gpu-compute: Fix off-by-one when creating an AddrRange

The end value of an AddrRange is already not included in the range,
so subtracting one from the end creates an off-by-one error.

This patch removes the extra -1 that was used when determining the
end of an AddrRange in allocateGpuVma

Change-Id: I75659e9a7fabd991bb37be9aa40f8e409eb21154
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/48020
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

src/gpu-compute/gpu_compute_driver.cc

@@ -985,7 +985,7 @@ void
 GPUComputeDriver::allocateGpuVma(Request::CacheCoherenceFlags mtype,
                                  Addr start, Addr length)
 {
-    AddrRange range = AddrRange(start, start + length - 1);
+    AddrRange range = AddrRange(start, start + length);
     DPRINTF(GPUDriver, "Registering [%p - %p] with MTYPE %d\n",
             range.start(), range.end(), mtype);
     fatal_if(gpuVmas.insert(range, mtype) == gpuVmas.end(),