A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 7, 16:30 UTC
[Hardware][Powerpc]Enable prefix caching and chunked prefill for ppc64le (#35081)
Signed-off-by: Akash kaothalkar <akash.kaothalkar@ibm.com> Co-authored-by: Akash kaothalkar <akash.kaothalkar@ibm.com>
Akash kaothalkar committed
Feb 26, 2026 at 15:51 UTC
e03ddcfbd4d686c91f6509c5451546437bbbf3e5
1 file changed
+5
-6
vllm/engine/arg_utils.py
+5
-6
@@ -2076,20 +2076,19 @@ class EngineArgs:
2076
)
2077
2078
# Disable chunked prefill and prefix caching for:
2079
- # POWER (ppc64le)/RISCV CPUs in V1
2079
+ # RISCV CPUs in V1
2080
if current_platform.is_cpu() and current_platform.get_cpu_architecture() in (
2081
- CpuArchEnum.POWERPC,
2081
CpuArchEnum.RISCV,
2082
):
2083
logger.info(
2085
- "Chunked prefill is not supported for POWER, "
2086
- "and RISC-V CPUs; "
2084
+ "Chunked prefill is not supported for"
2085
+ "RISC-V CPUs; "
2086
"disabling it for V1 backend."
2087
)
2088
self.enable_chunked_prefill = False
2089
logger.info(
2091
- "Prefix caching is not supported for POWER, "
2092
- "and RISC-V CPUs; "
2090
+ "Prefix caching is not supported for "
2091
+ "RISC-V CPUs; "
2092
"disabling it for V1 backend."
2093
)
2094
self.enable_prefix_caching = False