From 4506188e003c7f67e1323f1330da239f22441da7 Mon Sep 17 00:00:00 2001
From: Matthew Poremba <matthew.poremba@amd.com>
Date: Sat, 26 Aug 2023 13:09:28 -0500
Subject: [PATCH] gpu-compute: Fix private offset/size register indexes

According to the ABI documentation from LLVM, the *low* register of flat
scratch (maxSGPR - 4) is the offset and the high register (maxSGPR - 3)
is size. These are currently backwards, resulting in some gnarly
addresses being generated leading to page fault and/or incorrect data.

This commit fixes this by setting the order correctly.

Change-Id: I0b1d077c49c0ee2a4e59b0f6d85cdb8f17f9be61
---
 src/gpu-compute/gpu_dyn_inst.cc | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/gpu-compute/gpu_dyn_inst.cc b/src/gpu-compute/gpu_dyn_inst.cc
index 991ba2c07a..8dfae1dc90 100644
--- a/src/gpu-compute/gpu_dyn_inst.cc
+++ b/src/gpu-compute/gpu_dyn_inst.cc
@@ -901,12 +901,12 @@ GPUDynInst::resolveFlatSegment(const VectorMask &mask)
         uint32_t numSgprs = wavefront()->maxSgprs;
         uint32_t physSgprIdx =
             wavefront()->computeUnit->registerManager->mapSgpr(wavefront(),
-                                                          numSgprs - 3);
+                                                          numSgprs - 4);
         uint32_t offset =
             wavefront()->computeUnit->srf[simdId]->read(physSgprIdx);
         physSgprIdx =
             wavefront()->computeUnit->registerManager->mapSgpr(wavefront(),
-                                                          numSgprs - 4);
+                                                          numSgprs - 3);
         uint32_t size =
             wavefront()->computeUnit->srf[simdId]->read(physSgprIdx);
         for (int lane = 0; lane < wavefront()->computeUnit->wfSize(); ++lane) {