From bf8b022e60abd3008cc9309d837ef2727e6c0acd Mon Sep 17 00:00:00 2001
From: Mieszko Dziadowiec <mieszko.dziadowiec@intel.com>
Date: Fri, 3 Apr 2026 14:47:35 +0200
Subject: [PATCH] [Intel][Triton] Support `round_int8` for Intel backend
 (#38825)

Signed-off-by: Mieszko Dziadowiec <mdziadowiec@habana.ai>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Stefano Castagnetta <scastagnetta@nvidia.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Stefano Castagnetta <scastagnetta@nvidia.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
---
 .../model_executor/layers/quantization/utils/int8_utils.py | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/vllm/model_executor/layers/quantization/utils/int8_utils.py b/vllm/model_executor/layers/quantization/utils/int8_utils.py
index 020098dff..a98e29ffd 100644
--- a/vllm/model_executor/layers/quantization/utils/int8_utils.py
+++ b/vllm/model_executor/layers/quantization/utils/int8_utils.py
@@ -88,6 +88,13 @@ if current_platform.is_rocm():
     def round_int8(x):
         return tl.extra.hip.libdevice.round(x).to(tl.int8)
 
+
+elif current_platform.is_xpu():
+
+    @triton.jit
+    def round_int8(x):
+        return tl.extra.intel.libdevice.round(x).to(tl.int8)
+
 else:
 
     @triton.jit