diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -41,8 +41,8 @@
 
 `tasty-bench` is a native Haskell library and works everywhere, where GHC
 does. We support a full range of architectures (`i386`, `amd64`, `armhf`,
-`arm64`, `ppc64le`, `s390x`) and operating systems (Linux, Windows, MacOS,
-FreeBSD), plus any GHC from 7.0 to 9.4.
+`arm64`, `ppc64le`, `s390x`) and operating systems (Linux, Windows, macOS,
+FreeBSD, OpenBSD, NetBSD), plus any GHC from 7.0 to 9.6.
 
 ## How is it possible?
 
@@ -93,6 +93,8 @@
   type:          exitcode-stdio-1.0
   build-depends: base, tasty-bench
   ghc-options:   "-with-rtsopts=-A32m"
+  if impl(ghc >= 8.6)
+    ghc-options: -fproc-alignment=64
 ```
 
 And here is `BenchFibo.hs`:
@@ -387,27 +389,24 @@
 
 Finally, as an ultimate measure to reduce interference between benchmarks,
 one can run each of them in a separate process. We do not quite recommend
-this approach, but if you are desperate, here is how.
-
-Assuming that a benchmark is declared in `cabal` file as `benchmark my-bench` component,
-let's first find its executable:
+this approach, but if you are desperate, here is how:
 
 ```sh
-cabal build --enable-benchmarks my-bench
-MYBENCH=$(cabal list-bin my-bench) # available since cabal-3.4
+cabal run -v0 all:benches -- -l | sed -e 's/[\"]/\\\\\\&/g' | while read -r name; do cabal run -v0 all:benches -- -p '$0 == "'"$name"'"'; done
 ```
 
-Now list all benchmark names (hopefully, they do not contain newlines),
-escape quotes and slashes, and run each of them separately:
-
-```sh
-$MYBENCH -l | sed -e 's/[\"]/\\\\\\&/g' | while read -r name; do $MYBENCH -p '$0 == "'"$name"'"'; done
-```
+This assumes that there is a single benchmark suite in the project
+and that benchmark names do not contain newlines.
 
 ## Comparison against baseline
 
-One can compare benchmark results against an earlier baseline in an automatic way.
-To use this feature, first run `tasty-bench` with `--csv FILE` key
+One can compare benchmark results against an earlier run in an automatic way.
+
+When using this feature, it's especially important to compile benchmarks with
+`ghc-options: `[`-fproc-alignment`](https://downloads.haskell.org/ghc/latest/docs/users_guide/debugging.html#ghc-flag--fproc-alignment)`=64`, otherwise results could be skewed by
+intermittent changes in cache-line alignment.
+
+Firstly, run `tasty-bench` with `--csv FILE` key
 to dump results to `FILE` in CSV format
 (it could be a good idea to set smaller `--stdev`, if possible):
 
diff --git a/changelog.md b/changelog.md
--- a/changelog.md
+++ b/changelog.md
@@ -1,3 +1,8 @@
+# 0.3.3
+
+* Drop support of `tasty < 1.2.3`.
+* Make benchmarks immune to `-fspec-constr-count` limit.
+
 # 0.3.2
 
 * Add `locateBenchmark` and `mapLeafBenchmarks`.
diff --git a/compare_benches.sh b/compare_benches.sh
--- a/compare_benches.sh
+++ b/compare_benches.sh
@@ -4,17 +4,25 @@
     printf "Usage:\n  compare_benches oldCommit newCommit ...\nwhere ... is passed to benchmarks directly.\n"
     return 0
   fi
+
   OLD="$1"
   shift
   NEW="$1"
   shift
+
+  HEADREF=$(git rev-parse --verify HEAD)
+  OLDREF=$(git rev-parse --verify "$OLD")
+  NEWREF=$(git rev-parse --verify "$NEW")
+
   OLDCSV=$(echo "$OLD".csv | sed -e s#/##g)
   NEWCSV=$(echo "$NEW".csv | sed -e s#/##g)
   OLDVSNEWCSV=$(echo "$OLD"-vs-"$NEW".csv | sed -e s#/##g)
-  git checkout -q "$OLD" && \
+
+  git checkout -q "$OLDREF" && \
   cabal run -v0 benchmarks -- --csv "$OLDCSV" "$@" && \
-  git checkout -q "$NEW" && \
+  git checkout -q "$NEWREF" && \
   cabal run -v0 benchmarks -- --baseline "$OLDCSV" --csv "$NEWCSV" "$@" && \
-  git checkout -q "@{-2}" && \
+  git checkout -q "$HEADREF" && \
+
   awk 'BEGIN{FS=",";OFS=",";print "Name,'"$OLD"','"$NEW"',Ratio"}FNR==1{trueNF=NF;next}NF<trueNF{print "Benchmark names should not contain newlines";exit 1}FNR==NR{oldTime=$(NF-trueNF+2);NF-=trueNF-1;a[$0]=oldTime;next}{newTime=$(NF-trueNF+2);NF-=trueNF-1;print $0,a[$0],newTime,newTime/a[$0];gs+=log(newTime/a[$0]);gc++}END{if(gc>0)print "Geometric mean,,",exp(gs/gc)}' "$OLDCSV" "$NEWCSV" > "$OLDVSNEWCSV"
 }
diff --git a/src/Test/Tasty/Bench.hs b/src/Test/Tasty/Bench.hs
--- a/src/Test/Tasty/Bench.hs
+++ b/src/Test/Tasty/Bench.hs
@@ -24,8 +24,8 @@
 
 @tasty-bench@ is a native Haskell library and works everywhere, where GHC
 does. We support a full range of architectures (@i386@, @amd64@, @armhf@,
-@arm64@, @ppc64le@, @s390x@) and operating systems (Linux, Windows, MacOS,
-FreeBSD), plus any GHC from 7.0 to 9.4.
+@arm64@, @ppc64le@, @s390x@) and operating systems (Linux, Windows, macOS,
+FreeBSD, OpenBSD, NetBSD), plus any GHC from 7.0 to 9.6.
 
 === How is it possible?
 
@@ -75,6 +75,8 @@
 >   type:          exitcode-stdio-1.0
 >   build-depends: base, tasty-bench
 >   ghc-options:   "-with-rtsopts=-A32m"
+>   if impl(ghc >= 8.6)
+>     ghc-options: -fproc-alignment=64
 
 And here is @BenchFibo.hs@:
 
@@ -351,23 +353,22 @@
 
 Finally, as an ultimate measure to reduce interference between
 benchmarks, one can run each of them in a separate process. We do not
-quite recommend this approach, but if you are desperate, here is how.
+quite recommend this approach, but if you are desperate, here is how:
 
-Assuming that a benchmark is declared in @cabal@ file as
-@benchmark@ @my-bench@ component, let’s first find its executable:
+> cabal run -v0 all:benches -- -l | sed -e 's/[\"]/\\\\\\&/g' | while read -r name; do cabal run -v0 all:benches -- -p '$0 == "'"$name"'"'; done
 
-> cabal build --enable-benchmarks my-bench
-> MYBENCH=$(cabal list-bin my-bench) # available since cabal-3.4
+This assumes that there is a single benchmark suite in the project
+and that benchmark names do not contain newlines.
 
-Now list all benchmark names (hopefully, they do not contain newlines),
-escape quotes and slashes, and run each of them separately:
+=== Comparison against baseline
 
-> $MYBENCH -l | sed -e 's/[\"]/\\\\\\&/g' | while read -r name; do $MYBENCH -p '$0 == "'"$name"'"'; done
+One can compare benchmark results against an earlier run in an automatic way.
 
-=== Comparison against baseline
+When using this feature, it's especially important to compile benchmarks with
+@ghc-options:@ [@-fproc-alignment@](https://downloads.haskell.org/ghc/latest/docs/users_guide/debugging.html#ghc-flag--fproc-alignment)@=64@, otherwise results could be skewed by
+intermittent changes in cache-line alignment.
 
-One can compare benchmark results against an earlier baseline in an
-automatic way. To use this feature, first run @tasty-bench@ with
+Firstly, run @tasty-bench@ with
 @--csv@ @FILE@ key to dump results to @FILE@ in CSV format
 (it could be a good idea to set smaller @--stdev@, if possible):
 
@@ -570,6 +571,7 @@
 
 -}
 
+{-# LANGUAGE BangPatterns #-}
 {-# LANGUAGE CPP #-}
 {-# LANGUAGE DeriveDataTypeable #-}
 {-# LANGUAGE DeriveFunctor #-}
@@ -586,10 +588,8 @@
   , Benchmark
   , bench
   , bgroup
-#if MIN_VERSION_tasty(1,2,0)
   , bcompare
   , bcompareWithin
-#endif
   , env
   , envWithCleanup
   ,
@@ -617,9 +617,7 @@
   , SvgPath(..)
   , TimeMode(..)
   -- * Utils
-#if MIN_VERSION_tasty(1,0,0)
   , locateBenchmark
-#endif
   , mapLeafBenchmarks
 #else
   , Timeout(..)
@@ -683,11 +681,9 @@
 import Test.Tasty.Ingredients
 import Test.Tasty.Ingredients.ConsoleReporter
 import Test.Tasty.Options
-#if MIN_VERSION_tasty(1,0,0)
 import Test.Tasty.Patterns.Eval (eval, asB, withFields)
 import Test.Tasty.Patterns.Types (Expr (And, Field, IntLit, NF, StringLit, Sub))
 import qualified Test.Tasty.Patterns.Types as Patterns
-#endif
 import Test.Tasty.Providers
 import Test.Tasty.Runners
 #endif
@@ -696,6 +692,15 @@
 import Data.Word (Word32)
 #endif
 
+#if MIN_VERSION_ghc_prim(0,3,1)
+import GHC.Types (SPEC(..))
+#else
+import GHC.Exts (SpecConstrAnnotation(..))
+
+data SPEC = SPEC | SPEC2
+{-# ANN type SPEC ForceSpecConstr #-}
+#endif
+
 #ifndef MIN_VERSION_tasty
 data Timeout
   = Timeout
@@ -740,16 +745,13 @@
 -- > localOption WallTime (bgroup [...])
 --
 -- section of your cabal file.
+--
 -- @since 0.3.2
 data TimeMode = CpuTime
   -- ^ Measure CPU time.
 #ifdef MIN_VERSION_tasty
-#if MIN_VERSION_tasty(1,2,2)
   | WallTime
-  -- ^ Measure wall-clock time. This requires @tasty-1.2.2@, so if you use 'WallTime'
-  -- it is prudent to add @tasty >= 1.2.2@ to @build-depends@
-  -- section of your cabal file.
-#endif
+  -- ^ Measure wall-clock time.
 #endif
   deriving (Typeable)
 
@@ -812,20 +814,16 @@
   defaultValue = CpuTime
   parseValue v = case v of
     "cpu" -> Just CpuTime
-#if MIN_VERSION_tasty(1,2,2)
     "wall" -> Just WallTime
-#endif
     _ -> Nothing
   optionName = pure "time-mode"
   optionHelp = pure "Whether to measure CPU time (\"cpu\") or wall-clock time (\"wall\")"
 #if MIN_VERSION_tasty(1,3,0)
   showDefaultValue m = Just $ case m of
     CpuTime -> "cpu"
-#if MIN_VERSION_tasty(1,2,2)
     WallTime -> "wall"
 #endif
 #endif
-#endif
 
 -- | Something that can be benchmarked, produced by 'nf', 'whnf', 'nfIO', 'whnfIO',
 -- 'nfAppIO', 'whnfAppIO' below.
@@ -1005,10 +1003,8 @@
 getTimePicoSecs timeMode = case timeMode of
   CpuTime -> fromInteger <$> getCPUTime
 #ifdef MIN_VERSION_tasty
-#if MIN_VERSION_tasty(1,2,2)
   WallTime -> round . (1e12 *) <$> getTime
 #endif
-#endif
 
 measure :: TimeMode -> Word64 -> Benchmarkable -> IO Measurement
 measure timeMode n (Benchmarkable act) = do
@@ -1122,14 +1118,11 @@
 bgroup :: String -> [Benchmark] -> Benchmark
 bgroup = testGroup
 
-#if MIN_VERSION_tasty(1,2,0)
 -- | Compare benchmarks, reporting relative speed up or slow down.
 --
 -- This function is a vague reminiscence of @bcompare@, which existed in pre-1.0
 -- versions of @criterion@, but their types are incompatible. Under the hood
--- 'bcompare' is a thin wrapper over 'after' and requires @tasty-1.2@.
--- If you use 'bcompare', it is prudent to add @tasty >= 1.2@ to @build-depends@
--- section of your cabal file.
+-- 'bcompare' is a thin wrapper over 'after'.
 --
 -- Here is a basic example:
 --
@@ -1187,7 +1180,6 @@
 
 bcomparePrefix :: String
 bcomparePrefix = "tasty-bench"
-#endif
 
 -- | Benchmarks are actually just a regular 'Test.Tasty.TestTree' in disguise.
 --
@@ -1223,8 +1215,8 @@
 
 #endif
 
-funcToBench :: (b -> c) -> (a -> b) -> a -> Benchmarkable
-funcToBench frc = (Benchmarkable .) . benchLoop
+funcToBench :: forall a b c. (b -> c) -> (a -> b) -> a -> Benchmarkable
+funcToBench frc = (Benchmarkable .) . benchLoop SPEC
   where
     -- Here we rely on the fact that GHC (unless spurred by
     -- -fstatic-argument-transformation) is not smart enough:
@@ -1241,15 +1233,20 @@
     --
     -- This function is called `benchLoop` instead of, say, `go`,
     -- so it is easier to spot in Core dumps.
-    benchLoop f x n
+    --
+    -- Forcing SpecConst optimization with SPEC makes the behaviour of benchmarks
+    -- independent of -fspec-constr-count.
+    benchLoop :: SPEC -> (a -> b) -> a -> Word64 -> IO ()
+    benchLoop !_ f x n
       | n == 0    = pure ()
       | otherwise = do
         _ <- evaluate (frc (f x))
-        benchLoop f x (n - 1)
+        benchLoop SPEC f x (n - 1)
 {-# INLINE funcToBench #-}
 
 -- | 'nf' @f@ @x@ measures time to compute
--- a normal form (by means of 'force') of an application of @f@ to @x@.
+-- a normal form (by means of 'force', not 'Control.DeepSeq.rnf')
+-- of an application of @f@ to @x@.
 -- This does not include time to evaluate @f@ or @x@ themselves.
 -- Ideally @x@ should be a primitive data type like 'Data.Int.Int'.
 --
@@ -1276,11 +1273,16 @@
 -- such measurements are noisy and involve an overhead. Results are more reliable
 -- when @f@ @x@ takes at least several milliseconds.
 --
--- Note that forcing a normal form requires an additional
+-- Remember that forcing a normal form requires an additional
 -- traverse of the structure. In certain scenarios (imagine benchmarking 'tail'),
 -- especially when 'NFData' instance is badly written,
 -- this traversal may take non-negligible time and affect results.
 --
+-- 'nf' @f@ is equivalent to 'whnf' ('force' '.' @f@), but not to
+-- 'whnf' ('Control.DeepSeq.rnf' '.' @f@). The former retains the result
+-- in memory until it is fully evaluated, while the latter allows
+-- evaluated parts of the result to be garbage-collected immediately.
+--
 -- Drop-in replacement for @Criterion.@'Criterion.nf' and
 -- @Gauge.@'Gauge.nf'.
 --
@@ -1331,14 +1333,14 @@
 {-# INLINE ioToBench #-}
 
 -- | 'nfIO' @x@ measures time to evaluate side-effects of @x@
--- and compute its normal form (by means of 'force').
+-- and compute its normal form (by means of 'force', not 'Control.DeepSeq.rnf').
 --
 -- Pure subexpression of an effectful computation @x@
 -- may be evaluated only once and get cached.
 -- To avoid surprising results it is usually preferable
 -- to use 'nfAppIO' instead.
 --
--- Note that forcing a normal form requires an additional
+-- Remember that forcing a normal form requires an additional
 -- traverse of the structure. In certain scenarios,
 -- especially when 'NFData' instance is badly written,
 -- this traversal may take non-negligible time and affect results.
@@ -1392,8 +1394,8 @@
 {-# INLINE ioFuncToBench #-}
 
 -- | 'nfAppIO' @f@ @x@ measures time to evaluate side-effects of
--- an application of @f@ to @x@.
--- and compute its normal form (by means of 'force').
+-- an application of @f@ to @x@
+-- and compute its normal form (by means of 'force', not 'Control.DeepSeq.rnf').
 -- This does not include time to evaluate @f@ or @x@ themselves.
 -- Ideally @x@ should be a primitive data type like 'Data.Int.Int'.
 --
@@ -1404,7 +1406,7 @@
 -- but if @x@ is not a primitive data type, consider forcing its evaluation
 -- separately, e. g., via 'env' or 'withResource'.
 --
--- Note that forcing a normal form requires an additional
+-- Remember that forcing a normal form requires an additional
 -- traverse of the structure. In certain scenarios,
 -- especially when 'NFData' instance is badly written,
 -- this traversal may take non-negligible time and affect results.
@@ -1422,7 +1424,7 @@
 {-# INLINE nfAppIO #-}
 
 -- | 'whnfAppIO' @f@ @x@ measures time to evaluate side-effects of
--- an application of @f@ to @x@.
+-- an application of @f@ to @x@
 -- and compute its weak head normal form.
 -- This does not include time to evaluate @f@ or @x@ themselves.
 -- Ideally @x@ should be a primitive data type like 'Data.Int.Int'.
@@ -1902,12 +1904,9 @@
   , foldAfter  = const foldDeps
 #else
   , foldGroup  = map . first . (++) . (++ ".")
-#if MIN_VERSION_tasty(1,2,0)
   , foldAfter  = foldDeps
 #endif
-#endif
   }
-#if MIN_VERSION_tasty(1,2,0)
   where
     foldDeps :: DependencyType -> Expr -> [(a, Unique (WithLoHi IM.Key))] -> [(a, Unique (WithLoHi IM.Key))]
     foldDeps AllSucceed (And (StringLit xs) p)
@@ -1921,7 +1920,6 @@
   foldMap (\(k, v) -> if withFields v pat == Right True then Unique k else mempty) $ IM.assocs im
   where
     pat = eval pattern >>= asB
-#endif
 
 postprocessResult
     :: (TestName -> Maybe (WithLoHi Result) -> Result -> Result)
@@ -1998,8 +1996,11 @@
 -- This helper is useful for bulk application of 'bcompare'.
 -- See also 'locateBenchmark'.
 --
--- Real world example: https://hackage.haskell.org/package/text-builder-linear-0.1/src/bench/Main.hs
+-- Real world examples:
 --
+-- * https://hackage.haskell.org/package/chimera-0.3.3.0/src/bench/Bench.hs
+-- * https://hackage.haskell.org/package/text-builder-linear-0.1/src/bench/Main.hs
+--
 -- @since 0.3.2
 mapLeafBenchmarks :: ([String] -> Benchmark -> Benchmark) -> Benchmark -> Benchmark
 mapLeafBenchmarks processLeaf = go mempty
@@ -2011,11 +2012,8 @@
       PlusTestOptions g tt -> PlusTestOptions g (go path tt)
       WithResource res f   -> WithResource res (go path . f)
       AskOptions f         -> AskOptions (go path . f)
-#if MIN_VERSION_tasty(1,2,0)
       After dep expr tt    -> After dep expr (go path tt)
-#endif
 
-#if MIN_VERSION_tasty(1,0,0)
 -- | Construct an AWK expression to locate an individual element or elements in 'Benchmark'
 -- by the suffix of the path. Names are listed in reverse order:
 -- from 'bench'\'s own name to a name of the outermost 'bgroup'.
@@ -2024,11 +2022,10 @@
 -- 'bcompare' ('Test.Tasty.Patterns.Printer.printAwkExpr' ('locateBenchmark' @path@)).
 -- See also 'mapLeafBenchmarks'.
 --
--- This function requires @tasty-1.0@, so if you use 'locateBenchmark'
--- it is prudent to add @tasty >= 1.0@ to @build-depends@
--- section of your cabal file.
+-- Real world examples:
 --
--- Real world example: https://hackage.haskell.org/package/text-builder-linear-0.1/src/bench/Main.hs
+-- * https://hackage.haskell.org/package/chimera-0.3.3.0/src/bench/Bench.hs
+-- * https://hackage.haskell.org/package/text-builder-linear-0.1/src/bench/Main.hs
 --
 -- @since 0.3.2
 locateBenchmark :: [String] -> Expr
@@ -2037,5 +2034,4 @@
   = foldl1' And
   $ zipWith (\i name -> Patterns.EQ (Field (Sub NF (IntLit i))) (StringLit name)) [0..] path
 
-#endif
 #endif
diff --git a/tasty-bench.cabal b/tasty-bench.cabal
--- a/tasty-bench.cabal
+++ b/tasty-bench.cabal
@@ -1,5 +1,5 @@
 name:          tasty-bench
-version:       0.3.2
+version:       0.3.3
 cabal-version: 1.18
 build-type:    Simple
 license:       MIT
@@ -25,7 +25,7 @@
 extra-doc-files:
   example.svg
 
-tested-with: GHC == 9.4.1, GHC == 9.2.4, GHC==9.0.2, GHC==8.10.7, GHC==8.8.4, GHC==8.6.5, GHC==8.4.4, GHC==8.2.2, GHC==8.0.2, GHC==7.10.3, GHC==7.8.4, GHC==7.6.3, GHC==7.4.2, GHC==7.2.2, GHC==7.0.4
+tested-with: GHC == 9.6.1 GHC == 9.4.4, GHC == 9.2.5, GHC==9.0.2, GHC==8.10.7, GHC==8.8.4, GHC==8.6.5, GHC==8.4.4, GHC==8.2.2, GHC==8.0.2, GHC==7.10.3, GHC==7.8.4, GHC==7.6.3, GHC==7.4.2, GHC==7.2.2, GHC==7.0.4
 
 source-repository head
   type: git
@@ -56,11 +56,12 @@
 
   build-depends:
     base >= 4.3 && < 5,
-    deepseq >= 1.1
+    deepseq >= 1.1,
+    ghc-prim
   if flag(tasty)
     build-depends:
       containers >= 0.4,
-      tasty >= 0.11.3
+      tasty >= 1.2.3
   if impl(ghc < 7.8)
     build-depends:
       tagged >= 0.2