Group PTX fast-math optimization passes under :fastmath (#828)

Co-authored-by: Tim Besard <tim.besard@gmail.com>

Author: wsmoses

src/ptx.jl

@@ -291,9 +291,11 @@ function optimize_module!(@nospecialize(job::CompilerJob{PTXCompilerTarget}),
             register!(pb, NVVMReflectPass())
             add!(pb, NVVMReflectPass())
         end
-        add!(pb, PTXRSqrtFastPass())
-        add!(pb, PTXFDivFastPass())
-        add!(pb, PTXFSqrtFastPass())
+        if get(optimization_options(job), :fastmath, true)
+            add!(pb, PTXRSqrtFastPass())
+            add!(pb, PTXFDivFastPass())
+            add!(pb, PTXFSqrtFastPass())
+        end
 
         add!(pb, NewPMFunctionPassManager()) do fpm
             # needed by GemmKernels.jl-like code