From fd6829f7576da07e3bb90de8821834d3ce44610c Mon Sep 17 00:00:00 2001
From: Whatcookie <MalcolmJestadt@gmail.com>
Date: Sat, 29 Jul 2023 02:01:01 -0400
Subject: [PATCH] SPU LLVM: AVX-512 optimization for CFLTU (#14384)

- Takes advantage of vrangeps and the new float to uint instructions from AVX-512
- Down from 6 to 3 instructions

TODO: Somehow ensure that this is what llvm outputs using CreateFPToUI?
---
 rpcs3/Emu/Cell/SPURecompiler.cpp | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/rpcs3/Emu/Cell/SPURecompiler.cpp b/rpcs3/Emu/Cell/SPURecompiler.cpp
index 1dd9acab29..5c88bba127 100644
--- a/rpcs3/Emu/Cell/SPURecompiler.cpp
+++ b/rpcs3/Emu/Cell/SPURecompiler.cpp
@@ -9871,6 +9871,15 @@ public:
 				a = eval(a * s);
 
 			value_t<s32[4]> r;
+
+			if (m_use_avx512)
+			{
+				const auto sc = clamp_smax(a);
+				r.value = m_ir->CreateFPToUI(sc.value, get_type<s32[4]>());
+				set_vr(op.rt, r);
+				return;
+			}
+
 			r.value = m_ir->CreateFPToUI(a.value, get_type<s32[4]>());
 			set_vr(op.rt, select(bitcast<s32[4]>(a) > splat<s32[4]>(((32 + 127) << 23) - 1), splat<s32[4]>(-1), r & ~(bitcast<s32[4]>(a) >> 31)));
 		}