Separately version the CUDA compiler (#2801)

Author: maleadt

Project.toml

@@ -7,6 +7,7 @@ AbstractFFTs = "621f4979-c628-5d54-868e-fcf4e3e8185c"
 Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
 BFloat16s = "ab4f0b2a-ad5b-11e8-123f-65d77653426b"
 CEnum = "fa961155-64e5-5f13-b03f-caf6b980ea82"
+CUDA_Compiler_jll = "d1e2174e-dfdc-576e-b43e-73b79eb1aca8"
 CUDA_Driver_jll = "4ee394cb-3365-5eb0-8335-949819d2adfc"
 CUDA_Runtime_Discovery = "1af6417a-86b4-443c-805f-a4643ffb695f"
 CUDA_Runtime_jll = "76a88914-d11a-5bdc-97e0-2f5a05c973a2"
@@ -40,23 +41,24 @@ demumble_jll = "1e29f10c-031c-5a83-9565-69cddfc27673"
 [weakdeps]
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
-SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 SparseMatricesCSR = "a0a7dd2c-ebf4-11e9-1f05-cf50bc540ca1"
+SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 
 [extensions]
 ChainRulesCoreExt = "ChainRulesCore"
 EnzymeCoreExt = "EnzymeCore"
-SpecialFunctionsExt = "SpecialFunctions"
 SparseMatricesCSRExt = "SparseMatricesCSR"
+SpecialFunctionsExt = "SpecialFunctions"
 
 [compat]
 AbstractFFTs = "0.4, 0.5, 1.0"
 Adapt = "4"
 BFloat16s = "0.2, 0.3, 0.4, 0.5"
 CEnum = "0.2, 0.3, 0.4, 0.5"
-CUDA_Driver_jll = "0.13"
-CUDA_Runtime_Discovery = "0.3.3"
-CUDA_Runtime_jll = "0.17"
+CUDA_Compiler_jll = "0.1.0"
+CUDA_Driver_jll = "12"
+CUDA_Runtime_Discovery = "1"
+CUDA_Runtime_jll = "0.18"
 ChainRulesCore = "1"
 Crayons = "4"
 DataFrames = "1"
@@ -92,5 +94,5 @@ julia = "1.10"
 [extras]
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
-SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 SparseMatricesCSR = "a0a7dd2c-ebf4-11e9-1f05-cf50bc540ca1"
+SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"

lib/cudadrv/state.jl

@@ -239,13 +239,6 @@ function context(dev::CuDevice)
               maxlog=1, _id=devidx)
     end
 
-    # warn about some known bugs
-    if runtime_version() < v"11.5" && capability(dev) < v"7"
-        @warn("""There are known codegen bugs on CUDA 11.4 and earlier for older GPUs like your $(name(dev)).
-                 Please use CUDA 11.5 or later, or switch to a different device.""",
-              maxlog=1, _id=devidx)
-    end
-
     # configure the primary context
     pctx = CuPrimaryContext(dev)
     ctx = CuContext(pctx)

lib/cudadrv/version.jl

@@ -90,6 +90,16 @@ function reset_runtime_version!()
     @info "Reset CUDA.jl toolkit preference, please re-start Julia for this to take effect."
 end
 
+"""
+    compiler_version()
+
+Returns the CUDA toolkit version that is used to provide the CUDA compiler (`ptxas`) and
+other tools. This is version separately from the CUDA Runtime, in order to ensure
+compatibility with the driver, and make sure we use the latest compatible version regardless
+of the selected runtime.
+"""
+compiler_version() = CUDA_Compiler_jll.cuda_version
+
 
 ## helpers
 

src/CUDA.jl

@@ -24,6 +24,8 @@ using LLVMLoopInfo
 
 using CUDA_Driver_jll
 
+using CUDA_Compiler_jll
+
 import CUDA_Runtime_jll
 const local_toolkit = CUDA_Runtime_jll.host_platform["cuda_local"] == "true"
 const toolkit_version = if CUDA_Runtime_jll.host_platform["cuda"] == "none"

src/compatibility.jl

@@ -56,8 +56,8 @@ const cuda_cap_db = Dict(
     v"8.7"   => between(v"11.4", highest),
     v"8.9"   => between(v"11.8", highest),
     v"9.0"   => between(v"11.8", highest),
-    v"12.0" => between(v"12.8", highest),
-    v"12.1" => between(v"12.9", highest),
+    v"12.0"  => between(v"12.8", highest),
+    v"12.1"  => between(v"12.9", highest),
 )
 
 function cuda_cap_support(ver::VersionNumber)
@@ -293,14 +293,15 @@ function llvm_compat(version=LLVM.version())
     return (cap=cap_support, ptx=ptx_support)
 end
 
-function cuda_compat(driver=driver_version(), runtime=runtime_version())
+function cuda_compat(driver=driver_version(), compiler=compiler_version())
+    # devices have to be supported by both the compiler and the driver
     driver_cap_support = cuda_cap_support(driver)
-    toolkit_cap_support = cuda_cap_support(runtime)
-    cap_support = sort(collect(driver_cap_support ∩ toolkit_cap_support))
+    compiler_cap_support = cuda_cap_support(compiler)
+    cap_support = sort(collect(driver_cap_support ∩ compiler_cap_support))
 
-    driver_ptx_support = cuda_ptx_support(driver)
-    toolkit_ptx_support = cuda_ptx_support(runtime)
-    ptx_support = sort(collect(driver_ptx_support ∩ toolkit_ptx_support))
+    # PTX code only has to be supported by the compiler
+    compiler_ptx_support = cuda_ptx_support(compiler)
+    ptx_support = cuda_ptx_support(compiler)
 
     return (cap=cap_support, ptx=ptx_support)
 end

src/compiler/compilation.jl

@@ -137,7 +137,7 @@ function GPUCompiler.mcgen(@nospecialize(job::CUDACompilerJob), mod::LLVM.Module
         # only emitting `.loc` instructions...
         #
         # according to NVIDIA, "it is fine for PTX producers to produce debug info but not
-        # set `.target debug` and if `--device-debug` isn't passed, PTXAS will compile in
+        # set `.target debug` and if `--device-debug` isn't passed, ptxas will compile in
         # release mode".
         asm = replace(asm, r"(\.target .+), debug" => s"\1")
     end

src/utilities.jl

@@ -41,18 +41,21 @@ end
 function versioninfo(io::IO=stdout)
     @assert functional(true)
 
-    print(io, "CUDA runtime $(runtime_version().major).$(runtime_version().minor), ")
+    println(io, "CUDA toolchain: ")
+
+    print(io, "- runtime $(runtime_version().major).$(runtime_version().minor), ")
     if local_toolkit
         println(io, "local installation")
     else
         println(io, "artifact installation")
     end
-    println(io, "CUDA driver $(driver_version().major).$(driver_version().minor)")
     if has_nvml()
-        println(io, "NVIDIA driver $(NVML.driver_version())")
+        print(io, "- driver $(NVML.driver_version())")
     else
-        println(io, "Unknown NVIDIA driver")
+        print(io, "- unknown driver")
     end
+    println(io, " for $(driver_version().major).$(driver_version().minor)")
+    println(io, "- compiler $(compiler_version().major).$(compiler_version().minor)")
     println(io)
 
     println(io, "CUDA libraries: ")
@@ -66,7 +69,7 @@ function versioninfo(io::IO=stdout)
 
     println(io, "Julia packages: ")
     println(io, "- CUDA: $(Base.pkgversion(CUDA))")
-    for name in [:CUDA_Driver_jll, :CUDA_Runtime_jll, :CUDA_Runtime_Discovery]
+    for name in [:CUDA_Driver_jll, :CUDA_Compiler_jll, :CUDA_Runtime_jll, :CUDA_Runtime_Discovery]
         isdefined(CUDA, name) || continue
         mod = getfield(CUDA, name)
         println(io, "- $(name): $(Base.pkgversion(mod))")