[llvm] [AArch64][TTI] Fix extract cost for scalar fmul users (PR #212739)

[Anutosh Bhat via llvm-commits]

anutosh491 wrote:

On Branch 
```
 anutosh491 at Anutoshs-MacBook-Air llvm-project % TEST=llvm/test/Transforms/SLPVectorizer/AArch64/extract-fmul-cost.ll


 anutosh491 at Anutoshs-MacBook-Air llvm-project % build-assert/bin/opt \
  -passes=slp-vectorizer \
  -mtriple=aarch64-unknown-linux \
  -debug-only=SLP \
  -disable-output \
  "$TEST" 2>&1 |
  grep -A6 -B2 "Use: x1"
SLP: Computing cost for external use of TreeEntry 1 in lane 1
  User:  %mul = fmul double %x1, %x2
  Use: x1
  ExtractElement cost for double from <2 x double>: 0
SLP: Spill Cost = 0.
SLP: Extract Cost = 0.
SLP: Total Cost = -4.
SLP: Found cost = -4 for VF=2
SLP: Decided to vectorize cost = -4


 anutosh491 at Anutoshs-MacBook-Air llvm-project % build-assert/bin/opt \
  -passes=slp-vectorizer \
  -slp-threshold=3 \
  -mtriple=aarch64-unknown-linux \
  -S \
  "$TEST" |
  grep -E "fadd <2 x double>|store <2 x double>"
  %2 = fadd <2 x double> %0, %1
  %5 = fadd <2 x double> %3, %4
  store <2 x double> %2, ptr %o0, align 8
  store <2 x double> %5, ptr %o2, align 8
```

On Main
```
 anutosh491 at Anutoshs-MacBook-Air llvm-project % build-assert/bin/opt \      
  -passes=slp-vectorizer \
  -mtriple=aarch64-unknown-linux \
  -debug-only=SLP \
  -disable-output \
  "$TEST" 2>&1 |                                
  grep -A6 -B2 "Use: x1"
SLP: Computing cost for external use of TreeEntry 1 in lane 1
  User:  %mul = fmul double %x1, %x2
  Use: x1
  ExtractElement cost for double from <2 x double>: 2
SLP: Spill Cost = 0.
SLP: Extract Cost = 2.
SLP: Total Cost = -2.
SLP: Found cost = -2 for VF=2
SLP: Decided to vectorize cost = -2


 anutosh491 at Anutoshs-MacBook-Air llvm-project % build-assert/bin/opt \      
  -passes=slp-vectorizer \
  -slp-threshold=3 \
  -mtriple=aarch64-unknown-linux \
  -S \
  "$TEST" |
  grep -E "fadd <2 x double>|store <2 x double>"
  %2 = fadd <2 x double> %0, %1
  store <2 x double> %2, ptr %o2, align 8
```

https://github.com/llvm/llvm-project/pull/212739