This is an archive of the discontinued LLVM Phabricator instance.

Differential D19637

AMDGPU: Emit error if too much LDS is used
ClosedPublic

Authored by arsenm on Apr 27 2016, 3:46 PM.

Details

Reviewers
• tstellarAMD

Diff Detail

Event Timeline

arsenm updated this revision to Diff 55344.Apr 27 2016, 3:46 PM
arsenm retitled this revision from to AMDGPU: Emit error if too much LDS is used.
arsenm updated this object.
arsenm added a reviewer: • tstellarAMD.
arsenm added a subscriber: llvm-commits.
Herald added a subscriber: arsenm. · View Herald TranscriptApr 27 2016, 3:46 PM
arsenm updated this revision to Diff 55353.Apr 27 2016, 4:21 PM

Test fix

• tstellarAMD accepted this revision.Apr 28 2016, 12:14 PM
• tstellarAMD edited edge metadata.

LGTM.

This revision is now accepted and ready to land.Apr 28 2016, 12:14 PM
arsenm closed this revision.Apr 28 2016, 12:43 PM

r267922

Revision Contents

PathSize
lib/
Target/
AMDGPU/
5 lines
test/
CodeGen/
AMDGPU/
4 lines
14 lines

Diff 55353

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

Show First 20 Lines • Show All 498 Lines • ▼ Show 20 Lines if (STM.hasSGPRInitBug()) {
ProgInfo.NumSGPR = AMDGPUSubtarget::FIXED_SGPR_COUNT_FOR_INIT_BUG; ProgInfo.NumSGPR = AMDGPUSubtarget::FIXED_SGPR_COUNT_FOR_INIT_BUG;
} }
if (MFI->NumUserSGPRs > STM.getMaxNumUserSGPRs()) { if (MFI->NumUserSGPRs > STM.getMaxNumUserSGPRs()) {
LLVMContext &Ctx = MF.getFunction()->getContext(); LLVMContext &Ctx = MF.getFunction()->getContext();
Ctx.emitError("too many user SGPRs used"); Ctx.emitError("too many user SGPRs used");
} }
if (MFI->LDSSize > static_cast<unsigned>(STM.getLocalMemorySize())) {
LLVMContext &Ctx = MF.getFunction()->getContext();
Ctx.emitError("LDS size exceeds device maximum");
}
ProgInfo.VGPRBlocks = (ProgInfo.NumVGPR - 1) / 4; ProgInfo.VGPRBlocks = (ProgInfo.NumVGPR - 1) / 4;
ProgInfo.SGPRBlocks = (ProgInfo.NumSGPR - 1) / 8; ProgInfo.SGPRBlocks = (ProgInfo.NumSGPR - 1) / 8;
// Set the value to initialize FP_ROUND and FP_DENORM parts of the mode // Set the value to initialize FP_ROUND and FP_DENORM parts of the mode
// register. // register.
ProgInfo.FloatMode = getFPMode(MF); ProgInfo.FloatMode = getFPMode(MF);
ProgInfo.IEEEMode = 0; ProgInfo.IEEEMode = 0;
▲ Show 20 Lines • Show All 223 Lines • Show Last 20 Lines

test/CodeGen/AMDGPU/32-bit-local-address-space.ll

Show First 20 Lines • Show All 85 Lines • ▼ Show 20 Lines
define void @infer_ptr_alignment_global_offset(float addrspace(1)* %out, i32 %tid) { define void @infer_ptr_alignment_global_offset(float addrspace(1)* %out, i32 %tid) {
%val = load float, float addrspace(3)* @g_lds %val = load float, float addrspace(3)* @g_lds
store float %val, float addrspace(1)* %out store float %val, float addrspace(1)* %out
ret void ret void
} }
@ptr = addrspace(3) global i32 addrspace(3)* undef @ptr = addrspace(3) global i32 addrspace(3)* undef
@dst = addrspace(3) global [16384 x i32] undef @dst = addrspace(3) global [16383 x i32] undef
; FUNC-LABEL: {{^}}global_ptr: ; FUNC-LABEL: {{^}}global_ptr:
; SI: ds_write_b32 ; SI: ds_write_b32
define void @global_ptr() nounwind { define void @global_ptr() nounwind {
store i32 addrspace(3)* getelementptr ([16384 x i32], [16384 x i32] addrspace(3)* @dst, i32 0, i32 16), i32 addrspace(3)* addrspace(3)* @ptr store i32 addrspace(3)* getelementptr ([16383 x i32], [16383 x i32] addrspace(3)* @dst, i32 0, i32 16), i32 addrspace(3)* addrspace(3)* @ptr
ret void ret void
} }
; FUNC-LABEL: {{^}}local_address_store: ; FUNC-LABEL: {{^}}local_address_store:
; SI: ds_write_b32 ; SI: ds_write_b32
define void @local_address_store(i32 addrspace(3)* %out, i32 %val) { define void @local_address_store(i32 addrspace(3)* %out, i32 %val) {
store i32 %val, i32 addrspace(3)* %out store i32 %val, i32 addrspace(3)* %out
ret void ret void
Show All 32 Lines

test/CodeGen/AMDGPU/over-max-lds-size.ll

  • This file was added.
; RUN: not llc -march=amdgcn -mcpu=tahiti < %s 2>&1 | FileCheck -check-prefix=ERROR %s
; RUN: not llc -march=amdgcn -mcpu=hawaii < %s 2>&1 | FileCheck -check-prefix=ERROR %s
; RUN: not llc -march=amdgcn -mcpu=fiji < %s 2>&1 | FileCheck -check-prefix=ERROR %s
; ERROR: error: LDS size exceeds device maximum
@huge = internal unnamed_addr addrspace(3) global [100000 x i32] undef, align 4
define void @promote_alloca_size_256() {
entry:
%v0 = getelementptr inbounds [100000 x i32], [100000 x i32] addrspace(3)* @huge, i32 0, i32 0
store i32 0, i32 addrspace(3)* %v0
ret void
}