[llvm] [RISCV] Update latency of MUL & CPOP in SiFive P600's scheduling model (PR #115042)
Min-Yih Hsu via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 5 10:19:20 PST 2024
https://github.com/mshockwave created https://github.com/llvm/llvm-project/pull/115042
It should be 2 cycles rather than 3 cycles.
>From 7c4162c813ebf204722e6d992ac01cb3526bc413 Mon Sep 17 00:00:00 2001
From: Min-Yih Hsu <min.hsu at sifive.com>
Date: Mon, 4 Nov 2024 12:17:23 -0800
Subject: [PATCH] [RISCV] Update latency of MUL & CPOP in SiFive P600's
scheduling model
It should be 2 cycles rather than 3 cycles.
---
llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td | 2 +-
.../llvm-mca/RISCV/SiFiveP600/mul-cpop.s | 63 +++++++++++++++++++
2 files changed, 64 insertions(+), 1 deletion(-)
create mode 100644 llvm/test/tools/llvm-mca/RISCV/SiFiveP600/mul-cpop.s
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
index 51aa003139fbad..90bf9d36afb20b 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFiveP600.td
@@ -119,7 +119,7 @@ def P600WriteCMOV : SchedWriteRes<[SiFiveP600Branch, SiFiveP600CMOV]> {
}
def : InstRW<[P600WriteCMOV], (instrs PseudoCCMOVGPRNoX0)>;
-let Latency = 3 in {
+let Latency = 2 in {
// Integer multiplication
def : WriteRes<WriteIMul, [SiFiveP600MulI2F]>;
def : WriteRes<WriteIMul32, [SiFiveP600MulI2F]>;
diff --git a/llvm/test/tools/llvm-mca/RISCV/SiFiveP600/mul-cpop.s b/llvm/test/tools/llvm-mca/RISCV/SiFiveP600/mul-cpop.s
new file mode 100644
index 00000000000000..ec067ba6584033
--- /dev/null
+++ b/llvm/test/tools/llvm-mca/RISCV/SiFiveP600/mul-cpop.s
@@ -0,0 +1,63 @@
+# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
+# RUN: llvm-mca -mtriple=riscv64 -mcpu=sifive-p670 -iterations=1 < %s | FileCheck %s
+
+mul s6, s6, s7
+
+mulw s4, s4, a2
+
+cpop t1, t1
+
+cpopw t2, t2
+
+# CHECK: Iterations: 1
+# CHECK-NEXT: Instructions: 4
+# CHECK-NEXT: Total Cycles: 8
+# CHECK-NEXT: Total uOps: 4
+
+# CHECK: Dispatch Width: 4
+# CHECK-NEXT: uOps Per Cycle: 0.50
+# CHECK-NEXT: IPC: 0.50
+# CHECK-NEXT: Block RThroughput: 4.0
+
+# CHECK: Instruction Info:
+# CHECK-NEXT: [1]: #uOps
+# CHECK-NEXT: [2]: Latency
+# CHECK-NEXT: [3]: RThroughput
+# CHECK-NEXT: [4]: MayLoad
+# CHECK-NEXT: [5]: MayStore
+# CHECK-NEXT: [6]: HasSideEffects (U)
+
+# CHECK: [1] [2] [3] [4] [5] [6] Instructions:
+# CHECK-NEXT: 1 2 1.00 mul s6, s6, s7
+# CHECK-NEXT: 1 2 1.00 mulw s4, s4, a2
+# CHECK-NEXT: 1 2 1.00 cpop t1, t1
+# CHECK-NEXT: 1 2 1.00 cpopw t2, t2
+
+# CHECK: Resources:
+# CHECK-NEXT: [0] - SiFiveP600Div
+# CHECK-NEXT: [1] - SiFiveP600FEXQ0
+# CHECK-NEXT: [2] - SiFiveP600FEXQ1
+# CHECK-NEXT: [3] - SiFiveP600FloatDiv
+# CHECK-NEXT: [4] - SiFiveP600IEXQ0
+# CHECK-NEXT: [5] - SiFiveP600IEXQ1
+# CHECK-NEXT: [6] - SiFiveP600IEXQ2
+# CHECK-NEXT: [7] - SiFiveP600IEXQ3
+# CHECK-NEXT: [8.0] - SiFiveP600LDST
+# CHECK-NEXT: [8.1] - SiFiveP600LDST
+# CHECK-NEXT: [9] - SiFiveP600VDiv
+# CHECK-NEXT: [10] - SiFiveP600VEXQ0
+# CHECK-NEXT: [11] - SiFiveP600VEXQ1
+# CHECK-NEXT: [12] - SiFiveP600VFloatDiv
+# CHECK-NEXT: [13] - SiFiveP600VLD
+# CHECK-NEXT: [14] - SiFiveP600VST
+
+# CHECK: Resource pressure per iteration:
+# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8.0] [8.1] [9] [10] [11] [12] [13] [14]
+# CHECK-NEXT: - - - - - 4.00 - - - - - - - - - -
+
+# CHECK: Resource pressure by instruction:
+# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8.0] [8.1] [9] [10] [11] [12] [13] [14] Instructions:
+# CHECK-NEXT: - - - - - 1.00 - - - - - - - - - - mul s6, s6, s7
+# CHECK-NEXT: - - - - - 1.00 - - - - - - - - - - mulw s4, s4, a2
+# CHECK-NEXT: - - - - - 1.00 - - - - - - - - - - cpop t1, t1
+# CHECK-NEXT: - - - - - 1.00 - - - - - - - - - - cpopw t2, t2
More information about the llvm-commits
mailing list