[MLIR][AVX512] Add integration test for vp2intersect

[springerm]

Summary

This is the first integration test for AVX512. Many AVX512
instructions require specialized hardware, so tests like this one cannot be
run natively on most machines. Therefore, AVX512 tests are enabled via a CMake
build parameter. In addition, if the path to Intel SDE is specified, AVX512
instructions are emulated and the test can run on any machine.

[aartbik]

Neat to see AVX512 integration tests!

mlir/CMakeLists.txt

114

arch-specific (omit the period) reads a bit better

mlir/integration_test/Dialect/Vector/CPU/AVX512/lit.local.cfg

4

This particular vpsintersection requires tiger lake AVX512_vp2intersect extensions. Do we want to introduce particular ways of specifying the different AVX512_xxx subsets, or do we just assume that AVX512 implies all subsets?

(don't need to resolve that in this CL, just wondering)

mlir/integration_test/Dialect/Vector/CPU/AVX512/test-vp2intersect-i32.mlir

6

ust -> must

[springerm]

Fix spelling mistakes.

[springerm]

mlir/integration_test/Dialect/Vector/CPU/AVX512/lit.local.cfg

4

That may be a good idea, depending on how important it is to run the code natively. Maybe for performance experiments.

My CPU actually supports a few AVX512 instructions, but not vp2intersect. (With SDE, the test can run on any machine, but it is really slow.)

[nicolasvasilache]

mlir/integration_test/Dialect/Vector/CPU/AVX512/test-vp2intersect-i32.mlir

4

Nice, thanks for investigating and making this work!

[aartbik]

mlir/integration_test/Dialect/Vector/CPU/AVX512/test-vp2intersect-i32.mlir

44

I see this has been submitted, so I am a bit after the fact, but since this deals with the AVX512 dialect, I would have had a small preference for

integration_test/Dialect/AVX512/CPU instead of
integration_test/Dialect/Vector/CPU/AVX512

mlir/CMakeLists.txt

 # C API needs all dialects for registration, but should be built before tests.
 add_subdirectory(lib/CAPI)
 if (MLIR_INCLUDE_TESTS)
   add_definitions(-DMLIR_INCLUDE_TESTS)
   add_subdirectory(unittests)
   add_subdirectory(test)
 endif()
 if (MLIR_INCLUDE_INTEGRATION_TESTS)
+  set(INTEL_SDE_EXECUTABLE "" CACHE STRING
+      "If set, arch.-specific integration tests are run with Intel SDE.")
+  option(MLIR_RUN_AVX512_TESTS "Run AVX512 tests.")
   add_definitions(-DMLIR_INCLUDE_INTEGRATION_TESTS)
   add_subdirectory(integration_test)
 endif()
 # Tools needs to come late to ensure that MLIR_ALL_LIBS is populated.
 # Generally things after this point may depend on MLIR_ALL_LIBS or libMLIR.so.
 add_subdirectory(tools)
 
 if( LLVM_INCLUDE_EXAMPLES )

mlir/integration_test/CMakeLists.txt

 # Passed to lit.site.cfg.py.in to set up the path where to find the libraries.
 set(MLIR_INTEGRATION_TEST_DIR ${CMAKE_LIBRARY_OUTPUT_DIRECTORY})
 
 set(MLIR_INTEGRATION_TEST_DEPENDS
   FileCheck count not
+  lli
   mlir-cpu-runner
   mlir-opt
   mlir_runner_utils
   mlir_c_runner_utils
   )
 
 configure_lit_site_cfg(
   ${CMAKE_CURRENT_SOURCE_DIR}/lit.site.cfg.py.in

mlir/integration_test/Dialect/Vector/CPU/AVX512/lit.local.cfg

+import sys
+
+# AVX512 tests must be enabled via build flag.
+if config.mlir_run_avx512_tests != 'ON':
+    config.unsupported = True
+
+# No JIT on win32.
+if sys.platform == 'win32':
+    config.unsupported = True
+
+if config.intel_sde_executable:
+    # Run test in emulator (Intel SDE).
+    config.substitutions.append(('%lli', config.intel_sde_executable + ' -tgl -- lli'))
+else:
+    config.substitutions.append(('%lli', 'lli'))

mlir/integration_test/Dialect/Vector/CPU/AVX512/test-vp2intersect-i32.mlir

+// RUN: mlir-opt %s -convert-scf-to-std -convert-vector-to-llvm="enable-avx512" -convert-std-to-llvm  | \
+// RUN: mlir-translate  --avx512-mlir-to-llvmir | \
+// RUN: %lli --entry-function=entry --mattr="avx512bw,avx512vp2intersect" --dlopen=%mlir_integration_test_dir/libmlir_c_runner_utils%shlibext | \
+// RUN: FileCheck %s
+
+// Note: To run this test, your CPU ust support AVX512 vp2intersect.
+
+func @entry() -> i32 {
+  %i0 = constant 0 : i32
+  %i1 = constant 1: i32
+  %i2 = constant 2: i32
+  %i3 = constant 7: i32
+  %i4 = constant 12: i32
+  %i5 = constant -10: i32
+  %i6 = constant -219: i32
+
+  %v0 = vector.broadcast %i1 : i32 to vector<16xi32>
+  %v1 = vector.insert %i2, %v0[1] : i32 into vector<16xi32>
+  %v2 = vector.insert %i3, %v1[4] : i32 into vector<16xi32>
+  %v3 = vector.insert %i4, %v2[6] : i32 into vector<16xi32>
+  %v4 = vector.insert %i5, %v3[7] : i32 into vector<16xi32>
+  %v5 = vector.insert %i0, %v4[10] : i32 into vector<16xi32>
+  %v6 = vector.insert %i0, %v5[12] : i32 into vector<16xi32>
+  %v7 = vector.insert %i3, %v6[13] : i32 into vector<16xi32>
+  %v8 = vector.insert %i3, %v7[14] : i32 into vector<16xi32>
+  %v9 = vector.insert %i0, %v8[15] : i32 into vector<16xi32>
+  vector.print %v9 : vector<16xi32>
+  // CHECK: ( 1, 2, 1, 1, 7, 1, 12, -10, 1, 1, 0, 1, 0, 7, 7, 0 )
+
+  %w0 = vector.broadcast %i1 : i32 to vector<16xi32>
+  %w1 = vector.insert %i2, %w0[4] : i32 into vector<16xi32>
+  %w2 = vector.insert %i6, %w1[7] : i32 into vector<16xi32>
+  %w3 = vector.insert %i4, %w2[8] : i32 into vector<16xi32>
+  %w4 = vector.insert %i4, %w3[9] : i32 into vector<16xi32>
+  %w5 = vector.insert %i4, %w4[10] : i32 into vector<16xi32>
+  %w6 = vector.insert %i0, %w5[11] : i32 into vector<16xi32>
+  %w7 = vector.insert %i0, %w6[12] : i32 into vector<16xi32>
+  %w8 = vector.insert %i0, %w7[13] : i32 into vector<16xi32>
+  %w9 = vector.insert %i0, %w8[15] : i32 into vector<16xi32>
+  vector.print %w9 : vector<16xi32>
+  // CHECK: ( 1, 1, 1, 1, 2, 1, 1, -219, 12, 12, 12, 0, 0, 0, 1, 0 )
+
+  %k1, %k2 = avx512.vp2intersect %v9, %w9 : vector<16xi32>
+
+  vector.print %k1 : vector<16xi1>
+  // CHECK: ( 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1 )
+
+  vector.print %k2 : vector<16xi1>
+  // CHECK: ( 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1 )
+
+  return %i0 : i32
+}

mlir/integration_test/lit.site.cfg.py.in

 config.host_ldflags = '@HOST_LDFLAGS@'
 config.llvm_use_sanitizer = "@LLVM_USE_SANITIZER@"
 config.llvm_host_triple = '@LLVM_HOST_TRIPLE@'
 config.host_arch = "@HOST_ARCH@"
 config.mlir_src_root = "@MLIR_SOURCE_DIR@"
 config.mlir_obj_root = "@MLIR_BINARY_DIR@"
 config.mlir_tools_dir = "@MLIR_TOOLS_DIR@"
 config.mlir_integration_test_dir = "@MLIR_INTEGRATION_TEST_DIR@"
+config.intel_sde_executable = "@INTEL_SDE_EXECUTABLE@"
+config.mlir_run_avx512_tests = "@MLIR_RUN_AVX512_TESTS@"
 
 # Support substitution of the tools_dir with user parameters. This is
 # used when we can't determine the tool dir at configuration time.
 try:
     config.llvm_tools_dir = config.llvm_tools_dir % lit_config.params
     config.llvm_shlib_dir = config.llvm_shlib_dir % lit_config.params
 except KeyError:
     e = sys.exc_info()[1]
     key, = e.args
     lit_config.fatal("unable to find %r parameter, use '--param=%s=VALUE'" % (key,key))
 
 import lit.llvm
 lit.llvm.initialize(lit_config, config)
 
 # Let the main config do the real work.
 lit_config.load_config(config, "@MLIR_SOURCE_DIR@/integration_test/lit.cfg.py")