[AMDGPU] Add llvm.amdgcn.fmad.ftz intrinsic
This intrinsic selects v_mad_f32 regardless of fp32 denorm support.
Differential Revision: https://reviews.llvm.org/D48573
llvm-svn: 335654
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index 43a9e0a..c8c0bde 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -360,6 +360,12 @@
[IntrNoMem, IntrSpeculatable]
>;
+// v_mad_f32/v_mac_f32, selected regardless of denorm support.
+def int_amdgcn_fmad_ftz :
+ Intrinsic<[llvm_float_ty],
+ [llvm_float_ty, llvm_float_ty, llvm_float_ty],
+ [IntrNoMem, IntrSpeculatable]
+>;
// Fields should mirror atomicrmw
class AMDGPUAtomicIncIntrin : Intrinsic<[llvm_anyint_ty],
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index b8e6c11..e6b3bd1 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -4922,6 +4922,9 @@
return SDValue(DAG.getMachineNode(AMDGPU::WWM, DL, Src.getValueType(), Src),
0);
}
+ case Intrinsic::amdgcn_fmad_ftz:
+ return DAG.getNode(AMDGPUISD::FMAD_FTZ, DL, VT, Op.getOperand(1),
+ Op.getOperand(2), Op.getOperand(3));
default:
if (const AMDGPU::ImageDimIntrinsicInfo *ImageDimIntr =
AMDGPU::getImageDimIntrinsicInfo(IntrinsicID))
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fmad.ftz.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fmad.ftz.ll
new file mode 100644
index 0000000..c9c9ef1
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fmad.ftz.ll
@@ -0,0 +1,114 @@
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
+; RUN: llc -march=amdgcn -mcpu=tonga -mattr=+fp32-denormals -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=+fp32-denormals -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
+
+declare float @llvm.amdgcn.fmad.ftz(float %a, float %b, float %c)
+
+; GCN-LABEL: {{^}}mad_f32:
+; GCN: v_ma{{[dc]}}_f32
+define amdgpu_kernel void @mad_f32(
+ float addrspace(1)* %r,
+ float addrspace(1)* %a,
+ float addrspace(1)* %b,
+ float addrspace(1)* %c) {
+ %a.val = load float, float addrspace(1)* %a
+ %b.val = load float, float addrspace(1)* %b
+ %c.val = load float, float addrspace(1)* %c
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float %a.val, float %b.val, float %c.val)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+; GCN-LABEL: {{^}}mad_f32_imm_a:
+; GCN: v_mov_b32_e32 [[KA:v[0-9]+]], 0x41000000
+; GCN: v_ma{{[dc]}}_f32 {{v[0-9]+}}, [[KA]],
+define amdgpu_kernel void @mad_f32_imm_a(
+ float addrspace(1)* %r,
+ float addrspace(1)* %b,
+ float addrspace(1)* %c) {
+ %b.val = load float, float addrspace(1)* %b
+ %c.val = load float, float addrspace(1)* %c
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float 8.0, float %b.val, float %c.val)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+; GCN-LABEL: {{^}}mad_f32_imm_b:
+; GCN: v_mov_b32_e32 [[KB:v[0-9]+]], 0x41000000
+; GCN: v_ma{{[dc]}}_f32 {{v[0-9]+}}, {{[vs][0-9]+}}, [[KB]],
+define amdgpu_kernel void @mad_f32_imm_b(
+ float addrspace(1)* %r,
+ float addrspace(1)* %a,
+ float addrspace(1)* %c) {
+ %a.val = load float, float addrspace(1)* %a
+ %c.val = load float, float addrspace(1)* %c
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float %a.val, float 8.0, float %c.val)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+; GCN-LABEL: {{^}}mad_f32_imm_c:
+; GCN: v_mov_b32_e32 [[KC:v[0-9]+]], 0x41000000
+; GCN: v_ma{{[dc]}}_f32 {{v[0-9]+}}, {{[vs][0-9]+}}, {{v[0-9]+}}, [[KC]]{{$}}
+define amdgpu_kernel void @mad_f32_imm_c(
+ float addrspace(1)* %r,
+ float addrspace(1)* %a,
+ float addrspace(1)* %b) {
+ %a.val = load float, float addrspace(1)* %a
+ %b.val = load float, float addrspace(1)* %b
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float %a.val, float %b.val, float 8.0)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+; GCN-LABEL: {{^}}mad_f32_neg_b:
+; GCN: v_mad_f32 v{{[0-9]+}}, s{{[0-9]+}}, -v{{[0-9]+}}, v{{[0-9]+}}
+define amdgpu_kernel void @mad_f32_neg_b(
+ float addrspace(1)* %r,
+ float addrspace(1)* %a,
+ float addrspace(1)* %b,
+ float addrspace(1)* %c) {
+ %a.val = load float, float addrspace(1)* %a
+ %b.val = load float, float addrspace(1)* %b
+ %c.val = load float, float addrspace(1)* %c
+ %neg.b = fsub float -0.0, %b.val
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float %a.val, float %neg.b, float %c.val)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+; GCN-LABEL: {{^}}mad_f32_abs_b:
+; GCN: v_mad_f32 v{{[0-9]+}}, s{{[0-9]+}}, |v{{[0-9]+}}|, v{{[0-9]+}}
+define amdgpu_kernel void @mad_f32_abs_b(
+ float addrspace(1)* %r,
+ float addrspace(1)* %a,
+ float addrspace(1)* %b,
+ float addrspace(1)* %c) {
+ %a.val = load float, float addrspace(1)* %a
+ %b.val = load float, float addrspace(1)* %b
+ %c.val = load float, float addrspace(1)* %c
+ %abs.b = call float @llvm.fabs.f32(float %b.val)
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float %a.val, float %abs.b, float %c.val)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+; GCN-LABEL: {{^}}mad_f32_neg_abs_b:
+; GCN: v_mad_f32 v{{[0-9]+}}, s{{[0-9]+}}, -|v{{[0-9]+}}|, v{{[0-9]+}}
+define amdgpu_kernel void @mad_f32_neg_abs_b(
+ float addrspace(1)* %r,
+ float addrspace(1)* %a,
+ float addrspace(1)* %b,
+ float addrspace(1)* %c) {
+ %a.val = load float, float addrspace(1)* %a
+ %b.val = load float, float addrspace(1)* %b
+ %c.val = load float, float addrspace(1)* %c
+ %abs.b = call float @llvm.fabs.f32(float %b.val)
+ %neg.abs.b = fsub float -0.0, %abs.b
+ %r.val = call float @llvm.amdgcn.fmad.ftz(float %a.val, float %neg.abs.b, float %c.val)
+ store float %r.val, float addrspace(1)* %r
+ ret void
+}
+
+declare float @llvm.fabs.f32(float)