[RISCV] Fix copy/paste mistake in Sh3Add_UWPat. #161923
Merged
+3
−11
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 pattern appears to be a copy of the pattern in Sh3AddPat but using sh3add.uw instead of sh3add. I think this is mistake and the pattern should be the equivalent of the first pattern from Sh1Add_UWPat and Sh2Add_UWPat. These classes were created to share with Andes in a788a1a, but there was so many test changes in there that we must have overlooked the changes to Zba codegen.
|
@llvm/pr-subscribers-backend-risc-v Author: Craig Topper (topperc) ChangesThis pattern appears to be a copy of the pattern in Sh3AddPat but using sh3add.uw instead of sh3add. I think this is mistake and the pattern should be the equivalent of the first pattern from Sh1Add_UWPat and Sh2Add_UWPat. These classes were created to share with Andes in a788a1a, but there was so many test changes in there that we must have overlooked the changes to Zba codegen. Full diff: https://github.com/llvm/llvm-project/pull/161923.diff 2 Files Affected:
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td b/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
index ce21d831250b2..8d9b7776ea430 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoZb.td
@@ -808,9 +808,9 @@ multiclass Sh2Add_UWPat<Instruction sh2add_uw> {
}
multiclass Sh3Add_UWPat<Instruction sh3add_uw> {
- def : Pat<(i64 (add_like_non_imm12 (and GPR:$rs1, 0xFFFFFFF8),
+ def : Pat<(i64 (add_like_non_imm12 (and (shl GPR:$rs1, (i64 3)), 0x7FFFFFFFF),
(XLenVT GPR:$rs2))),
- (sh3add_uw (XLenVT (SRLIW GPR:$rs1, 3)), GPR:$rs2)>;
+ (sh3add_uw GPR:$rs1, GPR:$rs2)>;
// Use SRLI to clear the LSBs and SHXADD_UW to mask and shift.
def : Pat<(i64 (add_like_non_imm12 (and GPR:$rs1, 0x7FFFFFFF8),
(XLenVT GPR:$rs2))),
diff --git a/llvm/test/CodeGen/RISCV/rv64zba.ll b/llvm/test/CodeGen/RISCV/rv64zba.ll
index c028d25169749..7fd76262d547a 100644
--- a/llvm/test/CodeGen/RISCV/rv64zba.ll
+++ b/llvm/test/CodeGen/RISCV/rv64zba.ll
@@ -409,15 +409,11 @@ define i64 @sh3adduw_2(i64 %0, i64 %1) {
;
; RV64ZBA-LABEL: sh3adduw_2:
; RV64ZBA: # %bb.0:
-; RV64ZBA-NEXT: slli a0, a0, 3
-; RV64ZBA-NEXT: srli a0, a0, 3
; RV64ZBA-NEXT: sh3add.uw a0, a0, a1
; RV64ZBA-NEXT: ret
;
; RV64XANDESPERF-LABEL: sh3adduw_2:
; RV64XANDESPERF: # %bb.0:
-; RV64XANDESPERF-NEXT: slli a0, a0, 3
-; RV64XANDESPERF-NEXT: srli a0, a0, 3
; RV64XANDESPERF-NEXT: nds.lea.d.ze a0, a1, a0
; RV64XANDESPERF-NEXT: ret
%3 = shl i64 %0, 3
@@ -436,15 +432,11 @@ define i64 @sh3adduw_3(i64 %0, i64 %1) {
;
; RV64ZBA-LABEL: sh3adduw_3:
; RV64ZBA: # %bb.0:
-; RV64ZBA-NEXT: slli a0, a0, 3
-; RV64ZBA-NEXT: srli a0, a0, 3
; RV64ZBA-NEXT: sh3add.uw a0, a0, a1
; RV64ZBA-NEXT: ret
;
; RV64XANDESPERF-LABEL: sh3adduw_3:
; RV64XANDESPERF: # %bb.0:
-; RV64XANDESPERF-NEXT: slli a0, a0, 3
-; RV64XANDESPERF-NEXT: srli a0, a0, 3
; RV64XANDESPERF-NEXT: nds.lea.d.ze a0, a1, a0
; RV64XANDESPERF-NEXT: ret
%3 = shl i64 %0, 3
@@ -2681,7 +2673,7 @@ define i64 @srliw_3_sh3add(ptr %0, i32 signext %1) {
; RV64ZBA-LABEL: srliw_3_sh3add:
; RV64ZBA: # %bb.0:
; RV64ZBA-NEXT: srliw a1, a1, 3
-; RV64ZBA-NEXT: sh3add.uw a0, a1, a0
+; RV64ZBA-NEXT: sh3add a0, a1, a0
; RV64ZBA-NEXT: ld a0, 0(a0)
; RV64ZBA-NEXT: ret
;
|
topperc
commented
Oct 3, 2025
| ; RV64ZBA-LABEL: srliw_3_sh3add: | ||
| ; RV64ZBA: # %bb.0: | ||
| ; RV64ZBA-NEXT: srliw a1, a1, 3 | ||
| ; RV64ZBA-NEXT: sh3add.uw a0, a1, a0 |
There was a problem hiding this comment.
srliw always puts 0s in the upper 32 bits so the .uw here isn't necessary. So I don't think the mistake can cause a miscompile.
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/141/builds/12068 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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 pattern appears to be a copy of the pattern in Sh3AddPat but using sh3add.uw instead of sh3add. I think this is a mistake and the pattern should be the equivalent of the first pattern from Sh1Add_UWPat and Sh2Add_UWPat.
These classes were created to share with Andes in a788a1a, but there was so many test changes in there that we must have overlooked the changes to Zba codegen.