Skip to content

[NVPTX] Support BFloat Store Parameter #137074

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
Apr 24, 2025
Merged

[NVPTX] Support BFloat Store Parameter #137074

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
c386cc0
support bf16
stumpOS Apr 23, 2025
a5a3fe8
add test
stumpOS Apr 23, 2025
d1e77e0
revert unintentional white space changes
stumpOS Apr 23, 2025
9c20a8f
also guard against v2bf16
stumpOS Apr 23, 2025
33a1785
format
stumpOS Apr 23, 2025
b85acb2
remove vector type from conditional
stumpOS Apr 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion llvm/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1868,7 +1868,7 @@ bool NVPTXDAGToDAGISel::tryStoreParam(SDNode *N) {
case 1: {
MVT::SimpleValueType MemTy = Mem->getMemoryVT().getSimpleVT().SimpleTy;
SDValue Imm = Ops[0];
if (MemTy != MVT::f16 && MemTy != MVT::v2f16 &&
if (MemTy != MVT::f16 && MemTy != MVT::bf16 &&
(isa<ConstantSDNode>(Imm) || isa<ConstantFPSDNode>(Imm))) {
// Convert immediate to target constant
if (MemTy == MVT::f32 || MemTy == MVT::f64) {
Expand Down
24 changes: 24 additions & 0 deletions llvm/test/CodeGen/NVPTX/st-param-imm.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2000,3 +2000,27 @@ declare void @call_v4_i8(%struct.char4 alignstack(4))
declare void @call_v4_i16(%struct.short4 alignstack(8))
declare void @call_v4_i32(%struct.int4 alignstack(16))
declare void @call_v4_f32(%struct.float4 alignstack(16))

define void @st_param_bfloat() {
; CHECK-LABEL: st_param_bfloat(
; CHECK: {
; CHECK-NEXT: .reg .b16 %rs<2>;
; CHECK-EMPTY:
; CHECK-NEXT:// %bb.0:
; CHECK-NEXT: mov.b16 %rs1, 0x4100;
; CHECK-NEXT: { // callseq 83, 0
; CHECK-NEXT: .param .align 2 .b8 param0[2];
; CHECK-NEXT: st.param.b16 [param0], %rs1;
; CHECK-NEXT: call.uni
; CHECK-NEXT: call_bfloat,
; CHECK-NEXT: (
; CHECK-NEXT: param0
; CHECK-NEXT: );
; CHECK-NEXT: } // callseq 83
; CHECK-NEXT: ret;
%five = bitcast i16 16640 to bfloat
call void @call_bfloat(bfloat %five)
ret void
}

declare void @call_bfloat(bfloat)
Loading