| 1 | // RUN: %clang_cc1 -emit-llvm -o - %s | FileCheck %s |
|---|
| 2 | // RUN: %clang_cc1 -emit-llvm -cl-denorms-are-zero -o - %s | FileCheck -check-prefix=DENORM-ZERO %s |
|---|
| 3 | |
|---|
| 4 | // Slow FMAF and slow f32 denormals |
|---|
| 5 | // RUN: %clang_cc1 -emit-llvm -o - -triple amdgcn--amdhsa -target-cpu pitcairn %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH %s |
|---|
| 6 | // RUN: %clang_cc1 -emit-llvm -cl-denorms-are-zero -o - -triple amdgcn--amdhsa -target-cpu pitcairn %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH-OPT %s |
|---|
| 7 | |
|---|
| 8 | // Fast FMAF, but slow f32 denormals |
|---|
| 9 | // RUN: %clang_cc1 -emit-llvm -o - -triple amdgcn--amdhsa -target-cpu tahiti %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH %s |
|---|
| 10 | // RUN: %clang_cc1 -emit-llvm -cl-denorms-are-zero -o - -triple amdgcn--amdhsa -target-cpu tahiti %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH-OPT %s |
|---|
| 11 | |
|---|
| 12 | // Fast F32 denormals, but slow FMAF |
|---|
| 13 | // RUN: %clang_cc1 -emit-llvm -o - -triple amdgcn--amdhsa -target-cpu fiji %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH %s |
|---|
| 14 | // RUN: %clang_cc1 -emit-llvm -cl-denorms-are-zero -o - -triple amdgcn--amdhsa -target-cpu fiji %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH-OPT %s |
|---|
| 15 | |
|---|
| 16 | // Fast F32 denormals and fast FMAF |
|---|
| 17 | // RUN: %clang_cc1 -emit-llvm -o - -triple amdgcn--amdhsa -target-cpu gfx900 %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-DENORM %s |
|---|
| 18 | // RUN: %clang_cc1 -emit-llvm -cl-denorms-are-zero -o - -triple amdgcn--amdhsa -target-cpu gfx900 %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FLUSH-OPT %s |
|---|
| 19 | |
|---|
| 20 | // RUN: %clang_cc1 -emit-llvm -target-feature +fp32-denormals -target-feature -fp64-fp16-denormals -cl-denorms-are-zero -o - -triple amdgcn--amdhsa -target-cpu fiji %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FEATURE %s |
|---|
| 21 | // RUN: %clang_cc1 -emit-llvm -target-feature +fp32-denormals -target-feature -fp64-fp16-denormals -cl-denorms-are-zero -o - -triple amdgcn--amdhsa -target-cpu pitcairn %s | FileCheck -check-prefixes=AMDGCN,AMDGCN-FEATURE %s |
|---|
| 22 | |
|---|
| 23 | |
|---|
| 24 | |
|---|
| 25 | // For all targets 'denorms-are-zero' attribute is set to 'true' |
|---|
| 26 | // if '-cl-denorms-are-zero' was specified and to 'false' otherwise. |
|---|
| 27 | |
|---|
| 28 | // CHECK-LABEL: define {{(dso_local )?}}void @f() |
|---|
| 29 | // CHECK: attributes #{{[0-9]*}} = {{{[^}]*}} "denorms-are-zero"="false" |
|---|
| 30 | // |
|---|
| 31 | // DENORM-ZERO-LABEL: define {{(dso_local )?}}void @f() |
|---|
| 32 | // DENORM-ZERO: attributes #{{[0-9]*}} = {{{[^}]*}} "denorms-are-zero"="true" |
|---|
| 33 | |
|---|
| 34 | // For amdgcn target cpu fiji, fp32 should be flushed since fiji does not support fp32 denormals, unless +fp32-denormals is |
|---|
| 35 | // explicitly set. amdgcn target always do not flush fp64 denormals. The control for fp64 and fp16 denormals is the same. |
|---|
| 36 | |
|---|
| 37 | // AMDGCN-LABEL: define void @f() |
|---|
| 38 | |
|---|
| 39 | // AMDGCN-FLUSH: attributes #{{[0-9]*}} = {{{[^}]*}} "denorms-are-zero"="false" {{.*}} "target-features"="{{[^"]*}}+fp64-fp16-denormals,{{[^"]*}}-fp32-denormals{{[^"]*}}" |
|---|
| 40 | // AMDGCN-FLUSH-OPT: attributes #{{[0-9]*}} = {{{[^}]*}} "denorms-are-zero"="true" {{.*}} "target-features"="{{[^"]*}}+fp64-fp16-denormals,{{[^"]*}}-fp32-denormals{{[^"]*}}" |
|---|
| 41 | |
|---|
| 42 | // AMDGCN-DENORM: attributes #{{[0-9]*}} = {{{[^}]*}} "denorms-are-zero"="false" {{.*}} "target-features"="{{[^"]*}}+fp32-denormals,{{[^"]*}}+fp64-fp16-denormals{{[^"]*}}" |
|---|
| 43 | |
|---|
| 44 | // AMDGCN-FEATURE: attributes #{{[0-9]*}} = {{{[^}]*}} "denorms-are-zero"="true" {{.*}} "target-features"="{{[^"]*}}+fp32-denormals,{{[^"]*}}-fp64-fp16-denormals{{[^"]*}}" |
|---|
| 45 | void f() {} |
|---|
| 46 | |
|---|