int Src0Idx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
AMDGPU::OpName::src0);
const MachineOperand *Src0 = &MI.getOperand(Src0Idx);
+ if (!Src0->isReg() && !Src0->isImm())
+ return nullptr;
+
if (Src0->isImm() && !isInlineConstant(MI, Src0Idx, *Src0))
return nullptr;
+
break;
}
}
if (!Src0Mods && !Src1Mods && !Clamp && !Omod &&
// If we have an SGPR input, we will violate the constant bus restriction.
- !RI.isSGPRReg(MBB->getParent()->getRegInfo(), Src0->getReg())) {
+ (!Src0->isReg() || !RI.isSGPRReg(MBB->getParent()->getRegInfo(), Src0->getReg()))) {
if (auto Imm = getFoldableImm(Src2)) {
return BuildMI(*MBB, MI, MI.getDebugLoc(),
get(IsF16 ? AMDGPU::V_MADAK_F16 : AMDGPU::V_MADAK_F32))
%2 = V_MAC_F32_e32 killed %0, %1, %3, implicit %exec
...
+
+# This can still fold if this is an inline immediate.
+
+# GCN-LABEL: name: test_madak_inlineimm_src0_f32
+# GCN: %1 = V_MADMK_F32 1073741824, 1078523331, %2, implicit %exec
+
+---
+name: test_madak_inlineimm_src0_f32
+registers:
+ - { id: 0, class: vgpr_32}
+ - { id: 1, class: vgpr_32 }
+ - { id: 2, class: vgpr_32 }
+body: |
+ bb.0:
+
+ %0 = V_MOV_B32_e32 1078523331, implicit %exec
+ %1 = V_MAC_F32_e32 1073741824, %0, %2, implicit %exec
+
+...
+# Non-inline immediate uses constant bus already.
+
+# GCN-LABEL: name: test_madak_otherimm_src0_f32
+# GCN: %1 = V_MAC_F32_e32 1120403456, %0, %1, implicit %exec
+
+---
+name: test_madak_otherimm_src0_f32
+registers:
+ - { id: 0, class: vgpr_32}
+ - { id: 1, class: vgpr_32 }
+ - { id: 2, class: vgpr_32 }
+body: |
+ bb.0:
+
+ %0 = V_MOV_B32_e32 1078523331, implicit %exec
+ %1 = V_MAC_F32_e32 1120403456, %0, %2, implicit %exec
+
+...
+# Non-inline immediate uses constant bus already.
+
+# GCN-LABEL: name: test_madak_other_constantlike_src0_f32
+# GCN: %1 = V_MAC_F32_e32 %stack.0, %0, %1, implicit %exec
+---
+name: test_madak_other_constantlike_src0_f32
+registers:
+ - { id: 0, class: vgpr_32}
+ - { id: 1, class: vgpr_32 }
+ - { id: 2, class: vgpr_32 }
+stack:
+ - { id: 0, name: "", type: default, offset: 0, size: 128, alignment: 8,
+ callee-saved-register: '', local-offset: 0, di-variable: '', di-expression: '',
+ di-location: '' }
+body: |
+ bb.0:
+
+ %0 = V_MOV_B32_e32 1078523331, implicit %exec
+ %1 = V_MAC_F32_e32 %stack.0, %0, %2, implicit %exec
+
+...