Allow shifting instead of masks if the types are legal. #162732
Closed
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
|
@llvm/pr-subscribers-backend-amdgpu Author: AZero13 (AZero13) ChangesFull diff: https://github.com/llvm/llvm-project/pull/162732.diff 3 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index 88691b931a8d8..9865143edbdfb 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -839,8 +839,11 @@ class LLVM_ABI TargetLoweringBase {
/// Return true if the variant with 2 variable shifts is preferred.
/// Return false if there is no preference.
virtual bool shouldFoldMaskToVariableShiftPair(SDValue X) const {
- // By default, let's assume that no one prefers shifts.
- return false;
+ // By default, let's assume that no one prefers shifts for vectors
+ EVT VT = X.getValueType();
+
+ // Prefer shifts for legal types
+ return isOperationLegal(ISD::SHL, VT);
}
/// Return true if it is profitable to fold a pair of shifts into a mask.
diff --git a/llvm/test/CodeGen/AMDGPU/extract-lowbits.ll b/llvm/test/CodeGen/AMDGPU/extract-lowbits.ll
index 89bd5f1dc7969..59bc1dd019de3 100644
--- a/llvm/test/CodeGen/AMDGPU/extract-lowbits.ll
+++ b/llvm/test/CodeGen/AMDGPU/extract-lowbits.ll
@@ -103,16 +103,16 @@ define i32 @bzhi32_c0(i32 %val, i32 %numlowbits) nounwind {
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_sub_i32_e32 v1, vcc, 32, v1
-; SI-NEXT: v_lshr_b32_e32 v1, -1, v1
-; SI-NEXT: v_and_b32_e32 v0, v1, v0
+; SI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; SI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; SI-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: bzhi32_c0:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_sub_u32_e32 v1, vcc, 32, v1
-; VI-NEXT: v_lshrrev_b32_e64 v1, v1, -1
-; VI-NEXT: v_and_b32_e32 v0, v1, v0
+; VI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; VI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; VI-NEXT: s_setpc_b64 s[30:31]
%numhighbits = sub i32 32, %numlowbits
%mask = lshr i32 -1, %numhighbits
@@ -121,12 +121,23 @@ define i32 @bzhi32_c0(i32 %val, i32 %numlowbits) nounwind {
}
define i32 @bzhi32_c0_clamp(i32 %val, i32 %numlowbits) nounwind {
-; GCN-LABEL: bzhi32_c0_clamp:
-; GCN: ; %bb.0:
-; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: v_and_b32_e32 v1, 31, v1
-; GCN-NEXT: v_bfe_u32 v0, v0, 0, v1
-; GCN-NEXT: s_setpc_b64 s[30:31]
+; SI-LABEL: bzhi32_c0_clamp:
+; SI: ; %bb.0:
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 31, v1
+; SI-NEXT: v_sub_i32_e32 v1, vcc, 32, v1
+; SI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; SI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
+; SI-NEXT: s_setpc_b64 s[30:31]
+;
+; VI-LABEL: bzhi32_c0_clamp:
+; VI: ; %bb.0:
+; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT: v_and_b32_e32 v1, 31, v1
+; VI-NEXT: v_sub_u32_e32 v1, vcc, 32, v1
+; VI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; VI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
+; VI-NEXT: s_setpc_b64 s[30:31]
%low5bits = and i32 %numlowbits, 31
%numhighbits = sub i32 32, %low5bits
%mask = lshr i32 -1, %numhighbits
@@ -139,16 +150,16 @@ define i32 @bzhi32_c1_indexzext(i32 %val, i8 %numlowbits) nounwind {
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_sub_i32_e32 v1, vcc, 32, v1
-; SI-NEXT: v_lshr_b32_e32 v1, -1, v1
-; SI-NEXT: v_and_b32_e32 v0, v1, v0
+; SI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; SI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; SI-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: bzhi32_c1_indexzext:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_sub_u16_e32 v1, 32, v1
-; VI-NEXT: v_lshrrev_b32_e64 v1, v1, -1
-; VI-NEXT: v_and_b32_e32 v0, v1, v0
+; VI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; VI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; VI-NEXT: s_setpc_b64 s[30:31]
%numhighbits = sub i8 32, %numlowbits
%sh_prom = zext i8 %numhighbits to i32
@@ -162,16 +173,16 @@ define i32 @bzhi32_c4_commutative(i32 %val, i32 %numlowbits) nounwind {
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_sub_i32_e32 v1, vcc, 32, v1
-; SI-NEXT: v_lshr_b32_e32 v1, -1, v1
-; SI-NEXT: v_and_b32_e32 v0, v0, v1
+; SI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; SI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; SI-NEXT: s_setpc_b64 s[30:31]
;
; VI-LABEL: bzhi32_c4_commutative:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_sub_u32_e32 v1, vcc, 32, v1
-; VI-NEXT: v_lshrrev_b32_e64 v1, v1, -1
-; VI-NEXT: v_and_b32_e32 v0, v0, v1
+; VI-NEXT: v_lshlrev_b32_e32 v0, v1, v0
+; VI-NEXT: v_lshrrev_b32_e32 v0, v1, v0
; VI-NEXT: s_setpc_b64 s[30:31]
%numhighbits = sub i32 32, %numlowbits
%mask = lshr i32 -1, %numhighbits
diff --git a/llvm/test/CodeGen/AMDGPU/r600.extract-lowbits.ll b/llvm/test/CodeGen/AMDGPU/r600.extract-lowbits.ll
index 5b21a3652d11c..1055fc101a6c5 100644
--- a/llvm/test/CodeGen/AMDGPU/r600.extract-lowbits.ll
+++ b/llvm/test/CodeGen/AMDGPU/r600.extract-lowbits.ll
@@ -266,7 +266,7 @@ define amdgpu_kernel void @bzhi32_c1_indexzext(i32 %val, i8 %numlowbits, ptr add
; EG: ; %bb.0:
; EG-NEXT: ALU 0, @8, KC0[], KC1[]
; EG-NEXT: TEX 0 @6
-; EG-NEXT: ALU 8, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT: ALU 7, @9, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
@@ -279,9 +279,8 @@ define amdgpu_kernel void @bzhi32_c1_indexzext(i32 %val, i8 %numlowbits, ptr add
; EG-NEXT: 32(4.484155e-44), 0(0.000000e+00)
; EG-NEXT: AND_INT * T0.W, PV.W, literal.x,
; EG-NEXT: 255(3.573311e-43), 0(0.000000e+00)
-; EG-NEXT: LSHR * T0.W, literal.x, PV.W,
-; EG-NEXT: -1(nan), 0(0.000000e+00)
-; EG-NEXT: AND_INT T0.X, PV.W, KC0[2].Y,
+; EG-NEXT: LSHL * T1.W, KC0[2].Y, PV.W,
+; EG-NEXT: LSHR T0.X, PV.W, T0.W,
; EG-NEXT: LSHR * T1.X, KC0[2].W, literal.x,
; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
;
@@ -289,7 +288,7 @@ define amdgpu_kernel void @bzhi32_c1_indexzext(i32 %val, i8 %numlowbits, ptr add
; CM: ; %bb.0:
; CM-NEXT: ALU 0, @8, KC0[], KC1[]
; CM-NEXT: TEX 0 @6
-; CM-NEXT: ALU 8, @9, KC0[CB0:0-32], KC1[]
+; CM-NEXT: ALU 7, @9, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
@@ -302,9 +301,8 @@ define amdgpu_kernel void @bzhi32_c1_indexzext(i32 %val, i8 %numlowbits, ptr add
; CM-NEXT: 32(4.484155e-44), 0(0.000000e+00)
; CM-NEXT: AND_INT * T0.W, PV.W, literal.x,
; CM-NEXT: 255(3.573311e-43), 0(0.000000e+00)
-; CM-NEXT: LSHR * T0.W, literal.x, PV.W,
-; CM-NEXT: -1(nan), 0(0.000000e+00)
-; CM-NEXT: AND_INT * T0.X, PV.W, KC0[2].Y,
+; CM-NEXT: LSHL * T1.W, KC0[2].Y, PV.W,
+; CM-NEXT: LSHR * T0.X, PV.W, T0.W,
; CM-NEXT: LSHR * T1.X, KC0[2].W, literal.x,
; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
%numhighbits = sub i8 32, %numlowbits
|
arsenm
reviewed
Oct 10, 2025
Comment on lines
+126
to
+130
| ; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | ||
| ; SI-NEXT: v_and_b32_e32 v1, 31, v1 | ||
| ; SI-NEXT: v_sub_i32_e32 v1, vcc, 32, v1 | ||
| ; SI-NEXT: v_lshlrev_b32_e32 v0, v1, v0 | ||
| ; SI-NEXT: v_lshrrev_b32_e32 v0, v1, v0 |
There was a problem hiding this comment.
Regression, broke a BFE pattern?
Comment on lines
+106
to
+107
| ; SI-NEXT: v_lshlrev_b32_e32 v0, v1, v0 | ||
| ; SI-NEXT: v_lshrrev_b32_e32 v0, v1, v0 |
There was a problem hiding this comment.
This is a neutral change at best
|
Why? I don't see any tests that show a benefit. I guess it could conceivably reduce register pressure, since the mask has to be materialized in a register, but that would only help if the shift amount has more uses, and there aren't any tests cases for that. |
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.
No description provided.