Skip to content

Commit 3c62e28

Browse files
committed
- Address reviewer comments
1 parent 75b5ef0 commit 3c62e28

File tree

2 files changed

+40
-3
lines changed

2 files changed

+40
-3
lines changed

llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -46,7 +46,7 @@ static cl::opt<unsigned>
4646
HardClauseLengthLimit("amdgpu-hard-clause-length-limit",
4747
cl::desc("Maximum number of memory instructions to "
4848
"place in the same hard clause"),
49-
cl::init(255), cl::Hidden);
49+
cl::Hidden);
5050

5151
namespace {
5252

@@ -196,11 +196,11 @@ class SIInsertHardClauses {
196196
return false;
197197

198198
unsigned MaxClauseLength = MF.getFunction().getFnAttributeAsParsedInteger(
199-
"amdgpu-hard-clause-length-limit", HardClauseLengthLimit);
199+
"amdgpu-hard-clause-length-limit", 255);
200200
if (HardClauseLengthLimit.getNumOccurrences())
201201
MaxClauseLength = HardClauseLengthLimit;
202202
MaxClauseLength = std::min(MaxClauseLength, ST->maxHardClauseLength());
203-
if (!MaxClauseLength)
203+
if (MaxClauseLength <= 1)
204204
return false;
205205

206206
const SIInstrInfo *SII = ST->getInstrInfo();

llvm/test/CodeGen/AMDGPU/hard-clause-limit.mir

Lines changed: 37 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3,6 +3,7 @@
33
# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -run-pass si-insert-hard-clauses -amdgpu-hard-clause-length-limit=16 %s -o - | FileCheck %s -check-prefix=GFX11-16
44
# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -run-pass si-insert-hard-clauses -amdgpu-hard-clause-length-limit=10 %s -o - | FileCheck %s -check-prefix=GFX11-10
55
# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -run-pass si-insert-hard-clauses -amdgpu-hard-clause-length-limit=4 %s -o - | FileCheck %s -check-prefix=GFX11-4
6+
# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -run-pass si-insert-hard-clauses -amdgpu-hard-clause-length-limit=1 %s -o - | FileCheck %s -check-prefix=GFX11-1
67

78
---
89
name: long_clause
@@ -198,6 +199,42 @@ body: |
198199
; GFX11-4-NEXT: $vgpr31 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 124, 0, 0, implicit $exec
199200
; GFX11-4-NEXT: $vgpr32 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 128, 0, 0, implicit $exec
200201
; GFX11-4-NEXT: }
202+
;
203+
; GFX11-1-LABEL: name: long_clause
204+
; GFX11-1: liveins: $sgpr0_sgpr1_sgpr2_sgpr3, $vgpr0
205+
; GFX11-1-NEXT: {{ $}}
206+
; GFX11-1-NEXT: $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 4, 0, 0, implicit $exec
207+
; GFX11-1-NEXT: $vgpr2 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 8, 0, 0, implicit $exec
208+
; GFX11-1-NEXT: $vgpr3 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 12, 0, 0, implicit $exec
209+
; GFX11-1-NEXT: $vgpr4 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 16, 0, 0, implicit $exec
210+
; GFX11-1-NEXT: $vgpr5 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 20, 0, 0, implicit $exec
211+
; GFX11-1-NEXT: $vgpr6 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 24, 0, 0, implicit $exec
212+
; GFX11-1-NEXT: $vgpr7 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 28, 0, 0, implicit $exec
213+
; GFX11-1-NEXT: $vgpr8 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 32, 0, 0, implicit $exec
214+
; GFX11-1-NEXT: $vgpr9 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 36, 0, 0, implicit $exec
215+
; GFX11-1-NEXT: $vgpr10 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 40, 0, 0, implicit $exec
216+
; GFX11-1-NEXT: $vgpr11 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 44, 0, 0, implicit $exec
217+
; GFX11-1-NEXT: $vgpr12 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 48, 0, 0, implicit $exec
218+
; GFX11-1-NEXT: $vgpr13 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 52, 0, 0, implicit $exec
219+
; GFX11-1-NEXT: $vgpr14 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 56, 0, 0, implicit $exec
220+
; GFX11-1-NEXT: $vgpr15 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 60, 0, 0, implicit $exec
221+
; GFX11-1-NEXT: $vgpr16 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 64, 0, 0, implicit $exec
222+
; GFX11-1-NEXT: $vgpr17 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 68, 0, 0, implicit $exec
223+
; GFX11-1-NEXT: $vgpr18 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 72, 0, 0, implicit $exec
224+
; GFX11-1-NEXT: $vgpr19 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 76, 0, 0, implicit $exec
225+
; GFX11-1-NEXT: $vgpr20 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 80, 0, 0, implicit $exec
226+
; GFX11-1-NEXT: $vgpr21 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 84, 0, 0, implicit $exec
227+
; GFX11-1-NEXT: $vgpr22 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 88, 0, 0, implicit $exec
228+
; GFX11-1-NEXT: $vgpr23 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 92, 0, 0, implicit $exec
229+
; GFX11-1-NEXT: $vgpr24 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 96, 0, 0, implicit $exec
230+
; GFX11-1-NEXT: $vgpr25 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 100, 0, 0, implicit $exec
231+
; GFX11-1-NEXT: $vgpr26 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 104, 0, 0, implicit $exec
232+
; GFX11-1-NEXT: $vgpr27 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 108, 0, 0, implicit $exec
233+
; GFX11-1-NEXT: $vgpr28 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 112, 0, 0, implicit $exec
234+
; GFX11-1-NEXT: $vgpr29 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 116, 0, 0, implicit $exec
235+
; GFX11-1-NEXT: $vgpr30 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 120, 0, 0, implicit $exec
236+
; GFX11-1-NEXT: $vgpr31 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 124, 0, 0, implicit $exec
237+
; GFX11-1-NEXT: $vgpr32 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 128, 0, 0, implicit $exec
201238
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 4, 0, 0, implicit $exec
202239
$vgpr2 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 8, 0, 0, implicit $exec
203240
$vgpr3 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 12, 0, 0, implicit $exec

0 commit comments

Comments
 (0)