gpu-compute: Simplify LGKM decrementing for Flat instructions
authorKyle Roarty <kyleroarty1716@gmail.com>
Tue, 19 Jan 2021 20:00:58 +0000 (14:00 -0600)
committerKyle Roarty <kyleroarty1716@gmail.com>
Thu, 21 Jan 2021 21:41:05 +0000 (21:41 +0000)
commit68fe6fa6cbb31045e33c074566b71ff444c2cf76
tree864a0b62fce60e37f340ba26a8975ad72a2fdb40
parent927cd10cb48c73e5a4a092ec42d36674c73892fd
gpu-compute: Simplify LGKM decrementing for Flat instructions

This commit makes it so LGKM count is decremented in a single place
(after completeAcc), which fixes a couple of potential bugs

1. Data is only written by completeAcc, not after initiateAcc. LGKM
count is supposed to be decremented after data is written.
2. LGKM count is now properly decremented for atomics without return

Change-Id: Ic791af3b42e04f7baaa0ce50cb2a2c6286c54f5a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/39396
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Matthew Poremba <matthew.poremba@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
src/gpu-compute/global_memory_pipeline.cc