Skip to content

[AMDGPU] Fix FMA Combine #119217

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

Merged
merged 2 commits into from
Dec 10, 2024
Merged

[AMDGPU] Fix FMA Combine #119217

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1ebcb75
[AMDGPU] Fix FMA Combine
piotrAMD Dec 9, 2024
0f602ec
Update fdot2.ll instead of adding new test
piotrAMD Dec 9, 2024
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/lib/Target/AMDGPU/SIISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14696,7 +14696,7 @@ SDValue SITargetLowering::performFMACombine(SDNode *N,
EVT VT = N->getValueType(0);
SDLoc SL(N);

if (!Subtarget->hasDot7Insts() || VT != MVT::f32)
if (!Subtarget->hasDot10Insts() || VT != MVT::f32)
Copy link
Contributor

Choose a reason for hiding this comment

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

I still have no idea what these numbers mean

return SDValue();

// FMA((F32)S0.x, (F32)S1. x, FMA((F32)S0.y, (F32)S1.y, (F32)z)) ->
Expand Down
21 changes: 17 additions & 4 deletions llvm/test/CodeGen/AMDGPU/fdot2.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
; RUN: llc -mtriple=amdgcn -mcpu=gfx906 -denormal-fp-math-f32=preserve-sign -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,GFX906
; RUN: llc -mtriple=amdgcn -mcpu=gfx906 -denormal-fp-math=preserve-sign -fp-contract=fast -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,GFX906-CONTRACT
; RUN: llc -mtriple=amdgcn -mcpu=gfx906 -denormal-fp-math=ieee -fp-contract=fast -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,GFX906-DENORM-CONTRACT
; RUN: llc -mtriple=amdgcn -mcpu=gfx906 -denormal-fp-math-f32=preserve-sign -enable-unsafe-fp-math -mattr="+dot7-insts,-dot10-insts" -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,GFX906-DOT10-DISABLED
; (fadd (fmul S1.x, S2.x), (fadd (fmul (S1.y, S2.y), z))) -> (fdot2 S1, S2, z)

; Tests to make sure fdot2 is not generated when vector elements of dot-product expressions
Expand All @@ -21,6 +22,7 @@

; GFX906-CONTRACT: v_mac_f16_e32
; GFX906-DENORM-CONTRACT: v_fma_f16
; GFX906-DOT10-DISABLED: v_fma_f16
define amdgpu_kernel void @dotproduct_f16(ptr addrspace(1) %src1,
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
Expand All @@ -44,8 +46,11 @@ entry:
}


; We only want to generate fdot2 if vector element of dot product is converted from f16 to f32
; and the vectors are of type <2 x half>
; We only want to generate fdot2 if:
; - vector element of dot product is converted from f16 to f32, and
; - the vectors are of type <2 x half>, and
; - "dot10-insts" is enabled

; GCN-LABEL: {{^}}dotproduct_f16_f32
; GFX900: v_mad_mix_f32
; GFX900: v_mad_mix_f32
Expand All @@ -59,6 +64,7 @@ entry:
; GFX906-CONTRACT: v_dot2_f32_f16

; GFX906-DENORM-CONTRACT: v_dot2_f32_f16
; GFX906-DOT10-DISABLED: v_fma_mix_f32
define amdgpu_kernel void @dotproduct_f16_f32(ptr addrspace(1) %src1,
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
Expand All @@ -85,8 +91,11 @@ entry:
ret void
}

; We only want to generate fdot2 if vector element of dot product is converted from f16 to f32
; and the vectors are of type <2 x half>
; We only want to generate fdot2 if:
; - vector element of dot product is converted from f16 to f32, and
; - the vectors are of type <2 x half>, and
; - "dot10-insts" is enabled

; GCN-LABEL: {{^}}dotproduct_diffvecorder
; GFX900: v_mad_mix_f32
; GFX900: v_mad_mix_f32
Expand All @@ -99,6 +108,7 @@ entry:

; GFX906-CONTRACT: v_dot2_f32_f16
; GFX906-DENORM-CONTRACT: v_dot2_f32_f16
; GFX906-DOT10-DISABLED: v_fma_mix_f32
define amdgpu_kernel void @dotproduct_diffvecorder(ptr addrspace(1) %src1,
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
Expand Down Expand Up @@ -136,6 +146,7 @@ entry:

; GFX906-CONTRACT: v_fma_mix_f32
; GFX906-DENORM-CONTRACT: v_fma_mix_f32
; GFX906-DOT10-DISABLED: v_fma_mix_f32
define amdgpu_kernel void @dotproduct_v4f16(ptr addrspace(1) %src1,
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
Expand Down Expand Up @@ -173,6 +184,7 @@ entry:

; GFX906-CONTRACT: v_fma_mix_f32
; GFX906-DENORM-CONTRACT: v_fma_mix_f32
; GFX906-DOT10-DISABLED: v_fma_mix_f32
define amdgpu_kernel void @NotAdotproduct(ptr addrspace(1) %src1,
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
Expand Down Expand Up @@ -210,6 +222,7 @@ entry:

; GFX906-CONTRACT: v_fma_mix_f32
; GFX906-DENORM-CONTRACT: v_fma_mix_f32
; GFX906-DOT10-DISABLED: v_fma_mix_f32
define amdgpu_kernel void @Diff_Idx_NotAdotproduct(ptr addrspace(1) %src1,
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
Expand Down
Loading