caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[MLIR][AMDGPU] Fix bug in GatherToLDSOpLowering, get the correct MemRefType for destination (#142915)

Browse Source 
This PR fixes a bug in GatherToLDSOpLowering, we were getting the
MemRefType of source for the destination. Additionally, some related
typos are corrected.

CC: @krzysz00 @umangyadav @lialan
This commit is contained in:
Daniel Hernandez-Juarez
2025-06-13 18:33:51 +02:00
committed by GitHub
parent bd33eef7f1
commit 68b6f392ed
3 changed files with 14 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/docs/AMDGPUUsage.rst
View File

@@ -1215,12 +1215,12 @@ The AMDGPU backend implements the following LLVM IR intrinsics.
denormalization mode, enabled traps, and floating point exceptions.
The format is a 64-bit concatenation of the MODE and TRAPSTS registers.
:ref:`llvm.set.fpenv<int_set_fpenv>` Sets the floating point environment to the specifies state.
:ref:`llvm.set.fpenv<int_set_fpenv>` Sets the floating point environment to the specified state.
llvm.amdgcn.load.to.lds.p<1/7> Loads values from global memory (either in the form of a global
a raw fat buffer pointer) to LDS. The size of the data copied can be 1, 2,
or 4 bytes (and gfx950 also allows 12 or 16 bytes). The LDS pointer
argument should be wavefront-uniform; the global pointer need not be.
The LDS pointer is implicitly offset by 4 * lane_id bytes for sies <= 4 bytes
The LDS pointer is implicitly offset by 4 * lane_id bytes for size <= 4 bytes
and 16 * lane_id bytes for larger sizes. This lowers to `global_load_lds`,
`buffer_load_* ... lds`, or `global_load__* ... lds` depending on address
space and architecture. `amdgcn.global.load.lds` has the same semantics as

2
mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
View File

@@ -1101,7 +1101,7 @@ struct GatherToLDSOpLowering : public ConvertOpToLLVMPattern<GatherToLDSOp> {
Location loc = op.getLoc();
auto srcMemRefType = cast<MemRefType>(op.getSrc().getType());
auto dstMemRefType = cast<MemRefType>(op.getSrc().getType());
auto dstMemRefType = cast<MemRefType>(op.getDst().getType());
// TODO: instead of only transfering one element per thread, we could
// augment it to transfer multiple elements per thread by issuing multiple

20
mlir/test/Conversion/AMDGPUToROCDL/load_lds.mlir
View File

@@ -31,8 +31,8 @@ func.func @global_load_to_rocdl_f32(%global : memref<128x72xf32, #gpu_global_add
// CHECK: %[[GLOBAL_PTR:.*]] = llvm.getelementptr %[[GLOBAL_BASE]][%[[SRC_OFFSET]]]
// CHECK: %[[LDS_BASE:.*]] = llvm.extractvalue %[[LDS_DESC]][1]
// CHECK: %[[C72_1:.*]] = llvm.mlir.constant(72 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C72_1]] : i64
// CHECK: %[[C64:.*]] = llvm.mlir.constant(64 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C64]] : i64
// CHECK: %[[DST_OFFSET:.*]] = llvm.add %[[MUL_2]], %[[IC0]] : i64
// CHECK: %[[LDS_PTR:.*]] = llvm.getelementptr %[[LDS_BASE]][%[[DST_OFFSET]]]
@@ -65,8 +65,8 @@ func.func @global_load_to_rocdl_i8(%global : memref<128x72xi8, #gpu_global_addrs
// CHECK: %[[GLOBAL_PTR:.*]] = llvm.getelementptr %[[GLOBAL_BASE]][%[[SRC_OFFSET]]]
// CHECK: %[[LDS_BASE:.*]] = llvm.extractvalue %[[LDS_DESC]][1]
// CHECK: %[[C72_1:.*]] = llvm.mlir.constant(72 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C72_1]] : i64
// CHECK: %[[C64:.*]] = llvm.mlir.constant(64 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C64]] : i64
// CHECK: %[[DST_OFFSET:.*]] = llvm.add %[[MUL_2]], %[[IC0]] : i64
// CHECK: %[[LDS_PTR:.*]] = llvm.getelementptr %[[LDS_BASE]][%[[DST_OFFSET]]]
@@ -103,8 +103,8 @@ func.func @global_load_to_rocdl_vec(%global : memref<128x72xi16, #gpu_global_add
// CHECK: %[[GLOBAL_PTR:.*]] = llvm.getelementptr %[[GLOBAL_BASE]][%[[SRC_OFFSET]]]
// CHECK: %[[LDS_BASE:.*]] = llvm.extractvalue %[[LDS_DESC]][1]
// CHECK: %[[C72_1:.*]] = llvm.mlir.constant(72 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C72_1]] : i64
// CHECK: %[[C128:.*]] = llvm.mlir.constant(128 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C128]] : i64
// CHECK: %[[DST_OFFSET:.*]] = llvm.add %[[MUL_2]], %[[IC0]] : i64
// CHECK: %[[LDS_PTR:.*]] = llvm.getelementptr %[[LDS_BASE]][%[[DST_OFFSET]]]
@@ -130,7 +130,9 @@ func.func @global_load_to_rocdl_dynamic_indices(%global : memref<512xi32, #gpu_g
// CHECK: %[[GLOBAL_BASE:.*]] = llvm.extractvalue %[[GLOBAL_DESC]][1]
// CHECK: %[[GLOBAL_PTR:.*]] = llvm.getelementptr %[[GLOBAL_BASE]][%[[SRCIDX_CAST]]]
// CHECK: %[[LDS_BASE:.*]] = llvm.extractvalue %[[LDS_DESC]][1]
// CHECK: %[[LDS_PTR:.*]] = llvm.getelementptr %[[LDS_BASE]][%[[DSTIDX_CAST]]]
// CHECK: %[[C64:.*]] = llvm.mlir.constant(64 : index) : i64
// CHECK: %[[DSTIDX:.*]] = llvm.mul %[[DSTIDX_CAST]], %[[C64]] : i64
// CHECK: %[[LDS_PTR:.*]] = llvm.getelementptr %[[LDS_BASE]][%[[DSTIDX]]]
// CHECK: rocdl.load.to.lds %[[GLOBAL_PTR]], %[[LDS_PTR]], 4
%alloc = memref.alloc() : memref<4x64xi32, #gpu_lds_addrspace>
%c0 = arith.constant 0 : index
@@ -166,8 +168,8 @@ func.func @fat_buffer_load_to_rocdl_f32(%global : memref<128x72xf32, #amdgpu_fat
// CHECK: %[[GLOBAL_PTR:.*]] = llvm.getelementptr %[[GLOBAL_BASE]][%[[SRC_OFFSET]]]
// CHECK: %[[LDS_BASE:.*]] = llvm.extractvalue %[[LDS_DESC]][1]
// CHECK: %[[C72_1:.*]] = llvm.mlir.constant(72 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C72_1]] : i64
// CHECK: %[[C64:.*]] = llvm.mlir.constant(64 : index) : i64
// CHECK: %[[MUL_2:.*]] = llvm.mul %[[IC32]], %[[C64]] : i64
// CHECK: %[[DST_OFFSET:.*]] = llvm.add %[[MUL_2]], %[[IC0]] : i64
// CHECK: %[[LDS_PTR:.*]] = llvm.getelementptr %[[LDS_BASE]][%[[DST_OFFSET]]]