[llvm] [llvm-exegesis] Follow up of #127564: fix test. NFC. (PR #130072)
Sjoerd Meijer via llvm-commits
llvm-commits at lists.llvm.org
Thu Mar 6 03:00:04 PST 2025
https://github.com/sjoerdmeijer created https://github.com/llvm/llvm-project/pull/130072
We don't need to run llvm-exegesis for the regression tests, just dump the object code, so this adds option --benchmark-phase=assemble-measured-code to achieve that.
>From b467bd1c8d4e61443cd705e87ad3af5a369a3b13 Mon Sep 17 00:00:00 2001
From: Sjoerd Meijer <smeijer at nvidia.com>
Date: Thu, 6 Mar 2025 02:57:16 -0800
Subject: [PATCH] [llvm-exegesis] Follow up of #127564: fix test. NFC.
We don't need to run llvm-exegesis for the regression tests, just dump
the object code, so this adds option --benchmark-phase=assemble-measured-code
to achieve that.
---
llvm/test/tools/llvm-exegesis/AArch64/setReg_init_check.s | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/llvm/test/tools/llvm-exegesis/AArch64/setReg_init_check.s b/llvm/test/tools/llvm-exegesis/AArch64/setReg_init_check.s
index 1b69ac68a2b30..da28388c96094 100644
--- a/llvm/test/tools/llvm-exegesis/AArch64/setReg_init_check.s
+++ b/llvm/test/tools/llvm-exegesis/AArch64/setReg_init_check.s
@@ -3,7 +3,7 @@ REQUIRES: aarch64-registered-target
## PPR Register Class Initialization Testcase
## Ideally, we should use PTRUE_{B/H/S/D} instead of FADDV_VPZ_D for an isolated test case;
## However, exegesis does not yet support PTRUE_{B/H/S/D}.
-RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FADDV_VPZ_D 2>&1
+RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FADDV_VPZ_D --benchmark-phase=assemble-measured-code 2>&1
RUN: llvm-objdump -d %d > %t.s
RUN: FileCheck %s --check-prefix=PPR_ASM < %t.s
PPR_ASM: <foo>:
@@ -14,7 +14,7 @@ PPR_ASM-NEXT: faddv d{{[0-9]+}}, p{{[0-9]+}}, z{{[0-9]+}}
## ZPR Register Class Initialization Testcase
## Ideally, we should use DUP_ZI_{B/H/S/D} instead of FADDV_VPZ_D for an isolated test case;
## However, exegesis does not yet support DUP_ZI_{B/H/S/D}.
-RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FADDV_VPZ_D 2>&1
+RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=FADDV_VPZ_D --benchmark-phase=assemble-measured-code 2>&1
RUN: llvm-objdump -d %d > %t.s
RUN: FileCheck %s --check-prefix=ZPR_ASM < %t.s
ZPR_ASM: <foo>:
@@ -23,7 +23,7 @@ ZPR_ASM-NEXT: mov z{{[0-9]+}}.d, #0x0
ZPR_ASM-NEXT: faddv d{{[0-9]+}}, p{{[0-9]+}}, z{{[0-9]+}}
## FPR128 Register Class Initialization Testcase
-RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADDVv16i8v 2>&1
+RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADDVv16i8v --benchmark-phase=assemble-measured-code 2>&1
RUN: llvm-objdump -d %d > %t.s
RUN: FileCheck %s --check-prefix=FPR128-ASM < %t.s
FPR128-ASM: <foo>:
@@ -31,7 +31,7 @@ FPR128-ASM: movi v{{[0-9]+}}.2d, #0000000000000000
FPR128-ASM-NEXT: addv b{{[0-9]+}}, v{{[0-9]+}}.16b
## FPR64 Register Class Initialization Testcase
-RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADDVv4i16v 2>&1
+RUN: llvm-exegesis -mcpu=neoverse-v2 -mode=latency --dump-object-to-disk=%d --opcode-name=ADDVv4i16v --benchmark-phase=assemble-measured-code 2>&1
RUN: llvm-objdump -d %d > %t.s
RUN: FileCheck %s --check-prefix=FPR64-ASM < %t.s
FPR64-ASM: <foo>:
More information about the llvm-commits
mailing list