Skip to content

[AMDGPU] update LDS block size for gfx1250 #167614

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

[AMDGPU] update LDS block size for gfx1250 #167614

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
97f6ccd
[AMDGPU] update LDS block size for gfx1250
changpeng Nov 12, 2025
b9844d9
Update AMDGPU::getLdsDwGranularity
changpeng Nov 12, 2025
c17c39c
Update maximum lds size for hsa-gfx125*-v4.s
changpeng Nov 13, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion llvm/docs/AMDGPUUsage.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5855,7 +5855,7 @@ The fields used by CP for code objects before V3 also match those specified in
GFX950
roundup(lds-size / (320 * 4))
GFX125*
roundup(lds-size / (256 * 4))
roundup(lds-size / (512 * 4))

24 1 bit ENABLE_EXCEPTION_IEEE_754_FP Wavefront starts execution
_INVALID_OPERATION with specified exceptions
Expand Down
9 changes: 3 additions & 6 deletions llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1161,12 +1161,9 @@ void AMDGPUAsmPrinter::getSIProgramInfo(SIProgramInfo &ProgInfo,
ProgInfo.DX10Clamp = Mode.DX10Clamp;

unsigned LDSAlignShift;
if (STM.getFeatureBits().test(FeatureAddressableLocalMemorySize327680)) {
// LDS is allocated in 256 dword blocks.
LDSAlignShift = 10;
} else if (STM.getFeatureBits().test(
FeatureAddressableLocalMemorySize163840)) {
// LDS is allocated in 320 dword blocks.
if (STM.getFeatureBits().test(FeatureAddressableLocalMemorySize327680) ||
STM.getFeatureBits().test(FeatureAddressableLocalMemorySize163840)) {
Comment on lines +1164 to +1165
Copy link
Contributor

@arsenm arsenm Nov 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems like this should be modeled more directly rather than inferring the allocation size by the target that supports this allocation size

Copy link
Contributor Author

@changpeng changpeng Nov 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems like this should be modeled more directly rather than inferring the allocation size by the target that supports this allocation size

Do you mean something directly like?:
if (gfx1250)
LDSAlignShift = 11;

Copy link
Contributor Author

@changpeng changpeng Nov 12, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@arsenm : should we get the lds block size from the target directly?

// LDS is allocated in 512 or 320 dword blocks.
LDSAlignShift = 11;
} else if (STM.getFeatureBits().test(
FeatureAddressableLocalMemorySize65536)) {
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/CodeGen/AMDGPU/extra-lds-size.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,10 +31,10 @@
; GFX1200-MESA: .long 45100
; GFX1200-MESA-NEXT: .long 1024

; GFX1250-PAL: '0x2c0b (SPI_SHADER_PGM_RSRC2_PS)': 0x200
; GFX1250-PAL: '0x2c0b (SPI_SHADER_PGM_RSRC2_PS)': 0x100
Copy link
Contributor

@arsenm arsenm Nov 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure how this test ends up showing the allocation granularity


; GFX1250-MESA: .long 45100
; GFX1250-MESA-NEXT: .long 512
; GFX1250-MESA-NEXT: .long 256

@lds = internal addrspace(3) global [4096 x i8] poison

Expand Down
6 changes: 3 additions & 3 deletions llvm/test/CodeGen/AMDGPU/lds-size-hsa-gfx1250.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ define amdgpu_kernel void @test_lds_i32(i32 %val) {
; GCN-LABEL: test_lds_array_i8:
; GCN: .amdhsa_group_segment_fixed_size 327680
; GCN: ; LDSByteSize: 327680 bytes/workgroup
; MESA: granulated_lds_size = 320
; MESA: granulated_lds_size = 160
define amdgpu_kernel void @test_lds_array_i8() {
%gep = getelementptr inbounds [327679 x i8], ptr addrspace(3) @lds.array.i8, i32 0, i32 5
%val = load i8, ptr addrspace(3) %gep
Expand All @@ -52,7 +52,7 @@ define amdgpu_kernel void @test_lds_array_i8() {
; GCN-LABEL: test_lds_array_i16:
; GCN: .amdhsa_group_segment_fixed_size 327680
; GCN: ; LDSByteSize: 327680 bytes/workgroup
; MESA: granulated_lds_size = 320
; MESA: granulated_lds_size = 160
define amdgpu_kernel void @test_lds_array_i16() {
%gep = getelementptr inbounds [163839 x i16], ptr addrspace(3) @lds.array.i16, i32 0, i32 10
%val = load i16, ptr addrspace(3) %gep
Expand All @@ -63,7 +63,7 @@ define amdgpu_kernel void @test_lds_array_i16() {
; GCN-LABEL: test_lds_array_i32:
; GCN: .amdhsa_group_segment_fixed_size 327680
; GCN: ; LDSByteSize: 327680 bytes/workgroup
; MESA: granulated_lds_size = 320
; MESA: granulated_lds_size = 160
define amdgpu_kernel void @test_lds_array_i32() {
%gep = getelementptr inbounds [81919 x i32], ptr addrspace(3) @lds.array.i32, i32 0, i32 20
%val = load i32, ptr addrspace(3) %gep
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/AMDGPU/pal-metadata-3.0.gfx1250.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -126,7 +126,7 @@
; CHECK-NEXT: .entry_point: _amdgpu_hs
; CHECK-NEXT: .entry_point_symbol: hs_shader
; CHECK-NEXT: .forward_progress: true
; CHECK-NEXT: .lds_size: 0x1000
; CHECK-NEXT: .lds_size: 0x800
; CHECK-NEXT: .mem_ordered: true
; CHECK-NEXT: .scratch_en: false
; CHECK-NEXT: .scratch_memory_size: 0
Expand Down
Loading