[RISCV] Use LBU for anyext i8 atomic_load. #161588
Merged
+27
−25
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This matches what we do for regular i8 extload due to the lack of c.lb in Zbc. This only affects global isel because SelectionDAG won't create an anyext i8 atomic_load today.
|
@llvm/pr-subscribers-backend-risc-v @llvm/pr-subscribers-llvm-globalisel Author: Craig Topper (topperc) ChangesThis matches what we do for regular i8 extload due to the lack of c.lb in Zbc. This only affects global isel because SelectionDAG won't create an anyext i8 atomic_load today. Full diff: https://github.com/llvm/llvm-project/pull/161588.diff 3 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVGISel.td b/llvm/lib/Target/RISCV/RISCVGISel.td
index 2e5f30f8fe35a..19d5aff023d53 100644
--- a/llvm/lib/Target/RISCV/RISCVGISel.td
+++ b/llvm/lib/Target/RISCV/RISCVGISel.td
@@ -109,8 +109,9 @@ def : LdPat<extloadi8, LBU, i16>; // Prefer unsigned due to no c.lb in Zcb.
def : StPat<truncstorei8, SB, GPR, i16>;
let Predicates = [HasAtomicLdSt] in {
- def : LdPat<atomic_load_aext_8, LB, i16>;
- def : LdPat<atomic_load_nonext_16, LH, i16>;
+ // Prefer unsigned due to no c.lb in Zcb.
+ def : LdPat<atomic_load_aext_8, LBU, i16>;
+ def : LdPat<atomic_load_nonext_16, LH, i16>;
def : StPat<atomic_store_8, SB, GPR, i16>;
def : StPat<atomic_store_16, SH, GPR, i16>;
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoA.td b/llvm/lib/Target/RISCV/RISCVInstrInfoA.td
index 59f5aebf658d8..99992d196b43d 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoA.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoA.td
@@ -165,10 +165,11 @@ class seq_cst_store<PatFrag base>
// any ordering. This is necessary because AtomicExpandPass has added fences to
// atomic load/stores and changed them to unordered ones.
let Predicates = [HasAtomicLdSt] in {
- def : LdPat<relaxed_load<atomic_load_asext_8>, LB>;
+ // Use unsigned for aext due to no c.lb in Zcb.
+ def : LdPat<relaxed_load<atomic_load_sext_8>, LB>;
+ def : LdPat<relaxed_load<atomic_load_azext_8>, LBU>;
def : LdPat<relaxed_load<atomic_load_asext_16>, LH>;
- def : LdPat<relaxed_load<atomic_load_zext_8>, LBU>;
- def : LdPat<relaxed_load<atomic_load_zext_16>, LHU>;
+ def : LdPat<relaxed_load<atomic_load_zext_16>, LHU>;
def : StPat<relaxed_store<atomic_store_8>, SB, GPR, XLenVT>;
def : StPat<relaxed_store<atomic_store_16>, SH, GPR, XLenVT>;
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/atomic-load-store.ll b/llvm/test/CodeGen/RISCV/GlobalISel/atomic-load-store.ll
index 9a1ed8f115b35..1d5d918422b28 100644
--- a/llvm/test/CodeGen/RISCV/GlobalISel/atomic-load-store.ll
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/atomic-load-store.ll
@@ -37,7 +37,7 @@ define i8 @atomic_load_i8_unordered(ptr %a) nounwind {
;
; RV32IA-LABEL: atomic_load_i8_unordered:
; RV32IA: # %bb.0:
-; RV32IA-NEXT: lb a0, 0(a0)
+; RV32IA-NEXT: lbu a0, 0(a0)
; RV32IA-NEXT: ret
;
; RV64I-LABEL: atomic_load_i8_unordered:
@@ -52,7 +52,7 @@ define i8 @atomic_load_i8_unordered(ptr %a) nounwind {
;
; RV64IA-LABEL: atomic_load_i8_unordered:
; RV64IA: # %bb.0:
-; RV64IA-NEXT: lb a0, 0(a0)
+; RV64IA-NEXT: lbu a0, 0(a0)
; RV64IA-NEXT: ret
%1 = load atomic i8, ptr %a unordered, align 1
ret i8 %1
@@ -71,7 +71,7 @@ define i8 @atomic_load_i8_monotonic(ptr %a) nounwind {
;
; RV32IA-LABEL: atomic_load_i8_monotonic:
; RV32IA: # %bb.0:
-; RV32IA-NEXT: lb a0, 0(a0)
+; RV32IA-NEXT: lbu a0, 0(a0)
; RV32IA-NEXT: ret
;
; RV64I-LABEL: atomic_load_i8_monotonic:
@@ -86,7 +86,7 @@ define i8 @atomic_load_i8_monotonic(ptr %a) nounwind {
;
; RV64IA-LABEL: atomic_load_i8_monotonic:
; RV64IA: # %bb.0:
-; RV64IA-NEXT: lb a0, 0(a0)
+; RV64IA-NEXT: lbu a0, 0(a0)
; RV64IA-NEXT: ret
%1 = load atomic i8, ptr %a monotonic, align 1
ret i8 %1
@@ -105,13 +105,13 @@ define i8 @atomic_load_i8_acquire(ptr %a) nounwind {
;
; RV32IA-WMO-LABEL: atomic_load_i8_acquire:
; RV32IA-WMO: # %bb.0:
-; RV32IA-WMO-NEXT: lb a0, 0(a0)
+; RV32IA-WMO-NEXT: lbu a0, 0(a0)
; RV32IA-WMO-NEXT: fence r, rw
; RV32IA-WMO-NEXT: ret
;
; RV32IA-TSO-LABEL: atomic_load_i8_acquire:
; RV32IA-TSO: # %bb.0:
-; RV32IA-TSO-NEXT: lb a0, 0(a0)
+; RV32IA-TSO-NEXT: lbu a0, 0(a0)
; RV32IA-TSO-NEXT: ret
;
; RV64I-LABEL: atomic_load_i8_acquire:
@@ -126,35 +126,35 @@ define i8 @atomic_load_i8_acquire(ptr %a) nounwind {
;
; RV64IA-WMO-LABEL: atomic_load_i8_acquire:
; RV64IA-WMO: # %bb.0:
-; RV64IA-WMO-NEXT: lb a0, 0(a0)
+; RV64IA-WMO-NEXT: lbu a0, 0(a0)
; RV64IA-WMO-NEXT: fence r, rw
; RV64IA-WMO-NEXT: ret
;
; RV64IA-TSO-LABEL: atomic_load_i8_acquire:
; RV64IA-TSO: # %bb.0:
-; RV64IA-TSO-NEXT: lb a0, 0(a0)
+; RV64IA-TSO-NEXT: lbu a0, 0(a0)
; RV64IA-TSO-NEXT: ret
;
; RV32IA-WMO-TRAILING-FENCE-LABEL: atomic_load_i8_acquire:
; RV32IA-WMO-TRAILING-FENCE: # %bb.0:
-; RV32IA-WMO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV32IA-WMO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV32IA-WMO-TRAILING-FENCE-NEXT: fence r, rw
; RV32IA-WMO-TRAILING-FENCE-NEXT: ret
;
; RV32IA-TSO-TRAILING-FENCE-LABEL: atomic_load_i8_acquire:
; RV32IA-TSO-TRAILING-FENCE: # %bb.0:
-; RV32IA-TSO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV32IA-TSO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV32IA-TSO-TRAILING-FENCE-NEXT: ret
;
; RV64IA-WMO-TRAILING-FENCE-LABEL: atomic_load_i8_acquire:
; RV64IA-WMO-TRAILING-FENCE: # %bb.0:
-; RV64IA-WMO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV64IA-WMO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV64IA-WMO-TRAILING-FENCE-NEXT: fence r, rw
; RV64IA-WMO-TRAILING-FENCE-NEXT: ret
;
; RV64IA-TSO-TRAILING-FENCE-LABEL: atomic_load_i8_acquire:
; RV64IA-TSO-TRAILING-FENCE: # %bb.0:
-; RV64IA-TSO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV64IA-TSO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV64IA-TSO-TRAILING-FENCE-NEXT: ret
%1 = load atomic i8, ptr %a acquire, align 1
ret i8 %1
@@ -174,14 +174,14 @@ define i8 @atomic_load_i8_seq_cst(ptr %a) nounwind {
; RV32IA-WMO-LABEL: atomic_load_i8_seq_cst:
; RV32IA-WMO: # %bb.0:
; RV32IA-WMO-NEXT: fence rw, rw
-; RV32IA-WMO-NEXT: lb a0, 0(a0)
+; RV32IA-WMO-NEXT: lbu a0, 0(a0)
; RV32IA-WMO-NEXT: fence r, rw
; RV32IA-WMO-NEXT: ret
;
; RV32IA-TSO-LABEL: atomic_load_i8_seq_cst:
; RV32IA-TSO: # %bb.0:
; RV32IA-TSO-NEXT: fence rw, rw
-; RV32IA-TSO-NEXT: lb a0, 0(a0)
+; RV32IA-TSO-NEXT: lbu a0, 0(a0)
; RV32IA-TSO-NEXT: ret
;
; RV64I-LABEL: atomic_load_i8_seq_cst:
@@ -197,40 +197,40 @@ define i8 @atomic_load_i8_seq_cst(ptr %a) nounwind {
; RV64IA-WMO-LABEL: atomic_load_i8_seq_cst:
; RV64IA-WMO: # %bb.0:
; RV64IA-WMO-NEXT: fence rw, rw
-; RV64IA-WMO-NEXT: lb a0, 0(a0)
+; RV64IA-WMO-NEXT: lbu a0, 0(a0)
; RV64IA-WMO-NEXT: fence r, rw
; RV64IA-WMO-NEXT: ret
;
; RV64IA-TSO-LABEL: atomic_load_i8_seq_cst:
; RV64IA-TSO: # %bb.0:
; RV64IA-TSO-NEXT: fence rw, rw
-; RV64IA-TSO-NEXT: lb a0, 0(a0)
+; RV64IA-TSO-NEXT: lbu a0, 0(a0)
; RV64IA-TSO-NEXT: ret
;
; RV32IA-WMO-TRAILING-FENCE-LABEL: atomic_load_i8_seq_cst:
; RV32IA-WMO-TRAILING-FENCE: # %bb.0:
; RV32IA-WMO-TRAILING-FENCE-NEXT: fence rw, rw
-; RV32IA-WMO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV32IA-WMO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV32IA-WMO-TRAILING-FENCE-NEXT: fence r, rw
; RV32IA-WMO-TRAILING-FENCE-NEXT: ret
;
; RV32IA-TSO-TRAILING-FENCE-LABEL: atomic_load_i8_seq_cst:
; RV32IA-TSO-TRAILING-FENCE: # %bb.0:
; RV32IA-TSO-TRAILING-FENCE-NEXT: fence rw, rw
-; RV32IA-TSO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV32IA-TSO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV32IA-TSO-TRAILING-FENCE-NEXT: ret
;
; RV64IA-WMO-TRAILING-FENCE-LABEL: atomic_load_i8_seq_cst:
; RV64IA-WMO-TRAILING-FENCE: # %bb.0:
; RV64IA-WMO-TRAILING-FENCE-NEXT: fence rw, rw
-; RV64IA-WMO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV64IA-WMO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV64IA-WMO-TRAILING-FENCE-NEXT: fence r, rw
; RV64IA-WMO-TRAILING-FENCE-NEXT: ret
;
; RV64IA-TSO-TRAILING-FENCE-LABEL: atomic_load_i8_seq_cst:
; RV64IA-TSO-TRAILING-FENCE: # %bb.0:
; RV64IA-TSO-TRAILING-FENCE-NEXT: fence rw, rw
-; RV64IA-TSO-TRAILING-FENCE-NEXT: lb a0, 0(a0)
+; RV64IA-TSO-TRAILING-FENCE-NEXT: lbu a0, 0(a0)
; RV64IA-TSO-TRAILING-FENCE-NEXT: ret
%1 = load atomic i8, ptr %a seq_cst, align 1
ret i8 %1
|
mahesh-attarde
pushed a commit
to mahesh-attarde/llvm-project
that referenced
this pull request
Oct 3, 2025
This matches what we do for regular i8 extload due to the lack of c.lb in Zbc. This only affects global isel because SelectionDAG won't create an anyext i8 atomic_load today.
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This matches what we do for regular i8 extload due to the lack of c.lb in Zbc.
This only affects global isel because SelectionDAG won't create an anyext i8 atomic_load today.