transpiler: invalid memory access error finally fixed

package/test/TranspilerTests.jl

@@ -48,30 +48,70 @@ end
 
 	expr = Vector{Expr}()
 	push!(expr, expressions[1])
-	# @time Transpiler.evaluate(expr, variables, parameters)
+	@time Transpiler.evaluate(expr, variables, parameters)
 end
 
 #TODO: test performance of transpiler PTX generation when doing "return String(take!(buffer))" vs "return take!(buffer)"
 
+function test_kernel(results)
+	@inbounds results[1] = 10f0
+
+	return nothing
+end
+
 @testset "TEMP" begin
+return
+	results = CuArray{Float32}(undef, 2)
+	# @device_code_ptx @cuda test_kernel(results)
+
+
+	# println(CUDA.code_ptx(kernel.fun, ))
+	# return
+
 	ptx = "
-	.version 7.1
+	.version 8.5
 	.target sm_61
 	.address_size 64
 
 	.visible .entry ExpressionProcessing(
-	.param .u32 param_1)
+	.param .u64 param_1)
 	{
-		.reg .u32   %parameter<1>;
-		.reg .u32   %i<1>;
+		.reg .b64   %parameter<1>;
+		.reg .b64   %i<1>;
+		//.reg .b64 %rd<6>;
 
-		ld.param.u32   %i0, [param_1];
-		cvta.to.global.u32   %parameter0, %i0;
+		ld.param.u64   %i0, [param_1];
+		cvta.to.global.u64   %parameter0, %i0;
 
 		st.global.f32  [%parameter0], 10.0;
 		ret;
 	}"
 
+	ptx = ".version 8.5
+.target sm_61
+.address_size 64
+
+.visible .entry ExpressionProcessing(
+  .param .u64 param_1)
+{
+.reg .b64   %parameter<1>;
+.reg .b32   %r<4>;
+.reg .pred   %p<1>;
+.reg .b64   %i<1>;
+
+ld.param.u64   %i0, [param_1];
+cvta.to.global.u64   %parameter0, %i0;
+
+mov.u32    %r0, %ntid.x;
+mov.u32    %r1, %ctaid.x;
+mov.u32    %r2, %tid.x;
+mad.lo.s32     %r3, %r0, %r1, %r2;
+setp.gt.s32    %p0, %r3, 2;
+@%p0 bra    \$L__BB0_2;
+st.global.f32  [%parameter0], 10.0;
+\$L__BB0_2: ret;
+}"
+
 	linker = CuLink()
 	add_data!(linker, "ExpressionProcessing", ptx)
 	
@@ -90,11 +130,10 @@ end
 	threads = min(variableCols, config.threads)
 	blocks = cld(variableCols, threads)
 
-	cudacall(func, Tuple{CuPtr{Float32}}, cudaResults; threads=1, blocks=1)
+	cudacall(func, Tuple{CuPtr{Float32}}, cudaResults; threads=4, blocks=1)
 	# launch(func, cudaResults; threads=threads, blocks=blocks)
 
 	println(Array(cudaResults))
 end
 
-
 # TODO: University setup at 10.20.1.7