[CodeGen] Allow -enable-ext-tsp-block-placement and -apply-ext-tsp-for-size passed together

[SharonXSharon]

Currently, the asserts fires when both UseExtTspForPerf and UseExtTspForSize are true on a given function.

Ideally, we should allow -enable-ext-tsp-block-placement and -apply-ext-tsp-for-size passed together, meaning run the block placement for performance on hot functions, while run the placement for size on cold functions.

The diff makes UseExtTspForPerf and UseExtTspForSize mutually exclusive per-function: functions with the OptForSize attribute use ext-tsp block placement for size, while the others use ext-tsp block placement for perf.

[llvmbot]

@llvm/pr-subscribers-backend-x86

Author: None (SharonXSharon)

Changes

Currently, the asserts fires when both UseExtTspForPerf and UseExtTspForSize are true on a given function.

Ideally, we should allow -enable-ext-tsp-block-placement and -apply-ext-tsp-for-size passed together, meaning run the block placement for performance on hot functions, while run the placement for size on cold functions.

The diff makes UseExtTspForPerf and UseExtTspForSize mutually exclusive per-function: functions with the OptForSize attribute use ext-tsp block placement for size, while the others use ext-tsp block placement for perf.

---

Full diff: https://github.com/llvm/llvm-project/pull/183642.diff

2 Files Affected:

• (modified) llvm/lib/CodeGen/MachineBlockPlacement.cpp (+2-2)
• (added) llvm/test/CodeGen/X86/code_placement_ext_tsp_size_and_perf.ll (+91)

diff --git a/llvm/lib/CodeGen/MachineBlockPlacement.cpp b/llvm/lib/CodeGen/MachineBlockPlacement.cpp
index 625fba36b3d3a..f0e5e7efd132c 100644
--- a/llvm/lib/CodeGen/MachineBlockPlacement.cpp
+++ b/llvm/lib/CodeGen/MachineBlockPlacement.cpp
@@ -3598,10 +3598,10 @@ bool MachineBlockPlacement::run(MachineFunction &MF) {
   bool UseExtTspForPerf = false;
   bool UseExtTspForSize = false;
   if (3 <= MF.size() && MF.size() <= ExtTspBlockPlacementMaxBlocks) {
+    UseExtTspForSize = OptForSize && ApplyExtTspForSize;
     UseExtTspForPerf =
-        EnableExtTspBlockPlacement &&
+        !UseExtTspForSize && EnableExtTspBlockPlacement &&
         (ApplyExtTspWithoutProfile || MF.getFunction().hasProfileData());
-    UseExtTspForSize = OptForSize && ApplyExtTspForSize;
   }
 
   // Apply tail duplication.
diff --git a/llvm/test/CodeGen/X86/code_placement_ext_tsp_size_and_perf.ll b/llvm/test/CodeGen/X86/code_placement_ext_tsp_size_and_perf.ll
new file mode 100644
index 0000000000000..0980bba58f018
--- /dev/null
+++ b/llvm/test/CodeGen/X86/code_placement_ext_tsp_size_and_perf.ll
@@ -0,0 +1,91 @@
+; RUN: llc -mcpu=corei7 -mtriple=x86_64-linux -verify-machineinstrs -enable-ext-tsp-block-placement -apply-ext-tsp-for-size < %s | FileCheck %s
+
+;; Cold function with optsize: should use ext-tsp for size.
+;; The size-optimized layout keeps the original order (b0, b1, b2) since it
+;; minimizes code size by avoiding extra jumps.
+define void @cold_func() optsize !prof !2 {
+;
+; +-----+
+; | b0  | -+
+; +-----+  |
+;   |      |
+;   | 10   |
+;   v      |
+; +-----+  |
+; | b1  |  | 10000
+; +-----+  |
+;   |      |
+;   | 10   |
+;   v      |
+; +-----+  |
+; | b2  | <+
+; +-----+
+;
+; CHECK-LABEL: cold_func:
+; CHECK: %b0
+; CHECK: %b1
+; CHECK: %b2
+
+b0:
+  %call = call zeroext i1 @a()
+  br i1 %call, label %b1, label %b2, !prof !1
+
+b1:
+  call void @d()
+  call void @d()
+  call void @d()
+  br label %b2
+
+b2:
+  call void @e()
+  ret void
+}
+
+;; Hot function without optsize: should use ext-tsp for perf.
+;; The perf-optimized layout reorders blocks to place the most likely successor
+;; (b2) right after b0, giving b0 -> b2 -> b1.
+define void @hot_func() !prof !3 {
+;
+; +-----+
+; | b0  | -+
+; +-----+  |
+;   |      |
+;   | 40   |
+;   v      |
+; +-----+  |
+; | b1  |  | 100
+; +-----+  |
+;   |      |
+;   | 40   |
+;   v      |
+; +-----+  |
+; | b2  | <+
+; +-----+
+;
+; CHECK-LABEL: hot_func:
+; CHECK: %b0
+; CHECK: %b2
+; CHECK: %b1
+
+b0:
+  %call = call zeroext i1 @a()
+  br i1 %call, label %b1, label %b2, !prof !1
+
+b1:
+  call void @d()
+  call void @d()
+  call void @d()
+  br label %b2
+
+b2:
+  call void @e()
+  ret void
+}
+
+declare zeroext i1 @a()
+declare void @d()
+declare void @e()
+
+!1 = !{!"branch_weights", i32 10, i32 10000}
+!2 = !{!"function_entry_count", i64 1}
+!3 = !{!"function_entry_count", i64 2200}