AMDGPU: Remove AMDGPU.fract intrinsic
Mesa doesn't use this, and this is pattern matched already
from fsub x, (ffloor x)
llvm-svn: 258513
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.AMDGPU.fract.f64.ll b/llvm/test/CodeGen/AMDGPU/fract.f64.ll
similarity index 61%
rename from llvm/test/CodeGen/AMDGPU/llvm.AMDGPU.fract.f64.ll
rename to llvm/test/CodeGen/AMDGPU/fract.f64.ll
index 6049dca..a19259f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.AMDGPU.fract.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fract.f64.ll
@@ -1,9 +1,9 @@
-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
-declare double @llvm.fabs.f64(double %Val)
-declare double @llvm.AMDGPU.fract.f64(double) nounwind readnone
+declare double @llvm.fabs.f64(double) #0
+declare double @llvm.floor.f64(double) #0
; FUNC-LABEL: {{^}}fract_f64:
; GCN: v_fract_f64_e32 [[FRC:v\[[0-9]+:[0-9]+\]]], v{{\[}}[[LO:[0-9]+]]:[[HI:[0-9]+]]]
@@ -15,10 +15,11 @@
; SI: v_cndmask_b32_e64 v[[RESHI:[0-9]+]], v[[MINHI]], v[[HI]], [[COND]]
; SI: buffer_store_dwordx2 v{{\[}}[[RESLO]]:[[RESHI]]]
; CI: buffer_store_dwordx2 [[FRC]]
-define void @fract_f64(double addrspace(1)* %out, double addrspace(1)* %src) nounwind {
- %val = load double, double addrspace(1)* %src, align 4
- %fract = call double @llvm.AMDGPU.fract.f64(double %val) nounwind readnone
- store double %fract, double addrspace(1)* %out, align 4
+define void @fract_f64(double addrspace(1)* %out, double addrspace(1)* %src) #1 {
+ %x = load double, double addrspace(1)* %src
+ %floor.x = call double @llvm.floor.f64(double %x)
+ %fract = fsub double %x, %floor.x
+ store double %fract, double addrspace(1)* %out
ret void
}
@@ -32,11 +33,12 @@
; SI: v_cndmask_b32_e64 v[[RESHI:[0-9]+]], v[[MINHI]], v[[HI]], [[COND]]
; SI: buffer_store_dwordx2 v{{\[}}[[RESLO]]:[[RESHI]]]
; CI: buffer_store_dwordx2 [[FRC]]
-define void @fract_f64_neg(double addrspace(1)* %out, double addrspace(1)* %src) nounwind {
- %val = load double, double addrspace(1)* %src, align 4
- %neg = fsub double 0.0, %val
- %fract = call double @llvm.AMDGPU.fract.f64(double %neg) nounwind readnone
- store double %fract, double addrspace(1)* %out, align 4
+define void @fract_f64_neg(double addrspace(1)* %out, double addrspace(1)* %src) #1 {
+ %x = load double, double addrspace(1)* %src
+ %neg.x = fsub double -0.0, %x
+ %floor.neg.x = call double @llvm.floor.f64(double %neg.x)
+ %fract = fsub double %neg.x, %floor.neg.x
+ store double %fract, double addrspace(1)* %out
ret void
}
@@ -50,11 +52,15 @@
; SI: v_cndmask_b32_e64 v[[RESHI:[0-9]+]], v[[MINHI]], v[[HI]], [[COND]]
; SI: buffer_store_dwordx2 v{{\[}}[[RESLO]]:[[RESHI]]]
; CI: buffer_store_dwordx2 [[FRC]]
-define void @fract_f64_neg_abs(double addrspace(1)* %out, double addrspace(1)* %src) nounwind {
- %val = load double, double addrspace(1)* %src, align 4
- %abs = call double @llvm.fabs.f64(double %val)
- %neg = fsub double 0.0, %abs
- %fract = call double @llvm.AMDGPU.fract.f64(double %neg) nounwind readnone
- store double %fract, double addrspace(1)* %out, align 4
+define void @fract_f64_neg_abs(double addrspace(1)* %out, double addrspace(1)* %src) #1 {
+ %x = load double, double addrspace(1)* %src
+ %abs.x = call double @llvm.fabs.f64(double %x)
+ %neg.abs.x = fsub double -0.0, %abs.x
+ %floor.neg.abs.x = call double @llvm.floor.f64(double %neg.abs.x)
+ %fract = fsub double %neg.abs.x, %floor.neg.abs.x
+ store double %fract, double addrspace(1)* %out
ret void
}
+
+attributes #0 = { nounwind readnone }
+attributes #1 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/fract.ll b/llvm/test/CodeGen/AMDGPU/fract.ll
new file mode 100644
index 0000000..5434248
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fract.ll
@@ -0,0 +1,58 @@
+; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
+; XUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+
+declare float @llvm.fabs.f32(float) #0
+declare float @llvm.floor.f32(float) #0
+
+; FUNC-LABEL: {{^}}fract_f32:
+; CI: v_fract_f32_e32 [[RESULT:v[0-9]+]], [[INPUT:v[0-9]+]]
+; SI: v_floor_f32_e32 [[FLR:v[0-9]+]], [[INPUT:v[0-9]+]]
+; SI: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[FLR]], [[INPUT]]
+; GCN: buffer_store_dword [[RESULT]]
+
+; XEG: FRACT
+define void @fract_f32(float addrspace(1)* %out, float addrspace(1)* %src) #1 {
+ %x = load float, float addrspace(1)* %src
+ %floor.x = call float @llvm.floor.f32(float %x)
+ %fract = fsub float %x, %floor.x
+ store float %fract, float addrspace(1)* %out
+ ret void
+}
+
+; FUNC-LABEL: {{^}}fract_f32_neg:
+; CI: v_fract_f32_e64 [[RESULT:v[0-9]+]], -[[INPUT:v[0-9]+]]
+; SI: v_floor_f32_e64 [[FLR:v[0-9]+]], -[[INPUT:v[0-9]+]]
+; SI: v_sub_f32_e64 [[RESULT:v[0-9]+]], -[[INPUT]], [[FLR]]
+; GCN: buffer_store_dword [[RESULT]]
+
+; XEG: FRACT
+define void @fract_f32_neg(float addrspace(1)* %out, float addrspace(1)* %src) #1 {
+ %x = load float, float addrspace(1)* %src
+ %x.neg = fsub float -0.0, %x
+ %floor.x.neg = call float @llvm.floor.f32(float %x.neg)
+ %fract = fsub float %x.neg, %floor.x.neg
+ store float %fract, float addrspace(1)* %out
+ ret void
+}
+
+; FUNC-LABEL: {{^}}fract_f32_neg_abs:
+; CI: v_fract_f32_e64 [[RESULT:v[0-9]+]], -|[[INPUT:v[0-9]+]]|
+; SI: v_floor_f32_e64 [[FLR:v[0-9]+]], -|[[INPUT:v[0-9]+]]|
+; SI: v_sub_f32_e64 [[RESULT:v[0-9]+]], -|[[INPUT]]|, [[FLR]]
+; GCN: buffer_store_dword [[RESULT]]
+
+; XEG: FRACT
+define void @fract_f32_neg_abs(float addrspace(1)* %out, float addrspace(1)* %src) #1 {
+ %x = load float, float addrspace(1)* %src
+ %abs.x = call float @llvm.fabs.f32(float %x)
+ %neg.abs.x = fsub float -0.0, %abs.x
+ %floor.neg.abs.x = call float @llvm.floor.f32(float %neg.abs.x)
+ %fract = fsub float %neg.abs.x, %floor.neg.abs.x
+ store float %fract, float addrspace(1)* %out
+ ret void
+}
+
+attributes #0 = { nounwind readnone }
+attributes #1 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.AMDGPU.fract.ll b/llvm/test/CodeGen/AMDGPU/llvm.AMDGPU.fract.ll
deleted file mode 100644
index c53837a..0000000
--- a/llvm/test/CodeGen/AMDGPU/llvm.AMDGPU.fract.ll
+++ /dev/null
@@ -1,49 +0,0 @@
-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
-
-declare float @llvm.fabs.f32(float %Val)
-declare float @llvm.AMDGPU.fract.f32(float) nounwind readnone
-
-; FUNC-LABEL: {{^}}fract_f32:
-; CI: v_fract_f32_e32 [[RESULT:v[0-9]+]], [[INPUT:v[0-9]+]]
-; SI: v_floor_f32_e32 [[FLR:v[0-9]+]], [[INPUT:v[0-9]+]]
-; SI: v_subrev_f32_e32 [[RESULT:v[0-9]+]], [[FLR]], [[INPUT]]
-; GCN: buffer_store_dword [[RESULT]]
-; EG: FRACT
-define void @fract_f32(float addrspace(1)* %out, float addrspace(1)* %src) nounwind {
- %val = load float, float addrspace(1)* %src, align 4
- %fract = call float @llvm.AMDGPU.fract.f32(float %val) nounwind readnone
- store float %fract, float addrspace(1)* %out, align 4
- ret void
-}
-
-; FUNC-LABEL: {{^}}fract_f32_neg:
-; CI: v_fract_f32_e64 [[RESULT:v[0-9]+]], -[[INPUT:v[0-9]+]]
-; SI: v_floor_f32_e64 [[FLR:v[0-9]+]], -[[INPUT:v[0-9]+]]
-; SI: v_sub_f32_e64 [[RESULT:v[0-9]+]], -[[INPUT]], [[FLR]]
-; GCN: buffer_store_dword [[RESULT]]
-; EG: FRACT
-define void @fract_f32_neg(float addrspace(1)* %out, float addrspace(1)* %src) nounwind {
- %val = load float, float addrspace(1)* %src, align 4
- %neg = fsub float 0.0, %val
- %fract = call float @llvm.AMDGPU.fract.f32(float %neg) nounwind readnone
- store float %fract, float addrspace(1)* %out, align 4
- ret void
-}
-
-; FUNC-LABEL: {{^}}fract_f32_neg_abs:
-; CI: v_fract_f32_e64 [[RESULT:v[0-9]+]], -|[[INPUT:v[0-9]+]]|
-; SI: v_floor_f32_e64 [[FLR:v[0-9]+]], -|[[INPUT:v[0-9]+]]|
-; SI: v_sub_f32_e64 [[RESULT:v[0-9]+]], -|[[INPUT]]|, [[FLR]]
-; GCN: buffer_store_dword [[RESULT]]
-; EG: FRACT
-define void @fract_f32_neg_abs(float addrspace(1)* %out, float addrspace(1)* %src) nounwind {
- %val = load float, float addrspace(1)* %src, align 4
- %abs = call float @llvm.fabs.f32(float %val)
- %neg = fsub float 0.0, %abs
- %fract = call float @llvm.AMDGPU.fract.f32(float %neg) nounwind readnone
- store float %fract, float addrspace(1)* %out, align 4
- ret void
-}
diff --git a/llvm/test/CodeGen/AMDGPU/si-spill-cf.ll b/llvm/test/CodeGen/AMDGPU/si-spill-cf.ll
index 2cb919f..b6f1b82 100644
--- a/llvm/test/CodeGen/AMDGPU/si-spill-cf.ll
+++ b/llvm/test/CodeGen/AMDGPU/si-spill-cf.ll
@@ -256,7 +256,8 @@
ENDIF2582: ; preds = %ELSE2584, %IF2565
%213 = fadd float %1, undef
%214 = fadd float 0.000000e+00, %213
- %215 = call float @llvm.AMDGPU.fract.f32(float %214)
+ %floor = call float @llvm.floor.f32(float %214)
+ %215 = fsub float %214, %floor
br i1 undef, label %IF2589, label %ELSE2590
IF2589: ; preds = %ENDIF2582
@@ -480,7 +481,7 @@
declare float @llvm.SI.load.const(<16 x i8>, i32) #1
; Function Attrs: nounwind readnone
-declare float @llvm.AMDGPU.fract.f32(float) #1
+declare float @llvm.floor.f32(float) #1
; Function Attrs: nounwind readnone
declare float @llvm.sqrt.f32(float) #1