llvmbot wrote:
@llvm/pr-subscribers-llvm-ir
@llvm/pr-subscribers-clang
Author: Hassnaa Hamdi (hassnaaHamdi)
Changes
According to specifications in
[ARM-software/acle/pull/309](https://github.com/ARM-software/acle/pull/309)
Add following intrinsics:
```
// svmax single,multi
svbfloat16x2_
https://github.com/CarolineConcatto commented:
It looks good Hassnaa, I am just concern about how we build the llvm-ir
intrinsic. All the other BF intrinsics don't have the name b + the type size
https://github.com/llvm/llvm-project/pull/90105
___
cfe
https://github.com/CarolineConcatto edited
https://github.com/llvm/llvm-project/pull/90105
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
@@ -109,6 +109,19 @@ define { , }
@multi_vec_max_single_x2_u64(<
ret { , } %res
}
+; BFMAX (Single, x2)
+
+define { , }
@multi_vec_max_single_x2_bf16( %zdn1, %zdn2, %zm) {
+; CHECK-LABEL: multi_vec_max_single_x2_bf16:
+; CHECK: // %bb.0:
+; CHECK-NEXT:// kil
@@ -3387,7 +3387,7 @@ let TargetPrefix = "aarch64" in {
// Multi-vector floating point min/max number
//
- foreach instr = ["fmaxnm", "fminnm"] in {
+ foreach instr = ["fmaxnm", "bfmaxnm", "fminnm", "bfminnm"] in {
momchil-velikov wrote:
Likewise here.
@@ -3373,7 +3373,7 @@ let TargetPrefix = "aarch64" in {
// Multi-vector min/max
//
- foreach ty = ["f", "s", "u"] in {
+ foreach ty = ["bf", "f", "s", "u"] in {
momchil-velikov wrote:
You could just omit that part. Then the `bfloat` intrinsics would use
momchil-velikov wrote:
Typo in commit message: `bflaot16`
> Variations other than bfloat16 had been already supported.
-> Variations other than bfloat16 are already supported.
https://github.com/llvm/llvm-project/pull/90105
___
cfe-commits mailing li