diff --git a/benchmark/.gitignore b/benchmark/.gitignore
index 496420140..d7f132308 100644
--- a/benchmark/.gitignore
+++ b/benchmark/.gitignore
@@ -1,3 +1,4 @@
 Manifest.toml
 *.json
-*.o
\ No newline at end of file
+*.o
+jl_*
\ No newline at end of file
diff --git a/benchmark/Project.toml b/benchmark/Project.toml
index 493ea907e..4584513ac 100644
--- a/benchmark/Project.toml
+++ b/benchmark/Project.toml
@@ -1,4 +1,5 @@
 [deps]
+ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"
 BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
 Finch = "9177782c-1635-4eb9-9bfb-d9dfa25e6bce"
 MatrixDepot = "b51810bb-c9f3-55da-ae3c-350fc1fbce05"
@@ -6,4 +7,4 @@ MatrixMarket = "4d4711f2-db25-561a-b6b3-d35e7d4047d3"
 PkgBenchmark = "32113eaa-f34f-5b0d-bd6c-c81e245fc73d"
 
 [compat]
-BenchmarkTools = "1.5"
\ No newline at end of file
+BenchmarkTools = "1.5"
diff --git a/benchmark/benchmarks.jl b/benchmark/benchmarks.jl
index 63d60f1e1..7935d5308 100644
--- a/benchmark/benchmarks.jl
+++ b/benchmark/benchmarks.jl
@@ -10,6 +10,29 @@ using BenchmarkTools
 using MatrixDepot
 using SparseArrays
 using Random
+using ArgParse
+
+s = ArgParseSettings("Run Finch.jl benchmarks. By default, all tests are run unless --include or --exclude options are provided. 
+If the environment variable FINCH_BENCHMARK_ARGS is set, it will override the given arguments.")
+
+@add_arg_table! s begin
+    "--include", "-i"
+    nargs = '*'
+    default = []
+    help = "list of benchmark suites to include, e.g., --include high-level structures"
+
+    "--exclude", "-e"
+    nargs = '*'
+    default = []
+    help = "list of benchmark suites to exclude, e.g., --exclude compile graphs"
+end
+
+if "FINCH_BENCHMARK_ARGS" in keys(ENV)
+    ARGS = split(ENV["FINCH_BENCHMARK_ARGS"], " ")
+end
+
+parsed_args = parse_args(ARGS, s)
+
 include(joinpath(@__DIR__, "../docs/examples/bfs.jl"))
 include(joinpath(@__DIR__, "../docs/examples/pagerank.jl"))
 include(joinpath(@__DIR__, "../docs/examples/shortest_paths.jl"))
@@ -20,73 +43,97 @@ SUITE = BenchmarkGroup()
 
 SUITE["high-level"] = BenchmarkGroup()
 
-let
-    k = Ref(0.0)
-    A = Tensor(Dense(Sparse(Element(0.0))), fsprand(10000, 10000, 0.01))
-    x = rand(1)
-    y = rand(1)
-    SUITE["high-level"]["permutedims(Dense(Sparse()))"] = @benchmarkable(permutedims($A, (2, 1)))
-end
+for (scheduler_name, scheduler) in [
+    "default_scheduler" => Finch.default_scheduler(),
+    "galley_scheduler" => Finch.galley_scheduler(),
+]
+    Finch.with_scheduler(scheduler) do
+        let
+            A = Tensor(Dense(Sparse(Element(0.0))), fsprand(10000, 10000, 0.01))
+            SUITE["high-level"]["permutedims(Dense(Sparse()))"][scheduler_name] = @benchmarkable(permutedims($A, (2, 1)))
+        end
 
-let
-    k = Ref(0.0)
-    A = Tensor(Dense(Dense(Element(0.0))), rand(10000, 10000))
-    x = rand(1)
-    y = rand(1)
-    SUITE["high-level"]["permutedims(Dense(Dense()))"] = @benchmarkable(permutedims($A, (2, 1)))
-end
+        let
+            A = Tensor(Dense(Dense(Element(0.0))), rand(10000, 10000))
+            SUITE["high-level"]["permutedims(Dense(Dense()))"][scheduler_name] = @benchmarkable(permutedims($A, (2, 1)))
+        end
 
-let
-    k = Ref(0.0)
-    x = rand(1)
-    y = rand(1)
-    SUITE["high-level"]["einsum_spmv_compile_overhead"] = @benchmarkable(
-        begin
-            A, x, y = (A, $x, $y)
-            @einsum y[i] += A[i, j] * x[j]
-        end,
-        setup = (A = Tensor(Dense(SparseList(Element($k[] += 1))), fsprand(1, 1, 1)))
-    )
-end
+        let
+            k = Ref(0.0)
+            x = rand(1)
+            y = rand(1)
+            SUITE["high-level"]["einsum_spmv_compile_overhead"][scheduler_name] = @benchmarkable(
+                begin
+                    A, x, y = (A, $x, $y)
+                    @einsum y[i] += A[i, j] * x[j]
+                end,
+                setup = (A = Tensor(Dense(SparseList(Element($k[] += 1))), fsprand(1, 1, 1)))
+            )
+        end
 
-let
-    A = Tensor(Dense(SparseList(Element(0.0))), fsprand(1, 1, 1))
-    x = rand(1)
-    SUITE["high-level"]["einsum_spmv_call_overhead"] = @benchmarkable(
-        begin
-            A, x = ($A, $x)
-            @einsum y[i] += A[i, j] * x[j]
-        end,
-    )
-end
+        let
+            N = 10
+            P = 0.0001
+            C = 16.0
+            SUITE["high-level"]["einsum_matmul_adaptive_overhead"][scheduler_name] = @benchmarkable(
+                begin
+                    @einsum C[i, j] += A[i, k] * B[k, j]
+                end,
+                setup = begin
+                    (N, P, C) = ($N, $P, $C)
+                    n = floor(Int, N * C^(rand()))
+                    m = floor(Int, N * C^(rand()))
+                    l = floor(Int, N * C^(rand()))
+                    p = floor(Int, P * C^(rand()))
+                    q = floor(Int, P * C^(rand()))
+                    A = fsprand(n, l, p)
+                    B = fsprand(l, m, q)
+                end,
+                evals = 1
+            )
+        end
 
-let
-    N = 1_000
-    K = 1_000
-    p = 0.001
-    A = Tensor(Dense(Dense(Element(0.0))), rand(N, K))
-    B = Tensor(Dense(Dense(Element(0.0))), rand(K, N))
-    M = Tensor(Dense(SparseList(Element(0.0))), fsprand(N, N, p))
-
-    SUITE["high-level"]["sddmm_fused"] = @benchmarkable(
-        begin
-            M = lazy($M)
-            A = lazy($A)
-            B = lazy($B)
-            compute(M .* (A * B))
-        end,
-    )
+        let
+            A = Tensor(Dense(SparseList(Element(0.0))), fsprand(1, 1, 1))
+            x = rand(1)
+            SUITE["high-level"]["einsum_spmv_call_overhead"][scheduler_name] = @benchmarkable(
+                begin
+                    A, x = ($A, $x)
+                    @einsum y[i] += A[i, j] * x[j]
+                end,
+            )
+        end
 
-    SUITE["high-level"]["sddmm_unfused"] = @benchmarkable(
-        begin
-            M = $M
-            A = $A
-            B = $B
-            M .* (A * B)
-        end,
-    )
+        let
+            N = 1_000
+            K = 1_000
+            p = 0.001
+            A = Tensor(Dense(Dense(Element(0.0))), rand(N, K))
+            B = Tensor(Dense(Dense(Element(0.0))), rand(K, N))
+            M = Tensor(Dense(SparseList(Element(0.0))), fsprand(N, N, p))
+
+            SUITE["high-level"]["sddmm_fused"][scheduler_name] = @benchmarkable(
+                begin
+                    M = lazy($M)
+                    A = lazy($A)
+                    B = lazy($B)
+                    compute(M .* (A * B))
+                end,
+            )
+
+            SUITE["high-level"]["sddmm_unfused"][scheduler_name] = @benchmarkable(
+                begin
+                    M = $M
+                    A = $A
+                    B = $B
+                    M .* (A * B)
+                end,
+            )
+        end
+    end
 end
 
+
 eval(let
     A = Tensor(Dense(SparseList(Element(0.0))), fsprand(1, 1, 1))
     x = rand(1)
@@ -348,4 +395,13 @@ x = rand(N)
 
 SUITE["structure"]["banded"]["SparseList"] = @benchmarkable spmv_serial($A_ref, $x)
 SUITE["structure"]["banded"]["SparseBand"] = @benchmarkable spmv_serial($A, $x)
-SUITE["structure"]["banded"]["SparseInterval"] = @benchmarkable spmv_serial($A2, $x)
\ No newline at end of file
+SUITE["structure"]["banded"]["SparseInterval"] = @benchmarkable spmv_serial($A2, $x)
+
+if !isempty(parsed_args["include"])
+    inc = reduce((a, b) -> :($a || $b), parsed_args["include"])
+    SUITE = eval(:(SUITE[@tagged $inc]))
+end
+if !isempty(parsed_args["exclude"])
+    exc = reduce((a, b) -> :($a || $b), parsed_args["exclude"])
+    SUITE = eval(:(SUITE[@tagged !$exc]))
+end
\ No newline at end of file
diff --git a/benchmark/runbenchmarks.jl b/benchmark/runbenchmarks.jl
index 5741f752c..a1a50f30e 100755
--- a/benchmark/runbenchmarks.jl
+++ b/benchmark/runbenchmarks.jl
@@ -11,7 +11,7 @@ end
 using PkgBenchmark
 benchmarkpkg(
     dirname(@__DIR__),
-    BenchmarkConfig(env = Dict("JULIA_NUM_THREADS" => "8")),
+    BenchmarkConfig(env = Dict("JULIA_NUM_THREADS" => "8", "FINCH_BENCHMARK_ARGS" => get(ENV, "FINCH_BENCHMARK_ARGS", join(ARGS, " ")))),
     resultfile = joinpath(@__DIR__, "result.json"),
 )
 
diff --git a/benchmark/runjudge.jl b/benchmark/runjudge.jl
index 273a304cc..e306adb85 100755
--- a/benchmark/runjudge.jl
+++ b/benchmark/runjudge.jl
@@ -11,19 +11,25 @@ end
 using PkgBenchmark
 
 function mkconfig(; kwargs...)
-    return BenchmarkConfig(env = Dict("JULIA_NUM_THREADS" => "1"); kwargs...)
+    return BenchmarkConfig(env = Dict("JULIA_NUM_THREADS" => "1", "FINCH_BENCHMARK_ARGS" => get(ENV, "FINCH_BENCHMARK_ARGS", join(ARGS, " "))); kwargs...)
 end
 
+script = tempname(joinpath(@__DIR__))
+
+cp(joinpath(@__DIR__, "benchmarks.jl"), script)
+
 group_target = benchmarkpkg(
     dirname(@__DIR__),
     mkconfig(),
     resultfile = joinpath(@__DIR__, "result-target.json"),
+    script = script,
 )
 
 group_baseline = benchmarkpkg(
     dirname(@__DIR__),
     mkconfig(id = "main"),
     resultfile = joinpath(@__DIR__, "result-baseline.json"),
+    script = script,
 )
 
 judgement = judge(group_target, group_baseline)
diff --git a/test/runtests.jl b/test/runtests.jl
index b3e0205c5..66f481012 100755
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -9,21 +9,29 @@ end
 using Test
 using ArgParse
 
-s = ArgParseSettings("Run Finch.jl tests. All tests are run by default. Specific
-test suites may be specified as positional arguments. Finch compares to
-reference output which depends on the system word size (currently
-$(Sys.WORD_SIZE)-bit). To overwrite $(Sys.WORD_SIZE==32 ? 64 : 32)-bit output,
-run this with a $(Sys.WORD_SIZE==32 ? 64 : 32)-bit julia executable.")
+s = ArgParseSettings("Run Finch.jl tests. By default, all tests are run unless --include or --exclude options are provided. 
+Finch compares to reference output which depends on the system word size (currently $(Sys.WORD_SIZE)-bit).To overwrite $(Sys.WORD_SIZE==32 ? 64 : 32)-bit output, run this with a $(Sys.WORD_SIZE==32 ? 64 : 32)-bit julia executable.
+If the environment variable FINCH_TEST_ARGS is set, it will override the given arguments.")
 
 @add_arg_table! s begin
     "--overwrite", "-w"
     action = :store_true
     help = "overwrite reference output for $(Sys.WORD_SIZE)-bit systems"
-    "suites"
+    "--include", "-i"
     nargs = '*'
-    default = ["all"]
-    help = "names of test suites to run, from: print, constructors, representation, merges, index, typical, kernels, issues, interface, galley, continuous, continuousexamples, simple, examples, fileio, docs, parallel, algebra."
+    default = []
+    help = "list of test suites to include, e.g., --include constructors merges"
+
+    "--exclude", "-e"
+    nargs = '*'
+    default = []
+    help = "list of test suites to exclude, e.g., --exclude parallel algebra"
 end
+
+if "FINCH_TEST_ARGS" in keys(ENV)
+    ARGS = split(ENV["FINCH_TEST_ARGS"], " ")
+end
+
 parsed_args = parse_args(ARGS, s)
 
 """
@@ -61,7 +69,9 @@ end
 
 function should_run(name)
     global parsed_args
-    return ("all" in parsed_args["suites"] || name in parsed_args["suites"])
+    inc = parsed_args["include"]
+    exc = parsed_args["exclude"]
+    return (isempty(inc) || name in inc) && !(name in exc)
 end
 
 macro repl(io, ex, quiet = false)
@@ -109,7 +119,6 @@ include("utils.jl")
         end
     end
     if should_run("parallel") include("test_parallel.jl") end
-    #if should_run("continuous") include("test_continuous.jl") end
     #algebra goes at the end since it calls refresh()
     if should_run("algebra") include("test_algebra.jl") end
 end