1 files changed, 66 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.sr.pk.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.sr.pk.bf16.ll
new file mode 100644
index 0000000..82991ae
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.sr.pk.bf16.ll
@@ -0,0 +1,66 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefix=GCN %s
+; xUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefix=GCN %s
+
+; FIXME: GlobalISel does not work with bf16
+
+declare <2 x bfloat> @llvm.amdgcn.cvt.sr.pk.bf16.f32(float, float, i32) #0
+
+define amdgpu_ps float @cvt_sr_pk_bf16_f32_vvv(float %src0, float %src1, i32 %src2) #1 {
+; GCN-LABEL: cvt_sr_pk_bf16_f32_vvv:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_cvt_sr_pk_bf16_f32 v0, v0, v1, v2
+; GCN-NEXT:    ; return to shader part epilog
+  %cvt = call <2 x bfloat> @llvm.amdgcn.cvt.sr.pk.bf16.f32(float %src0, float %src1, i32 %src2) #0
+  %ret = bitcast <2 x bfloat> %cvt to float
+  ret float %ret
+}
+
+define amdgpu_ps float @cvt_sr_pk_bf16_f32_sss(float inreg %src0, float inreg %src1, i32 inreg %src2) #1 {
+; GCN-LABEL: cvt_sr_pk_bf16_f32_sss:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_mov_b32_e32 v0, s2
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_cvt_sr_pk_bf16_f32 v0, s0, s1, v0
+; GCN-NEXT:    ; return to shader part epilog
+  %cvt = call <2 x bfloat> @llvm.amdgcn.cvt.sr.pk.bf16.f32(float %src0, float %src1, i32 %src2) #0
+  %ret = bitcast <2 x bfloat> %cvt to float
+  ret float %ret
+}
+
+define amdgpu_ps float @cvt_sr_pk_bf16_f32_vvi(float %src0, float %src1) #1 {
+; GCN-LABEL: cvt_sr_pk_bf16_f32_vvi:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_cvt_sr_pk_bf16_f32 v0, v0, v1, 0x10002
+; GCN-NEXT:    ; return to shader part epilog
+  %cvt = call <2 x bfloat> @llvm.amdgcn.cvt.sr.pk.bf16.f32(float %src0, float %src1, i32 65538) #0
+  %ret = bitcast <2 x bfloat> %cvt to float
+  ret float %ret
+}
+
+define amdgpu_ps float @cvt_sr_pk_bf16_f32_vvi_mods(float %src0, float %src1) #1 {
+; GCN-LABEL: cvt_sr_pk_bf16_f32_vvi_mods:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_cvt_sr_pk_bf16_f32 v0, -v0, |v1|, 1
+; GCN-NEXT:    ; return to shader part epilog
+  %s0 = fneg float %src0
+  %s1 = call float @llvm.fabs.f32(float %src1) #0
+  %cvt = call <2 x bfloat> @llvm.amdgcn.cvt.sr.pk.bf16.f32(float %s0, float %s1, i32 1) #0
+  %ret = bitcast <2 x bfloat> %cvt to float
+  ret float %ret
+}
+
+define amdgpu_ps float @cvt_sr_pk_bf16_f32_ssi(float inreg %src0, float inreg %src1) #1 {
+; GCN-LABEL: cvt_sr_pk_bf16_f32_ssi:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_cvt_sr_pk_bf16_f32 v0, s0, s1, 1
+; GCN-NEXT:    ; return to shader part epilog
+  %cvt = call <2 x bfloat> @llvm.amdgcn.cvt.sr.pk.bf16.f32(float %src0, float %src1, i32 1) #0
+  %ret = bitcast <2 x bfloat> %cvt to float
+  ret float %ret
+}
+
+declare float @llvm.fabs.f32(float) #0
+
+attributes #0 = { nounwind readnone }
+attributes #1 = { nounwind }