[RISCV] Support fastcc passing scalable vectors indirectly with no free GPRs.

[topperc]

We can still pass indirectly by putting the pointer on the stack. This is what we do in the normal calling convention.

[llvmbot]

@llvm/pr-subscribers-backend-risc-v

Author: Craig Topper (topperc)

Changes

We can still pass indirectly by putting the pointer on the stack. This is what we do in the normal calling convention.

---

Full diff: https://github.com/llvm/llvm-project/pull/107623.diff

2 Files Affected:

• (modified) llvm/lib/Target/RISCV/RISCVCallingConv.cpp (+27-17)
• (modified) llvm/test/CodeGen/RISCV/rvv/calling-conv-fastcc.ll (+116)

diff --git a/llvm/lib/Target/RISCV/RISCVCallingConv.cpp b/llvm/lib/Target/RISCV/RISCVCallingConv.cpp
index c6a66e69401a6f..76fed279482468 100644
--- a/llvm/lib/Target/RISCV/RISCVCallingConv.cpp
+++ b/llvm/lib/Target/RISCV/RISCVCallingConv.cpp
@@ -603,25 +603,35 @@ bool llvm::CC_RISCV_FastCC(unsigned ValNo, MVT ValVT, MVT LocVT,
       if (ValVT.isFixedLengthVector())
         LocVT = TLI.getContainerForFixedLengthVector(LocVT);
       State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
-    } else {
-      // Try and pass the address via a "fast" GPR.
-      if (MCRegister GPRReg = State.AllocateReg(getFastCCArgGPRs(ABI))) {
-        LocInfo = CCValAssign::Indirect;
-        LocVT = Subtarget.getXLenVT();
-        State.addLoc(CCValAssign::getReg(ValNo, ValVT, GPRReg, LocVT, LocInfo));
-      } else if (ValVT.isFixedLengthVector()) {
-        auto StackAlign =
-            MaybeAlign(ValVT.getScalarSizeInBits() / 8).valueOrOne();
-        unsigned StackOffset =
-            State.AllocateStack(ValVT.getStoreSize(), StackAlign);
-        State.addLoc(
-            CCValAssign::getMem(ValNo, ValVT, StackOffset, LocVT, LocInfo));
-      } else {
-        // Can't pass scalable vectors on the stack.
-        return true;
-      }
+      return false;
     }
 
+    // Try and pass the address via a "fast" GPR.
+    if (MCRegister GPRReg = State.AllocateReg(getFastCCArgGPRs(ABI))) {
+      LocInfo = CCValAssign::Indirect;
+      LocVT = Subtarget.getXLenVT();
+      State.addLoc(CCValAssign::getReg(ValNo, ValVT, GPRReg, LocVT, LocInfo));
+      return false;
+    }
+
+    // Pass scalable vectors indirectly by storing the pointer on the stack.
+    if (ValVT.isScalableVector()) {
+      LocInfo = CCValAssign::Indirect;
+      LocVT = Subtarget.getXLenVT();
+      unsigned XLen = Subtarget.getXLen();
+      unsigned StackOffset = State.AllocateStack(XLen / 8, Align(XLen / 8));
+      State.addLoc(
+          CCValAssign::getMem(ValNo, ValVT, StackOffset, LocVT, LocInfo));
+      return false;
+    }
+
+    // Pass fixed-length vectors on the stack.
+    auto StackAlign =
+        MaybeAlign(ValVT.getScalarSizeInBits() / 8).valueOrOne();
+    unsigned StackOffset =
+        State.AllocateStack(ValVT.getStoreSize(), StackAlign);
+    State.addLoc(
+        CCValAssign::getMem(ValNo, ValVT, StackOffset, LocVT, LocInfo));
     return false;
   }
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/calling-conv-fastcc.ll b/llvm/test/CodeGen/RISCV/rvv/calling-conv-fastcc.ll
index fb84a2528778a3..427ce9d097135b 100644
--- a/llvm/test/CodeGen/RISCV/rvv/calling-conv-fastcc.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/calling-conv-fastcc.ll
@@ -633,3 +633,119 @@ define fastcc <vscale x 32 x i32> @pass_vector_arg_indirect_stack(<vscale x 32 x
   %s = call fastcc <vscale x 32 x i32> @vector_arg_indirect_stack(i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, <vscale x 32 x i32> zeroinitializer, <vscale x 32 x i32> zeroinitializer, <vscale x 32 x i32> zeroinitializer, i32 8)
   ret <vscale x 32 x i32> %s
 }
+
+; Test case where we are out of registers for the vector and all GPRs are used.
+define fastcc <vscale x 16 x i32> @vector_arg_indirect_stack_no_gpr(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8, i32 %9, i32 %10, i32 %11, <vscale x 16 x i32> %x, <vscale x 16 x i32> %y, <vscale x 16 x i32> %z) {
+; RV32-LABEL: vector_arg_indirect_stack_no_gpr:
+; RV32:       # %bb.0:
+; RV32-NEXT:    lw a0, 0(sp)
+; RV32-NEXT:    vl8re32.v v16, (a0)
+; RV32-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
+; RV32-NEXT:    vadd.vv v8, v8, v16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vector_arg_indirect_stack_no_gpr:
+; RV64:       # %bb.0:
+; RV64-NEXT:    ld a0, 0(sp)
+; RV64-NEXT:    vl8re32.v v16, (a0)
+; RV64-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
+; RV64-NEXT:    vadd.vv v8, v8, v16
+; RV64-NEXT:    ret
+  %s = add <vscale x 16 x i32> %x, %z
+  ret <vscale x 16 x i32> %s
+}
+
+; Calling the function above. Ensure we pass the arguments correctly.
+define fastcc <vscale x 16 x i32> @pass_vector_arg_indirect_stack_no_gpr(<vscale x 16 x i32> %x, <vscale x 16 x i32> %y, <vscale x 16 x i32> %z) {
+; RV32-LABEL: pass_vector_arg_indirect_stack_no_gpr:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -80
+; RV32-NEXT:    .cfi_def_cfa_offset 80
+; RV32-NEXT:    sw ra, 76(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw s0, 72(sp) # 4-byte Folded Spill
+; RV32-NEXT:    sw s1, 68(sp) # 4-byte Folded Spill
+; RV32-NEXT:    .cfi_offset ra, -4
+; RV32-NEXT:    .cfi_offset s0, -8
+; RV32-NEXT:    .cfi_offset s1, -12
+; RV32-NEXT:    addi s0, sp, 80
+; RV32-NEXT:    .cfi_def_cfa s0, 0
+; RV32-NEXT:    csrr a0, vlenb
+; RV32-NEXT:    slli a0, a0, 3
+; RV32-NEXT:    sub sp, sp, a0
+; RV32-NEXT:    andi sp, sp, -64
+; RV32-NEXT:    mv s1, sp
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
+; RV32-NEXT:    vmv.v.i v8, 0
+; RV32-NEXT:    addi a0, s1, 64
+; RV32-NEXT:    vs8r.v v8, (a0)
+; RV32-NEXT:    li a1, 1
+; RV32-NEXT:    li a2, 2
+; RV32-NEXT:    li a3, 3
+; RV32-NEXT:    li a4, 4
+; RV32-NEXT:    li a5, 5
+; RV32-NEXT:    li a6, 6
+; RV32-NEXT:    li a7, 7
+; RV32-NEXT:    li t3, 8
+; RV32-NEXT:    li t4, 9
+; RV32-NEXT:    li t5, 10
+; RV32-NEXT:    li t6, 11
+; RV32-NEXT:    sw a0, 0(sp)
+; RV32-NEXT:    li a0, 0
+; RV32-NEXT:    vmv.v.i v16, 0
+; RV32-NEXT:    call vector_arg_indirect_stack_no_gpr
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    addi sp, s0, -80
+; RV32-NEXT:    lw ra, 76(sp) # 4-byte Folded Reload
+; RV32-NEXT:    lw s0, 72(sp) # 4-byte Folded Reload
+; RV32-NEXT:    lw s1, 68(sp) # 4-byte Folded Reload
+; RV32-NEXT:    addi sp, sp, 80
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: pass_vector_arg_indirect_stack_no_gpr:
+; RV64:       # %bb.0:
+; RV64-NEXT:    addi sp, sp, -96
+; RV64-NEXT:    .cfi_def_cfa_offset 96
+; RV64-NEXT:    sd ra, 88(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd s0, 80(sp) # 8-byte Folded Spill
+; RV64-NEXT:    sd s1, 72(sp) # 8-byte Folded Spill
+; RV64-NEXT:    .cfi_offset ra, -8
+; RV64-NEXT:    .cfi_offset s0, -16
+; RV64-NEXT:    .cfi_offset s1, -24
+; RV64-NEXT:    addi s0, sp, 96
+; RV64-NEXT:    .cfi_def_cfa s0, 0
+; RV64-NEXT:    csrr a0, vlenb
+; RV64-NEXT:    slli a0, a0, 3
+; RV64-NEXT:    sub sp, sp, a0
+; RV64-NEXT:    andi sp, sp, -64
+; RV64-NEXT:    mv s1, sp
+; RV64-NEXT:    addi sp, sp, -16
+; RV64-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
+; RV64-NEXT:    vmv.v.i v8, 0
+; RV64-NEXT:    addi a0, s1, 64
+; RV64-NEXT:    vs8r.v v8, (a0)
+; RV64-NEXT:    li a1, 1
+; RV64-NEXT:    li a2, 2
+; RV64-NEXT:    li a3, 3
+; RV64-NEXT:    li a4, 4
+; RV64-NEXT:    li a5, 5
+; RV64-NEXT:    li a6, 6
+; RV64-NEXT:    li a7, 7
+; RV64-NEXT:    li t3, 8
+; RV64-NEXT:    li t4, 9
+; RV64-NEXT:    li t5, 10
+; RV64-NEXT:    li t6, 11
+; RV64-NEXT:    sd a0, 0(sp)
+; RV64-NEXT:    li a0, 0
+; RV64-NEXT:    vmv.v.i v16, 0
+; RV64-NEXT:    call vector_arg_indirect_stack_no_gpr
+; RV64-NEXT:    addi sp, sp, 16
+; RV64-NEXT:    addi sp, s0, -96
+; RV64-NEXT:    ld ra, 88(sp) # 8-byte Folded Reload
+; RV64-NEXT:    ld s0, 80(sp) # 8-byte Folded Reload
+; RV64-NEXT:    ld s1, 72(sp) # 8-byte Folded Reload
+; RV64-NEXT:    addi sp, sp, 96
+; RV64-NEXT:    ret
+  %s = call fastcc <vscale x 16 x i32> @vector_arg_indirect_stack_no_gpr(i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, <vscale x 16 x i32> zeroinitializer, <vscale x 16 x i32> zeroinitializer, <vscale x 16 x i32> zeroinitializer)
+  ret <vscale x 16 x i32> %s
+}

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[michaelmaitland]

LGTM