wasm-micro-runtime/samples/workload/XNNPACK/xnnpack.patch

diff --git a/.bazelrc b/.bazelrc
index ec740f38..2c193244 100644
--- a/.bazelrc
+++ b/.bazelrc
@@ -49,4 +49,10 @@ build:ios_fat --watchos_cpus=armv7k
 build:macos --apple_platform_type=macos
 
 build:macos_arm64 --config=macos
-build:macos_arm64 --cpu=darwin_arm64
\ No newline at end of file
+build:macos_arm64 --cpu=darwin_arm64
+
+build:wasm --copt=-msimd128
+build:wasm --cpu=wasm
+build:wasm --crosstool_top=@emsdk//emscripten_toolchain:everything
+build:wasm --host_crosstool_top=@bazel_tools//tools/cpp:toolchain
+
diff --git a/BUILD.bazel b/BUILD.bazel
index 1f2b15a8..e7abf838 100644
--- a/BUILD.bazel
+++ b/BUILD.bazel
@@ -4996,7 +4996,7 @@ xnnpack_cc_library(
 ######################### Benchmarks for micro-kernels #########################
 
 xnnpack_benchmark(
-    name = "qs8_gemm_bench",
+    name = "qs8_gemm_bench.wasm",
     srcs = [
         "bench/gemm.h",
         "bench/qs8-gemm.cc",
@@ -5007,7 +5007,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "qs8_requantization_bench",
+    name = "qs8_requantization_bench.wasm",
     srcs = [
         "bench/qs8-requantization.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5017,7 +5017,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "qu8_gemm_bench",
+    name = "qu8_gemm_bench.wasm",
     srcs = [
         "bench/gemm.h",
         "bench/qu8-gemm.cc",
@@ -5028,7 +5028,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "qu8_requantization_bench",
+    name = "qu8_requantization_bench.wasm",
     srcs = [
         "bench/qu8-requantization.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5038,11 +5038,10 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f16_igemm_bench",
+    name = "f16_igemm_bench.wasm",
     srcs = [
         "bench/f16-igemm.cc",
         "bench/conv.h",
-        "bench/google/conv.h",
         "src/xnnpack/AlignedAllocator.h",
     ] + MICROKERNEL_BENCHMARK_HDRS,
     deps = MICROKERNEL_BENCHMARK_DEPS + [
@@ -5052,7 +5051,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f16_gemm_bench",
+    name = "f16_gemm_bench.wasm",
     srcs = [
         "bench/f16-gemm.cc",
         "bench/gemm.h",
@@ -5064,7 +5063,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f16_spmm_bench",
+    name = "f16_spmm_bench.wasm",
     srcs = [
         "bench/f16-spmm.cc",
         "bench/spmm.h",
@@ -5074,7 +5073,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_igemm_bench",
+    name = "f32_igemm_bench.wasm",
     srcs = [
         "bench/f32-igemm.cc",
         "bench/conv.h",
@@ -5087,7 +5086,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f16_relu_bench",
+    name = "f16_relu_bench.wasm",
     srcs = [
         "bench/f16-relu.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5096,7 +5095,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_conv_hwc_bench",
+    name = "f32_conv_hwc_bench.wasm",
     srcs = [
         "bench/f32-conv-hwc.cc",
         "bench/dconv.h",
@@ -5108,7 +5107,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_conv_hwc2chw_bench",
+    name = "f32_conv_hwc2chw_bench.wasm",
     srcs = [
         "bench/f32-conv-hwc2chw.cc",
         "bench/dconv.h",
@@ -5120,11 +5119,10 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f16_dwconv_bench",
+    name = "f16_dwconv_bench.wasm",
     srcs = [
         "bench/f16-dwconv.cc",
         "bench/dwconv.h",
-        "bench/google/dwconv.h",
         "src/xnnpack/AlignedAllocator.h",
     ] + MICROKERNEL_BENCHMARK_HDRS,
     deps = MICROKERNEL_BENCHMARK_DEPS + [
@@ -5134,7 +5132,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_dwconv_bench",
+    name = "f32_dwconv_bench.wasm",
     srcs = [
         "bench/f32-dwconv.cc",
         "bench/dwconv.h",
@@ -5147,7 +5145,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_dwconv2d_chw_bench",
+    name = "f32_dwconv2d_chw_bench.wasm",
     srcs = [
         "bench/f32-dwconv2d-chw.cc",
         "bench/dwconv.h",
@@ -5160,7 +5158,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_gemm_bench",
+    name = "f32_gemm_bench.wasm",
     srcs = [
         "bench/f32-gemm.cc",
         "bench/gemm.h",
@@ -5171,7 +5169,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_hswish_bench",
+    name = "f32_hswish_bench.wasm",
     srcs = [
         "bench/f32-hswish.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5180,7 +5178,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_raddexpminusmax_bench",
+    name = "f32_raddexpminusmax_bench.wasm",
     srcs = [
         "bench/f32-raddexpminusmax.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5189,7 +5187,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_raddextexp_bench",
+    name = "f32_raddextexp_bench.wasm",
     srcs = [
         "bench/f32-raddextexp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5198,7 +5196,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_raddstoreexpminusmax_bench",
+    name = "f32_raddstoreexpminusmax_bench.wasm",
     srcs = [
         "bench/f32-raddstoreexpminusmax.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5207,7 +5205,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_relu_bench",
+    name = "f32_relu_bench.wasm",
     srcs = [
         "bench/f32-relu.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5216,7 +5214,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_rmax_bench",
+    name = "f32_rmax_bench.wasm",
     srcs = [
         "bench/f32-rmax.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5225,7 +5223,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_sigmoid_bench",
+    name = "f32_sigmoid_bench.wasm",
     srcs = [
         "bench/f32-sigmoid.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5234,7 +5232,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_spmm_bench",
+    name = "f32_spmm_bench.wasm",
     srcs = [
         "bench/f32-spmm.cc",
         "bench/spmm.h",
@@ -5244,7 +5242,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_softmax_bench",
+    name = "f32_softmax_bench.wasm",
     srcs = [
         "bench/f32-softmax.cc",
     ] + MICROKERNEL_BENCHMARK_HDRS,
@@ -5253,7 +5251,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_velu_bench",
+    name = "f32_velu_bench.wasm",
     srcs = [
         "bench/f32-velu.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5262,7 +5260,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_vscaleexpminusmax_bench",
+    name = "f32_vscaleexpminusmax_bench.wasm",
     srcs = [
         "bench/f32-vscaleexpminusmax.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5271,7 +5269,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_vscaleextexp_bench",
+    name = "f32_vscaleextexp_bench.wasm",
     srcs = [
         "bench/f32-vscaleextexp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5280,7 +5278,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_vsqrt_bench",
+    name = "f32_vsqrt_bench.wasm",
     srcs = [
         "bench/f32-vsqrt.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5289,7 +5287,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_im2col_gemm_bench",
+    name = "f32_im2col_gemm_bench.wasm",
     srcs = [
         "bench/f32-im2col-gemm.cc",
         "bench/conv.h",
@@ -5302,7 +5300,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "rounding_bench",
+    name = "rounding_bench.wasm",
     srcs = [
         "bench/rounding.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5314,7 +5312,7 @@ xnnpack_benchmark(
 ########################### Benchmarks for operators ###########################
 
 xnnpack_benchmark(
-    name = "average_pooling_bench",
+    name = "average_pooling_bench.wasm",
     srcs = ["bench/average-pooling.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5322,7 +5320,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "bankers_rounding_bench",
+    name = "bankers_rounding_bench.wasm",
     srcs = ["bench/bankers-rounding.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5330,7 +5328,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "ceiling_bench",
+    name = "ceiling_bench.wasm",
     srcs = ["bench/ceiling.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5338,13 +5336,13 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "channel_shuffle_bench",
+    name = "channel_shuffle_bench.wasm",
     srcs = ["bench/channel-shuffle.cc"],
     deps = OPERATOR_BENCHMARK_DEPS,
 )
 
 xnnpack_benchmark(
-    name = "convolution_bench",
+    name = "convolution_bench.wasm",
     srcs = ["bench/convolution.cc"],
     copts = xnnpack_optional_tflite_copts() + xnnpack_optional_armcl_copts(),
     tags = ["nowin32"],
@@ -5352,7 +5350,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "deconvolution_bench",
+    name = "deconvolution_bench.wasm",
     srcs = ["bench/deconvolution.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5360,7 +5358,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "elu_bench",
+    name = "elu_bench.wasm",
     srcs = ["bench/elu.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5368,7 +5366,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "floor_bench",
+    name = "floor_bench.wasm",
     srcs = ["bench/floor.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5376,13 +5374,13 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "global_average_pooling_bench",
+    name = "global_average_pooling_bench.wasm",
     srcs = ["bench/global-average-pooling.cc"],
     deps = OPERATOR_BENCHMARK_DEPS,
 )
 
 xnnpack_benchmark(
-    name = "hardswish_bench",
+    name = "hardswish_bench.wasm",
     srcs = ["bench/hardswish.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5390,13 +5388,13 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "max_pooling_bench",
+    name = "max_pooling_bench.wasm",
     srcs = ["bench/max-pooling.cc"],
     deps = OPERATOR_BENCHMARK_DEPS,
 )
 
 xnnpack_benchmark(
-    name = "sigmoid_bench",
+    name = "sigmoid_bench.wasm",
     srcs = ["bench/sigmoid.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5404,7 +5402,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "prelu_bench",
+    name = "prelu_bench.wasm",
     srcs = ["bench/prelu.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5412,7 +5410,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "softmax_bench",
+    name = "softmax_bench.wasm",
     srcs = ["bench/softmax.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5420,7 +5418,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "square_root_bench",
+    name = "square_root_bench.wasm",
     srcs = ["bench/square-root.cc"],
     copts = xnnpack_optional_tflite_copts(),
     tags = ["nowin32"],
@@ -5428,7 +5426,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "truncation_bench",
+    name = "truncation_bench.wasm",
     srcs = ["bench/truncation.cc"],
     deps = OPERATOR_BENCHMARK_DEPS,
 )
@@ -5620,7 +5618,7 @@ cc_library(
 )
 
 xnnpack_benchmark(
-    name = "f32_dwconv_e2e_bench",
+    name = "f32_dwconv_e2e_bench.wasm",
     srcs = [
         "bench/f32-dwconv-e2e.cc",
         "bench/end2end.h",
@@ -5635,7 +5633,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_gemm_e2e_bench",
+    name = "f32_gemm_e2e_bench.wasm",
     srcs = [
         "bench/f32-gemm-e2e.cc",
         "bench/end2end.h",
@@ -5650,7 +5648,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "qs8_gemm_e2e_bench",
+    name = "qs8_gemm_e2e_bench.wasm",
     srcs = [
         "bench/qs8-gemm-e2e.cc",
         "bench/end2end.h",
@@ -5663,7 +5661,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "end2end_bench",
+    name = "end2end_bench.wasm",
     srcs = ["bench/end2end.cc"],
     deps = [
         ":XNNPACK",
@@ -5690,7 +5688,7 @@ xnnpack_benchmark(
 #################### Accuracy evaluation for math functions ####################
 
 xnnpack_benchmark(
-    name = "f32_exp_ulp_eval",
+    name = "f32_exp_ulp_eval.wasm",
     srcs = [
         "eval/f32-exp-ulp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5702,7 +5700,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_expminus_ulp_eval",
+    name = "f32_expminus_ulp_eval.wasm",
     srcs = [
         "eval/f32-expminus-ulp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5714,7 +5712,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_expm1minus_ulp_eval",
+    name = "f32_expm1minus_ulp_eval.wasm",
     srcs = [
         "eval/f32-expm1minus-ulp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5726,7 +5724,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_extexp_ulp_eval",
+    name = "f32_extexp_ulp_eval.wasm",
     srcs = [
         "eval/f32-extexp-ulp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5738,7 +5736,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_sigmoid_ulp_eval",
+    name = "f32_sigmoid_ulp_eval.wasm",
     srcs = [
         "eval/f32-sigmoid-ulp.cc",
         "src/xnnpack/AlignedAllocator.h",
@@ -5750,7 +5748,7 @@ xnnpack_benchmark(
 )
 
 xnnpack_benchmark(
-    name = "f32_sqrt_ulp_eval",
+    name = "f32_sqrt_ulp_eval.wasm",
     srcs = [
         "eval/f32-sqrt-ulp.cc",
         "src/xnnpack/AlignedAllocator.h",
diff --git a/WORKSPACE b/WORKSPACE
index 4fa1aa2f..6181aab2 100644
--- a/WORKSPACE
+++ b/WORKSPACE
@@ -5,8 +5,8 @@ load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
 # Bazel rule definitions
 http_archive(
     name = "rules_cc",
-    strip_prefix = "rules_cc-master",
-    urls = ["https://github.com/bazelbuild/rules_cc/archive/master.zip"],
+    strip_prefix = "rules_cc-main",
+    urls = ["https://github.com/bazelbuild/rules_cc/archive/main.zip"],
 )
 
 # Google Test framework, used by most unit-tests.
@@ -19,8 +19,8 @@ http_archive(
 # Google Benchmark library, used in micro-benchmarks.
 http_archive(
     name = "com_google_benchmark",
-    strip_prefix = "benchmark-master",
-    urls = ["https://github.com/google/benchmark/archive/master.zip"],
+    strip_prefix = "benchmark-1.5.3",
+    urls = ["https://github.com/google/benchmark/archive/refs/tags/v1.5.3.zip"],
 )
 
 # FP16 library, used for half-precision conversions
@@ -89,3 +89,18 @@ android_ndk_repository(name = "androidndk")
 
 # Android SDK location and API is auto-detected from $ANDROID_HOME environment variable
 android_sdk_repository(name = "androidsdk")
+
+# emscripten library
+http_archive(
+    name = "emsdk",
+    strip_prefix = "emsdk-c1589b55641787d55d53e883852035beea9aec3f/bazel",
+    url = "https://github.com/emscripten-core/emsdk/archive/c1589b55641787d55d53e883852035beea9aec3f.tar.gz",
+    sha256 = "7a58a9996b113d3e0675df30b5f17e28aa47de2e684a844f05394fe2f6f12e8e",
+)
+
+load("@emsdk//:deps.bzl", emsdk_deps = "deps")
+emsdk_deps()
+
+load("@emsdk//:emscripten_deps.bzl", emsdk_emscripten_deps = "emscripten_deps")
+emsdk_emscripten_deps()
+
diff --git a/build_defs.bzl b/build_defs.bzl
index 10345032..0e926fca 100644
--- a/build_defs.bzl
+++ b/build_defs.bzl
@@ -1,6 +1,6 @@
 """Build definitions and rules for XNNPACK."""
 
-load(":emscripten.bzl", "xnnpack_emscripten_benchmark_linkopts", "xnnpack_emscripten_deps", "xnnpack_emscripten_minimal_linkopts", "xnnpack_emscripten_test_linkopts")
+load(":emscripten.bzl", "xnnpack_emscripten_benchmark_linkopts", "xnnpack_emscripten_deps", "xnnpack_emscripten_minimal_linkopts", "xnnpack_emscripten_test_linkopts", "xnnpack_emscripten_benchmark_copts")
 
 def xnnpack_visibility():
     """Visibility of :XNNPACK target.
@@ -424,10 +424,15 @@ def xnnpack_benchmark(name, srcs, copts = [], deps = [], tags = []):
             ":windows_x86_64_mingw": ["-Wno-unused-function"],
             ":windows_x86_64_msys": ["-Wno-unused-function"],
             ":windows_x86_64": [],
+            ":emscripten": xnnpack_emscripten_benchmark_copts(),
+            ":emscripten_wasm": xnnpack_emscripten_benchmark_copts(),
+            ":emscripten_wasmsimd": xnnpack_emscripten_benchmark_copts(),
             "//conditions:default": ["-Wno-unused-function"],
         }) + copts,
         linkopts = select({
             ":emscripten": xnnpack_emscripten_benchmark_linkopts(),
+            ":emscripten_wasm": xnnpack_emscripten_benchmark_linkopts(),
+            ":emscripten_wasmsimd": xnnpack_emscripten_benchmark_linkopts(),
             ":windows_x86_64_mingw": ["-lshlwapi"],
             ":windows_x86_64_msys": ["-lshlwapi"],
             "//conditions:default": [],
diff --git a/emscripten.bzl b/emscripten.bzl
index 0a0caedf..aafe3199 100644
--- a/emscripten.bzl
+++ b/emscripten.bzl
@@ -6,6 +6,7 @@ def xnnpack_emscripten_minimal_linkopts():
         "-s ASSERTIONS=0",
         "-s ERROR_ON_UNDEFINED_SYMBOLS=1",
         "-s EXIT_RUNTIME=1",
+        "--oformat=wasm",
     ]
 
 def xnnpack_emscripten_test_linkopts():
@@ -17,21 +18,36 @@ def xnnpack_emscripten_test_linkopts():
         "-s EXIT_RUNTIME=1",
         "-s ALLOW_MEMORY_GROWTH=1",
         "--pre-js $(location :preamble.js.lds)",
+        "--oformat=wasm",
     ]
 
 def xnnpack_emscripten_benchmark_linkopts():
     """Emscripten-specific linkopts for benchmarks."""
     return [
         "-s ASSERTIONS=1",
-        "-s ERROR_ON_UNDEFINED_SYMBOLS=1",
-        "-s EXIT_RUNTIME=1",
-        "-s ALLOW_MEMORY_GROWTH=1",
+        "-s ERROR_ON_UNDEFINED_SYMBOLS=0",
+        "-s ALLOW_MEMORY_GROWTH=0",
         "-s TOTAL_MEMORY=436207616",  # 416M
-        "--pre-js $(location :preamble.js.lds)",
+        "-s USE_PTHREADS=0",
+        "-s STANDALONE_WASM=1",
+        "-Wno-unused",
+        "-Wl,--export=__heap_base",
+        "-Wl,--export=__data_end",
+        "-Wl,--export=malloc",
+        "-Wl,--export=free",
+        "--oformat=wasm",
     ]
 
 def xnnpack_emscripten_deps():
     """Emscripten-specific dependencies for unit tests and benchmarks."""
+    return []
+
+def xnnpack_emscripten_benchmark_copts():
     return [
-        ":preamble.js.lds",
+        "-s ASSERTIONS=1",
+        "-s ERROR_ON_UNDEFINED_SYMBOLS=0",
+        "-s ALLOW_MEMORY_GROWTH=0",
+        "-s USE_PTHREADS=0",
+        "-s STANDALONE_WASM=1",
+        "-Wno-unused",
     ]
diff --git a/third_party/cpuinfo.BUILD b/third_party/cpuinfo.BUILD
index 128d683e..f6c287c4 100644
--- a/third_party/cpuinfo.BUILD
+++ b/third_party/cpuinfo.BUILD
@@ -343,5 +343,5 @@ config_setting(
 
 config_setting(
     name = "emscripten",
-    values = {"crosstool_top": "//toolchain:emscripten"},
+    values = {"crosstool_top": "@emsdk//emscripten_toolchain:everything"},
 )