added support for gfx940,gfx941,gfx942 (#335)

* added support for gfx940,gfx941,gfx942 * Allow CI to pick up dependencies from same branch * Adjust SpMV tests for gfx940 (#588) * Adjust tests for gfx940 * Adjust tests for gfx940 spmv_csc --------- Co-authored-by: jsandham <james.sandham@amd.com> * Adjust tests for rounding errors on gfx940 * Fix rounding errors encountered on gfx940 * Adjust failing csrmv test on gfx940 --------- Co-authored-by: Eiden Yoshida <47196116+eidenyoshida@users.noreply.github.com> Co-authored-by: jsandham <james.sandham@amd.com> Co-authored-by: James Sandham <33790278+jsandham@users.noreply.github.com>
ROCm · Jul 7, 2023 · 2ac2e27 · 2ac2e27
1 parent 8a971a8
commit 2ac2e27
Show file tree

Hide file tree

Showing 8 changed files with 238 additions and 7 deletions.
diff --git a/.jenkins/common.groovy b/.jenkins/common.groovy
@@ -15,7 +15,7 @@ def runCompileCommand(platform, project, jobName, boolean sameOrg=false)
     {
         project.libraryDependencies.each
         { libraryName ->
-            getDependenciesCommand += auxiliary.getLibrary(libraryName, platform.jenkinsLabel, 'develop', sameOrg)
+            getDependenciesCommand += auxiliary.getLibrary(libraryName, platform.jenkinsLabel, null, sameOrg)
         }
     }
 

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,9 +2,10 @@
 
 Full documentation for rocSPARSE is available at [rocsparse.readthedocs.io](https://rocsparse.readthedocs.io/en/latest/).
 
-## rocSPARSE 2.5.3 for ROCm 5.7.0
+## rocSPARSE 2.5.4 for ROCm 5.7.0
 ### Added
 - Added more mixed precisions for SpMV, (matrix: float, vectors: double, calculation: double) and (matrix: rocsparse_float_complex, vectors: rocsparse_double_complex, calculation: rocsparse_double_complex)
+- Added support for gfx940, gfx941 and gfx942
 ### Improved
 - Fixed a bug in csrsm and bsrsm
 ### Known Issues

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -147,7 +147,7 @@ endif()
 #Set the AMDGPU_TARGETS with backward compatibility
  if(COMMAND rocm_check_target_ids)
     rocm_check_target_ids(DEFAULT_AMDGPU_TARGETS
-        TARGETS "gfx803;gfx900:xnack-;gfx906:xnack-;gfx908:xnack-;gfx90a:xnack-;gfx90a:xnack+;gfx1030;gfx1100;gfx1101;gfx1102"
+        TARGETS "gfx803;gfx900:xnack-;gfx906:xnack-;gfx908:xnack-;gfx90a:xnack-;gfx90a:xnack+;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx1101;gfx1102"
    )
 else()
     # Use target ID syntax if supported for AMDGPU_TARGETS
@@ -179,7 +179,7 @@ if( CMAKE_CXX_COMPILER_ID MATCHES "Clang" )
 endif( )
 
 # Setup version
-set(VERSION_STRING "2.5.3")
+set(VERSION_STRING "2.5.4")
 rocm_setup_version(VERSION ${VERSION_STRING})
 set(rocsparse_SOVERSION 0.1)
 

diff --git a/clients/tests/test_bsrmv.yaml b/clients/tests/test_bsrmv.yaml
@@ -146,6 +146,7 @@ Tests:
 
 - name: bsrmv_file
   category: pre_checkin
+  hardware: [gfx900, gfx906, gfx908, gfx90a, gfx1030]
   function: bsrmv
   precision: *single_double_precisions
   M: 1
@@ -162,6 +163,25 @@ Tests:
              nos5,
              nos7]
 
+- name: bsrmv_file
+  category: pre_checkin
+  hardware: [gfx940]
+  function: bsrmv
+  precision: *double_only_precisions
+  M: 1
+  N: 1
+  block_dim: [2, 15]
+  alpha_beta: *alpha_beta_range_checkin
+  direction: [rocsparse_direction_column]
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_one]
+  matrix: [rocsparse_matrix_file_rocalution]
+  storage: [rocsparse_storage_mode_sorted, rocsparse_storage_mode_unsorted]
+  filename: [scircuit,
+             nos3,
+             nos5,
+             nos7]
+
 - name: bsrmv_file
   category: pre_checkin
   function: bsrmv
@@ -270,6 +290,7 @@ Tests:
 
 - name: bsrmv_file
   category: nightly
+  hardware: [gfx900, gfx906, gfx908, gfx90a, gfx1030]
   function: bsrmv
   precision: *single_double_precisions
   M: 1
@@ -286,6 +307,41 @@ Tests:
              sme3Dc,
              shipsec1,
              webbase-1M]
+
+- name: bsrmv_file
+  category: nightly
+  hardware: [gfx940]
+  function: bsrmv
+  precision: *single_double_precisions
+  M: 1
+  N: 1
+  block_dim: [2, 12]
+  alpha_beta: *alpha_beta_range_nightly
+  direction: [rocsparse_direction_row]
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_zero]
+  matrix: [rocsparse_matrix_file_rocalution]
+  storage: [rocsparse_storage_mode_sorted, rocsparse_storage_mode_unsorted]
+  filename: [amazon0312,
+             Chebyshev4,
+             webbase-1M]
+
+- name: bsrmv_file
+  category: nightly
+  hardware: [gfx940]
+  function: bsrmv
+  precision: *double_only_precisions
+  M: 1
+  N: 1
+  block_dim: [2, 12]
+  alpha_beta: *alpha_beta_range_nightly
+  direction: [rocsparse_direction_row]
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_zero]
+  matrix: [rocsparse_matrix_file_rocalution]
+  storage: [rocsparse_storage_mode_sorted, rocsparse_storage_mode_unsorted]
+  filename: [sme3Dc,
+             shipsec1]
 
 - name: bsrmv_file
   category: nightly

diff --git a/clients/tests/test_csrmv.yaml b/clients/tests/test_csrmv.yaml
@@ -188,6 +188,7 @@ Tests:
 
 - name: csrmv_file
   category: pre_checkin
+  hardware: [gfx900, gfx906, gfx908, gfx90a, gfx1030]
   function: csrmv
   precision: *single_double_precisions
   M: 1
@@ -204,6 +205,25 @@ Tests:
              nos5,
              nos7]
 
+- name: csrmv_file
+  category: pre_checkin
+  hardware: [gfx940]
+  function: csrmv
+  precision: *double_only_precisions
+  M: 1
+  N: 1
+  alpha_beta: *alpha_beta_range_checkin
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_zero]
+  matrix: [rocsparse_matrix_file_rocalution]
+  matrix_type: [rocsparse_matrix_type_general]
+  spmv_alg: [rocsparse_spmv_alg_csr_adaptive]
+  filename: [scircuit,
+             nos1,
+             nos3,
+             nos5,
+             nos7]
+
 - name: csrmv_file
   category: nightly
   function: csrmv
@@ -215,14 +235,34 @@ Tests:
   baseA: [rocsparse_index_base_one]
   matrix: [rocsparse_matrix_file_rocalution]
   matrix_type: [rocsparse_matrix_type_general]
-  spmv_alg: [rocsparse_spmv_alg_csr_adaptive, rocsparse_spmv_alg_csr_stream]
+  spmv_alg: [rocsparse_spmv_alg_csr_adaptive]
   filename: [bibd_22_8,
              bmwcra_1,
              amazon0312,
              ASIC_320k,
              sme3Dc,
              shipsec1]
 
+- name: csrmv_file
+  category: nightly
+  hardware: [gfx940]
+  function: csrmv
+  precision: *double_only_precisions
+  M: 1
+  N: 1
+  alpha_beta: *alpha_beta_range_nightly
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_one]
+  matrix: [rocsparse_matrix_file_rocalution]
+  matrix_type: [rocsparse_matrix_type_general]
+  spmv_alg: [rocsparse_spmv_alg_csr_stream]
+  filename: [bibd_22_8,
+             bmwcra_1,
+             amazon0312,
+             ASIC_320k,
+             sme3Dc,
+             mac_econ_fwd500]
+
 - name: csrmv_file
   category: quick
   function: csrmv
@@ -240,6 +280,7 @@ Tests:
 
 - name: csrmv_file
   category: pre_checkin
+  hardware: [gfx900, gfx906, gfx908, gfx90a, gfx1030]
   function: csrmv
   precision: *single_double_precisions_complex
   M: 1
@@ -253,6 +294,22 @@ Tests:
   filename: [mplate,
              Chevron2]
 
+- name: csrmv_file
+  category: pre_checkin
+  hardware: [gfx940]
+  function: csrmv
+  precision: *double_only_precisions_complex
+  M: 1
+  N: 1
+  alpha_beta: *alpha_beta_range_checkin
+  transA: [rocsparse_operation_none, rocsparse_operation_conjugate_transpose]
+  baseA: [rocsparse_index_base_one]
+  matrix: [rocsparse_matrix_file_rocalution]
+  matrix_type: [rocsparse_matrix_type_general]
+  spmv_alg: [rocsparse_spmv_alg_csr_adaptive, rocsparse_spmv_alg_csr_stream]
+  filename: [mplate,
+             Chevron2]
+
 - name: csrmv_file
   category: nightly
   function: csrmv

diff --git a/clients/tests/test_gebsrmv.yaml b/clients/tests/test_gebsrmv.yaml
@@ -386,9 +386,40 @@ Tests:
   baseA: [rocsparse_index_base_zero]
   matrix: [rocsparse_matrix_file_rocalution]
   filename: [amazon0312,
-             shipsec1,
              webbase-1M]
 
+- name: gebsrmv_file
+  category: nightly
+  hardware: [gfx900, gfx906, gfx908, gfx90a, gfx1030]
+  function: gebsrmv
+  precision: *single_double_precisions
+  M: 1
+  N: 1
+  row_block_dimA: [4]
+  col_block_dimA: [2, 5, 8]
+  alpha_beta: *alpha_beta_range_1
+  direction: [rocsparse_direction_row]
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_zero]
+  matrix: [rocsparse_matrix_file_rocalution]
+  filename: [shipsec1]
+
+- name: gebsrmv_file
+  category: nightly
+  hardware: [gfx940]
+  function: gebsrmv
+  precision: *double_only_precisions
+  M: 1
+  N: 1
+  row_block_dimA: [4]
+  col_block_dimA: [2, 5, 8]
+  alpha_beta: *alpha_beta_range_1
+  direction: [rocsparse_direction_row]
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_zero]
+  matrix: [rocsparse_matrix_file_rocalution]
+  filename: [shipsec1]
+
 - name: gebsrmv_file
   category: nightly
   function: gebsrmv

diff --git a/clients/tests/test_spmv_csc.yaml b/clients/tests/test_spmv_csc.yaml
@@ -252,6 +252,7 @@ Tests:
 
 - name: spmv_csc
   category: quick
+  hardware: [gfx900, gfx906, gfx908, gfx90a, gfx1030]
   function: spmv_csc
   indextype: *i32i32_i64i32_i64i64
   precision: *single_double_precisions_complex_real
@@ -264,6 +265,23 @@ Tests:
   uplo: [rocsparse_fill_mode_lower]
   spmv_alg: [rocsparse_spmv_alg_csr_adaptive, rocsparse_spmv_alg_csr_stream]
 
+- name: spmv_csc
+  category: quick
+  hardware: [gfx940]
+  function: spmv_csc
+  indextype: *i32i32_i64i32_i64i64
+  precision: *single_double_precisions_complex_real
+  M_N: *M_N_range_quick
+  alpha_beta: *alpha_beta_range_quick
+  transA: [rocsparse_operation_none]
+  baseA: [rocsparse_index_base_zero]
+  matrix: [rocsparse_matrix_random]
+  matrix_type: [rocsparse_matrix_type_symmetric, rocsparse_matrix_type_triangular]
+  uplo: [rocsparse_fill_mode_lower]
+  spmv_alg: [rocsparse_spmv_alg_csr_adaptive, rocsparse_spmv_alg_csr_stream]
+
+
+
 - name: spmv_csc
   category: pre_checkin
   function: spmv_csc