diff --git a/Manifest.toml b/Manifest.toml
index 6c5e435e02..65ef0b2d31 100644
--- a/Manifest.toml
+++ b/Manifest.toml
@@ -1,19 +1,32 @@
 # This file is machine-generated - editing it directly is not advised
 
 [[AbstractFFTs]]
-deps = ["ChainRulesCore", "LinearAlgebra", "Test"]
+deps = ["LinearAlgebra"]
 git-tree-sha1 = "d92ad398961a3ed262d8bf04a1a2b8340f915fef"
 uuid = "621f4979-c628-5d54-868e-fcf4e3e8185c"
 version = "1.5.0"
 
+    [AbstractFFTs.extensions]
+    AbstractFFTsChainRulesCoreExt = "ChainRulesCore"
+    AbstractFFTsTestExt = "Test"
+
+    [AbstractFFTs.weakdeps]
+    ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+    Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
 [[Adapt]]
 deps = ["LinearAlgebra", "Requires"]
 git-tree-sha1 = "76289dc51920fdc6e0013c872ba9551d54961c24"
 uuid = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
 version = "3.6.2"
+weakdeps = ["StaticArrays"]
+
+    [Adapt.extensions]
+    AdaptStaticArraysExt = "StaticArrays"
 
 [[ArgTools]]
 uuid = "0dad84c5-d112-42e6-8d28-ef12dabb789f"
+version = "1.1.1"
 
 [[Artifacts]]
 uuid = "56f22d72-fd6d-98f1-02f0-08ddc0907c33"
@@ -56,27 +69,20 @@ git-tree-sha1 = "5248d9c45712e51e27ba9b30eebec65658c6ce29"
 uuid = "76a88914-d11a-5bdc-97e0-2f5a05c973a2"
 version = "0.6.0+0"
 
-[[ChainRulesCore]]
-deps = ["Compat", "LinearAlgebra", "SparseArrays"]
-git-tree-sha1 = "e30f2f4e20f7f186dc36529910beaedc60cfa644"
-uuid = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
-version = "1.16.0"
-
-[[ChangesOfVariables]]
-deps = ["InverseFunctions", "LinearAlgebra", "Test"]
-git-tree-sha1 = "2fba81a302a7be671aefe194f0525ef231104e7f"
-uuid = "9e997f8a-9a97-42d5-a9f1-ce6bfc15e2c0"
-version = "0.1.8"
-
 [[Compat]]
-deps = ["Dates", "LinearAlgebra", "UUIDs"]
+deps = ["UUIDs"]
 git-tree-sha1 = "e460f044ca8b99be31d35fe54fc33a5c33dd8ed7"
 uuid = "34da2185-b29b-5c13-b0c7-acf172513d20"
 version = "4.9.0"
+weakdeps = ["Dates", "LinearAlgebra"]
+
+    [Compat.extensions]
+    CompatLinearAlgebraExt = "LinearAlgebra"
 
 [[CompilerSupportLibraries_jll]]
 deps = ["Artifacts", "Libdl"]
 uuid = "e66e0078-7015-5450-92f7-15fbd957f2ae"
+version = "1.0.2+0"
 
 [[Crayons]]
 git-tree-sha1 = "249fe38abf76d48563e2f4556bebd215aa317e15"
@@ -116,14 +122,25 @@ uuid = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 version = "0.9.3"
 
 [[Downloads]]
-deps = ["ArgTools", "LibCURL", "NetworkOptions"]
+deps = ["ArgTools", "FileWatching", "LibCURL", "NetworkOptions"]
 uuid = "f43a241f-c20a-4ad4-852c-f6b1247861c6"
+version = "1.6.0"
 
 [[ExprTools]]
 git-tree-sha1 = "27415f162e6028e81c72b82ef756bf321213b6ec"
 uuid = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
 version = "0.1.10"
 
+[[FastmathOverlay]]
+git-tree-sha1 = "f4babf7d075349104366b0f68c9492d3c82ed021"
+repo-rev = "main"
+repo-url = "https://github.com/vchuravy/FastmathOverlay.jl"
+uuid = "32340bd4-0178-4ab8-8fce-9b94bc594446"
+version = "0.1.0"
+
+[[FileWatching]]
+uuid = "7b1f6079-737a-58dc-b8bc-7a2ca5c1b5ee"
+
 [[Future]]
 deps = ["Random"]
 uuid = "9fa8497b-333b-5362-9e8d-4d0656e87820"
@@ -142,9 +159,11 @@ version = "0.1.5"
 
 [[GPUCompiler]]
 deps = ["ExprTools", "InteractiveUtils", "LLVM", "Libdl", "Logging", "Scratch", "TimerOutputs", "UUIDs"]
-git-tree-sha1 = "72b2e3c2ba583d1a7aa35129e56cf92e07c083e3"
+git-tree-sha1 = "7b56a06e5b8ff4e4ecebe070cc90662497f16858"
+repo-rev = "vc/mtv"
+repo-url = "https://github.com/JuliaGPU/GPUCompiler.jl.git"
 uuid = "61eb1bfa-7361-4325-ad38-22787b887f55"
-version = "0.21.4"
+version = "0.22.0"
 
 [[InlineStrings]]
 deps = ["Parsers"]
@@ -156,12 +175,6 @@ version = "1.4.0"
 deps = ["Markdown"]
 uuid = "b77e0a4c-d291-57a0-90e8-8db25a27a240"
 
-[[InverseFunctions]]
-deps = ["Test"]
-git-tree-sha1 = "68772f49f54b479fa88ace904f6127f0a3bb2e46"
-uuid = "3587e190-3f89-42d0-90ee-14403ec27112"
-version = "0.1.12"
-
 [[InvertedIndices]]
 git-tree-sha1 = "0dc7b50b8d436461be01300fd8cd45aa0274b038"
 uuid = "41ab1584-1d38-5bbf-9106-f11c6c58b48f"
@@ -178,10 +191,10 @@ uuid = "82899510-4779-5014-852e-03e436cf321d"
 version = "1.0.0"
 
 [[JLLWrappers]]
-deps = ["Preferences"]
-git-tree-sha1 = "abc9885a7ca2052a736a600f7fa66209f96506e1"
+deps = ["Artifacts", "Preferences"]
+git-tree-sha1 = "a7e91ef94114d5bc8952bcaa8d6ff952cf709808"
 uuid = "692b3bcd-3c85-4b1f-b108-f13ce0eb3210"
-version = "1.4.1"
+version = "1.4.2"
 
 [[KernelAbstractions]]
 deps = ["Adapt", "Atomix", "InteractiveUtils", "LinearAlgebra", "MacroTools", "PrecompileTools", "Requires", "SparseArrays", "StaticArrays", "UUIDs", "UnsafeAtomics", "UnsafeAtomicsLLVM"]
@@ -189,6 +202,12 @@ git-tree-sha1 = "4c5875e4c228247e1c2b087669846941fb6e0118"
 uuid = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
 version = "0.9.8"
 
+    [KernelAbstractions.extensions]
+    EnzymeExt = "EnzymeCore"
+
+    [KernelAbstractions.weakdeps]
+    EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
+
 [[LLVM]]
 deps = ["CEnum", "LLVMExtra_jll", "Libdl", "Printf", "Unicode"]
 git-tree-sha1 = "8695a49bfe05a2dc0feeefd06b4ca6361a018729"
@@ -213,10 +232,12 @@ uuid = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
 [[LibCURL]]
 deps = ["LibCURL_jll", "MozillaCACerts_jll"]
 uuid = "b27032c2-a3e7-50c8-80cd-2d36dbcbfd21"
+version = "0.6.3"
 
 [[LibCURL_jll]]
 deps = ["Artifacts", "LibSSH2_jll", "Libdl", "MbedTLS_jll", "Zlib_jll", "nghttp2_jll"]
 uuid = "deac9b47-8bc7-5906-a0fe-35ac56dc84c0"
+version = "7.84.0+0"
 
 [[LibGit2]]
 deps = ["Base64", "NetworkOptions", "Printf", "SHA"]
@@ -225,20 +246,31 @@ uuid = "76f85450-5226-5b5a-8eaa-529ad045b433"
 [[LibSSH2_jll]]
 deps = ["Artifacts", "Libdl", "MbedTLS_jll"]
 uuid = "29816b5a-b9ab-546f-933c-edad1886dfa8"
+version = "1.10.2+0"
 
 [[Libdl]]
 uuid = "8f399da3-3557-5675-b5ff-fb832c97cbdb"
 
 [[LinearAlgebra]]
-deps = ["Libdl"]
+deps = ["Libdl", "OpenBLAS_jll", "libblastrampoline_jll"]
 uuid = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 
 [[LogExpFunctions]]
-deps = ["ChainRulesCore", "ChangesOfVariables", "DocStringExtensions", "InverseFunctions", "IrrationalConstants", "LinearAlgebra"]
+deps = ["DocStringExtensions", "IrrationalConstants", "LinearAlgebra"]
 git-tree-sha1 = "c3ce8e7420b3a6e071e0fe4745f5d4300e37b13f"
 uuid = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
 version = "0.3.24"
 
+    [LogExpFunctions.extensions]
+    LogExpFunctionsChainRulesCoreExt = "ChainRulesCore"
+    LogExpFunctionsChangesOfVariablesExt = "ChangesOfVariables"
+    LogExpFunctionsInverseFunctionsExt = "InverseFunctions"
+
+    [LogExpFunctions.weakdeps]
+    ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
+    ChangesOfVariables = "9e997f8a-9a97-42d5-a9f1-ce6bfc15e2c0"
+    InverseFunctions = "3587e190-3f89-42d0-90ee-14403ec27112"
+
 [[Logging]]
 uuid = "56ddb016-857b-54e1-b83d-db4d58db5568"
 
@@ -255,6 +287,7 @@ uuid = "d6f4376e-aef5-505a-96c1-9c027394607a"
 [[MbedTLS_jll]]
 deps = ["Artifacts", "Libdl"]
 uuid = "c8ffd9c3-330d-5841-b78e-0817d7145fa1"
+version = "2.28.2+0"
 
 [[Missings]]
 deps = ["DataAPI"]
@@ -264,13 +297,21 @@ version = "1.1.0"
 
 [[MozillaCACerts_jll]]
 uuid = "14a3606d-f60d-562e-9121-12d972cd8159"
+version = "2022.10.11"
 
 [[NetworkOptions]]
 uuid = "ca575930-c2e3-43a9-ace4-1e988b2c1908"
+version = "1.2.0"
+
+[[OpenBLAS_jll]]
+deps = ["Artifacts", "CompilerSupportLibraries_jll", "Libdl"]
+uuid = "4536629a-c528-5b80-bd46-f80d51c5b363"
+version = "0.3.21+4"
 
 [[OpenLibm_jll]]
 deps = ["Artifacts", "Libdl"]
 uuid = "05823500-19ac-5b8b-9628-191a04bc5112"
+version = "0.8.1+0"
 
 [[OpenSpecFun_jll]]
 deps = ["Artifacts", "CompilerSupportLibraries_jll", "JLLWrappers", "Libdl", "Pkg"]
@@ -290,8 +331,9 @@ uuid = "69de0a69-1ddd-5017-9359-2bf0b02dc9f0"
 version = "2.7.2"
 
 [[Pkg]]
-deps = ["Artifacts", "Dates", "Downloads", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
+deps = ["Artifacts", "Dates", "Downloads", "FileWatching", "LibGit2", "Libdl", "Logging", "Markdown", "Printf", "REPL", "Random", "SHA", "Serialization", "TOML", "Tar", "UUIDs", "p7zip_jll"]
 uuid = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+version = "1.9.0"
 
 [[PooledArrays]]
 deps = ["DataAPI", "Future"]
@@ -326,7 +368,7 @@ deps = ["InteractiveUtils", "Markdown", "Sockets", "Unicode"]
 uuid = "3fa0cd96-eef1-5676-8a61-b3b8758bbffb"
 
 [[Random]]
-deps = ["Serialization"]
+deps = ["SHA", "Serialization"]
 uuid = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 
 [[Random123]]
@@ -354,6 +396,7 @@ version = "1.3.0"
 
 [[SHA]]
 uuid = "ea8e919c-243c-51af-8825-aaa63cd721ce"
+version = "0.7.0"
 
 [[Scratch]]
 deps = ["Dates"]
@@ -380,20 +423,30 @@ uuid = "a2af1166-a08f-5f64-846c-94a0d3cef48c"
 version = "1.1.1"
 
 [[SparseArrays]]
-deps = ["LinearAlgebra", "Random"]
+deps = ["Libdl", "LinearAlgebra", "Random", "Serialization", "SuiteSparse_jll"]
 uuid = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 
 [[SpecialFunctions]]
-deps = ["ChainRulesCore", "IrrationalConstants", "LogExpFunctions", "OpenLibm_jll", "OpenSpecFun_jll"]
-git-tree-sha1 = "7beb031cf8145577fbccacd94b8a8f4ce78428d3"
+deps = ["IrrationalConstants", "LogExpFunctions", "OpenLibm_jll", "OpenSpecFun_jll"]
+git-tree-sha1 = "e2cfc4012a19088254b3950b85c3c1d8882d864d"
 uuid = "276daf66-3868-5448-9aa4-cd146d93841b"
-version = "2.3.0"
+version = "2.3.1"
+
+    [SpecialFunctions.extensions]
+    SpecialFunctionsChainRulesCoreExt = "ChainRulesCore"
+
+    [SpecialFunctions.weakdeps]
+    ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 
 [[StaticArrays]]
-deps = ["LinearAlgebra", "Random", "StaticArraysCore", "Statistics"]
+deps = ["LinearAlgebra", "Random", "StaticArraysCore"]
 git-tree-sha1 = "9cabadf6e7cd2349b6cf49f1915ad2028d65e881"
 uuid = "90137ffa-7385-5640-81b9-e52037218182"
 version = "1.6.2"
+weakdeps = ["Statistics"]
+
+    [StaticArrays.extensions]
+    StaticArraysStatisticsExt = "Statistics"
 
 [[StaticArraysCore]]
 git-tree-sha1 = "36b3d696ce6366023a0ea192b4cd442268995a0d"
@@ -403,15 +456,22 @@ version = "1.4.2"
 [[Statistics]]
 deps = ["LinearAlgebra", "SparseArrays"]
 uuid = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+version = "1.9.0"
 
 [[StringManipulation]]
 git-tree-sha1 = "46da2434b41f41ac3594ee9816ce5541c6096123"
 uuid = "892a3eda-7b42-436c-8928-eab12a02cf0e"
 version = "0.3.0"
 
+[[SuiteSparse_jll]]
+deps = ["Artifacts", "Libdl", "Pkg", "libblastrampoline_jll"]
+uuid = "bea87d4a-7f5b-5778-9afe-8cc45184846c"
+version = "5.10.1+6"
+
 [[TOML]]
 deps = ["Dates"]
 uuid = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+version = "1.0.3"
 
 [[TableTraits]]
 deps = ["IteratorInterfaceExtensions"]
@@ -428,6 +488,7 @@ version = "1.10.1"
 [[Tar]]
 deps = ["ArgTools", "SHA"]
 uuid = "a4e569a6-e804-4fa4-b0f3-eef7a1d5b13e"
+version = "1.10.0"
 
 [[Test]]
 deps = ["InteractiveUtils", "Logging", "Random", "Serialization"]
@@ -460,11 +521,19 @@ version = "0.1.3"
 [[Zlib_jll]]
 deps = ["Libdl"]
 uuid = "83775a58-1f1d-513f-b197-d71354ab007a"
+version = "1.2.13+0"
+
+[[libblastrampoline_jll]]
+deps = ["Artifacts", "Libdl"]
+uuid = "8e850b90-86db-534c-a0d3-1478176c7d93"
+version = "5.8.0+0"
 
 [[nghttp2_jll]]
 deps = ["Artifacts", "Libdl"]
 uuid = "8e850ede-7688-5339-a07c-302acd2aaf8d"
+version = "1.48.0+0"
 
 [[p7zip_jll]]
 deps = ["Artifacts", "Libdl"]
 uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"
+version = "17.4.0+0"
diff --git a/Project.toml b/Project.toml
index 1c9fea40d9..4335ffce2c 100644
--- a/Project.toml
+++ b/Project.toml
@@ -13,6 +13,7 @@ CUDA_Runtime_jll = "76a88914-d11a-5bdc-97e0-2f5a05c973a2"
 Crayons = "a8cc5b0e-0ffa-5ad4-8c14-923d3ee1735f"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
 ExprTools = "e2ba6199-217a-4e67-a87a-7c52f15ade04"
+FastmathOverlay = "32340bd4-0178-4ab8-8fce-9b94bc594446"
 GPUArrays = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
 GPUCompiler = "61eb1bfa-7361-4325-ad38-22787b887f55"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
@@ -46,7 +47,7 @@ Crayons = "4"
 DataFrames = "1"
 ExprTools = "0.1"
 GPUArrays = "8.6"
-GPUCompiler = "0.21"
+GPUCompiler = "0.22"
 KernelAbstractions = "0.9.2"
 LLVM = "6"
 Preferences = "1"
diff --git a/src/compiler/compilation.jl b/src/compiler/compilation.jl
index c46f031bcd..284f63ca7a 100644
--- a/src/compiler/compilation.jl
+++ b/src/compiler/compilation.jl
@@ -1,6 +1,8 @@
 ## gpucompiler interface implementation
 
-struct CUDACompilerParams <: AbstractCompilerParams end
+Base.@kwdef struct CUDACompilerParams <: AbstractCompilerParams
+    contract::Bool=false
+end
 const CUDACompilerConfig = CompilerConfig{PTXCompilerTarget, CUDACompilerParams}
 const CUDACompilerJob = CompilerJob{PTXCompilerTarget,CUDACompilerParams}
 
@@ -38,7 +40,19 @@ function GPUCompiler.link_libraries!(@nospecialize(job::CUDACompilerJob), mod::L
     return
 end
 
-GPUCompiler.method_table(@nospecialize(job::CUDACompilerJob)) = method_table
+import FastmathOverlay
+if FastmathOverlay.functional()
+    function GPUCompiler.method_table_view(@nospecialize(job::CUDACompilerJob))
+        if job.config.params.contract
+            mtv = FastmathOverlay.contract(job.world, method_table)
+            return mtv
+        else
+            return Core.Compiler.OverlayMethodTable(job.world, method_table)
+        end
+    end
+else
+    GPUCompiler.method_table(@nospecialize(job::CUDACompilerJob)) = method_table
+end
 
 GPUCompiler.kernel_state_type(job::CUDACompilerJob) = KernelState
 
@@ -68,7 +82,7 @@ function compiler_config(dev; kwargs...)
     end
     return config
 end
-@noinline function _compiler_config(dev; kernel=true, name=nothing, always_inline=false, kwargs...)
+@noinline function _compiler_config(dev; kernel=true, name=nothing, always_inline=false, contract=false, kwargs...)
     # determine the toolchain (cached, because this is slow)
     if !isassigned(_toolchain)
         _toolchain[] = supported_toolchain()
@@ -92,7 +106,7 @@ end
 
     # create GPUCompiler objects
     target = PTXCompilerTarget(; cap, ptx, debuginfo, kwargs...)
-    params = CUDACompilerParams()
+    params = CUDACompilerParams(; contract)
     CompilerConfig(target, params; kernel, name, always_inline)
 end
 
diff --git a/src/compiler/execution.jl b/src/compiler/execution.jl
index 5859de2d88..4b19d4327d 100644
--- a/src/compiler/execution.jl
+++ b/src/compiler/execution.jl
@@ -6,7 +6,7 @@ export @cuda, cudaconvert, cufunction, dynamic_cufunction, nextwarp, prevwarp
 ## high-level @cuda interface
 
 const MACRO_KWARGS = [:dynamic, :launch]
-const COMPILER_KWARGS = [:kernel, :name, :always_inline, :minthreads, :maxthreads, :blocks_per_sm, :maxregs]
+const COMPILER_KWARGS = [:kernel, :name, :always_inline, :minthreads, :maxthreads, :blocks_per_sm, :maxregs, :fastmath, :contract]
 const LAUNCH_KWARGS = [:cooperative, :blocks, :threads, :shmem, :stream]
 
 
@@ -306,6 +306,7 @@ The following keyword arguments are supported:
   supported on LLVM 4.0+)
 - `name`: override the name that the kernel will have in the generated code
 - `always_inline`: inline all function calls in the kernel
+- `fastmath`: use less precise square roots and flush denormals
 
 The output of this function is automatically cached, i.e. you can simply call `cufunction`
 in a hot path without degrading performance. New code will be generated automatically, when
diff --git a/test/core/codegen.jl b/test/core/codegen.jl
index e948972407..8ebeff0558 100644
--- a/test/core/codegen.jl
+++ b/test/core/codegen.jl
@@ -112,6 +112,34 @@ end
     @test cpu(input) == gpu(input)
 end
 
+@testset "contract" begin
+    # FIXME: Overlay table is not used for top-level function
+    ir = sprint(io->CUDA.code_llvm(io, *, Tuple{Float64, Float64}, contract=true))
+    @test_broken contains(ir, "fmul nsz contract double %0, %1")
+
+    f(a, b) = a * b
+    g(a, b) = a + b
+    h(a, b) = a - b
+
+    ir = sprint(io->CUDA.code_llvm(io, f, Tuple{Float64, Float64}, contract=true))
+    @test_broken contains(ir, "fmul nsz contract double %0, %1")
+
+    ir = sprint(io->CUDA.code_llvm(io, f, Tuple{Float32, Float32}, contract=true))
+    @test contains(ir, "fmul nsz contract float %0, %1")
+
+    ir = sprint(io->CUDA.code_llvm(io, g, Tuple{Float64, Float64}, contract=true))
+    @test contains(ir, "fadd nsz contract double %0, %1")
+
+    ir = sprint(io->CUDA.code_llvm(io, g, Tuple{Float32, Float32}, contract=true))
+    @test contains(ir, "fadd nsz contract float %0, %1")
+
+    ir = sprint(io->CUDA.code_llvm(io, h, Tuple{Float64, Float64}, contract=true))
+    @test contains(ir, "fsub nsz contract double %0, %1")
+
+    ir = sprint(io->CUDA.code_llvm(io, h, Tuple{Float32, Float32}, contract=true))
+    @test contains(ir, "fsub nsz contract float %0, %1")
+end
+
 end
 
 ############################################################################################
@@ -157,6 +185,29 @@ end
     @test !occursin(".local", asm)
 end
 
+@testset "fastmath" begin
+    function sqrt_kernel(x)
+        i = threadIdx().x
+        @inbounds x[i] = sqrt(x[i])
+        return
+    end
+
+    function div_kernel(x)
+        i = threadIdx().x
+        @fastmath @inbounds x[i] = 1 / x[i]
+        return
+    end
+
+    asm = sprint(io->CUDA.code_ptx(io, sqrt_kernel, Tuple{CuDeviceArray{Float32,1,AS.Global}}))
+    @test occursin("sqrt.r", asm)
+
+    asm = sprint(io->CUDA.code_ptx(io, sqrt_kernel, Tuple{CuDeviceArray{Float32,1,AS.Global}}; fastmath=true))
+    @test occursin("sqrt.approx.ftz", asm)
+
+    asm = sprint(io->CUDA.code_ptx(io, div_kernel, Tuple{CuDeviceArray{Float32,1,AS.Global}}; fastmath=true))
+    @test occursin("div.approx.ftz", asm)
+end
+
 end
 
 ############################################################################################