ZarkoCA updated this revision to Diff 376630.
ZarkoCA edited the summary of this revision.
ZarkoCA added a comment.
Herald added a subscriber: steven.zhang.
- Only truncate **b** and not result
- Add a proper test case
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D106409/new/
https://reviews.llvm.org/D106409
Files:
clang/lib/Headers/altivec.h
clang/test/CodeGen/ppc-vec_ct-truncate.c
Index: clang/test/CodeGen/ppc-vec_ct-truncate.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/ppc-vec_ct-truncate.c
@@ -0,0 +1,90 @@
+// REQUIRES: powerpc-registered-target
+// RUN: %clang_cc1 -target-feature +altivec -target-feature +vsx \
+// RUN: -triple powerpc64-ibm-aix-xcoff -emit-llvm %s -o - \
+// RUN: -target-cpu pwr8 | FileCheck %s
+// RUN: %clang_cc1 -target-feature +altivec -target-feature +vsx \
+// RUN: -triple powerpc64-unknown-linux-gnu -emit-llvm %s -o - \
+// RUN: -target-cpu pwr8 | FileCheck %s
+// RUN: %clang_cc1 -target-feature +altivec -target-feature +vsx \
+// RUN: -triple powerpc64le-unknown-linux-gnu -emit-llvm %s -o - \
+// RUN: -target-cpu pwr8 | FileCheck %s
+#include <altivec.h>
+vector double a1 = {-1.234e-5, 1.2345};
+vector double a2 = {-1.234e-5, 1.2345};
+vector signed int vd1, vd2;
+vector double vd3 = {0.234, 1.234};
+vector float vf1 = {0.234, 1.234, 2.345, 3.456};
+vector float vf2 = {0.234, 1.234, 2.345, 3.456};
+vector float vf3 = {0.234, 1.234, 2.345, 3.456};
+vector float vf4 = {0.234, 1.234, 2.345, 3.456};
+vector float vf5 = {0.234, 1.234, 2.345, 3.456};
+vector signed int vsi1 = {1, 2, 3, 4};
+vector signed int vsi2 = {1, 2, 3, 4};
+vector signed int vsi3 = {1, 2, 3, 4};
+vector signed int vsi4 = {1, 2, 3, 4};
+vector long long int vl1 = {5000000000, 100000000000000000};
+vector long long int vl2 = {5000000000, 100000000000000000};
+vector double vd;
+vector float vf;
+vector signed long long vsll;
+vector unsigned int vsi5;
+vector unsigned int vsi6;
+vector unsigned long long vull;
+
+void test() {
+ // CHECK-LABEL: @test(
+ // CHECK-NEXT: entry:
+ // CHECK-LE-LABEL: @test(
+ // CHECK-LE-NEXT: entry:
+
+ vd1 = vec_cts(a1, 31);
+ // CHECK: [[TMP0:%.*]] = load <2 x double>, <2 x double>* @a1, align 16
+ // CHECK-NEXT: fmul <2 x double> [[TMP0]], <double 0x41E0000000000000, double 0x41E0000000000000>
+
+ vd2 = vec_cts(a2, 500);
+ // CHECK: [[TMP4:%.*]] = load <2 x double>, <2 x double>* @a2, align 16
+ // CHECK-NEXT: fmul <2 x double> [[TMP4]], <double 0x4130000000000000, double 0x4130000000000000>
+
+ vsi3 = vec_ctu(vf1, 31);
+ // CHECK: [[TMP8:%.*]] = load <4 x float>, <4 x float>* @vf1, align 16
+ // CHECK-NEXT: call <4 x i32> @llvm.ppc.altivec.vctuxs(<4 x float> [[TMP8]], i32 31)
+
+ vsi4 = vec_ctu(vf2, 500);
+ // CHECK: [[TMP10:%.*]] = load <4 x float>, <4 x float>* @vf2, align 16
+ // CHECK-NEXT: call <4 x i32> @llvm.ppc.altivec.vctuxs(<4 x float> [[TMP10]], i32 20)
+
+ vull = vec_ctul(vf1, 31);
+ // CHECK: [[TMP12:%.*]] = load <4 x float>, <4 x float>* @vf1, align 16
+ // CHECK-NEXT: fmul <4 x float> [[TMP12]], <float 0x41E0000000000000, float 0x41E0000000000000, float 0x41E0000000000000, float 0x41E0000000000000>
+
+ vull = vec_ctul(vf3, 500);
+ // CHECK: [[TMP21:%.*]] = load <4 x float>, <4 x float>* @vf3, align 16
+ // CHECK-NEXT: fmul <4 x float> [[TMP21]], <float 0x4130000000000000, float 0x4130000000000000, float 0x4130000000000000, float 0x4130000000000000>
+
+ vsll = vec_ctsl(vf4, 31);
+ // CHECK: [[TMP30:%.*]] = load <4 x float>, <4 x float>* @vf4, align 16
+ // CHECK-NEXT: fmul <4 x float> [[TMP30]], <float 0x41E0000000000000, float 0x41E0000000000000, float 0x41E0000000000000, float 0x41E0000000000000>
+
+ vsll = vec_ctsl(vf5, 500);
+ // CHECK: [[TMP39:%.*]] = load <4 x float>, <4 x float>* @vf5, align 16
+ // CHECK-NEXT: fmul <4 x float> [[TMP39]], <float 0x4130000000000000, float 0x4130000000000000, float 0x4130000000000000, float 0x4130000000000000>
+
+ vf = vec_ctf(vsi1, 31);
+ // CHECK: [[TMP48:%.*]] = load <4 x i32>, <4 x i32>* @vsi1, align 16
+ // CHECK-NEXT: call <4 x float> @llvm.ppc.altivec.vcfsx(<4 x i32> [[TMP48]], i32 31)
+
+ vf = vec_ctf(vsi2, 500);
+ // CHECK: [[TMP50:%.*]] = load <4 x i32>, <4 x i32>* @vsi2, align 16
+ // CHECK-NEXT: call <4 x float> @llvm.ppc.altivec.vcfsx(<4 x i32> [[TMP50]], i32 20)
+
+ vd = vec_ctd(vsi3, 31);
+ // CHECK: [[TMP53:%.*]] = load <4 x i32>, <4 x i32>* @vsi3, align 16
+ // CHECK: [[TMP83:%.*]] = call <2 x double> @llvm.ppc.vsx.xvcvsxwdp(<4 x i32> [[TMP82:%.*]])
+ // CHECK-NEXT: fmul <2 x double> [[TMP83]], <double 0x3E00000000000000, double 0x3E00000000000000>
+
+ vd = vec_ctd(vsi4, 500);
+
+ // CHECK: [[TMP84:%.*]] = load <4 x i32>, <4 x i32>* @vsi4, align 16
+ // CHECK: [[TMP115:%.*]] = call <2 x double> @llvm.ppc.vsx.xvcvsxwdp(<4 x i32> [[TMP114:%.*]])
+ // CHECK-NEXT: fmul <2 x double> [[TMP115]], <double 0x3EB0000000000000, double 0x3EB0000000000000>
+}
Index: clang/lib/Headers/altivec.h
===================================================================
--- clang/lib/Headers/altivec.h
+++ clang/lib/Headers/altivec.h
@@ -3178,64 +3178,69 @@
#ifdef __XL_COMPAT_ALTIVEC__
#define vec_ctf(__a, __b) \
_Generic((__a), vector int \
- : (vector float)__builtin_altivec_vcfsx((vector int)(__a), (__b)), \
+ : (vector float)__builtin_altivec_vcfsx((vector int)(__a), \
+ ((__b)&0x1F)), \
vector unsigned int \
: (vector float)__builtin_altivec_vcfux((vector unsigned int)(__a), \
- (__b)), \
+ ((__b)&0x1F)), \
vector unsigned long long \
: (__builtin_vsx_xvcvuxdsp((vector unsigned long long)(__a)) * \
- (vector float)(vector unsigned)((0x7f - (__b)) << 23)), \
+ (vector float)(vector unsigned)((0x7f - ((__b)&0x1F)) << 23)), \
vector signed long long \
: (__builtin_vsx_xvcvsxdsp((vector signed long long)(__a)) * \
- (vector float)(vector unsigned)((0x7f - (__b)) << 23)))
+ (vector float)(vector unsigned)((0x7f - ((__b)&0x1F)) << 23)))
#else // __XL_COMPAT_ALTIVEC__
#define vec_ctf(__a, __b) \
- _Generic((__a), vector int \
- : (vector float)__builtin_altivec_vcfsx((vector int)(__a), (__b)), \
- vector unsigned int \
- : (vector float)__builtin_altivec_vcfux((vector unsigned int)(__a), \
- (__b)), \
- vector unsigned long long \
- : (__builtin_convertvector((vector unsigned long long)(__a), \
- vector double) * \
- (vector double)(vector unsigned long long)((0x3ffULL - (__b)) \
- << 52)), \
- vector signed long long \
- : (__builtin_convertvector((vector signed long long)(__a), \
- vector double) * \
- (vector double)(vector unsigned long long)((0x3ffULL - (__b)) \
- << 52)))
+ _Generic( \
+ (__a), vector int \
+ : (vector float)__builtin_altivec_vcfsx((vector int)(__a), \
+ ((__b)&0x1F)), \
+ vector unsigned int \
+ : (vector float)__builtin_altivec_vcfux((vector unsigned int)(__a), \
+ ((__b)&0x1F)), \
+ vector unsigned long long \
+ : (__builtin_convertvector((vector unsigned long long)(__a), \
+ vector double) * \
+ (vector double)(vector unsigned long long)((0x3ffULL - ((__b)&0x1F)) \
+ << 52)), \
+ vector signed long long \
+ : (__builtin_convertvector((vector signed long long)(__a), \
+ vector double) * \
+ (vector double)(vector unsigned long long)((0x3ffULL - ((__b)&0x1F)) \
+ << 52)))
#endif // __XL_COMPAT_ALTIVEC__
#else
#define vec_ctf(__a, __b) \
_Generic((__a), vector int \
- : (vector float)__builtin_altivec_vcfsx((vector int)(__a), (__b)), \
+ : (vector float)__builtin_altivec_vcfsx((vector int)(__a), \
+ ((__b)&0x1F)), \
vector unsigned int \
: (vector float)__builtin_altivec_vcfux((vector unsigned int)(__a), \
- (__b)))
+ ((__b)&0x1F)))
#endif
/* vec_ctd */
#ifdef __VSX__
#define vec_ctd(__a, __b) \
- _Generic((__a), vector signed int \
- : (vec_doublee((vector signed int)(__a)) * \
- (vector double)(vector unsigned long long)((0x3ffULL - (__b)) \
- << 52)), \
- vector unsigned int \
- : (vec_doublee((vector unsigned int)(__a)) * \
- (vector double)(vector unsigned long long)((0x3ffULL - (__b)) \
- << 52)), \
- vector unsigned long long \
- : (__builtin_convertvector((vector unsigned long long)(__a), \
- vector double) * \
- (vector double)(vector unsigned long long)((0x3ffULL - (__b)) \
- << 52)), \
- vector signed long long \
- : (__builtin_convertvector((vector signed long long)(__a), \
- vector double) * \
- (vector double)(vector unsigned long long)((0x3ffULL - (__b)) \
- << 52)))
+ _Generic( \
+ (__a), vector signed int \
+ : (vec_doublee((vector signed int)(__a)) * \
+ (vector double)(vector unsigned long long)((0x3ffULL - ((__b)&0x1F)) \
+ << 52)), \
+ vector unsigned int \
+ : (vec_doublee((vector unsigned int)(__a)) * \
+ (vector double)(vector unsigned long long)((0x3ffULL - ((__b)&0x1F)) \
+ << 52)), \
+ vector unsigned long long \
+ : (__builtin_convertvector((vector unsigned long long)(__a), \
+ vector double) * \
+ (vector double)(vector unsigned long long)((0x3ffULL - ((__b)&0x1F)) \
+ << 52)), \
+ vector signed long long \
+ : (__builtin_convertvector((vector signed long long)(__a), \
+ vector double) * \
+ (vector double)(vector unsigned long long)((0x3ffULL - ((__b)&0x1F)) \
+ << 52)))
#endif // __VSX__
/* vec_vcfsx */
@@ -3251,24 +3256,26 @@
#ifdef __XL_COMPAT_ALTIVEC__
#define vec_cts(__a, __b) \
_Generic((__a), vector float \
- : __builtin_altivec_vctsxs((vector float)(__a), (__b)), \
+ : __builtin_altivec_vctsxs((vector float)(__a), ((__b)&0x1F)), \
vector double \
: __extension__({ \
vector double __ret = \
(vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + (__b)) \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
<< 52); \
__builtin_vsx_xvcvdpsxws(__ret); \
}))
#else // __XL_COMPAT_ALTIVEC__
#define vec_cts(__a, __b) \
_Generic((__a), vector float \
- : __builtin_altivec_vctsxs((vector float)(__a), (__b)), \
+ : __builtin_altivec_vctsxs((vector float)(__a), ((__b)&0x1F)), \
vector double \
: __extension__({ \
vector double __ret = \
(vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + (__b)) \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
<< 52); \
__builtin_convertvector(__ret, vector signed long long); \
}))
@@ -3287,24 +3294,26 @@
#ifdef __XL_COMPAT_ALTIVEC__
#define vec_ctu(__a, __b) \
_Generic((__a), vector float \
- : __builtin_altivec_vctuxs((vector float)(__a), (__b)), \
+ : __builtin_altivec_vctuxs((vector float)(__a), ((__b)&0x1F)), \
vector double \
: __extension__({ \
vector double __ret = \
(vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + __b) \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
<< 52); \
__builtin_vsx_xvcvdpuxws(__ret); \
}))
#else // __XL_COMPAT_ALTIVEC__
#define vec_ctu(__a, __b) \
_Generic((__a), vector float \
- : __builtin_altivec_vctuxs((vector float)(__a), (__b)), \
+ : __builtin_altivec_vctuxs((vector float)(__a), ((__b)&0x1F)), \
vector double \
: __extension__({ \
vector double __ret = \
(vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + __b) \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
<< 52); \
__builtin_convertvector(__ret, vector unsigned long long); \
}))
@@ -3318,42 +3327,44 @@
#ifdef __VSX__
#define vec_ctsl(__a, __b) \
- _Generic((__a), vector float \
- : __extension__({ \
- vector float __ret = \
- (vector float)(__a) * \
- (vector float)(vector unsigned)((0x7f + (__b)) << 23); \
- __builtin_vsx_xvcvspsxds( \
- __builtin_vsx_xxsldwi(__ret, __ret, 1)); \
- }), \
- vector double \
- : __extension__({ \
- vector double __ret = \
- (vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + __b) \
- << 52); \
- __builtin_convertvector(__ret, vector signed long long); \
- }))
+ _Generic( \
+ (__a), vector float \
+ : __extension__({ \
+ vector float __ret = \
+ (vector float)(__a) * \
+ (vector float)(vector unsigned)((0x7f + ((__b)&0x1F)) << 23); \
+ __builtin_vsx_xvcvspsxds(__builtin_vsx_xxsldwi(__ret, __ret, 1)); \
+ }), \
+ vector double \
+ : __extension__({ \
+ vector double __ret = \
+ (vector double)(__a) * \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
+ << 52); \
+ __builtin_convertvector(__ret, vector signed long long); \
+ }))
/* vec_ctul */
#define vec_ctul(__a, __b) \
- _Generic((__a), vector float \
- : __extension__({ \
- vector float __ret = \
- (vector float)(__a) * \
- (vector float)(vector unsigned)((0x7f + (__b)) << 23); \
- __builtin_vsx_xvcvspuxds( \
- __builtin_vsx_xxsldwi(__ret, __ret, 1)); \
- }), \
- vector double \
- : __extension__({ \
- vector double __ret = \
- (vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + __b) \
- << 52); \
- __builtin_convertvector(__ret, vector unsigned long long); \
- }))
+ _Generic( \
+ (__a), vector float \
+ : __extension__({ \
+ vector float __ret = \
+ (vector float)(__a) * \
+ (vector float)(vector unsigned)((0x7f + ((__b)&0x1F)) << 23); \
+ __builtin_vsx_xvcvspuxds(__builtin_vsx_xxsldwi(__ret, __ret, 1)); \
+ }), \
+ vector double \
+ : __extension__({ \
+ vector double __ret = \
+ (vector double)(__a) * \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
+ << 52); \
+ __builtin_convertvector(__ret, vector unsigned long long); \
+ }))
#endif
#else // __LITTLE_ENDIAN__
/* vec_ctsl */
@@ -3364,14 +3375,16 @@
: __extension__({ \
vector float __ret = \
(vector float)(__a) * \
- (vector float)(vector unsigned)((0x7f + (__b)) << 23); \
+ (vector float)(vector unsigned)((0x7f + ((__b)&0x1F)) \
+ << 23); \
__builtin_vsx_xvcvspsxds(__ret); \
}), \
vector double \
: __extension__({ \
vector double __ret = \
(vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + __b) \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b & 0x1F))) \
<< 52); \
__builtin_convertvector(__ret, vector signed long long); \
}))
@@ -3383,14 +3396,16 @@
: __extension__({ \
vector float __ret = \
(vector float)(__a) * \
- (vector float)(vector unsigned)((0x7f + (__b)) << 23); \
+ (vector float)(vector unsigned)((0x7f + ((__b)&0x1F)) \
+ << 23); \
__builtin_vsx_xvcvspuxds(__ret); \
}), \
vector double \
: __extension__({ \
vector double __ret = \
(vector double)(__a) * \
- (vector double)(vector unsigned long long)((0x3ffULL + __b) \
+ (vector double)(vector unsigned long long)((0x3ffULL + \
+ ((__b)&0x1F)) \
<< 52); \
__builtin_convertvector(__ret, vector unsigned long long); \
}))
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
