 2001 From: Cedric Nugteren Date: Wed, 8 Feb 2017 21:14:38 +0100 Subject: Added tuning results for Titan X (Pascal version) --- src/database/kernels/xgemm.hpp | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'src/database/kernels/xgemm.hpp') diff --git a/src/database/kernels/xgemm.hpp b/src/database/kernels/xgemm.hpp index d9414f8b..5f62672b 100644 --- a/src/database/kernels/xgemm.hpp +++ b/src/database/kernels/xgemm.hpp @@ -95,9 +95,10 @@ const Database::DatabaseEntry XgemmSingle = { { "GeForce GTX TITAN", { {"KWG",16}, {"KWI",8}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",2}, {"VWN",2} } }, { "GeForce GTX TITAN Black", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",4}, {"VWN",1} } }, { "GeForce GTX TITAN X", { {"KWG",16}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",16}, {"MWG",128}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",128}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",4}, {"VWN",8} } }, + { "TITAN X (Pascal)", { {"KWG",16}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",128}, {"NDIMB",32}, {"NDIMC",8}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",4}, {"VWN",2} } }, { "Tesla K20m", { {"KWG",16}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",8}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",2}, {"VWN",4} } }, { "Tesla K40m", { {"KWG",16}, {"KWI",8}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",128}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",2}, {"VWN",4} } }, - { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",4}, {"VWN",2} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",4}, {"VWN",4} } }, } }, { // Default @@ -172,9 +173,10 @@ const Database::DatabaseEntry XgemmComplexSingle = { { "GeForce GTX TITAN", { {"KWG",16}, {"KWI",8}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",32}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "GeForce GTX TITAN Black", { {"KWG",16}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",2} } }, { "GeForce GTX TITAN X", { {"KWG",16}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",1}, {"SB",0}, {"STRM",1}, {"STRN",1}, {"VWM",1}, {"VWN",4} } }, + { "TITAN X (Pascal)", { {"KWG",32}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",32}, {"NDIMC",16}, {"NWG",32}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",1}, {"VWN",1} } }, { "Tesla K20m", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",4} } }, { "Tesla K40m", { {"KWG",16}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",32}, {"MWG",32}, {"NDIMB",32}, {"NDIMC",8}, {"NWG",64}, {"SA",0}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, - { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",32}, {"NDIMC",32}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, } }, { // Default @@ -236,6 +238,7 @@ const Database::DatabaseEntry XgemmDouble = { { "GeForce GTX TITAN", { {"KWG",16}, {"KWI",8}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",32}, {"NWG",128}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",2}, {"VWN",2} } }, { "GeForce GTX TITAN Black", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",8}, {"NWG",16}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "GeForce GTX TITAN X", { {"KWG",16}, {"KWI",8}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",16}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, + { "TITAN X (Pascal)", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",32}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",4}, {"VWN",1} } }, { "Tesla K20m", { {"KWG",16}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Tesla K40m", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",32}, {"NWG",128}, {"SA",1}, {"SB",0}, {"STRM",1}, {"STRN",1}, {"VWM",2}, {"VWN",4} } }, { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",4} } }, @@ -299,9 +302,10 @@ const Database::DatabaseEntry XgemmComplexDouble = { { "GeForce GTX 980", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",16}, {"NWG",128}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",1}, {"VWM",2}, {"VWN",2} } }, { "GeForce GTX TITAN Black", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",1}, {"VWN",1} } }, { "GeForce GTX TITAN X", { {"KWG",32}, {"KWI",8}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",128}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, + { "TITAN X (Pascal)", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",16}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Tesla K20m", { {"KWG",32}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Tesla K40m", { {"KWG",16}, {"KWI",8}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",32}, {"NDIMC",16}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, - { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",32}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",32}, {"NDIMC",32}, {"NWG",64}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",1} } }, } }, { // Default -- cgit v1.2.3 From 0643a29af51f9eb13e2b276d0a0e74590c699d3b Mon Sep 17 00:00:00 2001 From: Cedric Nugteren Date: Sat, 18 Feb 2017 13:59:10 +0100 Subject: Added tuning parameters for the AMD RX480 GPU (Ellesmere) --- README.md | 1 + src/database/kernels/copy.hpp | 22 ++++++++++++++++------ src/database/kernels/pad.hpp | 14 ++++++++++++-- src/database/kernels/padtranspose.hpp | 12 +++++++++++- src/database/kernels/transpose.hpp | 14 ++++++++++++-- src/database/kernels/xaxpy.hpp | 16 +++++++++++++--- src/database/kernels/xdot.hpp | 12 +++++++++++- src/database/kernels/xgemm.hpp | 16 +++++++++++++--- src/database/kernels/xgemm_direct.hpp | 17 +++++++++++++---- src/database/kernels/xgemv.hpp | 10 ++++++++++ src/database/kernels/xgemv_fast.hpp | 10 ++++++++++ src/database/kernels/xgemv_fast_rot.hpp | 10 ++++++++++ src/database/kernels/xger.hpp | 20 +++++++++++++++----- 13 files changed, 147 insertions(+), 27 deletions(-) (limited to 'src/database/kernels/xgemm.hpp') diff --git a/README.md b/README.md index 0e11e719..ceebc588 100644 --- a/README.md +++ b/README.md @@ -124,6 +124,7 @@ The CLBlast library will be tuned in the future for the most commonly used OpenC - Tesla K40m * AMD GPUs: - AMD Radeon R9 M370X Compute Engine + - Ellesmere - Hawaii - Oland - Pitcairn diff --git a/src/database/kernels/copy.hpp b/src/database/kernels/copy.hpp index 9bc613b9..b2aa736f 100644 --- a/src/database/kernels/copy.hpp +++ b/src/database/kernels/copy.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry CopyHalf = { "Copy", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"COPY_DIMX",16}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",4} } }, + { "default", { {"COPY_DIMX",16}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",4} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"COPY_DIMX",8}, {"COPY_DIMY",16}, {"COPY_VW",8}, {"COPY_WPT",4} } }, @@ -26,7 +32,7 @@ const Database::DatabaseEntry CopyHalf = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"COPY_DIMX",8}, {"COPY_DIMY",32}, {"COPY_VW",4}, {"COPY_WPT",8} } }, + { "default", { {"COPY_DIMX",16}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",4} } }, } }, } @@ -39,13 +45,14 @@ const Database::DatabaseEntry CopySingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",1} } }, + { "Ellesmere", { {"COPY_DIMX",8}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",8} } }, { "Hawaii", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",2} } }, { "Oland", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",2} } }, { "Pitcairn", { {"COPY_DIMX",8}, {"COPY_DIMY",16}, {"COPY_VW",4}, {"COPY_WPT",1} } }, { "Tahiti", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",2} } }, { "Tonga", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",4} } }, { "Turks", { {"COPY_DIMX",8}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",2} } }, - { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",2} } }, + { "default", { {"COPY_DIMX",8}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",1} } }, } }, { // ARM GPUs @@ -104,7 +111,7 @@ const Database::DatabaseEntry CopySingle = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",2} } }, + { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",4} } }, } }, } @@ -117,13 +124,14 @@ const Database::DatabaseEntry CopyComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",1} } }, + { "Ellesmere", { {"COPY_DIMX",16}, {"COPY_DIMY",16}, {"COPY_VW",1}, {"COPY_WPT",4} } }, { "Hawaii", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",2} } }, { "Oland", { {"COPY_DIMX",8}, {"COPY_DIMY",16}, {"COPY_VW",1}, {"COPY_WPT",1} } }, { "Pitcairn", { {"COPY_DIMX",8}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",2} } }, { "Tahiti", { {"COPY_DIMX",8}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",2} } }, { "Tonga", { {"COPY_DIMX",8}, {"COPY_DIMY",32}, {"COPY_VW",1}, {"COPY_WPT",2} } }, { "Turks", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",4}, {"COPY_WPT",1} } }, - { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",4} } }, + { "default", { {"COPY_DIMX",16}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",1} } }, } }, { // Intel CPUs @@ -174,7 +182,7 @@ const Database::DatabaseEntry CopyComplexSingle = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"COPY_DIMX",16}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",2} } }, + { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",1} } }, } }, } @@ -187,12 +195,13 @@ const Database::DatabaseEntry CopyDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",1} } }, + { "Ellesmere", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",4} } }, { "Hawaii", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",2} } }, { "Oland", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",8} } }, { "Pitcairn", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",1} } }, { "Tahiti", { {"COPY_DIMX",8}, {"COPY_DIMY",32}, {"COPY_VW",2}, {"COPY_WPT",1} } }, { "Tonga", { {"COPY_DIMX",8}, {"COPY_DIMY",32}, {"COPY_VW",2}, {"COPY_WPT",4} } }, - { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",2} } }, + { "default", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",4} } }, } }, { // ARM GPUs @@ -252,6 +261,7 @@ const Database::DatabaseEntry CopyComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"COPY_DIMX",8}, {"COPY_DIMY",16}, {"COPY_VW",1}, {"COPY_WPT",1} } }, + { "Ellesmere", { {"COPY_DIMX",8}, {"COPY_DIMY",32}, {"COPY_VW",1}, {"COPY_WPT",2} } }, { "Hawaii", { {"COPY_DIMX",32}, {"COPY_DIMY",8}, {"COPY_VW",2}, {"COPY_WPT",8} } }, { "Oland", { {"COPY_DIMX",8}, {"COPY_DIMY",16}, {"COPY_VW",1}, {"COPY_WPT",1} } }, { "Pitcairn", { {"COPY_DIMX",16}, {"COPY_DIMY",8}, {"COPY_VW",1}, {"COPY_WPT",1} } }, diff --git a/src/database/kernels/pad.hpp b/src/database/kernels/pad.hpp index f22399eb..e4005527 100644 --- a/src/database/kernels/pad.hpp +++ b/src/database/kernels/pad.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry PadHalf = { "Pad", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"PAD_DIMX",16}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, + { "default", { {"PAD_DIMX",16}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"PAD_DIMX",8}, {"PAD_DIMY",8}, {"PAD_WPTX",4}, {"PAD_WPTY",1} } }, @@ -39,6 +45,7 @@ const Database::DatabaseEntry PadSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, + { "Ellesmere", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",2}, {"PAD_WPTY",2} } }, { "Hawaii", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",4} } }, { "Oland", { {"PAD_DIMX",8}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, { "Pitcairn", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, @@ -117,6 +124,7 @@ const Database::DatabaseEntry PadComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, + { "Ellesmere", { {"PAD_DIMX",16}, {"PAD_DIMY",16}, {"PAD_WPTX",2}, {"PAD_WPTY",4} } }, { "Hawaii", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, { "Oland", { {"PAD_DIMX",8}, {"PAD_DIMY",32}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, { "Pitcairn", { {"PAD_DIMX",8}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, @@ -182,7 +190,7 @@ const Database::DatabaseEntry PadComplexSingle = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",2}, {"PAD_WPTY",1} } }, + { "default", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, } }, } @@ -195,12 +203,13 @@ const Database::DatabaseEntry PadDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, + { "Ellesmere", { {"PAD_DIMX",8}, {"PAD_DIMY",32}, {"PAD_WPTX",2}, {"PAD_WPTY",1} } }, { "Hawaii", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, { "Oland", { {"PAD_DIMX",8}, {"PAD_DIMY",32}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, { "Pitcairn", { {"PAD_DIMX",8}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, { "Tahiti", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, { "Tonga", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",4}, {"PAD_WPTY",1} } }, - { "default", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, + { "default", { {"PAD_DIMX",16}, {"PAD_DIMY",16}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, } }, { // ARM GPUs @@ -260,6 +269,7 @@ const Database::DatabaseEntry PadComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PAD_DIMX",16}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, + { "Ellesmere", { {"PAD_DIMX",8}, {"PAD_DIMY",16}, {"PAD_WPTX",1}, {"PAD_WPTY",2} } }, { "Hawaii", { {"PAD_DIMX",32}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, { "Oland", { {"PAD_DIMX",8}, {"PAD_DIMY",16}, {"PAD_WPTX",2}, {"PAD_WPTY",1} } }, { "Pitcairn", { {"PAD_DIMX",16}, {"PAD_DIMY",8}, {"PAD_WPTX",1}, {"PAD_WPTY",1} } }, diff --git a/src/database/kernels/padtranspose.hpp b/src/database/kernels/padtranspose.hpp index ce40914c..92aa4f7b 100644 --- a/src/database/kernels/padtranspose.hpp +++ b/src/database/kernels/padtranspose.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry PadtransposeHalf = { "Padtranspose", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, + { "default", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",1} } }, @@ -39,6 +45,7 @@ const Database::DatabaseEntry PadtransposeSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, + { "Ellesmere", { {"PADTRA_PAD",1}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, { "Hawaii", { {"PADTRA_PAD",1}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, { "Oland", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, { "Pitcairn", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, @@ -117,6 +124,7 @@ const Database::DatabaseEntry PadtransposeComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, + { "Ellesmere", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, { "Hawaii", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",2} } }, { "Oland", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, { "Pitcairn", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, @@ -195,6 +203,7 @@ const Database::DatabaseEntry PadtransposeDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, + { "Ellesmere", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, { "Hawaii", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",2} } }, { "Oland", { {"PADTRA_PAD",0}, {"PADTRA_TILE",16}, {"PADTRA_WPT",4} } }, { "Pitcairn", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, @@ -247,7 +256,7 @@ const Database::DatabaseEntry PadtransposeDouble = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"PADTRA_PAD",1}, {"PADTRA_TILE",16}, {"PADTRA_WPT",2} } }, + { "default", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, } }, } @@ -260,6 +269,7 @@ const Database::DatabaseEntry PadtransposeComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, + { "Ellesmere", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, { "Hawaii", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, { "Oland", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, { "Pitcairn", { {"PADTRA_PAD",0}, {"PADTRA_TILE",8}, {"PADTRA_WPT",4} } }, diff --git a/src/database/kernels/transpose.hpp b/src/database/kernels/transpose.hpp index b80565b9..0ee2d83b 100644 --- a/src/database/kernels/transpose.hpp +++ b/src/database/kernels/transpose.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry TransposeHalf = { "Transpose", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",8} } }, + { "default", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",8} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"TRA_DIM",8}, {"TRA_PAD",1}, {"TRA_SHUFFLE",1}, {"TRA_WPT",8} } }, @@ -26,7 +32,7 @@ const Database::DatabaseEntry TransposeHalf = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"TRA_DIM",8}, {"TRA_PAD",1}, {"TRA_SHUFFLE",0}, {"TRA_WPT",8} } }, + { "default", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, } }, } @@ -39,13 +45,14 @@ const Database::DatabaseEntry TransposeSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",8} } }, + { "Ellesmere", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, { "Hawaii", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",8} } }, { "Oland", { {"TRA_DIM",8}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, { "Pitcairn", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, { "Tahiti", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, { "Tonga", { {"TRA_DIM",8}, {"TRA_PAD",1}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, { "Turks", { {"TRA_DIM",8}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, - { "default", { {"TRA_DIM",8}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, + { "default", { {"TRA_DIM",8}, {"TRA_PAD",1}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, } }, { // ARM GPUs @@ -117,6 +124,7 @@ const Database::DatabaseEntry TransposeComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"TRA_DIM",4}, {"TRA_PAD",1}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, + { "Ellesmere", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, { "Hawaii", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, { "Oland", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, { "Pitcairn", { {"TRA_DIM",8}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, @@ -189,6 +197,7 @@ const Database::DatabaseEntry TransposeDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, + { "Ellesmere", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",4} } }, { "Hawaii", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, { "Oland", { {"TRA_DIM",8}, {"TRA_PAD",1}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, { "Pitcairn", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, @@ -254,6 +263,7 @@ const Database::DatabaseEntry TransposeComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"TRA_DIM",4}, {"TRA_PAD",1}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, + { "Ellesmere", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, { "Hawaii", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",2} } }, { "Oland", { {"TRA_DIM",16}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, { "Pitcairn", { {"TRA_DIM",4}, {"TRA_PAD",0}, {"TRA_SHUFFLE",1}, {"TRA_WPT",1} } }, diff --git a/src/database/kernels/xaxpy.hpp b/src/database/kernels/xaxpy.hpp index 5fefb5c3..4b747f0a 100644 --- a/src/database/kernels/xaxpy.hpp +++ b/src/database/kernels/xaxpy.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XaxpyHalf = { "Xaxpy", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"VW",4}, {"WGS",128}, {"WPT",4} } }, + { "default", { {"VW",4}, {"WGS",128}, {"WPT",4} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"VW",1}, {"WGS",64}, {"WPT",1} } }, @@ -26,7 +32,7 @@ const Database::DatabaseEntry XaxpyHalf = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"VW",8}, {"WGS",64}, {"WPT",1} } }, + { "default", { {"VW",8}, {"WGS",256}, {"WPT",4} } }, } }, } @@ -39,13 +45,14 @@ const Database::DatabaseEntry XaxpySingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, + { "Ellesmere", { {"VW",1}, {"WGS",64}, {"WPT",4} } }, { "Hawaii", { {"VW",2}, {"WGS",64}, {"WPT",2} } }, { "Oland", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, { "Pitcairn", { {"VW",2}, {"WGS",128}, {"WPT",1} } }, { "Tahiti", { {"VW",2}, {"WGS",64}, {"WPT",1} } }, { "Tonga", { {"VW",1}, {"WGS",256}, {"WPT",8} } }, { "Turks", { {"VW",2}, {"WGS",256}, {"WPT",1} } }, - { "default", { {"VW",2}, {"WGS",64}, {"WPT",2} } }, + { "default", { {"VW",2}, {"WGS",256}, {"WPT",1} } }, } }, { // ARM GPUs @@ -117,6 +124,7 @@ const Database::DatabaseEntry XaxpyComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW",2}, {"WGS",64}, {"WPT",8} } }, + { "Ellesmere", { {"VW",2}, {"WGS",256}, {"WPT",1} } }, { "Hawaii", { {"VW",1}, {"WGS",128}, {"WPT",2} } }, { "Oland", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, { "Pitcairn", { {"VW",1}, {"WGS",64}, {"WPT",1} } }, @@ -195,6 +203,7 @@ const Database::DatabaseEntry XaxpyDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW",1}, {"WGS",256}, {"WPT",1} } }, + { "Ellesmere", { {"VW",2}, {"WGS",64}, {"WPT",4} } }, { "Hawaii", { {"VW",1}, {"WGS",64}, {"WPT",2} } }, { "Oland", { {"VW",1}, {"WGS",64}, {"WPT",1} } }, { "Pitcairn", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, @@ -247,7 +256,7 @@ const Database::DatabaseEntry XaxpyDouble = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"VW",1}, {"WGS",128}, {"WPT",2} } }, + { "default", { {"VW",2}, {"WGS",256}, {"WPT",1} } }, } }, } @@ -260,6 +269,7 @@ const Database::DatabaseEntry XaxpyComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, + { "Ellesmere", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, { "Hawaii", { {"VW",2}, {"WGS",64}, {"WPT",1} } }, { "Oland", { {"VW",1}, {"WGS",256}, {"WPT",1} } }, { "Pitcairn", { {"VW",1}, {"WGS",128}, {"WPT",1} } }, diff --git a/src/database/kernels/xdot.hpp b/src/database/kernels/xdot.hpp index 67360b76..ea154b6e 100644 --- a/src/database/kernels/xdot.hpp +++ b/src/database/kernels/xdot.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XdotHalf = { "Xdot", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"WGS1",256}, {"WGS2",64} } }, + { "default", { {"WGS1",256}, {"WGS2",64} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"WGS1",256}, {"WGS2",32} } }, @@ -39,12 +45,13 @@ const Database::DatabaseEntry XdotSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",128}, {"WGS2",32} } }, + { "Ellesmere", { {"WGS1",128}, {"WGS2",32} } }, { "Oland", { {"WGS1",256}, {"WGS2",32} } }, { "Pitcairn", { {"WGS1",128}, {"WGS2",32} } }, { "Tahiti", { {"WGS1",128}, {"WGS2",32} } }, { "Tonga", { {"WGS1",64}, {"WGS2",32} } }, { "Turks", { {"WGS1",128}, {"WGS2",64} } }, - { "default", { {"WGS1",128}, {"WGS2",64} } }, + { "default", { {"WGS1",128}, {"WGS2",32} } }, } }, { // Intel CPUs @@ -99,6 +106,7 @@ const Database::DatabaseEntry XdotComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WGS2",32} } }, + { "Ellesmere", { {"WGS1",256}, {"WGS2",32} } }, { "Oland", { {"WGS1",128}, {"WGS2",32} } }, { "Pitcairn", { {"WGS1",256}, {"WGS2",32} } }, { "Tahiti", { {"WGS1",64}, {"WGS2",32} } }, @@ -159,6 +167,7 @@ const Database::DatabaseEntry XdotDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WGS2",128} } }, + { "Ellesmere", { {"WGS1",128}, {"WGS2",64} } }, { "Oland", { {"WGS1",256}, {"WGS2",32} } }, { "Pitcairn", { {"WGS1",128}, {"WGS2",32} } }, { "Tahiti", { {"WGS1",256}, {"WGS2",32} } }, @@ -207,6 +216,7 @@ const Database::DatabaseEntry XdotComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WGS2",32} } }, + { "Ellesmere", { {"WGS1",256}, {"WGS2",32} } }, { "Oland", { {"WGS1",256}, {"WGS2",32} } }, { "Pitcairn", { {"WGS1",256}, {"WGS2",32} } }, { "Tahiti", { {"WGS1",256}, {"WGS2",32} } }, diff --git a/src/database/kernels/xgemm.hpp b/src/database/kernels/xgemm.hpp index 5f62672b..751f403b 100644 --- a/src/database/kernels/xgemm.hpp +++ b/src/database/kernels/xgemm.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XgemmHalf = { "Xgemm", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",4}, {"VWN",4} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",4}, {"VWN",4} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics Skylake ULT GT2", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",4}, {"VWN",4} } }, @@ -38,6 +44,7 @@ const Database::DatabaseEntry XgemmSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",16}, {"NWG",128}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",8} } }, + { "Ellesmere", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",2} } }, { "Hawaii", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",32}, {"MWG",128}, {"NDIMB",32}, {"NDIMC",8}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",4}, {"VWN",2} } }, { "Oland", { {"KWG",16}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",32}, {"NDIMC",16}, {"NWG",128}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",2}, {"VWN",4} } }, { "Pitcairn", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",128}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, @@ -116,6 +123,7 @@ const Database::DatabaseEntry XgemmComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWG",32}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",32}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",64}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",1}, {"VWM",2}, {"VWN",8} } }, + { "Ellesmere", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",4} } }, { "Hawaii", { {"KWG",32}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",16}, {"NWG",32}, {"SA",1}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Oland", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",32}, {"NDIMC",32}, {"NWG",128}, {"SA",1}, {"SB",0}, {"STRM",0}, {"STRN",1}, {"VWM",2}, {"VWN",4} } }, { "Pitcairn", { {"KWG",16}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",4}, {"VWN",2} } }, @@ -181,7 +189,7 @@ const Database::DatabaseEntry XgemmComplexSingle = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",1} } }, } }, } @@ -194,6 +202,7 @@ const Database::DatabaseEntry XgemmDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",4}, {"VWN",4} } }, + { "Ellesmere", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",64}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",2} } }, { "Hawaii", { {"KWG",16}, {"KWI",8}, {"MDIMA",32}, {"MDIMC",8}, {"MWG",128}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",4} } }, { "Oland", { {"KWG",16}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",16}, {"NDIMC",8}, {"NWG",16}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",1}, {"VWM",1}, {"VWN",1} } }, { "Pitcairn", { {"KWG",32}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",64}, {"NDIMB",8}, {"NDIMC",16}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",2} } }, @@ -259,12 +268,13 @@ const Database::DatabaseEntry XgemmComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWG",32}, {"KWI",8}, {"MDIMA",8}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",1}, {"VWM",2}, {"VWN",2} } }, + { "Ellesmere", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",16}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Hawaii", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",32}, {"SA",1}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",2} } }, { "Oland", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",32}, {"NWG",128}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",4} } }, { "Pitcairn", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",32}, {"NWG",32}, {"SA",0}, {"SB",1}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Tahiti", { {"KWG",16}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",16}, {"SA",0}, {"SB",0}, {"STRM",1}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, { "Tonga", { {"KWG",16}, {"KWI",2}, {"MDIMA",32}, {"MDIMC",16}, {"MWG",32}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",16}, {"SA",1}, {"SB",1}, {"STRM",1}, {"STRN",1}, {"VWM",1}, {"VWN",1} } }, - { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",16}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",2} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",16}, {"MDIMC",16}, {"MWG",16}, {"NDIMB",16}, {"NDIMC",16}, {"NWG",16}, {"SA",1}, {"SB",1}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } }, } }, { // ARM GPUs @@ -310,7 +320,7 @@ const Database::DatabaseEntry XgemmComplexDouble = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",64}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",2}, {"VWN",2} } }, + { "default", { {"KWG",32}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",16}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",32}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",2} } }, } }, } diff --git a/src/database/kernels/xgemm_direct.hpp b/src/database/kernels/xgemm_direct.hpp index bec0164f..4984475f 100644 --- a/src/database/kernels/xgemm_direct.hpp +++ b/src/database/kernels/xgemm_direct.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XgemmDirectHalf = { "XgemmDirect", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"KWID",8}, {"MDIMAD",32}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",32}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",32} } }, + { "default", { {"KWID",8}, {"MDIMAD",32}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",32}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",32} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics Skylake ULT GT2", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",8} } }, @@ -25,7 +31,7 @@ const Database::DatabaseEntry XgemmDirectHalf = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",8} } }, + { "default", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, } }, } @@ -38,9 +44,10 @@ const Database::DatabaseEntry XgemmDirectSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",2}, {"VWND",2}, {"WGD",32} } }, + { "Ellesmere", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",32}, {"NDIMCD",32}, {"PADA",1}, {"PADB",1}, {"VWMD",2}, {"VWND",1}, {"WGD",32} } }, { "Tonga", { {"KWID",16}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",32}, {"NDIMCD",8}, {"PADA",0}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",32} } }, { "Turks", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, - { "default", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, + { "default", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",2}, {"VWND",2}, {"WGD",32} } }, } }, { // Intel CPUs @@ -68,7 +75,7 @@ const Database::DatabaseEntry XgemmDirectSingle = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",4}, {"VWND",2}, {"WGD",32} } }, + { "default", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",4}, {"VWND",4}, {"WGD",32} } }, } }, } @@ -124,6 +131,7 @@ const Database::DatabaseEntry XgemmDirectDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, + { "Ellesmere", { {"KWID",8}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",8}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",2}, {"VWND",1}, {"WGD",32} } }, { "Tonga", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",32} } }, { "default", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, } @@ -159,6 +167,7 @@ const Database::DatabaseEntry XgemmDirectComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, + { "Ellesmere", { {"KWID",16}, {"MDIMAD",32}, {"MDIMCD",32}, {"NDIMBD",16}, {"NDIMCD",8}, {"PADA",0}, {"PADB",0}, {"VWMD",1}, {"VWND",1}, {"WGD",32} } }, { "Tonga", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, { "default", { {"KWID",2}, {"MDIMAD",16}, {"MDIMCD",16}, {"NDIMBD",16}, {"NDIMCD",16}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, } @@ -181,7 +190,7 @@ const Database::DatabaseEntry XgemmDirectComplexDouble = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",1}, {"WGD",16} } }, + { "default", { {"KWID",2}, {"MDIMAD",8}, {"MDIMCD",8}, {"NDIMBD",8}, {"NDIMCD",8}, {"PADA",1}, {"PADB",1}, {"VWMD",1}, {"VWND",2}, {"WGD",16} } }, } }, } diff --git a/src/database/kernels/xgemv.hpp b/src/database/kernels/xgemv.hpp index 3bb31dc2..05f47554 100644 --- a/src/database/kernels/xgemv.hpp +++ b/src/database/kernels/xgemv.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XgemvHalf = { "Xgemv", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"WGS1",256}, {"WPT1",1} } }, + { "default", { {"WGS1",256}, {"WPT1",1} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"WGS1",64}, {"WPT1",1} } }, @@ -39,6 +45,7 @@ const Database::DatabaseEntry XgemvSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",128}, {"WPT1",1} } }, + { "Ellesmere", { {"WGS1",256}, {"WPT1",1} } }, { "Hawaii", { {"WGS1",128}, {"WPT1",1} } }, { "Oland", { {"WGS1",128}, {"WPT1",1} } }, { "Pitcairn", { {"WGS1",256}, {"WPT1",1} } }, @@ -110,6 +117,7 @@ const Database::DatabaseEntry XgemvComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WPT1",1} } }, + { "Ellesmere", { {"WGS1",32}, {"WPT1",1} } }, { "Hawaii", { {"WGS1",64}, {"WPT1",1} } }, { "Oland", { {"WGS1",64}, {"WPT1",1} } }, { "Pitcairn", { {"WGS1",64}, {"WPT1",1} } }, @@ -177,6 +185,7 @@ const Database::DatabaseEntry XgemvDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WPT1",1} } }, + { "Ellesmere", { {"WGS1",32}, {"WPT1",1} } }, { "Hawaii", { {"WGS1",128}, {"WPT1",1} } }, { "Oland", { {"WGS1",256}, {"WPT1",1} } }, { "Pitcairn", { {"WGS1",256}, {"WPT1",1} } }, @@ -235,6 +244,7 @@ const Database::DatabaseEntry XgemvComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WPT1",1} } }, + { "Ellesmere", { {"WGS1",32}, {"WPT1",1} } }, { "Hawaii", { {"WGS1",64}, {"WPT1",1} } }, { "Oland", { {"WGS1",256}, {"WPT1",1} } }, { "Pitcairn", { {"WGS1",256}, {"WPT1",1} } }, diff --git a/src/database/kernels/xgemv_fast.hpp b/src/database/kernels/xgemv_fast.hpp index b9a2eba2..efc64d0c 100644 --- a/src/database/kernels/xgemv_fast.hpp +++ b/src/database/kernels/xgemv_fast.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XgemvFastHalf = { "XgemvFast", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"VW2",1}, {"WGS2",32}, {"WPT2",1} } }, + { "default", { {"VW2",1}, {"WGS2",32}, {"WPT2",1} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"VW2",1}, {"WGS2",16}, {"WPT2",1} } }, @@ -39,6 +45,7 @@ const Database::DatabaseEntry XgemvFastSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW2",1}, {"WGS2",128}, {"WPT2",1} } }, + { "Ellesmere", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Hawaii", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Oland", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Pitcairn", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, @@ -110,6 +117,7 @@ const Database::DatabaseEntry XgemvFastComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW2",2}, {"WGS2",256}, {"WPT2",2} } }, + { "Ellesmere", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Hawaii", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Oland", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Pitcairn", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, @@ -171,6 +179,7 @@ const Database::DatabaseEntry XgemvFastDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW2",1}, {"WGS2",256}, {"WPT2",1} } }, + { "Ellesmere", { {"VW2",1}, {"WGS2",128}, {"WPT2",1} } }, { "Hawaii", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Oland", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Pitcairn", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, @@ -229,6 +238,7 @@ const Database::DatabaseEntry XgemvFastComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW2",1}, {"WGS2",256}, {"WPT2",1} } }, + { "Ellesmere", { {"VW2",1}, {"WGS2",16}, {"WPT2",1} } }, { "Hawaii", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, { "Oland", { {"VW2",1}, {"WGS2",256}, {"WPT2",1} } }, { "Pitcairn", { {"VW2",1}, {"WGS2",64}, {"WPT2",1} } }, diff --git a/src/database/kernels/xgemv_fast_rot.hpp b/src/database/kernels/xgemv_fast_rot.hpp index ee3cebdc..aff83ff2 100644 --- a/src/database/kernels/xgemv_fast_rot.hpp +++ b/src/database/kernels/xgemv_fast_rot.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XgemvFastRotHalf = { "XgemvFastRot", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"VW3",8}, {"WGS3",32}, {"WPT3",32} } }, + { "default", { {"VW3",8}, {"WGS3",32}, {"WPT3",32} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics Skylake ULT GT2", { {"VW3",8}, {"WGS3",128}, {"WPT3",32} } }, @@ -38,6 +44,7 @@ const Database::DatabaseEntry XgemvFastRotSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW3",8}, {"WGS3",64}, {"WPT3",32} } }, + { "Ellesmere", { {"VW3",8}, {"WGS3",32}, {"WPT3",32} } }, { "Tonga", { {"VW3",8}, {"WGS3",128}, {"WPT3",32} } }, { "Turks", { {"VW3",8}, {"WGS3",128}, {"WPT3",16} } }, { "default", { {"VW3",8}, {"WGS3",128}, {"WPT3",32} } }, @@ -86,6 +93,7 @@ const Database::DatabaseEntry XgemvFastRotComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW3",8}, {"WGS3",16}, {"WPT3",16} } }, + { "Ellesmere", { {"VW3",2}, {"WGS3",32}, {"WPT3",16} } }, { "Tonga", { {"VW3",4}, {"WGS3",32}, {"WPT3",32} } }, { "Turks", { {"VW3",4}, {"WGS3",32}, {"WPT3",8} } }, { "default", { {"VW3",8}, {"WGS3",32}, {"WPT3",32} } }, @@ -124,6 +132,7 @@ const Database::DatabaseEntry XgemvFastRotDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW3",4}, {"WGS3",16}, {"WPT3",16} } }, + { "Ellesmere", { {"VW3",4}, {"WGS3",16}, {"WPT3",16} } }, { "Tonga", { {"VW3",4}, {"WGS3",16}, {"WPT3",16} } }, { "default", { {"VW3",4}, {"WGS3",16}, {"WPT3",16} } }, } @@ -161,6 +170,7 @@ const Database::DatabaseEntry XgemvFastRotComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"VW3",4}, {"WGS3",32}, {"WPT3",16} } }, + { "Ellesmere", { {"VW3",4}, {"WGS3",16}, {"WPT3",16} } }, { "Tonga", { {"VW3",4}, {"WGS3",16}, {"WPT3",8} } }, { "default", { {"VW3",8}, {"WGS3",32}, {"WPT3",16} } }, } diff --git a/src/database/kernels/xger.hpp b/src/database/kernels/xger.hpp index ade9dcbf..ab528800 100644 --- a/src/database/kernels/xger.hpp +++ b/src/database/kernels/xger.hpp @@ -17,6 +17,12 @@ namespace database { const Database::DatabaseEntry XgerHalf = { "Xger", Precision::kHalf, { + { // AMD GPUs + kDeviceTypeGPU, "AMD", { + { "Ellesmere", { {"WGS1",64}, {"WGS2",1}, {"WPT",2} } }, + { "default", { {"WGS1",64}, {"WGS2",1}, {"WPT",2} } }, + } + }, { // Intel GPUs kDeviceTypeGPU, "Intel", { { "Intel(R) HD Graphics 5500 BroadWell U-Processor GT2", { {"WGS1",256}, {"WGS2",1}, {"WPT",2} } }, @@ -26,7 +32,7 @@ const Database::DatabaseEntry XgerHalf = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"WGS1",4}, {"WGS2",8}, {"WPT",2} } }, + { "default", { {"WGS1",64}, {"WGS2",1}, {"WPT",2} } }, } }, } @@ -39,13 +45,14 @@ const Database::DatabaseEntry XgerSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",256}, {"WGS2",1}, {"WPT",1} } }, + { "Ellesmere", { {"WGS1",64}, {"WGS2",4}, {"WPT",2} } }, { "Hawaii", { {"WGS1",64}, {"WGS2",2}, {"WPT",1} } }, { "Oland", { {"WGS1",32}, {"WGS2",4}, {"WPT",2} } }, { "Pitcairn", { {"WGS1",64}, {"WGS2",1}, {"WPT",1} } }, { "Tahiti", { {"WGS1",256}, {"WGS2",1}, {"WPT",1} } }, { "Tonga", { {"WGS1",256}, {"WGS2",1}, {"WPT",2} } }, { "Turks", { {"WGS1",64}, {"WGS2",4}, {"WPT",2} } }, - { "default", { {"WGS1",32}, {"WGS2",8}, {"WPT",1} } }, + { "default", { {"WGS1",32}, {"WGS2",8}, {"WPT",2} } }, } }, { // ARM GPUs @@ -92,7 +99,7 @@ const Database::DatabaseEntry XgerSingle = { }, { // Default kDeviceTypeAll, "default", { - { "default", { {"WGS1",32}, {"WGS2",4}, {"WPT",2} } }, + { "default", { {"WGS1",32}, {"WGS2",8}, {"WPT",2} } }, } }, } @@ -105,13 +112,14 @@ const Database::DatabaseEntry XgerComplexSingle = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WGS2",4}, {"WPT",1} } }, + { "Ellesmere", { {"WGS1",16}, {"WGS2",8}, {"WPT",2} } }, { "Hawaii", { {"WGS1",64}, {"WGS2",1}, {"WPT",2} } }, { "Oland", { {"WGS1",4}, {"WGS2",8}, {"WPT",1} } }, { "Pitcairn", { {"WGS1",128}, {"WGS2",2}, {"WPT",1} } }, { "Tahiti", { {"WGS1",64}, {"WGS2",2}, {"WPT",1} } }, { "Tonga", { {"WGS1",64}, {"WGS2",1}, {"WPT",1} } }, { "Turks", { {"WGS1",128}, {"WGS2",2}, {"WPT",1} } }, - { "default", { {"WGS1",128}, {"WGS2",2}, {"WPT",1} } }, + { "default", { {"WGS1",64}, {"WGS2",4}, {"WPT",1} } }, } }, { // ARM GPUs @@ -171,12 +179,13 @@ const Database::DatabaseEntry XgerDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",32}, {"WGS2",4}, {"WPT",1} } }, + { "Ellesmere", { {"WGS1",64}, {"WGS2",1}, {"WPT",4} } }, { "Hawaii", { {"WGS1",32}, {"WGS2",4}, {"WPT",2} } }, { "Oland", { {"WGS1",128}, {"WGS2",1}, {"WPT",2} } }, { "Pitcairn", { {"WGS1",64}, {"WGS2",1}, {"WPT",1} } }, { "Tahiti", { {"WGS1",64}, {"WGS2",2}, {"WPT",1} } }, { "Tonga", { {"WGS1",8}, {"WGS2",16}, {"WPT",2} } }, - { "default", { {"WGS1",32}, {"WGS2",8}, {"WPT",1} } }, + { "default", { {"WGS1",128}, {"WGS2",2}, {"WPT",1} } }, } }, { // ARM GPUs @@ -225,6 +234,7 @@ const Database::DatabaseEntry XgerComplexDouble = { { // AMD GPUs kDeviceTypeGPU, "AMD", { { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",64}, {"WGS2",1}, {"WPT",1} } }, + { "Ellesmere", { {"WGS1",8}, {"WGS2",16}, {"WPT",1} } }, { "Hawaii", { {"WGS1",128}, {"WGS2",1}, {"WPT",1} } }, { "Oland", { {"WGS1",16}, {"WGS2",16}, {"WPT",2} } }, { "Pitcairn", { {"WGS1",64}, {"WGS2",4}, {"WPT",1} } }, -- cgit v1.2.3