[llvm] [CostModel][X86] Price a vXi1 mask broadcast instead of scalarizing it (PR #222467)

[Simon Pilgrim via llvm-commits]

================
@@ -0,0 +1,61 @@
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=skylake-avx512 -passes='print<cost-model>' -cost-kind=all -disable-output 2>&1 | FileCheck %s --check-prefixes=COST,COST64,DQ
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=skylake-avx512 -mattr=-avx512dq -passes='print<cost-model>' -cost-kind=all -disable-output 2>&1 | FileCheck %s --check-prefixes=COST,COST64,NODQ
+; RUN: opt < %s -mtriple=i386-unknown-linux-gnu -mcpu=skylake-avx512 -passes='print<cost-model>' -cost-kind=all -disable-output 2>&1 | FileCheck %s --check-prefixes=COST,COST32,DQ
+
+; Cost tuples below are reciprocal throughput, code size, latency and size+latency.
----------------
RKSimon wrote:

please use the update scripts - costs tests are a nightmare to maintain without them

https://github.com/llvm/llvm-project/pull/222467