derek/gem5
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

misc: Move MemPool based calls to the SEWorkload.

Browse Source 
These currently proxy to the System object, but this is one step towards
moving the MemPool-s out of the System and into the SEWorkload where
they really should have been from the start.

Change-Id: Id27e7b874c283abf07bd892c8467a9cc52e2fdff
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/50342
Reviewed-by: Matthew Poremba <matthew.poremba@amd.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This commit is contained in:
Gabe Black
2021-09-13 17:13:06 -07:00
parent 9da405ce96
commit bec16fbc31
7 changed files with 43 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/gpu-compute/gpu_compute_driver.cc
View File

@@ -52,6 +52,7 @@
#include "params/GPUComputeDriver.hh"
#include "sim/full_system.hh"
#include "sim/process.hh"
#include "sim/se_workload.hh"
#include "sim/syscall_emul_buf.hh"
namespace gem5
@@ -758,7 +759,8 @@ GPUComputeDriver::ioctl(ThreadContext *tc, unsigned req, Addr ioc_buf)
// this as uncacheable from the CPU so that we can implement
// direct CPU framebuffer access similar to what we currently
// offer in real HW through the so-called Large BAR feature.
pa_addr = process->system->allocPhysPages(npages, dGPUPoolID);
pa_addr = process->seWorkload->allocPhysPages(
npages, dGPUPoolID);
//
// TODO: Uncacheable accesses need to be supported by the
// CPU-side protocol for this to work correctly. I believe
@@ -773,7 +775,7 @@ GPUComputeDriver::ioctl(ThreadContext *tc, unsigned req, Addr ioc_buf)
mmap_offset = args->mmap_offset;
// USERPTR allocations are system memory mapped into GPUVM
// space. The user provides the driver with the pointer.
pa_addr = process->system->allocPhysPages(npages);
pa_addr = process->seWorkload->allocPhysPages(npages);
DPRINTF(GPUDriver, "Mapping VA %p to framebuffer PA %p size "
"%d\n", args->va_addr, pa_addr, args->size);
@@ -794,7 +796,7 @@ GPUComputeDriver::ioctl(ThreadContext *tc, unsigned req, Addr ioc_buf)
// We will lazily map it into host memory on first touch. The
// fixupFault will find the original SVM aperture mapped to the
// host.
pa_addr = process->system->allocPhysPages(npages);
pa_addr = process->seWorkload->allocPhysPages(npages);
DPRINTF(GPUDriver, "Mapping VA %p to framebuffer PA %p size "
"%d\n", args->va_addr, pa_addr, args->size);