This is an archive of the discontinued LLVM Phabricator instance.

Differential D19093

AMDGPU: Include LDS size in printed comment
ClosedPublic

Authored by arsenm on Apr 13 2016, 7:32 PM.

Download Raw Diff

Details

Reviewers

• tstellarAMD

Diff Detail

Event Timeline

arsenm updated this revision to Diff 53654.Apr 13 2016, 7:32 PM

arsenm retitled this revision from to AMDGPU: Include LDS size in printed comment.

arsenm updated this object.

arsenm added a reviewer: • tstellarAMD.

arsenm added a subscriber: llvm-commits.

Herald added a subscriber: arsenm. · View Herald TranscriptApr 13 2016, 7:32 PM

LGTM.

This revision is now accepted and ready to land.Apr 14 2016, 7:18 AM

r266382

Revision Contents

Path

Size

lib/

Target/

AMDGPU/

AMDGPUAsmPrinter.cpp

2 lines

test/

CodeGen/

AMDGPU/

lds-size.ll

14 lines

Diff 53654

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

Show First 20 Lines • Show All 226 Lines • ▼ Show 20 Lines	if (STM.getGeneration() >= AMDGPUSubtarget::SOUTHERN_ISLANDS) {
OutStreamer->emitRawComment(" NumVgprs: " + Twine(KernelInfo.NumVGPR),		OutStreamer->emitRawComment(" NumVgprs: " + Twine(KernelInfo.NumVGPR),
false);		false);
OutStreamer->emitRawComment(" FloatMode: " + Twine(KernelInfo.FloatMode),		OutStreamer->emitRawComment(" FloatMode: " + Twine(KernelInfo.FloatMode),
false);		false);
OutStreamer->emitRawComment(" IeeeMode: " + Twine(KernelInfo.IEEEMode),		OutStreamer->emitRawComment(" IeeeMode: " + Twine(KernelInfo.IEEEMode),
false);		false);
OutStreamer->emitRawComment(" ScratchSize: " + Twine(KernelInfo.ScratchSize),		OutStreamer->emitRawComment(" ScratchSize: " + Twine(KernelInfo.ScratchSize),
false);		false);
		OutStreamer->emitRawComment(" LDSByteSize: " + Twine(KernelInfo.LDSSize) +
		" bytes/workgroup (compile time only)", false);

OutStreamer->emitRawComment(" COMPUTE_PGM_RSRC2:USER_SGPR: " +		OutStreamer->emitRawComment(" COMPUTE_PGM_RSRC2:USER_SGPR: " +
Twine(G_00B84C_USER_SGPR(KernelInfo.ComputePGMRSrc2)),		Twine(G_00B84C_USER_SGPR(KernelInfo.ComputePGMRSrc2)),
false);		false);
OutStreamer->emitRawComment(" COMPUTE_PGM_RSRC2:TGID_X_EN: " +		OutStreamer->emitRawComment(" COMPUTE_PGM_RSRC2:TGID_X_EN: " +
Twine(G_00B84C_TGID_X_EN(KernelInfo.ComputePGMRSrc2)),		Twine(G_00B84C_TGID_X_EN(KernelInfo.ComputePGMRSrc2)),
false);		false);
OutStreamer->emitRawComment(" COMPUTE_PGM_RSRC2:TGID_Y_EN: " +		OutStreamer->emitRawComment(" COMPUTE_PGM_RSRC2:TGID_Y_EN: " +
▲ Show 20 Lines • Show All 476 Lines • Show Last 20 Lines

test/CodeGen/AMDGPU/lds-size.ll

	; RUN: llc < %s -march=r600 -mcpu=redwood \| FileCheck %s			; RUN: llc -march=amdgcn < %s \| FileCheck -check-prefix=ALL -check-prefix=GCN %s
				; RUN: llc -march=r600 -mcpu=redwood < %s \| FileCheck -check-prefix=ALL -check-prefix=EG %s

	; This test makes sure we do not double count global values when they are			; This test makes sure we do not double count global values when they are
	; used in different basic blocks.			; used in different basic blocks.

	; CHECK: .long 166120			; GCN: .long 47180
	; CHECK-NEXT: .long 1			; GCN-NEXT: .long 32900
	; CHECK-LABEL: {{^}}test:
				; EG: .long 166120
				; EG-NEXT: .long 1
				; ALL: {{^}}test:

				; GCN: ; LDSByteSize: 4 bytes/workgroup (compile time only)
	@lds = internal unnamed_addr addrspace(3) global i32 undef, align 4			@lds = internal unnamed_addr addrspace(3) global i32 undef, align 4

	define void @test(i32 addrspace(1)* %out, i32 %cond) {			define void @test(i32 addrspace(1)* %out, i32 %cond) {
	entry:			entry:
	%0 = icmp eq i32 %cond, 0			%0 = icmp eq i32 %cond, 0
	br i1 %0, label %if, label %else			br i1 %0, label %if, label %else

	if:			if:
	Show All 10 Lines