[llvm] [RISCV] Promote fixed-length bf16 arith vector ops with zvfbfmin (PR #112393)

Tue Oct 15 15:17:19 PDT 2024

llvm-ci wrote:

LLVM Buildbot has detected a new failure on builder `fuchsia-x86_64-linux` running on `fuchsia-debian-64-us-central1-a-1` while building `llvm` at step 4 "annotate".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/11/builds/6654

<details>
<summary>Here is the relevant piece of the build log for the reference</summary>

```
Step 4 (annotate) failure: 'python ../llvm-zorg/zorg/buildbot/builders/annotated/fuchsia-linux.py ...' (failure)
...
[1326/1331] Building CXX object unittests/Transforms/Scalar/CMakeFiles/ScalarTests.dir/LoopPassManagerTest.cpp.o
clang++: warning: optimization flag '-ffat-lto-objects' is not supported [-Wignored-optimization-argument]
[1327/1331] Linking CXX executable unittests/Transforms/Scalar/ScalarTests
[1327/1331] Running the LLVM regression tests
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using ld.lld: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/ld.lld
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using lld-link: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/lld-link
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using ld64.lld: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/ld64.lld
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using wasm-ld: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/wasm-ld
-- Testing: 56105 tests, 60 workers --
Testing: 
FAIL: LLVM :: Analysis/CostModel/RISCV/reduce-fmul.ll (397 of 56105)
******************** TEST 'LLVM :: Analysis/CostModel/RISCV/reduce-fmul.ll' FAILED ********************
Exit Code: 1

Command Output (stderr):
--
RUN: at line 2: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/opt < /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+zfbfmin,+zvfbfmin -passes="print<cost-model>" -cost-kind=throughput 2>&1 -disable-output | /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/FileCheck /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll --check-prefixes=FP-REDUCE,FP-REDUCE-ZVFH
+ /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/opt -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+zfbfmin,+zvfbfmin '-passes=print<cost-model>' -cost-kind=throughput -disable-output
+ /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/FileCheck /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll --check-prefixes=FP-REDUCE,FP-REDUCE-ZVFH
/var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll:9:19: error: FP-REDUCE-NEXT: expected string not found in input
; FP-REDUCE-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V2 = call fast bfloat @llvm.vector.reduce.fmul.v2bf16(bfloat 0xR0000, <2 x bfloat> undef)
                  ^
<stdin>:2:149: note: scanning from here
Cost Model: Found an estimated cost of 1 for instruction: %V1 = call fast bfloat @llvm.vector.reduce.fmul.v1bf16(bfloat 0xR0000, <1 x bfloat> undef)
                                                                                                                                                    ^
<stdin>:3:1: note: possible intended match here
Cost Model: Found an estimated cost of 8 for instruction: %V2 = call fast bfloat @llvm.vector.reduce.fmul.v2bf16(bfloat 0xR0000, <2 x bfloat> undef)
^

Input file: <stdin>
Check file: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll

-dump-input=help explains the following input dump.

Input was:
<<<<<<
          1: Printing analysis 'Cost Model Analysis' for function 'reduce_fmul_bfloat': 
          2: Cost Model: Found an estimated cost of 1 for instruction: %V1 = call fast bfloat @llvm.vector.reduce.fmul.v1bf16(bfloat 0xR0000, <1 x bfloat> undef) 
next:9'0                                                                                                                                                         X error: no match found
          3: Cost Model: Found an estimated cost of 8 for instruction: %V2 = call fast bfloat @llvm.vector.reduce.fmul.v2bf16(bfloat 0xR0000, <2 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
next:9'1     ?                                                                                                                                                     possible intended match
          4: Cost Model: Found an estimated cost of 15 for instruction: %V4 = call fast bfloat @llvm.vector.reduce.fmul.v4bf16(bfloat 0xR0000, <4 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
          5: Cost Model: Found an estimated cost of 28 for instruction: %V8 = call fast bfloat @llvm.vector.reduce.fmul.v8bf16(bfloat 0xR0000, <8 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
          6: Cost Model: Found an estimated cost of 73 for instruction: %V16 = call fast bfloat @llvm.vector.reduce.fmul.v16bf16(bfloat 0xR0000, <16 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
          7: Cost Model: Found an estimated cost of 211 for instruction: %v32 = call fast bfloat @llvm.vector.reduce.fmul.v32bf16(bfloat 0xR0000, <32 x bfloat> undef) 
Step 7 (check) failure: check (failure)
...
[1326/1331] Building CXX object unittests/Transforms/Scalar/CMakeFiles/ScalarTests.dir/LoopPassManagerTest.cpp.o
clang++: warning: optimization flag '-ffat-lto-objects' is not supported [-Wignored-optimization-argument]
[1327/1331] Linking CXX executable unittests/Transforms/Scalar/ScalarTests
[1327/1331] Running the LLVM regression tests
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using ld.lld: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/ld.lld
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using lld-link: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/lld-link
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using ld64.lld: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/ld64.lld
llvm-lit: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/utils/lit/lit/llvm/config.py:506: note: using wasm-ld: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/wasm-ld
-- Testing: 56105 tests, 60 workers --
Testing: 
FAIL: LLVM :: Analysis/CostModel/RISCV/reduce-fmul.ll (397 of 56105)
******************** TEST 'LLVM :: Analysis/CostModel/RISCV/reduce-fmul.ll' FAILED ********************
Exit Code: 1

Command Output (stderr):
--
RUN: at line 2: /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/opt < /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+zfbfmin,+zvfbfmin -passes="print<cost-model>" -cost-kind=throughput 2>&1 -disable-output | /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/FileCheck /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll --check-prefixes=FP-REDUCE,FP-REDUCE-ZVFH
+ /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/opt -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+zfbfmin,+zvfbfmin '-passes=print<cost-model>' -cost-kind=throughput -disable-output
+ /var/lib/buildbot/fuchsia-x86_64-linux/build/llvm-build-2b3iocta/bin/FileCheck /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll --check-prefixes=FP-REDUCE,FP-REDUCE-ZVFH
/var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll:9:19: error: FP-REDUCE-NEXT: expected string not found in input
; FP-REDUCE-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %V2 = call fast bfloat @llvm.vector.reduce.fmul.v2bf16(bfloat 0xR0000, <2 x bfloat> undef)
                  ^
<stdin>:2:149: note: scanning from here
Cost Model: Found an estimated cost of 1 for instruction: %V1 = call fast bfloat @llvm.vector.reduce.fmul.v1bf16(bfloat 0xR0000, <1 x bfloat> undef)
                                                                                                                                                    ^
<stdin>:3:1: note: possible intended match here
Cost Model: Found an estimated cost of 8 for instruction: %V2 = call fast bfloat @llvm.vector.reduce.fmul.v2bf16(bfloat 0xR0000, <2 x bfloat> undef)
^

Input file: <stdin>
Check file: /var/lib/buildbot/fuchsia-x86_64-linux/llvm-project/llvm/test/Analysis/CostModel/RISCV/reduce-fmul.ll

-dump-input=help explains the following input dump.

Input was:
<<<<<<
          1: Printing analysis 'Cost Model Analysis' for function 'reduce_fmul_bfloat': 
          2: Cost Model: Found an estimated cost of 1 for instruction: %V1 = call fast bfloat @llvm.vector.reduce.fmul.v1bf16(bfloat 0xR0000, <1 x bfloat> undef) 
next:9'0                                                                                                                                                         X error: no match found
          3: Cost Model: Found an estimated cost of 8 for instruction: %V2 = call fast bfloat @llvm.vector.reduce.fmul.v2bf16(bfloat 0xR0000, <2 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
next:9'1     ?                                                                                                                                                     possible intended match
          4: Cost Model: Found an estimated cost of 15 for instruction: %V4 = call fast bfloat @llvm.vector.reduce.fmul.v4bf16(bfloat 0xR0000, <4 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
          5: Cost Model: Found an estimated cost of 28 for instruction: %V8 = call fast bfloat @llvm.vector.reduce.fmul.v8bf16(bfloat 0xR0000, <8 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
          6: Cost Model: Found an estimated cost of 73 for instruction: %V16 = call fast bfloat @llvm.vector.reduce.fmul.v16bf16(bfloat 0xR0000, <16 x bfloat> undef) 
next:9'0     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
          7: Cost Model: Found an estimated cost of 211 for instruction: %v32 = call fast bfloat @llvm.vector.reduce.fmul.v32bf16(bfloat 0xR0000, <32 x bfloat> undef) 

```

</details>

https://github.com/llvm/llvm-project/pull/112393