[AMD][MI35X] Qwen3.5-fp4 add ROCM_QUICK_REDUCE_QUANTIZATION=INT8, drop aiter-allreduce-fusion

Add ROCM_QUICK_REDUCE_QUANTIZATION=INT8 and remove --enable-aiter-allreduce-fusion
for Qwen3.5-397B-A17B-MXFP4 mi355x single-node benchmarks (non-MTP and MTP).

Author: yichiche

benchmarks/single_node/fixed_seq_len/qwen3.5_fp4_mi355x.sh

@@ -21,6 +21,7 @@ export SGLANG_USE_AITER=1
 export SGLANG_USE_AITER_UNIFIED_ATTN=1
 export AITER_FLYDSL_FORCE=1
 export SGLANG_MAMBA_SSM_DTYPE=bfloat16
+export ROCM_QUICK_REDUCE_QUANTIZATION=INT8
 
 SERVER_LOG=/workspace/server.log
 MEM_FRAC_STATIC=${MEM_FRAC_STATIC:-0.8}
@@ -41,7 +42,7 @@ python3 -m sglang.launch_server --model-path=$MODEL --trust-remote-code \
 --model-loader-extra-config '{"enable_multithread_load": true}' \
 --watchdog-timeout 1200  \
 --disable-radix-cache \
---enable-aiter-allreduce-fusion --max-running-requests $CONC \
+--max-running-requests $CONC \
 --page-size 16 \
 > $SERVER_LOG 2>&1 &
 

benchmarks/single_node/fixed_seq_len/qwen3.5_fp4_mi355x_mtp.sh

@@ -21,6 +21,7 @@ export SGLANG_USE_AITER=1
 export SGLANG_USE_AITER_UNIFIED_ATTN=1
 export AITER_FLYDSL_FORCE=1
 export SGLANG_MAMBA_SSM_DTYPE=bfloat16
+export ROCM_QUICK_REDUCE_QUANTIZATION=INT8
 
 SERVER_LOG=/workspace/server.log
 MEM_FRAC_STATIC=${MEM_FRAC_STATIC:-0.8}
@@ -41,7 +42,7 @@ python3 -m sglang.launch_server --model-path=$MODEL --trust-remote-code \
 --model-loader-extra-config '{"enable_multithread_load": true}' \
 --watchdog-timeout 1200  \
 --disable-radix-cache \
---enable-aiter-allreduce-fusion --max-running-requests $CONC \
+--max-running-requests $CONC \
 --page-size 16 \
 --speculative-algorithm EAGLE \
 --speculative-num-steps 3 \

perf-changelog.yaml

@@ -4885,7 +4885,15 @@
   description:
     - "Bump image to lmsysorg/sglang-rocm:v0.5.15-rocm720-mi35x-20260713"
     - "Add SGLANG_MAMBA_SSM_DTYPE=bfloat16 in both non-MTP and MTP benchmark scripts"
-  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2201  
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2201
+
+- config-keys:
+    - qwen3.5-fp4-mi355x-sglang
+    - qwen3.5-fp4-mi355x-sglang-mtp
+  description:
+    - "Add ROCM_QUICK_REDUCE_QUANTIZATION=INT8 in both non-MTP and MTP benchmark scripts"
+    - "Remove --enable-aiter-allreduce-fusion from both non-MTP and MTP launch commands"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2202  
 
 - config-keys:
     - dsv4-fp4-b200-vllm-agentic