From b5105c025cfacc1d7eaed9736c510c3b31eaee80 Mon Sep 17 00:00:00 2001
From: Nick Fitzgerald <fitzgen@gmail.com>
Date: Wed, 10 Nov 2021 15:45:43 -0800
Subject: [PATCH] MachInst: always rematerialize constants, rather than assign
 them registers

There were a few previous code paths that attempted to handle this, but this new
check handles it for all callers.

Rematerializing constants, rather than assigning and reusing a register, allows
for lower register pressure.
---
 cranelift/codegen/src/machinst/lower.rs       | 32 ++++++++++--
 .../filetests/filetests/isa/x64/i128.clif     | 50 ++++++++++---------
 2 files changed, 55 insertions(+), 27 deletions(-)

diff --git a/cranelift/codegen/src/machinst/lower.rs b/cranelift/codegen/src/machinst/lower.rs
index e6e608ec3d..8f65449cde 100644
--- a/cranelift/codegen/src/machinst/lower.rs
+++ b/cranelift/codegen/src/machinst/lower.rs
@@ -16,8 +16,9 @@ use crate::ir::{
     ValueDef, ValueLabelAssignments, ValueLabelStart,
 };
 use crate::machinst::{
-    writable_value_regs, ABICallee, BlockIndex, BlockLoweringOrder, LoweredBlock, MachLabel, VCode,
-    VCodeBuilder, VCodeConstant, VCodeConstantData, VCodeConstants, VCodeInst, ValueRegs,
+    non_writable_value_regs, writable_value_regs, ABICallee, BlockIndex, BlockLoweringOrder,
+    LoweredBlock, MachLabel, VCode, VCodeBuilder, VCodeConstant, VCodeConstantData, VCodeConstants,
+    VCodeInst, ValueRegs,
 };
 use crate::CodegenResult;
 use alloc::boxed::Box;
@@ -1197,7 +1198,32 @@ impl<'func, I: VCodeInst> LowerCtx for Lower<'func, I> {
 
     fn put_value_in_regs(&mut self, val: Value) -> ValueRegs<Reg> {
         let val = self.f.dfg.resolve_aliases(val);
-        log::trace!("put_value_in_reg: val {}", val);
+        log::trace!("put_value_in_regs: val {}", val);
+
+        // If the value is a constant, then (re)materialize it at each use. This
+        // lowers register pressure.
+        if let Some(c) = self
+            .f
+            .dfg
+            .value_def(val)
+            .inst()
+            .and_then(|inst| self.get_constant(inst))
+        {
+            let ty = self.f.dfg.value_type(val);
+
+            let regs = self.alloc_tmp(ty);
+            log::trace!(" -> regs {:?}", regs);
+            assert!(regs.is_valid());
+
+            let insts = I::gen_constant(regs, c.into(), ty, |ty| {
+                self.alloc_tmp(ty).only_reg().unwrap()
+            });
+            for inst in insts {
+                self.emit(inst);
+            }
+            return non_writable_value_regs(regs);
+        }
+
         let mut regs = self.value_regs[val];
         log::trace!(" -> regs {:?}", regs);
         assert!(regs.is_valid());
diff --git a/cranelift/filetests/filetests/isa/x64/i128.clif b/cranelift/filetests/filetests/isa/x64/i128.clif
index 3bb3989d9d..fb49847e92 100644
--- a/cranelift/filetests/filetests/isa/x64/i128.clif
+++ b/cranelift/filetests/filetests/isa/x64/i128.clif
@@ -697,32 +697,34 @@ block2(v6: i128):
     return v8
 
 ; check: Block 0:
-; check:  pushq   %rbp
-; nextln: movq    %rsp, %rbp
-; nextln: xorq    %rdi, %rdi
-; nextln: xorq    %rsi, %rsi
-; nextln: testb   $$1, %dl
-; nextln: jnz     label1; j label2
+; check:   pushq   %rbp
+; nextln:  movq    %rsp, %rbp
+; nextln:  testb   $$1, %dl
+; nextln:  jnz     label1; j label2
 ; check: Block 1:
-; check:  movl    $$1, %ecx
-; nextln: xorq    %rax, %rax
-; nextln: addq    %rcx, %rdi
-; nextln: adcq    %rax, %rsi
-; nextln: movq    %rdi, %rax
-; nextln: movq    %rsi, %rdx
-; nextln: movq    %rbp, %rsp
-; nextln: popq    %rbp
-; nextln: ret
+; check:   movl    $$0, %edi
+; nextln:  movl    $$0, %esi
+; nextln:  movl    $$1, %ecx
+; nextln:  movl    $$0, %eax
+; nextln:  addq    %rcx, %rdi
+; nextln:  adcq    %rax, %rsi
+; nextln:  movq    %rdi, %rax
+; nextln:  movq    %rsi, %rdx
+; nextln:  movq    %rbp, %rsp
+; nextln:  popq    %rbp
+; nextln:  ret
 ; check: Block 2:
-; check:  movl    $$2, %ecx
-; nextln: xorq    %rax, %rax
-; nextln: addq    %rcx, %rdi
-; nextln: adcq    %rax, %rsi
-; nextln: movq    %rdi, %rax
-; nextln: movq    %rsi, %rdx
-; nextln: movq    %rbp, %rsp
-; nextln: popq    %rbp
-; nextln: ret
+; check:   movl    $$0, %edi
+; nextln:  movl    $$0, %esi
+; nextln:  movl    $$2, %ecx
+; nextln:  movl    $$0, %eax
+; nextln:  addq    %rcx, %rdi
+; nextln:  adcq    %rax, %rsi
+; nextln:  movq    %rdi, %rax
+; nextln:  movq    %rsi, %rdx
+; nextln:  movq    %rbp, %rsp
+; nextln:  popq    %rbp
+; nextln:  ret
 
 }