From 3445dcec31a4e8a2b2da5ff420059d106af2bfab Mon Sep 17 00:00:00 2001 From: Andrew Senkevich Date: Mon, 15 Feb 2016 15:55:39 +0300 Subject: Fix and test for *finite workaround in libmvec. [BZ #19590] modified: math/Makefile modified: sysdeps/x86_64/fpu/Makefile modified: sysdeps/x86_64/fpu/svml_finite_alias.S new file: sysdeps/x86_64/fpu/test-libmvec-alias-main.c new file: sysdeps/x86_64/fpu/test-libmvec-alias.c --- math/Makefile | 12 +++- sysdeps/x86_64/fpu/Makefile | 7 ++ sysdeps/x86_64/fpu/svml_finite_alias.S | 3 +- sysdeps/x86_64/fpu/test-libmvec-alias-main.c | 56 +++++++++++++++ sysdeps/x86_64/fpu/test-libmvec-alias.c | 101 +++++++++++++++++++++++++++ 5 files changed, 176 insertions(+), 3 deletions(-) create mode 100644 sysdeps/x86_64/fpu/test-libmvec-alias-main.c create mode 100644 sysdeps/x86_64/fpu/test-libmvec-alias.c diff --git a/math/Makefile b/math/Makefile index 7d573a0074..b2bc67c705 100644 --- a/math/Makefile +++ b/math/Makefile @@ -126,15 +126,17 @@ test-longdouble-yes = test-ldouble test-ildoubl test-ldouble-finite ifneq (no,$(PERL)) libm-vec-tests = $(addprefix test-,$(libmvec-tests)) +libm-vec-alias-tests = $(addprefix test-,$(libmvec-alias-tests)) libm-tests = test-float test-double $(test-longdouble-$(long-double-fcts)) \ test-ifloat test-idouble test-float-finite test-double-finite \ - $(libm-vec-tests) + $(libm-vec-tests) $(libm-vec-alias-tests) libm-tests.o = $(addsuffix .o,$(libm-tests)) tests += $(libm-tests) libm-tests-generated = libm-test-ulps.h libm-have-vector-test.h libm-test.c generated += $(libm-tests-generated) libm-test.stmp + # This is needed for dependencies before-compile += $(objpfx)libm-test.c @@ -234,6 +236,7 @@ include ../Rules ifneq (no,$(PERL)) # This must come after the inclusion of sysdeps Makefiles via Rules. $(addprefix $(objpfx), $(addsuffix .o, $(libm-vec-tests))): $(objpfx)libm-test.stmp +$(addprefix $(objpfx), $(addsuffix .os, $(libmvec-alias-tests))): $(objpfx)libm-test.stmp # Run the math programs to automatically generate ULPs files. .PHONY: regen-ulps @@ -301,6 +304,13 @@ $(addprefix $(objpfx),$(tests-static)): $(objpfx)libm.a $(addprefix $(objpfx), $(libm-vec-tests)): $(objpfx)%: $(objpfx)%-wrappers.o $(libm) \ $(libmvec) +# $(objpfx)$(libm-vec-alias-tests): $($(addprefix $(objpfx),$(libm-vec-alias-tests)):%=%.os) $(objpfx)../mathvec/libmvec_nonshared.a $(libmvec) + +$(addprefix $(objpfx), $(libm-vec-alias-tests)): $(addprefix $(objpfx), $(addsuffix -main.o, $(libm-vec-alias-tests))) $(objpfx)../mathvec/libmvec_nonshared.a $(libmvec) + +# $(objpfx)test-libmvec-alias.so: $(addprefix $(objpfx),addsuffix .os, $(libm-vec-alias-tests)) $(objpfx)../mathvec/libmvec_nonshared.a $(libmvec) + + gmp-objs = $(patsubst %,$(common-objpfx)stdlib/%.o,\ add_n sub_n cmp addmul_1 mul_1 mul_n divmod_1 \ lshift rshift mp_clz_tab udiv_qrnnd inlines \ diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile index 88742faff1..807c208853 100644 --- a/sysdeps/x86_64/fpu/Makefile +++ b/sysdeps/x86_64/fpu/Makefile @@ -30,9 +30,11 @@ ifeq ($(subdir),math) ifeq ($(build-mathvec),yes) libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \ float-vlen4 float-vlen8 float-vlen8-avx2 +libmvec-alias-tests = libmvec-alias ifeq (yes,$(config-cflags-avx512)) libmvec-tests += double-vlen8 float-vlen16 +libmvec-alias-tests += libmvec-alias-avx512 endif double-vlen4-arch-ext-cflags = -mavx @@ -43,6 +45,11 @@ float-vlen8-arch-ext-cflags = -mavx float-vlen8-arch-ext2-cflags = -mavx2 float-vlen16-arch-ext-cflags = -mavx512f +CFLAGS-test-libmvec-alias.c = -fPIC -Wl,-shared $(double-vlen4-arch-ext2-cflags) +LDFLAGS-test-libmvec-alias.so = -shared +CFLAGS-test-libmvec-alias-avx512.c = -fPIC -Wl,-shared $(double-vlen8-arch-ext-cflags) +LDFLAGS-test-libmvec-alias-avx512.so = -shared + CFLAGS-test-double-vlen4-avx2.c = $(libm-test-vec-cflags) CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags) diff --git a/sysdeps/x86_64/fpu/svml_finite_alias.S b/sysdeps/x86_64/fpu/svml_finite_alias.S index 0062fe49ff..8314cf4a9a 100644 --- a/sysdeps/x86_64/fpu/svml_finite_alias.S +++ b/sysdeps/x86_64/fpu/svml_finite_alias.S @@ -23,8 +23,7 @@ #define ALIAS_IMPL(alias, target) \ ENTRY (alias); \ - call target; \ - ret; \ + jmp target@PLT; \ END (alias) .text diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias-main.c b/sysdeps/x86_64/fpu/test-libmvec-alias-main.c new file mode 100644 index 0000000000..acd465b0d7 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-libmvec-alias-main.c @@ -0,0 +1,56 @@ + +extern double log_vlen2 (double); +extern double exp_vlen2 (double); +extern double pow_vlen2 (double, double); + +extern double log_vlen4 (double); +extern double exp_vlen4 (double); +extern double pow_vlen4 (double, double); + +extern double log_vlen4_avx2 (double); +extern double exp_vlen4_avx2 (double); +extern double pow_vlen4_avx2 (double, double); + +extern float logf_vlen4 (float); +extern float expf_vlen4 (float); +extern float powf_vlen4 (float, float); + +extern float logf_vlen8 (float); +extern float expf_vlen8 (float); +extern float powf_vlen8 (float, float); + +extern float logf_vlen8_avx2 (float); +extern float expf_vlen8_avx2 (float); +extern float powf_vlen8_avx2 (float, float); + +int main( void) +{ + double res = 0.0; + float resf = 0.0; + + res += log_vlen2(1.0); + res += exp_vlen2(1.0); + res += pow_vlen2(1.1, 1.1); + + res += log_vlen4(1.0); + res += exp_vlen4(1.0); + res += pow_vlen4(1.1, 1.1); + + res += log_vlen4_avx2(1.0); + res += exp_vlen4_avx2(1.0); + res += pow_vlen4_avx2(1.1, 1.1); + + resf += logf_vlen4(1.0); + resf += expf_vlen4(1.0); + resf += powf_vlen4(1.1, 1.1); + + resf += logf_vlen8(1.0); + resf += expf_vlen8(1.0); + resf += powf_vlen8(1.1, 1.1); + + resf += logf_vlen8_avx2(1.0); + resf += expf_vlen8_avx2(1.0); + resf += powf_vlen8_avx2(1.1, 1.1); + + return 0; +} diff --git a/sysdeps/x86_64/fpu/test-libmvec-alias.c b/sysdeps/x86_64/fpu/test-libmvec-alias.c new file mode 100644 index 0000000000..6a6a103165 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-libmvec-alias.c @@ -0,0 +1,101 @@ +#include + +#include "test-double-vlen2.h" +#define VEC_TYPE __m128d + +VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVbN2v___log_finite) +VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVbN2v___exp_finite) +VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVbN2vv___pow_finite) + +#undef VEC_TYPE +#undef VECTOR_WRAPPER +#undef VECTOR_WRAPPER_ff +#undef VEC_SUFF +#undef VEC_LEN + +#include "test-double-vlen4.h" +#define VEC_TYPE __m256d + +VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVcN4v___log_finite) +VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVcN4v___exp_finite) +VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVcN4vv___pow_finite) + +#undef VEC_SUFF +#define VEC_SUFF _vlen4_avx2 + +VECTOR_WRAPPER (WRAPPER_NAME (log), _ZGVdN4v___log_finite) +VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVdN4v___exp_finite) +VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVdN4vv___pow_finite) + +#undef FUNC +#undef FLOAT +#undef BUILD_COMPLEX +#undef TEST_MSG +#undef CHOOSE +#undef FUNC_TEST +#undef VEC_TYPE +#undef VECTOR_WRAPPER +#undef VECTOR_WRAPPER_ff +#undef VEC_SUFF +#undef VEC_LEN + +#include "test-float-vlen4.h" +#define VEC_TYPE __m128 + +VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVbN4v___logf_finite) +VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVbN4v___expf_finite) +VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVbN4vv___powf_finite) + +#undef VEC_TYPE +#undef VECTOR_WRAPPER +#undef VECTOR_WRAPPER_ff +#undef VEC_SUFF +#undef VEC_LEN + +#include "test-float-vlen8.h" +#define VEC_TYPE __m256 + +VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVcN8v___logf_finite) +VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVcN8v___expf_finite) +VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVcN8vv___powf_finite) + +#undef VEC_SUFF +#define VEC_SUFF _vlen8_avx2 + +VECTOR_WRAPPER (WRAPPER_NAME (logf), _ZGVdN8v___logf_finite) +VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVdN8v___expf_finite) +VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVdN8vv___powf_finite) + +/* +int main( void) +{ + double res = 0.0; + float resf = 0.0; + + res += log_vlen2(1.0); + res += exp_vlen2(1.0); + res += pow_vlen2(1.1, 1.1); + + res += log_vlen4(1.0); + res += exp_vlen4(1.0); + res += pow_vlen4(1.1, 1.1); + + res += log_vlen4_avx2(1.0); + res += exp_vlen4_avx2(1.0); + res += pow_vlen4_avx2(1.1, 1.1); + + resf += logf_vlen4(1.0); + resf += expf_vlen4(1.0); + resf += powf_vlen4(1.1, 1.1); + + resf += logf_vlen8(1.0); + resf += expf_vlen8(1.0); + resf += powf_vlen8(1.1, 1.1); + + resf += logf_vlen8_avx2(1.0); + resf += expf_vlen8_avx2(1.0); + resf += powf_vlen8_avx2(1.1, 1.1); + + return 0; +} +*/ -- cgit v1.2.1