mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-25 20:23:11 +01:00
4d8c0e830b
According to IR LangRef, the FMF flag: contract Allow floating-point contraction (e.g. fusing a multiply followed by an addition into a fused multiply-and-add). reassoc Allow reassociation transformations for floating-point instructions. This may dramatically change results in floating-point. My understanding is that these two flags shouldn't imply each other, as we might have a SDNode that can be reassociated with others, but not contractble. eg: We may want following fmul/fad/fsub to freely reassoc, but don't want fma being generated here. %F = fmul reassoc double %A, %B ; <double> [#uses=1] %G = fmul reassoc double %C, %D ; <double> [#uses=1] %H = fadd reassoc double %F, %G ; <double> [#uses=1] %I = fsub reassoc double %H, %E ; <double> [#uses=1] Before https://reviews.llvm.org/D45710, `reassoc` flag actually did not imply isContratable either. The current implementation also only check the flag in fadd node, ignoring fmul node, this patch update that as well. Reviewed By: spatel, qiucf Differential Revision: https://reviews.llvm.org/D104247
197 lines
9.1 KiB
LLVM
197 lines
9.1 KiB
LLVM
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,SI %s
|
|
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,SI %s
|
|
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
|
|
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
|
|
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,VI %s
|
|
; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,VI %s
|
|
|
|
; GCN-LABEL: {{^}}fmuladd_f64:
|
|
; GCN: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
|
|
define amdgpu_kernel void @fmuladd_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2, double addrspace(1)* %in3) #0 {
|
|
%r0 = load double, double addrspace(1)* %in1
|
|
%r1 = load double, double addrspace(1)* %in2
|
|
%r2 = load double, double addrspace(1)* %in3
|
|
%r3 = tail call double @llvm.fmuladd.f64(double %r0, double %r1, double %r2)
|
|
store double %r3, double addrspace(1)* %out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fmul_fadd_f64:
|
|
; GCN-CONTRACT: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
|
|
|
|
; GCN-STRICT: v_mul_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
|
|
; GCN-STRICT: v_add_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
|
|
define amdgpu_kernel void @fmul_fadd_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2, double addrspace(1)* %in3) #0 {
|
|
%r0 = load double, double addrspace(1)* %in1
|
|
%r1 = load double, double addrspace(1)* %in2
|
|
%r2 = load double, double addrspace(1)* %in3
|
|
%tmp = fmul double %r0, %r1
|
|
%r3 = fadd double %tmp, %r2
|
|
store double %r3, double addrspace(1)* %out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fmul_fadd_contract_f64:
|
|
; GCN: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
|
|
|
|
define amdgpu_kernel void @fmul_fadd_contract_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2, double addrspace(1)* %in3) #0 {
|
|
%r0 = load double, double addrspace(1)* %in1
|
|
%r1 = load double, double addrspace(1)* %in2
|
|
%r2 = load double, double addrspace(1)* %in3
|
|
%tmp = fmul contract double %r0, %r1
|
|
%r3 = fadd contract double %tmp, %r2
|
|
store double %r3, double addrspace(1)* %out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fadd_a_a_b_f64:
|
|
; GCN: {{buffer|flat}}_load_dwordx2 [[R1:v\[[0-9]+:[0-9]+\]]],
|
|
; GCN: {{buffer|flat}}_load_dwordx2 [[R2:v\[[0-9]+:[0-9]+\]]],
|
|
|
|
; GCN-STRICT: v_add_f64 [[TMP:v\[[0-9]+:[0-9]+\]]], [[R1]], [[R1]]
|
|
; GCN-STRICT: v_add_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[TMP]], [[R2]]
|
|
|
|
; GCN-CONTRACT: v_fma_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[R1]], 2.0, [[R2]]
|
|
|
|
; SI: buffer_store_dwordx2 [[RESULT]]
|
|
; VI: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]]
|
|
define amdgpu_kernel void @fadd_a_a_b_f64(double addrspace(1)* %out,
|
|
double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2) #0 {
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
|
|
%gep.0 = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
%gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
|
|
%gep.out = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
|
|
%r0 = load volatile double, double addrspace(1)* %gep.0
|
|
%r1 = load volatile double, double addrspace(1)* %gep.1
|
|
|
|
%add.0 = fadd double %r0, %r0
|
|
%add.1 = fadd double %add.0, %r1
|
|
store double %add.1, double addrspace(1)* %gep.out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fadd_b_a_a_f64:
|
|
; GCN: {{buffer|flat}}_load_dwordx2 [[R1:v\[[0-9]+:[0-9]+\]]],
|
|
; GCN: {{buffer|flat}}_load_dwordx2 [[R2:v\[[0-9]+:[0-9]+\]]],
|
|
|
|
; GCN-STRICT: v_add_f64 [[TMP:v\[[0-9]+:[0-9]+\]]], [[R1]], [[R1]]
|
|
; GCN-STRICT: v_add_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[R2]], [[TMP]]
|
|
|
|
; GCN-CONTRACT: v_fma_f64 [[RESULT:v\[[0-9]+:[0-9]+\]]], [[R1]], 2.0, [[R2]]
|
|
|
|
; SI: buffer_store_dwordx2 [[RESULT]]
|
|
; VI: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, [[RESULT]]
|
|
define amdgpu_kernel void @fadd_b_a_a_f64(double addrspace(1)* %out,
|
|
double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2) #0 {
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
|
|
%gep.0 = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
%gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
|
|
%gep.out = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
|
|
%r0 = load volatile double, double addrspace(1)* %gep.0
|
|
%r1 = load volatile double, double addrspace(1)* %gep.1
|
|
|
|
%add.0 = fadd double %r0, %r0
|
|
%add.1 = fadd double %r1, %add.0
|
|
store double %add.1, double addrspace(1)* %gep.out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}mad_sub_f64:
|
|
; GCN-STRICT: v_mul_f64 v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}
|
|
; GCN-STRICT: v_add_f64 v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}, -v{{\[[0-9]+:[0-9]+\]}}
|
|
|
|
; GCN-CONTRACT: v_fma_f64 v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}, v{{\[[0-9]+:[0-9]+\]}}, -v{{\[[0-9]+:[0-9]+\]}}
|
|
define amdgpu_kernel void @mad_sub_f64(double addrspace(1)* noalias nocapture %out, double addrspace(1)* noalias nocapture readonly %ptr) #1 {
|
|
%tid = tail call i32 @llvm.amdgcn.workitem.id.x() #0
|
|
%tid.ext = sext i32 %tid to i64
|
|
%gep0 = getelementptr double, double addrspace(1)* %ptr, i64 %tid.ext
|
|
%add1 = add i64 %tid.ext, 1
|
|
%gep1 = getelementptr double, double addrspace(1)* %ptr, i64 %add1
|
|
%add2 = add i64 %tid.ext, 2
|
|
%gep2 = getelementptr double, double addrspace(1)* %ptr, i64 %add2
|
|
%outgep = getelementptr double, double addrspace(1)* %out, i64 %tid.ext
|
|
%a = load volatile double, double addrspace(1)* %gep0, align 8
|
|
%b = load volatile double, double addrspace(1)* %gep1, align 8
|
|
%c = load volatile double, double addrspace(1)* %gep2, align 8
|
|
%mul = fmul double %a, %b
|
|
%sub = fsub double %mul, %c
|
|
store double %sub, double addrspace(1)* %outgep, align 8
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fadd_a_a_b_f64_fast_add0:
|
|
; GCN-STRICT: v_add_f64
|
|
; GCN-STRICT: v_add_f64
|
|
|
|
; GCN-CONTRACT: v_fma_f64
|
|
define amdgpu_kernel void @fadd_a_a_b_f64_fast_add0(double addrspace(1)* %out,
|
|
double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2) #0 {
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
|
|
%gep.0 = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
%gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
|
|
%gep.out = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
|
|
%r0 = load volatile double, double addrspace(1)* %gep.0
|
|
%r1 = load volatile double, double addrspace(1)* %gep.1
|
|
|
|
%add.0 = fadd fast double %r0, %r0
|
|
%add.1 = fadd double %add.0, %r1
|
|
store double %add.1, double addrspace(1)* %gep.out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fadd_a_a_b_f64_fast_add1:
|
|
; GCN-STRICT: v_add_f64
|
|
; GCN-STRICT: v_add_f64
|
|
|
|
; GCN-CONTRACT: v_fma_f64
|
|
define amdgpu_kernel void @fadd_a_a_b_f64_fast_add1(double addrspace(1)* %out,
|
|
double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2) #0 {
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
|
|
%gep.0 = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
%gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
|
|
%gep.out = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
|
|
%r0 = load volatile double, double addrspace(1)* %gep.0
|
|
%r1 = load volatile double, double addrspace(1)* %gep.1
|
|
|
|
%add.0 = fadd double %r0, %r0
|
|
%add.1 = fadd fast double %add.0, %r1
|
|
store double %add.1, double addrspace(1)* %gep.out
|
|
ret void
|
|
}
|
|
|
|
; GCN-LABEL: {{^}}fadd_a_a_b_f64_fast:
|
|
; GCN: v_fma_f64
|
|
define amdgpu_kernel void @fadd_a_a_b_f64_fast(double addrspace(1)* %out,
|
|
double addrspace(1)* %in1,
|
|
double addrspace(1)* %in2) #0 {
|
|
%tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
|
|
%gep.0 = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
%gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
|
|
%gep.out = getelementptr double, double addrspace(1)* %out, i32 %tid
|
|
|
|
%r0 = load volatile double, double addrspace(1)* %gep.0
|
|
%r1 = load volatile double, double addrspace(1)* %gep.1
|
|
|
|
%add.0 = fadd fast double %r0, %r0
|
|
%add.1 = fadd fast double %add.0, %r1
|
|
store double %add.1, double addrspace(1)* %gep.out
|
|
ret void
|
|
}
|
|
|
|
declare i32 @llvm.amdgcn.workitem.id.x() #1
|
|
declare double @llvm.fmuladd.f64(double, double, double) #1
|
|
|
|
attributes #0 = { nounwind }
|
|
attributes #1 = { nounwind readnone }
|