* [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig
@ 2021-10-15 13:12 H.J. Lu
2021-10-20 16:00 ` PING " H.J. Lu
0 siblings, 1 reply; 6+ messages in thread
From: H.J. Lu @ 2021-10-15 13:12 UTC (permalink / raw)
To: libc-alpha
1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which
contains libmvec ABI test dependencies and CFLAGS, in the build directory.
2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS.
Tested on SSE4, AVX, AVX2 and AVX512 machines.
---
sysdeps/x86_64/fpu/Makeconfig | 92 ++++++++++++++++++
sysdeps/x86_64/fpu/Makefile | 134 +++++++++-----------------
sysdeps/x86_64/fpu/multiarch/Makefile | 68 ++++---------
3 files changed, 155 insertions(+), 139 deletions(-)
create mode 100644 sysdeps/x86_64/fpu/Makeconfig
diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig
new file mode 100644
index 0000000000..0a9341cdd9
--- /dev/null
+++ b/sysdeps/x86_64/fpu/Makeconfig
@@ -0,0 +1,92 @@
+# Generate libmvec.mk for libmvec ABI tests.
+
+# Copyright (C) 2021 Free Software Foundation, Inc.
+# This file is part of the GNU C Library.
+
+# The GNU C Library is free software; you can redistribute it and/or
+# modify it under the terms of the GNU Lesser General Public
+# License as published by the Free Software Foundation; either
+# version 2.1 of the License, or (at your option) any later version.
+
+# The GNU C Library is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+# Lesser General Public License for more details.
+
+# You should have received a copy of the GNU Lesser General Public
+# License along with the GNU C Library; if not, see
+# <https://www.gnu.org/licenses/>.
+
+-include $(common-objpfx)libmvec.mk
+postclean-generated += libmvec.mk
+
+# Define for both math and mathvec directories.
+libmvec-funcs = \
+ cos \
+ exp \
+ log \
+ pow \
+ sin \
+ sincos \
+
+# FIXME: Add ABI tests for all libmvec functions and replace
+# libmvec-abi-funcs with libmvec-funcs.
+libmvec-abi-funcs = \
+ sincos
+
+# The base libmvec ABI tests.
+libmvec-abi-func-tests = \
+ $(addprefix test-double-libmvec-,$(libmvec-abi-funcs)) \
+ $(addsuffix f,$(addprefix test-float-libmvec-,$(libmvec-abi-funcs)))
+
+# The AVX libmvec ABI tests.
+libmvec-abi-func-avx-tests = \
+ $(addsuffix -avx,$(libmvec-abi-func-tests))
+#
+# The AVX2 libmvec ABI tests.
+libmvec-abi-func-avx2-tests = \
+ $(addsuffix -avx2,$(libmvec-abi-func-tests))
+#
+# The AVX512F libmvec ABI tests.
+libmvec-abi-func-avx512f-tests = \
+ $(addsuffix -avx512f,$(libmvec-abi-func-tests))
+
+$(common-objpfx)libmvec.mk: $(common-objpfx)config.make
+ (echo "ifeq (\$$(subdir)\$$(build-mathvec),mathyes)"; \
+ echo; \
+ for i in "" "-avx" "-avx2" "-avx512f"; do \
+ for t in $(libmvec-abi-func-tests); do \
+ echo "\$$(objpfx)$$t$$i: \\"; \
+ echo " \$$(objpfx)$$t$$i.o \\"; \
+ echo " \$$(objpfx)test-libmvec$$i.o \$$(libmvec)"; \
+ echo; \
+ done; \
+ done; \
+ for t in $(libmvec-abi-funcs); do \
+ echo "CFLAGS-test-double-libmvec-$$t.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags)"; \
+ echo "CFLAGS-test-double-libmvec-$$t-avx.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags) \\"; \
+ echo " \$$(double-vlen4-arch-ext-cflags)"; \
+ echo "CFLAGS-test-double-libmvec-$$t-avx2.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags) \\"; \
+ echo " \$$(double-vlen4-arch-ext2-cflags)"; \
+ echo "CFLAGS-test-double-libmvec-$$t-avx512f.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags) \\"; \
+ echo " \$$(double-vlen8-arch-ext-cflags)"; \
+ echo; \
+ echo "CFLAGS-test-float-libmvec-$${t}f.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags)"; \
+ echo "CFLAGS-test-float-libmvec-$${t}f-avx.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags) \\"; \
+ echo " \$$(float-vlen8-arch-ext-cflags)"; \
+ echo "CFLAGS-test-float-libmvec-$${t}f-avx2.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags) \\"; \
+ echo " \$$(float-vlen8-arch-ext2-cflags)"; \
+ echo "CFLAGS-test-float-libmvec-$${t}f-avx512f.c = \\"; \
+ echo " \$$(libmvec-abi-test-cflags) \\"; \
+ echo " \$$(float-vlen16-arch-ext-cflags)"; \
+ echo; \
+ done; \
+ echo "endif") > $@T
+ mv -f $@T $@
diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
index e5a65750c0..d172ae815d 100644
--- a/sysdeps/x86_64/fpu/Makefile
+++ b/sysdeps/x86_64/fpu/Makefile
@@ -1,87 +1,48 @@
ifeq ($(subdir),mathvec)
-libmvec-support += svml_d_cos2_core svml_d_cos4_core_avx \
- svml_d_cos4_core svml_d_cos8_core \
- svml_d_sin2_core svml_d_sin4_core_avx \
- svml_d_sin4_core svml_d_sin8_core svml_d_trig_data \
- svml_s_cosf4_core svml_s_cosf8_core_avx \
- svml_s_cosf8_core svml_s_cosf16_core svml_s_trig_data \
- svml_s_sinf4_core svml_s_sinf8_core_avx \
- svml_s_sinf8_core svml_s_sinf16_core \
- svml_d_sincos2_core svml_d_sincos4_core_avx \
- svml_d_sincos4_core svml_d_sincos8_core \
- svml_d_log2_core svml_d_log4_core_avx svml_d_log4_core \
- svml_d_log8_core svml_d_log_data svml_s_logf4_core \
- svml_s_logf8_core_avx svml_s_logf8_core svml_s_logf16_core \
- svml_s_logf_data svml_d_exp2_core svml_d_exp4_core_avx \
- svml_d_exp4_core svml_d_exp8_core svml_d_exp_data \
- svml_s_expf4_core svml_s_expf8_core_avx svml_s_expf8_core \
- svml_s_expf16_core svml_s_expf_data svml_d_pow2_core \
- svml_d_pow4_core_avx svml_d_pow4_core svml_d_pow8_core \
- svml_d_pow_data svml_s_powf4_core svml_s_powf8_core_avx \
- svml_s_powf8_core svml_s_powf16_core svml_s_powf_data \
- svml_s_sincosf4_core svml_s_sincosf8_core_avx \
- svml_s_sincosf8_core svml_s_sincosf16_core
+libmvec-double-func-list = \
+ 2_core \
+ 4_core \
+ 4_core_avx \
+ 8_core
+libmvec-float-func-list = \
+ f4_core \
+ f8_core \
+ f8_core_avx \
+ f16_core
+libmvec-support += \
+ svml_d_exp_data \
+ svml_d_log_data \
+ svml_d_pow_data \
+ svml_d_trig_data \
+ svml_s_expf_data \
+ svml_s_logf_data \
+ svml_s_powf_data \
+ svml_s_trig_data \
+ $(foreach l,$(libmvec-double-func-list), \
+ $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
+ $(foreach l,$(libmvec-float-func-list), \
+ $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
endif
# Variables for libmvec tests.
-ifeq ($(subdir),math)
-ifeq ($(build-mathvec),yes)
+ifeq ($(subdir)$(build-mathvec),mathyes)
libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
- float-vlen4 float-vlen8 float-vlen8-avx2
-tests += test-double-libmvec-sincos test-double-libmvec-sincos-avx \
- test-double-libmvec-sincos-avx2 test-float-libmvec-sincosf \
- test-float-libmvec-sincosf-avx test-float-libmvec-sincosf-avx2
-test-extras += test-libmvec test-libmvec-avx \
- test-libmvec-avx2 test-libmvec-avx512f
-extra-test-objs += test-libmvec.o test-libmvec-avx.o \
- test-libmvec-avx2.o test-libmvec-avx512f.o
-
-$(objpfx)test-double-libmvec-sincos: \
- $(objpfx)test-double-libmvec-sincos.o \
- $(objpfx)test-libmvec.o $(libmvec)
-
-$(objpfx)test-double-libmvec-sincos-avx: \
- $(objpfx)test-double-libmvec-sincos-avx.o \
- $(objpfx)test-libmvec-avx.o $(libmvec)
-
-$(objpfx)test-double-libmvec-sincos-avx2: \
- $(objpfx)test-double-libmvec-sincos-avx2.o \
- $(objpfx)test-libmvec-avx2.o $(libmvec)
-
-$(objpfx)test-float-libmvec-sincosf: \
- $(objpfx)test-float-libmvec-sincosf.o \
- $(objpfx)test-libmvec.o $(libmvec)
-
-$(objpfx)test-float-libmvec-sincosf-avx: \
- $(objpfx)test-float-libmvec-sincosf-avx.o \
- $(objpfx)test-libmvec-avx.o $(libmvec)
-
-$(objpfx)test-float-libmvec-sincosf-avx2: \
- $(objpfx)test-float-libmvec-sincosf-avx2.o \
- $(objpfx)test-libmvec-avx2.o $(libmvec)
-
-libmvec-tests += double-vlen8 float-vlen16
-tests += test-double-libmvec-sincos-avx512f \
- test-float-libmvec-sincosf-avx512f
-test-extras += test-libmvec-avx512f
-extra-test-objs += test-libmvec-avx512f.o
-
-$(objpfx)test-double-libmvec-sincos-avx512f: \
- $(objpfx)test-double-libmvec-sincos-avx512f.o \
- $(objpfx)test-libmvec-avx512f.o $(libmvec)
-
-$(objpfx)test-float-libmvec-sincosf-avx512f: \
- $(objpfx)test-float-libmvec-sincosf-avx512f.o \
- $(objpfx)test-libmvec-avx512f.o $(libmvec)
-
-double-vlen2-funcs = cos exp log pow sin sincos
-double-vlen4-funcs = cos exp log pow sin sincos
-double-vlen4-avx2-funcs = cos exp log pow sin sincos
-double-vlen8-funcs = cos exp log pow sin sincos
-float-vlen4-funcs = cos exp log pow sin sincos
-float-vlen8-funcs = cos exp log pow sin sincos
-float-vlen8-avx2-funcs = cos exp log pow sin sincos
-float-vlen16-funcs = cos exp log pow sin sincos
+ float-vlen4 float-vlen8 float-vlen8-avx2 \
+ double-vlen8 float-vlen16
+tests += \
+ $(libmvec-abi-func-tests) \
+ $(libmvec-abi-func-avx-tests) \
+ $(libmvec-abi-func-avx2-tests) \
+ $(libmvec-abi-func-avx512f-tests)
+
+double-vlen2-funcs = $(libmvec-funcs)
+double-vlen4-funcs = $(libmvec-funcs)
+double-vlen4-avx2-funcs = $(libmvec-funcs)
+double-vlen8-funcs = $(libmvec-funcs)
+float-vlen4-funcs = $(libmvec-funcs)
+float-vlen8-funcs = $(libmvec-funcs)
+float-vlen8-avx2-funcs = $(libmvec-funcs)
+float-vlen16-funcs = $(libmvec-funcs)
double-vlen4-arch-ext-cflags = -mavx
double-vlen4-arch-ext2-cflags = -mavx2
@@ -91,22 +52,13 @@ float-vlen8-arch-ext-cflags = -mavx
float-vlen8-arch-ext2-cflags = -mavx2
float-vlen16-arch-ext-cflags = -mavx512f
-libmvec-sincos-cflags = $(libm-test-fast-math-cflags) -fno-inline -fopenmp -Wno-unknown-pragmas
+libmvec-abi-test-cflags = \
+ $(libm-test-fast-math-cflags) \
+ -fno-inline -fopenmp -Wno-unknown-pragmas
CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)
CFLAGS-test-float-vlen8-avx2-wrappers.c = $(float-vlen8-arch-ext2-cflags)
-
-CFLAGS-test-double-libmvec-sincos.c = $(libmvec-sincos-cflags)
-CFLAGS-test-double-libmvec-sincos-avx.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext-cflags)
-CFLAGS-test-double-libmvec-sincos-avx2.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext2-cflags)
-CFLAGS-test-double-libmvec-sincos-avx512f.c = $(libmvec-sincos-cflags) $(double-vlen8-arch-ext-cflags)
-
-CFLAGS-test-float-libmvec-sincosf.c = $(libmvec-sincos-cflags)
-CFLAGS-test-float-libmvec-sincosf-avx.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext-cflags)
-CFLAGS-test-float-libmvec-sincosf-avx2.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext2-cflags)
-CFLAGS-test-float-libmvec-sincosf-avx512f.c = $(libmvec-sincos-cflags) $(float-vlen16-arch-ext-cflags)
-endif
endif
ifeq ($(subdir)$(config-cflags-mprefer-vector-width),mathyes)
diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
index d425ffd6d3..ec796277a5 100644
--- a/sysdeps/x86_64/fpu/multiarch/Makefile
+++ b/sysdeps/x86_64/fpu/multiarch/Makefile
@@ -59,52 +59,24 @@ CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
endif
ifeq ($(subdir),mathvec)
-libmvec-sysdep_routines += svml_d_cos2_core_sse4 svml_d_cos4_core_avx2 \
- svml_d_cos8_core_avx512 svml_d_sin2_core_sse4 \
- svml_d_sin4_core_avx2 svml_d_sin8_core_avx512 \
- svml_d_log2_core_sse4 svml_d_log4_core_avx2 \
- svml_d_log8_core_avx512 svml_d_sincos2_core_sse4 \
- svml_d_sincos4_core_avx2 svml_d_sincos8_core_avx512 \
- svml_s_cosf4_core_sse4 svml_s_cosf8_core_avx2 \
- svml_s_cosf16_core_avx512 svml_s_sinf4_core_sse4 \
- svml_s_sinf8_core_avx2 svml_s_sinf16_core_avx512 \
- svml_s_logf4_core_sse4 svml_s_logf8_core_avx2 \
- svml_s_logf16_core_avx512 svml_d_exp2_core_sse4 \
- svml_d_exp4_core_avx2 svml_d_exp8_core_avx512 \
- svml_s_expf4_core_sse4 svml_s_expf8_core_avx2 \
- svml_s_expf16_core_avx512 svml_d_pow2_core_sse4 \
- svml_d_pow4_core_avx2 svml_d_pow8_core_avx512 \
- svml_s_powf4_core_sse4 svml_s_powf8_core_avx2 \
- svml_s_powf16_core_avx512 svml_s_sincosf4_core_sse4 \
- svml_s_sincosf8_core_avx2 \
- svml_s_sincosf16_core_avx512 \
- svml_d_cos2_core-sse2 svml_d_cos4_core-sse \
- svml_d_cos8_core-avx2 svml_d_exp2_core-sse2 \
- svml_d_exp4_core-sse svml_d_exp8_core-avx2 \
- svml_d_log2_core-sse2 svml_d_log4_core-sse \
- svml_d_log8_core-avx2 svml_d_pow2_core-sse2 \
- svml_d_pow4_core-sse svml_d_pow8_core-avx2 \
- svml_d_sin2_core-sse2 svml_d_sin4_core-sse \
- svml_d_sin8_core-avx2 \
- svml_d_sincos2_core-sse2 \
- svml_d_sincos4_core-sse \
- svml_d_sincos8_core-avx2 \
- svml_s_cosf16_core-avx2 \
- svml_s_cosf4_core-sse2 \
- svml_s_cosf8_core-sse \
- svml_s_expf16_core-avx2 \
- svml_s_expf4_core-sse2 \
- svml_s_expf8_core-sse \
- svml_s_logf16_core-avx2 \
- svml_s_logf4_core-sse2 \
- svml_s_logf8_core-sse \
- svml_s_powf16_core-avx2 \
- svml_s_powf4_core-sse2 \
- svml_s_powf8_core-sse \
- svml_s_sincosf16_core-avx2 \
- svml_s_sincosf4_core-sse2 \
- svml_s_sincosf8_core-sse \
- svml_s_sinf16_core-avx2 \
- svml_s_sinf4_core-sse2 \
- svml_s_sinf8_core-sse
+libmvec-multiarch-double-func-list = \
+ 2_core-sse2 \
+ 2_core_sse4 \
+ 4_core-sse \
+ 4_core_avx2 \
+ 8_core-avx2 \
+ 8_core_avx512
+libmvec-multiarch-float-func-list = \
+ f4_core-sse2 \
+ f4_core_sse4 \
+ f8_core-sse \
+ f8_core_avx2 \
+ f16_core-avx2 \
+ f16_core_avx512
+
+libmvec-sysdep_routines += \
+ $(foreach l,$(libmvec-multiarch-double-func-list), \
+ $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
+ $(foreach l,$(libmvec-multiarch-float-func-list), \
+ $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
endif
--
2.31.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* PING [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig
2021-10-15 13:12 [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig H.J. Lu
@ 2021-10-20 16:00 ` H.J. Lu
2021-10-20 18:07 ` Noah Goldstein
0 siblings, 1 reply; 6+ messages in thread
From: H.J. Lu @ 2021-10-20 16:00 UTC (permalink / raw)
To: GNU C Library
On Fri, Oct 15, 2021 at 6:12 AM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> 1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which
> contains libmvec ABI test dependencies and CFLAGS, in the build directory.
> 2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS.
>
> Tested on SSE4, AVX, AVX2 and AVX512 machines.
> ---
> sysdeps/x86_64/fpu/Makeconfig | 92 ++++++++++++++++++
> sysdeps/x86_64/fpu/Makefile | 134 +++++++++-----------------
> sysdeps/x86_64/fpu/multiarch/Makefile | 68 ++++---------
> 3 files changed, 155 insertions(+), 139 deletions(-)
> create mode 100644 sysdeps/x86_64/fpu/Makeconfig
>
> diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig
> new file mode 100644
> index 0000000000..0a9341cdd9
> --- /dev/null
> +++ b/sysdeps/x86_64/fpu/Makeconfig
> @@ -0,0 +1,92 @@
> +# Generate libmvec.mk for libmvec ABI tests.
> +
> +# Copyright (C) 2021 Free Software Foundation, Inc.
> +# This file is part of the GNU C Library.
> +
> +# The GNU C Library is free software; you can redistribute it and/or
> +# modify it under the terms of the GNU Lesser General Public
> +# License as published by the Free Software Foundation; either
> +# version 2.1 of the License, or (at your option) any later version.
> +
> +# The GNU C Library is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> +# Lesser General Public License for more details.
> +
> +# You should have received a copy of the GNU Lesser General Public
> +# License along with the GNU C Library; if not, see
> +# <https://www.gnu.org/licenses/>.
> +
> +-include $(common-objpfx)libmvec.mk
> +postclean-generated += libmvec.mk
> +
> +# Define for both math and mathvec directories.
> +libmvec-funcs = \
> + cos \
> + exp \
> + log \
> + pow \
> + sin \
> + sincos \
> +
> +# FIXME: Add ABI tests for all libmvec functions and replace
> +# libmvec-abi-funcs with libmvec-funcs.
> +libmvec-abi-funcs = \
> + sincos
> +
> +# The base libmvec ABI tests.
> +libmvec-abi-func-tests = \
> + $(addprefix test-double-libmvec-,$(libmvec-abi-funcs)) \
> + $(addsuffix f,$(addprefix test-float-libmvec-,$(libmvec-abi-funcs)))
> +
> +# The AVX libmvec ABI tests.
> +libmvec-abi-func-avx-tests = \
> + $(addsuffix -avx,$(libmvec-abi-func-tests))
> +#
> +# The AVX2 libmvec ABI tests.
> +libmvec-abi-func-avx2-tests = \
> + $(addsuffix -avx2,$(libmvec-abi-func-tests))
> +#
> +# The AVX512F libmvec ABI tests.
> +libmvec-abi-func-avx512f-tests = \
> + $(addsuffix -avx512f,$(libmvec-abi-func-tests))
> +
> +$(common-objpfx)libmvec.mk: $(common-objpfx)config.make
> + (echo "ifeq (\$$(subdir)\$$(build-mathvec),mathyes)"; \
> + echo; \
> + for i in "" "-avx" "-avx2" "-avx512f"; do \
> + for t in $(libmvec-abi-func-tests); do \
> + echo "\$$(objpfx)$$t$$i: \\"; \
> + echo " \$$(objpfx)$$t$$i.o \\"; \
> + echo " \$$(objpfx)test-libmvec$$i.o \$$(libmvec)"; \
> + echo; \
> + done; \
> + done; \
> + for t in $(libmvec-abi-funcs); do \
> + echo "CFLAGS-test-double-libmvec-$$t.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags)"; \
> + echo "CFLAGS-test-double-libmvec-$$t-avx.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags) \\"; \
> + echo " \$$(double-vlen4-arch-ext-cflags)"; \
> + echo "CFLAGS-test-double-libmvec-$$t-avx2.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags) \\"; \
> + echo " \$$(double-vlen4-arch-ext2-cflags)"; \
> + echo "CFLAGS-test-double-libmvec-$$t-avx512f.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags) \\"; \
> + echo " \$$(double-vlen8-arch-ext-cflags)"; \
> + echo; \
> + echo "CFLAGS-test-float-libmvec-$${t}f.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags)"; \
> + echo "CFLAGS-test-float-libmvec-$${t}f-avx.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags) \\"; \
> + echo " \$$(float-vlen8-arch-ext-cflags)"; \
> + echo "CFLAGS-test-float-libmvec-$${t}f-avx2.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags) \\"; \
> + echo " \$$(float-vlen8-arch-ext2-cflags)"; \
> + echo "CFLAGS-test-float-libmvec-$${t}f-avx512f.c = \\"; \
> + echo " \$$(libmvec-abi-test-cflags) \\"; \
> + echo " \$$(float-vlen16-arch-ext-cflags)"; \
> + echo; \
> + done; \
> + echo "endif") > $@T
> + mv -f $@T $@
> diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
> index e5a65750c0..d172ae815d 100644
> --- a/sysdeps/x86_64/fpu/Makefile
> +++ b/sysdeps/x86_64/fpu/Makefile
> @@ -1,87 +1,48 @@
> ifeq ($(subdir),mathvec)
> -libmvec-support += svml_d_cos2_core svml_d_cos4_core_avx \
> - svml_d_cos4_core svml_d_cos8_core \
> - svml_d_sin2_core svml_d_sin4_core_avx \
> - svml_d_sin4_core svml_d_sin8_core svml_d_trig_data \
> - svml_s_cosf4_core svml_s_cosf8_core_avx \
> - svml_s_cosf8_core svml_s_cosf16_core svml_s_trig_data \
> - svml_s_sinf4_core svml_s_sinf8_core_avx \
> - svml_s_sinf8_core svml_s_sinf16_core \
> - svml_d_sincos2_core svml_d_sincos4_core_avx \
> - svml_d_sincos4_core svml_d_sincos8_core \
> - svml_d_log2_core svml_d_log4_core_avx svml_d_log4_core \
> - svml_d_log8_core svml_d_log_data svml_s_logf4_core \
> - svml_s_logf8_core_avx svml_s_logf8_core svml_s_logf16_core \
> - svml_s_logf_data svml_d_exp2_core svml_d_exp4_core_avx \
> - svml_d_exp4_core svml_d_exp8_core svml_d_exp_data \
> - svml_s_expf4_core svml_s_expf8_core_avx svml_s_expf8_core \
> - svml_s_expf16_core svml_s_expf_data svml_d_pow2_core \
> - svml_d_pow4_core_avx svml_d_pow4_core svml_d_pow8_core \
> - svml_d_pow_data svml_s_powf4_core svml_s_powf8_core_avx \
> - svml_s_powf8_core svml_s_powf16_core svml_s_powf_data \
> - svml_s_sincosf4_core svml_s_sincosf8_core_avx \
> - svml_s_sincosf8_core svml_s_sincosf16_core
> +libmvec-double-func-list = \
> + 2_core \
> + 4_core \
> + 4_core_avx \
> + 8_core
> +libmvec-float-func-list = \
> + f4_core \
> + f8_core \
> + f8_core_avx \
> + f16_core
> +libmvec-support += \
> + svml_d_exp_data \
> + svml_d_log_data \
> + svml_d_pow_data \
> + svml_d_trig_data \
> + svml_s_expf_data \
> + svml_s_logf_data \
> + svml_s_powf_data \
> + svml_s_trig_data \
> + $(foreach l,$(libmvec-double-func-list), \
> + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> + $(foreach l,$(libmvec-float-func-list), \
> + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> endif
>
> # Variables for libmvec tests.
> -ifeq ($(subdir),math)
> -ifeq ($(build-mathvec),yes)
> +ifeq ($(subdir)$(build-mathvec),mathyes)
> libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
> - float-vlen4 float-vlen8 float-vlen8-avx2
> -tests += test-double-libmvec-sincos test-double-libmvec-sincos-avx \
> - test-double-libmvec-sincos-avx2 test-float-libmvec-sincosf \
> - test-float-libmvec-sincosf-avx test-float-libmvec-sincosf-avx2
> -test-extras += test-libmvec test-libmvec-avx \
> - test-libmvec-avx2 test-libmvec-avx512f
> -extra-test-objs += test-libmvec.o test-libmvec-avx.o \
> - test-libmvec-avx2.o test-libmvec-avx512f.o
> -
> -$(objpfx)test-double-libmvec-sincos: \
> - $(objpfx)test-double-libmvec-sincos.o \
> - $(objpfx)test-libmvec.o $(libmvec)
> -
> -$(objpfx)test-double-libmvec-sincos-avx: \
> - $(objpfx)test-double-libmvec-sincos-avx.o \
> - $(objpfx)test-libmvec-avx.o $(libmvec)
> -
> -$(objpfx)test-double-libmvec-sincos-avx2: \
> - $(objpfx)test-double-libmvec-sincos-avx2.o \
> - $(objpfx)test-libmvec-avx2.o $(libmvec)
> -
> -$(objpfx)test-float-libmvec-sincosf: \
> - $(objpfx)test-float-libmvec-sincosf.o \
> - $(objpfx)test-libmvec.o $(libmvec)
> -
> -$(objpfx)test-float-libmvec-sincosf-avx: \
> - $(objpfx)test-float-libmvec-sincosf-avx.o \
> - $(objpfx)test-libmvec-avx.o $(libmvec)
> -
> -$(objpfx)test-float-libmvec-sincosf-avx2: \
> - $(objpfx)test-float-libmvec-sincosf-avx2.o \
> - $(objpfx)test-libmvec-avx2.o $(libmvec)
> -
> -libmvec-tests += double-vlen8 float-vlen16
> -tests += test-double-libmvec-sincos-avx512f \
> - test-float-libmvec-sincosf-avx512f
> -test-extras += test-libmvec-avx512f
> -extra-test-objs += test-libmvec-avx512f.o
> -
> -$(objpfx)test-double-libmvec-sincos-avx512f: \
> - $(objpfx)test-double-libmvec-sincos-avx512f.o \
> - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> -
> -$(objpfx)test-float-libmvec-sincosf-avx512f: \
> - $(objpfx)test-float-libmvec-sincosf-avx512f.o \
> - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> -
> -double-vlen2-funcs = cos exp log pow sin sincos
> -double-vlen4-funcs = cos exp log pow sin sincos
> -double-vlen4-avx2-funcs = cos exp log pow sin sincos
> -double-vlen8-funcs = cos exp log pow sin sincos
> -float-vlen4-funcs = cos exp log pow sin sincos
> -float-vlen8-funcs = cos exp log pow sin sincos
> -float-vlen8-avx2-funcs = cos exp log pow sin sincos
> -float-vlen16-funcs = cos exp log pow sin sincos
> + float-vlen4 float-vlen8 float-vlen8-avx2 \
> + double-vlen8 float-vlen16
> +tests += \
> + $(libmvec-abi-func-tests) \
> + $(libmvec-abi-func-avx-tests) \
> + $(libmvec-abi-func-avx2-tests) \
> + $(libmvec-abi-func-avx512f-tests)
> +
> +double-vlen2-funcs = $(libmvec-funcs)
> +double-vlen4-funcs = $(libmvec-funcs)
> +double-vlen4-avx2-funcs = $(libmvec-funcs)
> +double-vlen8-funcs = $(libmvec-funcs)
> +float-vlen4-funcs = $(libmvec-funcs)
> +float-vlen8-funcs = $(libmvec-funcs)
> +float-vlen8-avx2-funcs = $(libmvec-funcs)
> +float-vlen16-funcs = $(libmvec-funcs)
>
> double-vlen4-arch-ext-cflags = -mavx
> double-vlen4-arch-ext2-cflags = -mavx2
> @@ -91,22 +52,13 @@ float-vlen8-arch-ext-cflags = -mavx
> float-vlen8-arch-ext2-cflags = -mavx2
> float-vlen16-arch-ext-cflags = -mavx512f
>
> -libmvec-sincos-cflags = $(libm-test-fast-math-cflags) -fno-inline -fopenmp -Wno-unknown-pragmas
> +libmvec-abi-test-cflags = \
> + $(libm-test-fast-math-cflags) \
> + -fno-inline -fopenmp -Wno-unknown-pragmas
>
> CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)
>
> CFLAGS-test-float-vlen8-avx2-wrappers.c = $(float-vlen8-arch-ext2-cflags)
> -
> -CFLAGS-test-double-libmvec-sincos.c = $(libmvec-sincos-cflags)
> -CFLAGS-test-double-libmvec-sincos-avx.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext-cflags)
> -CFLAGS-test-double-libmvec-sincos-avx2.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext2-cflags)
> -CFLAGS-test-double-libmvec-sincos-avx512f.c = $(libmvec-sincos-cflags) $(double-vlen8-arch-ext-cflags)
> -
> -CFLAGS-test-float-libmvec-sincosf.c = $(libmvec-sincos-cflags)
> -CFLAGS-test-float-libmvec-sincosf-avx.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext-cflags)
> -CFLAGS-test-float-libmvec-sincosf-avx2.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext2-cflags)
> -CFLAGS-test-float-libmvec-sincosf-avx512f.c = $(libmvec-sincos-cflags) $(float-vlen16-arch-ext-cflags)
> -endif
> endif
>
> ifeq ($(subdir)$(config-cflags-mprefer-vector-width),mathyes)
> diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
> index d425ffd6d3..ec796277a5 100644
> --- a/sysdeps/x86_64/fpu/multiarch/Makefile
> +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
> @@ -59,52 +59,24 @@ CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
> endif
>
> ifeq ($(subdir),mathvec)
> -libmvec-sysdep_routines += svml_d_cos2_core_sse4 svml_d_cos4_core_avx2 \
> - svml_d_cos8_core_avx512 svml_d_sin2_core_sse4 \
> - svml_d_sin4_core_avx2 svml_d_sin8_core_avx512 \
> - svml_d_log2_core_sse4 svml_d_log4_core_avx2 \
> - svml_d_log8_core_avx512 svml_d_sincos2_core_sse4 \
> - svml_d_sincos4_core_avx2 svml_d_sincos8_core_avx512 \
> - svml_s_cosf4_core_sse4 svml_s_cosf8_core_avx2 \
> - svml_s_cosf16_core_avx512 svml_s_sinf4_core_sse4 \
> - svml_s_sinf8_core_avx2 svml_s_sinf16_core_avx512 \
> - svml_s_logf4_core_sse4 svml_s_logf8_core_avx2 \
> - svml_s_logf16_core_avx512 svml_d_exp2_core_sse4 \
> - svml_d_exp4_core_avx2 svml_d_exp8_core_avx512 \
> - svml_s_expf4_core_sse4 svml_s_expf8_core_avx2 \
> - svml_s_expf16_core_avx512 svml_d_pow2_core_sse4 \
> - svml_d_pow4_core_avx2 svml_d_pow8_core_avx512 \
> - svml_s_powf4_core_sse4 svml_s_powf8_core_avx2 \
> - svml_s_powf16_core_avx512 svml_s_sincosf4_core_sse4 \
> - svml_s_sincosf8_core_avx2 \
> - svml_s_sincosf16_core_avx512 \
> - svml_d_cos2_core-sse2 svml_d_cos4_core-sse \
> - svml_d_cos8_core-avx2 svml_d_exp2_core-sse2 \
> - svml_d_exp4_core-sse svml_d_exp8_core-avx2 \
> - svml_d_log2_core-sse2 svml_d_log4_core-sse \
> - svml_d_log8_core-avx2 svml_d_pow2_core-sse2 \
> - svml_d_pow4_core-sse svml_d_pow8_core-avx2 \
> - svml_d_sin2_core-sse2 svml_d_sin4_core-sse \
> - svml_d_sin8_core-avx2 \
> - svml_d_sincos2_core-sse2 \
> - svml_d_sincos4_core-sse \
> - svml_d_sincos8_core-avx2 \
> - svml_s_cosf16_core-avx2 \
> - svml_s_cosf4_core-sse2 \
> - svml_s_cosf8_core-sse \
> - svml_s_expf16_core-avx2 \
> - svml_s_expf4_core-sse2 \
> - svml_s_expf8_core-sse \
> - svml_s_logf16_core-avx2 \
> - svml_s_logf4_core-sse2 \
> - svml_s_logf8_core-sse \
> - svml_s_powf16_core-avx2 \
> - svml_s_powf4_core-sse2 \
> - svml_s_powf8_core-sse \
> - svml_s_sincosf16_core-avx2 \
> - svml_s_sincosf4_core-sse2 \
> - svml_s_sincosf8_core-sse \
> - svml_s_sinf16_core-avx2 \
> - svml_s_sinf4_core-sse2 \
> - svml_s_sinf8_core-sse
> +libmvec-multiarch-double-func-list = \
> + 2_core-sse2 \
> + 2_core_sse4 \
> + 4_core-sse \
> + 4_core_avx2 \
> + 8_core-avx2 \
> + 8_core_avx512
> +libmvec-multiarch-float-func-list = \
> + f4_core-sse2 \
> + f4_core_sse4 \
> + f8_core-sse \
> + f8_core_avx2 \
> + f16_core-avx2 \
> + f16_core_avx512
> +
> +libmvec-sysdep_routines += \
> + $(foreach l,$(libmvec-multiarch-double-func-list), \
> + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> + $(foreach l,$(libmvec-multiarch-float-func-list), \
> + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> endif
> --
> 2.31.1
>
PING.
--
H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: PING [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig
2021-10-20 16:00 ` PING " H.J. Lu
@ 2021-10-20 18:07 ` Noah Goldstein
2021-10-20 18:18 ` H.J. Lu
0 siblings, 1 reply; 6+ messages in thread
From: Noah Goldstein @ 2021-10-20 18:07 UTC (permalink / raw)
To: H.J. Lu; +Cc: GNU C Library
On Wed, Oct 20, 2021 at 11:01 AM H.J. Lu via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> On Fri, Oct 15, 2021 at 6:12 AM H.J. Lu <hjl.tools@gmail.com> wrote:
> >
> > 1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which
> > contains libmvec ABI test dependencies and CFLAGS, in the build directory.
> > 2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS.
> >
> > Tested on SSE4, AVX, AVX2 and AVX512 machines.
> > ---
> > sysdeps/x86_64/fpu/Makeconfig | 92 ++++++++++++++++++
> > sysdeps/x86_64/fpu/Makefile | 134 +++++++++-----------------
> > sysdeps/x86_64/fpu/multiarch/Makefile | 68 ++++---------
> > 3 files changed, 155 insertions(+), 139 deletions(-)
> > create mode 100644 sysdeps/x86_64/fpu/Makeconfig
> >
> > diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig
> > new file mode 100644
> > index 0000000000..0a9341cdd9
> > --- /dev/null
> > +++ b/sysdeps/x86_64/fpu/Makeconfig
> > @@ -0,0 +1,92 @@
> > +# Generate libmvec.mk for libmvec ABI tests.
> > +
> > +# Copyright (C) 2021 Free Software Foundation, Inc.
> > +# This file is part of the GNU C Library.
> > +
> > +# The GNU C Library is free software; you can redistribute it and/or
> > +# modify it under the terms of the GNU Lesser General Public
> > +# License as published by the Free Software Foundation; either
> > +# version 2.1 of the License, or (at your option) any later version.
> > +
> > +# The GNU C Library is distributed in the hope that it will be useful,
> > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > +# Lesser General Public License for more details.
> > +
> > +# You should have received a copy of the GNU Lesser General Public
> > +# License along with the GNU C Library; if not, see
> > +# <https://www.gnu.org/licenses/>.
> > +
> > +-include $(common-objpfx)libmvec.mk
> > +postclean-generated += libmvec.mk
> > +
> > +# Define for both math and mathvec directories.
> > +libmvec-funcs = \
> > + cos \
> > + exp \
> > + log \
> > + pow \
> > + sin \
> > + sincos \
> > +
> > +# FIXME: Add ABI tests for all libmvec functions and replace
> > +# libmvec-abi-funcs with libmvec-funcs.
> > +libmvec-abi-funcs = \
> > + sincos
What if we add a function that only supports some extensions (i.e just
AVX2 but not AVX and AVX512)?
> > +
> > +# The base libmvec ABI tests.
> > +libmvec-abi-func-tests = \
> > + $(addprefix test-double-libmvec-,$(libmvec-abi-funcs)) \
> > + $(addsuffix f,$(addprefix test-float-libmvec-,$(libmvec-abi-funcs)))
> > +
> > +# The AVX libmvec ABI tests.
> > +libmvec-abi-func-avx-tests = \
> > + $(addsuffix -avx,$(libmvec-abi-func-tests))
> > +#
> > +# The AVX2 libmvec ABI tests.
> > +libmvec-abi-func-avx2-tests = \
> > + $(addsuffix -avx2,$(libmvec-abi-func-tests))
> > +#
> > +# The AVX512F libmvec ABI tests.
> > +libmvec-abi-func-avx512f-tests = \
> > + $(addsuffix -avx512f,$(libmvec-abi-func-tests))
> > +
> > +$(common-objpfx)libmvec.mk: $(common-objpfx)config.make
> > + (echo "ifeq (\$$(subdir)\$$(build-mathvec),mathyes)"; \
> > + echo; \
> > + for i in "" "-avx" "-avx2" "-avx512f"; do \
> > + for t in $(libmvec-abi-func-tests); do \
> > + echo "\$$(objpfx)$$t$$i: \\"; \
> > + echo " \$$(objpfx)$$t$$i.o \\"; \
> > + echo " \$$(objpfx)test-libmvec$$i.o \$$(libmvec)"; \
> > + echo; \
> > + done; \
> > + done; \
> > + for t in $(libmvec-abi-funcs); do \
> > + echo "CFLAGS-test-double-libmvec-$$t.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags)"; \
> > + echo "CFLAGS-test-double-libmvec-$$t-avx.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > + echo " \$$(double-vlen4-arch-ext-cflags)"; \
> > + echo "CFLAGS-test-double-libmvec-$$t-avx2.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > + echo " \$$(double-vlen4-arch-ext2-cflags)"; \
> > + echo "CFLAGS-test-double-libmvec-$$t-avx512f.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > + echo " \$$(double-vlen8-arch-ext-cflags)"; \
> > + echo; \
> > + echo "CFLAGS-test-float-libmvec-$${t}f.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags)"; \
> > + echo "CFLAGS-test-float-libmvec-$${t}f-avx.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > + echo " \$$(float-vlen8-arch-ext-cflags)"; \
> > + echo "CFLAGS-test-float-libmvec-$${t}f-avx2.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > + echo " \$$(float-vlen8-arch-ext2-cflags)"; \
> > + echo "CFLAGS-test-float-libmvec-$${t}f-avx512f.c = \\"; \
> > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > + echo " \$$(float-vlen16-arch-ext-cflags)"; \
> > + echo; \
> > + done; \
> > + echo "endif") > $@T
> > + mv -f $@T $@
> > diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
> > index e5a65750c0..d172ae815d 100644
> > --- a/sysdeps/x86_64/fpu/Makefile
> > +++ b/sysdeps/x86_64/fpu/Makefile
> > @@ -1,87 +1,48 @@
> > ifeq ($(subdir),mathvec)
> > -libmvec-support += svml_d_cos2_core svml_d_cos4_core_avx \
> > - svml_d_cos4_core svml_d_cos8_core \
> > - svml_d_sin2_core svml_d_sin4_core_avx \
> > - svml_d_sin4_core svml_d_sin8_core svml_d_trig_data \
> > - svml_s_cosf4_core svml_s_cosf8_core_avx \
> > - svml_s_cosf8_core svml_s_cosf16_core svml_s_trig_data \
> > - svml_s_sinf4_core svml_s_sinf8_core_avx \
> > - svml_s_sinf8_core svml_s_sinf16_core \
> > - svml_d_sincos2_core svml_d_sincos4_core_avx \
> > - svml_d_sincos4_core svml_d_sincos8_core \
> > - svml_d_log2_core svml_d_log4_core_avx svml_d_log4_core \
> > - svml_d_log8_core svml_d_log_data svml_s_logf4_core \
> > - svml_s_logf8_core_avx svml_s_logf8_core svml_s_logf16_core \
> > - svml_s_logf_data svml_d_exp2_core svml_d_exp4_core_avx \
> > - svml_d_exp4_core svml_d_exp8_core svml_d_exp_data \
> > - svml_s_expf4_core svml_s_expf8_core_avx svml_s_expf8_core \
> > - svml_s_expf16_core svml_s_expf_data svml_d_pow2_core \
> > - svml_d_pow4_core_avx svml_d_pow4_core svml_d_pow8_core \
> > - svml_d_pow_data svml_s_powf4_core svml_s_powf8_core_avx \
> > - svml_s_powf8_core svml_s_powf16_core svml_s_powf_data \
> > - svml_s_sincosf4_core svml_s_sincosf8_core_avx \
> > - svml_s_sincosf8_core svml_s_sincosf16_core
> > +libmvec-double-func-list = \
> > + 2_core \
> > + 4_core \
> > + 4_core_avx \
> > + 8_core
> > +libmvec-float-func-list = \
> > + f4_core \
> > + f8_core \
> > + f8_core_avx \
> > + f16_core
> > +libmvec-support += \
> > + svml_d_exp_data \
> > + svml_d_log_data \
> > + svml_d_pow_data \
> > + svml_d_trig_data \
> > + svml_s_expf_data \
> > + svml_s_logf_data \
> > + svml_s_powf_data \
> > + svml_s_trig_data \
> > + $(foreach l,$(libmvec-double-func-list), \
> > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > + $(foreach l,$(libmvec-float-func-list), \
> > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > endif
> >
> > # Variables for libmvec tests.
> > -ifeq ($(subdir),math)
> > -ifeq ($(build-mathvec),yes)
> > +ifeq ($(subdir)$(build-mathvec),mathyes)
> > libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
> > - float-vlen4 float-vlen8 float-vlen8-avx2
> > -tests += test-double-libmvec-sincos test-double-libmvec-sincos-avx \
> > - test-double-libmvec-sincos-avx2 test-float-libmvec-sincosf \
> > - test-float-libmvec-sincosf-avx test-float-libmvec-sincosf-avx2
> > -test-extras += test-libmvec test-libmvec-avx \
> > - test-libmvec-avx2 test-libmvec-avx512f
> > -extra-test-objs += test-libmvec.o test-libmvec-avx.o \
> > - test-libmvec-avx2.o test-libmvec-avx512f.o
> > -
> > -$(objpfx)test-double-libmvec-sincos: \
> > - $(objpfx)test-double-libmvec-sincos.o \
> > - $(objpfx)test-libmvec.o $(libmvec)
> > -
> > -$(objpfx)test-double-libmvec-sincos-avx: \
> > - $(objpfx)test-double-libmvec-sincos-avx.o \
> > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > -
> > -$(objpfx)test-double-libmvec-sincos-avx2: \
> > - $(objpfx)test-double-libmvec-sincos-avx2.o \
> > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > -
> > -$(objpfx)test-float-libmvec-sincosf: \
> > - $(objpfx)test-float-libmvec-sincosf.o \
> > - $(objpfx)test-libmvec.o $(libmvec)
> > -
> > -$(objpfx)test-float-libmvec-sincosf-avx: \
> > - $(objpfx)test-float-libmvec-sincosf-avx.o \
> > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > -
> > -$(objpfx)test-float-libmvec-sincosf-avx2: \
> > - $(objpfx)test-float-libmvec-sincosf-avx2.o \
> > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > -
> > -libmvec-tests += double-vlen8 float-vlen16
> > -tests += test-double-libmvec-sincos-avx512f \
> > - test-float-libmvec-sincosf-avx512f
> > -test-extras += test-libmvec-avx512f
> > -extra-test-objs += test-libmvec-avx512f.o
> > -
> > -$(objpfx)test-double-libmvec-sincos-avx512f: \
> > - $(objpfx)test-double-libmvec-sincos-avx512f.o \
> > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > -
> > -$(objpfx)test-float-libmvec-sincosf-avx512f: \
> > - $(objpfx)test-float-libmvec-sincosf-avx512f.o \
> > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > -
> > -double-vlen2-funcs = cos exp log pow sin sincos
> > -double-vlen4-funcs = cos exp log pow sin sincos
> > -double-vlen4-avx2-funcs = cos exp log pow sin sincos
> > -double-vlen8-funcs = cos exp log pow sin sincos
> > -float-vlen4-funcs = cos exp log pow sin sincos
> > -float-vlen8-funcs = cos exp log pow sin sincos
> > -float-vlen8-avx2-funcs = cos exp log pow sin sincos
> > -float-vlen16-funcs = cos exp log pow sin sincos
> > + float-vlen4 float-vlen8 float-vlen8-avx2 \
> > + double-vlen8 float-vlen16
> > +tests += \
> > + $(libmvec-abi-func-tests) \
> > + $(libmvec-abi-func-avx-tests) \
> > + $(libmvec-abi-func-avx2-tests) \
> > + $(libmvec-abi-func-avx512f-tests)
> > +
> > +double-vlen2-funcs = $(libmvec-funcs)
> > +double-vlen4-funcs = $(libmvec-funcs)
> > +double-vlen4-avx2-funcs = $(libmvec-funcs)
> > +double-vlen8-funcs = $(libmvec-funcs)
> > +float-vlen4-funcs = $(libmvec-funcs)
> > +float-vlen8-funcs = $(libmvec-funcs)
> > +float-vlen8-avx2-funcs = $(libmvec-funcs)
> > +float-vlen16-funcs = $(libmvec-funcs)
> >
> > double-vlen4-arch-ext-cflags = -mavx
> > double-vlen4-arch-ext2-cflags = -mavx2
> > @@ -91,22 +52,13 @@ float-vlen8-arch-ext-cflags = -mavx
> > float-vlen8-arch-ext2-cflags = -mavx2
> > float-vlen16-arch-ext-cflags = -mavx512f
> >
> > -libmvec-sincos-cflags = $(libm-test-fast-math-cflags) -fno-inline -fopenmp -Wno-unknown-pragmas
> > +libmvec-abi-test-cflags = \
> > + $(libm-test-fast-math-cflags) \
> > + -fno-inline -fopenmp -Wno-unknown-pragmas
> >
> > CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)
> >
> > CFLAGS-test-float-vlen8-avx2-wrappers.c = $(float-vlen8-arch-ext2-cflags)
> > -
> > -CFLAGS-test-double-libmvec-sincos.c = $(libmvec-sincos-cflags)
> > -CFLAGS-test-double-libmvec-sincos-avx.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext-cflags)
> > -CFLAGS-test-double-libmvec-sincos-avx2.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext2-cflags)
> > -CFLAGS-test-double-libmvec-sincos-avx512f.c = $(libmvec-sincos-cflags) $(double-vlen8-arch-ext-cflags)
> > -
> > -CFLAGS-test-float-libmvec-sincosf.c = $(libmvec-sincos-cflags)
> > -CFLAGS-test-float-libmvec-sincosf-avx.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext-cflags)
> > -CFLAGS-test-float-libmvec-sincosf-avx2.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext2-cflags)
> > -CFLAGS-test-float-libmvec-sincosf-avx512f.c = $(libmvec-sincos-cflags) $(float-vlen16-arch-ext-cflags)
> > -endif
> > endif
> >
> > ifeq ($(subdir)$(config-cflags-mprefer-vector-width),mathyes)
> > diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
> > index d425ffd6d3..ec796277a5 100644
> > --- a/sysdeps/x86_64/fpu/multiarch/Makefile
> > +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
> > @@ -59,52 +59,24 @@ CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
> > endif
> >
> > ifeq ($(subdir),mathvec)
> > -libmvec-sysdep_routines += svml_d_cos2_core_sse4 svml_d_cos4_core_avx2 \
> > - svml_d_cos8_core_avx512 svml_d_sin2_core_sse4 \
> > - svml_d_sin4_core_avx2 svml_d_sin8_core_avx512 \
> > - svml_d_log2_core_sse4 svml_d_log4_core_avx2 \
> > - svml_d_log8_core_avx512 svml_d_sincos2_core_sse4 \
> > - svml_d_sincos4_core_avx2 svml_d_sincos8_core_avx512 \
> > - svml_s_cosf4_core_sse4 svml_s_cosf8_core_avx2 \
> > - svml_s_cosf16_core_avx512 svml_s_sinf4_core_sse4 \
> > - svml_s_sinf8_core_avx2 svml_s_sinf16_core_avx512 \
> > - svml_s_logf4_core_sse4 svml_s_logf8_core_avx2 \
> > - svml_s_logf16_core_avx512 svml_d_exp2_core_sse4 \
> > - svml_d_exp4_core_avx2 svml_d_exp8_core_avx512 \
> > - svml_s_expf4_core_sse4 svml_s_expf8_core_avx2 \
> > - svml_s_expf16_core_avx512 svml_d_pow2_core_sse4 \
> > - svml_d_pow4_core_avx2 svml_d_pow8_core_avx512 \
> > - svml_s_powf4_core_sse4 svml_s_powf8_core_avx2 \
> > - svml_s_powf16_core_avx512 svml_s_sincosf4_core_sse4 \
> > - svml_s_sincosf8_core_avx2 \
> > - svml_s_sincosf16_core_avx512 \
> > - svml_d_cos2_core-sse2 svml_d_cos4_core-sse \
> > - svml_d_cos8_core-avx2 svml_d_exp2_core-sse2 \
> > - svml_d_exp4_core-sse svml_d_exp8_core-avx2 \
> > - svml_d_log2_core-sse2 svml_d_log4_core-sse \
> > - svml_d_log8_core-avx2 svml_d_pow2_core-sse2 \
> > - svml_d_pow4_core-sse svml_d_pow8_core-avx2 \
> > - svml_d_sin2_core-sse2 svml_d_sin4_core-sse \
> > - svml_d_sin8_core-avx2 \
> > - svml_d_sincos2_core-sse2 \
> > - svml_d_sincos4_core-sse \
> > - svml_d_sincos8_core-avx2 \
> > - svml_s_cosf16_core-avx2 \
> > - svml_s_cosf4_core-sse2 \
> > - svml_s_cosf8_core-sse \
> > - svml_s_expf16_core-avx2 \
> > - svml_s_expf4_core-sse2 \
> > - svml_s_expf8_core-sse \
> > - svml_s_logf16_core-avx2 \
> > - svml_s_logf4_core-sse2 \
> > - svml_s_logf8_core-sse \
> > - svml_s_powf16_core-avx2 \
> > - svml_s_powf4_core-sse2 \
> > - svml_s_powf8_core-sse \
> > - svml_s_sincosf16_core-avx2 \
> > - svml_s_sincosf4_core-sse2 \
> > - svml_s_sincosf8_core-sse \
> > - svml_s_sinf16_core-avx2 \
> > - svml_s_sinf4_core-sse2 \
> > - svml_s_sinf8_core-sse
> > +libmvec-multiarch-double-func-list = \
> > + 2_core-sse2 \
> > + 2_core_sse4 \
> > + 4_core-sse \
> > + 4_core_avx2 \
> > + 8_core-avx2 \
> > + 8_core_avx512
> > +libmvec-multiarch-float-func-list = \
> > + f4_core-sse2 \
> > + f4_core_sse4 \
> > + f8_core-sse \
> > + f8_core_avx2 \
> > + f16_core-avx2 \
> > + f16_core_avx512
> > +
> > +libmvec-sysdep_routines += \
> > + $(foreach l,$(libmvec-multiarch-double-func-list), \
> > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > + $(foreach l,$(libmvec-multiarch-float-func-list), \
> > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > endif
> > --
> > 2.31.1
> >
>
> PING.
>
> --
> H.J.
How can I test this patch?
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: PING [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig
2021-10-20 18:07 ` Noah Goldstein
@ 2021-10-20 18:18 ` H.J. Lu
2021-10-20 18:47 ` Noah Goldstein
0 siblings, 1 reply; 6+ messages in thread
From: H.J. Lu @ 2021-10-20 18:18 UTC (permalink / raw)
To: Noah Goldstein; +Cc: GNU C Library
On Wed, Oct 20, 2021 at 11:07 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> On Wed, Oct 20, 2021 at 11:01 AM H.J. Lu via Libc-alpha
> <libc-alpha@sourceware.org> wrote:
> >
> > On Fri, Oct 15, 2021 at 6:12 AM H.J. Lu <hjl.tools@gmail.com> wrote:
> > >
> > > 1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which
> > > contains libmvec ABI test dependencies and CFLAGS, in the build directory.
> > > 2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS.
> > >
> > > Tested on SSE4, AVX, AVX2 and AVX512 machines.
> > > ---
> > > sysdeps/x86_64/fpu/Makeconfig | 92 ++++++++++++++++++
> > > sysdeps/x86_64/fpu/Makefile | 134 +++++++++-----------------
> > > sysdeps/x86_64/fpu/multiarch/Makefile | 68 ++++---------
> > > 3 files changed, 155 insertions(+), 139 deletions(-)
> > > create mode 100644 sysdeps/x86_64/fpu/Makeconfig
> > >
> > > diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig
> > > new file mode 100644
> > > index 0000000000..0a9341cdd9
> > > --- /dev/null
> > > +++ b/sysdeps/x86_64/fpu/Makeconfig
> > > @@ -0,0 +1,92 @@
> > > +# Generate libmvec.mk for libmvec ABI tests.
> > > +
> > > +# Copyright (C) 2021 Free Software Foundation, Inc.
> > > +# This file is part of the GNU C Library.
> > > +
> > > +# The GNU C Library is free software; you can redistribute it and/or
> > > +# modify it under the terms of the GNU Lesser General Public
> > > +# License as published by the Free Software Foundation; either
> > > +# version 2.1 of the License, or (at your option) any later version.
> > > +
> > > +# The GNU C Library is distributed in the hope that it will be useful,
> > > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > > +# Lesser General Public License for more details.
> > > +
> > > +# You should have received a copy of the GNU Lesser General Public
> > > +# License along with the GNU C Library; if not, see
> > > +# <https://www.gnu.org/licenses/>.
> > > +
> > > +-include $(common-objpfx)libmvec.mk
> > > +postclean-generated += libmvec.mk
> > > +
> > > +# Define for both math and mathvec directories.
> > > +libmvec-funcs = \
> > > + cos \
> > > + exp \
> > > + log \
> > > + pow \
> > > + sin \
> > > + sincos \
> > > +
> > > +# FIXME: Add ABI tests for all libmvec functions and replace
> > > +# libmvec-abi-funcs with libmvec-funcs.
> > > +libmvec-abi-funcs = \
> > > + sincos
>
> What if we add a function that only supports some extensions (i.e just
> AVX2 but not AVX and AVX512)?
All libmvec functions must support SSE, AVX and AVX512.
> > > +
> > > +# The base libmvec ABI tests.
> > > +libmvec-abi-func-tests = \
> > > + $(addprefix test-double-libmvec-,$(libmvec-abi-funcs)) \
> > > + $(addsuffix f,$(addprefix test-float-libmvec-,$(libmvec-abi-funcs)))
> > > +
> > > +# The AVX libmvec ABI tests.
> > > +libmvec-abi-func-avx-tests = \
> > > + $(addsuffix -avx,$(libmvec-abi-func-tests))
> > > +#
> > > +# The AVX2 libmvec ABI tests.
> > > +libmvec-abi-func-avx2-tests = \
> > > + $(addsuffix -avx2,$(libmvec-abi-func-tests))
> > > +#
> > > +# The AVX512F libmvec ABI tests.
> > > +libmvec-abi-func-avx512f-tests = \
> > > + $(addsuffix -avx512f,$(libmvec-abi-func-tests))
> > > +
> > > +$(common-objpfx)libmvec.mk: $(common-objpfx)config.make
> > > + (echo "ifeq (\$$(subdir)\$$(build-mathvec),mathyes)"; \
> > > + echo; \
> > > + for i in "" "-avx" "-avx2" "-avx512f"; do \
> > > + for t in $(libmvec-abi-func-tests); do \
> > > + echo "\$$(objpfx)$$t$$i: \\"; \
> > > + echo " \$$(objpfx)$$t$$i.o \\"; \
> > > + echo " \$$(objpfx)test-libmvec$$i.o \$$(libmvec)"; \
> > > + echo; \
> > > + done; \
> > > + done; \
> > > + for t in $(libmvec-abi-funcs); do \
> > > + echo "CFLAGS-test-double-libmvec-$$t.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags)"; \
> > > + echo "CFLAGS-test-double-libmvec-$$t-avx.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > + echo " \$$(double-vlen4-arch-ext-cflags)"; \
> > > + echo "CFLAGS-test-double-libmvec-$$t-avx2.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > + echo " \$$(double-vlen4-arch-ext2-cflags)"; \
> > > + echo "CFLAGS-test-double-libmvec-$$t-avx512f.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > + echo " \$$(double-vlen8-arch-ext-cflags)"; \
> > > + echo; \
> > > + echo "CFLAGS-test-float-libmvec-$${t}f.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags)"; \
> > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > + echo " \$$(float-vlen8-arch-ext-cflags)"; \
> > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx2.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > + echo " \$$(float-vlen8-arch-ext2-cflags)"; \
> > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx512f.c = \\"; \
> > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > + echo " \$$(float-vlen16-arch-ext-cflags)"; \
> > > + echo; \
> > > + done; \
> > > + echo "endif") > $@T
> > > + mv -f $@T $@
> > > diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
> > > index e5a65750c0..d172ae815d 100644
> > > --- a/sysdeps/x86_64/fpu/Makefile
> > > +++ b/sysdeps/x86_64/fpu/Makefile
> > > @@ -1,87 +1,48 @@
> > > ifeq ($(subdir),mathvec)
> > > -libmvec-support += svml_d_cos2_core svml_d_cos4_core_avx \
> > > - svml_d_cos4_core svml_d_cos8_core \
> > > - svml_d_sin2_core svml_d_sin4_core_avx \
> > > - svml_d_sin4_core svml_d_sin8_core svml_d_trig_data \
> > > - svml_s_cosf4_core svml_s_cosf8_core_avx \
> > > - svml_s_cosf8_core svml_s_cosf16_core svml_s_trig_data \
> > > - svml_s_sinf4_core svml_s_sinf8_core_avx \
> > > - svml_s_sinf8_core svml_s_sinf16_core \
> > > - svml_d_sincos2_core svml_d_sincos4_core_avx \
> > > - svml_d_sincos4_core svml_d_sincos8_core \
> > > - svml_d_log2_core svml_d_log4_core_avx svml_d_log4_core \
> > > - svml_d_log8_core svml_d_log_data svml_s_logf4_core \
> > > - svml_s_logf8_core_avx svml_s_logf8_core svml_s_logf16_core \
> > > - svml_s_logf_data svml_d_exp2_core svml_d_exp4_core_avx \
> > > - svml_d_exp4_core svml_d_exp8_core svml_d_exp_data \
> > > - svml_s_expf4_core svml_s_expf8_core_avx svml_s_expf8_core \
> > > - svml_s_expf16_core svml_s_expf_data svml_d_pow2_core \
> > > - svml_d_pow4_core_avx svml_d_pow4_core svml_d_pow8_core \
> > > - svml_d_pow_data svml_s_powf4_core svml_s_powf8_core_avx \
> > > - svml_s_powf8_core svml_s_powf16_core svml_s_powf_data \
> > > - svml_s_sincosf4_core svml_s_sincosf8_core_avx \
> > > - svml_s_sincosf8_core svml_s_sincosf16_core
> > > +libmvec-double-func-list = \
> > > + 2_core \
> > > + 4_core \
> > > + 4_core_avx \
> > > + 8_core
> > > +libmvec-float-func-list = \
> > > + f4_core \
> > > + f8_core \
> > > + f8_core_avx \
> > > + f16_core
> > > +libmvec-support += \
> > > + svml_d_exp_data \
> > > + svml_d_log_data \
> > > + svml_d_pow_data \
> > > + svml_d_trig_data \
> > > + svml_s_expf_data \
> > > + svml_s_logf_data \
> > > + svml_s_powf_data \
> > > + svml_s_trig_data \
> > > + $(foreach l,$(libmvec-double-func-list), \
> > > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > > + $(foreach l,$(libmvec-float-func-list), \
> > > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > > endif
> > >
> > > # Variables for libmvec tests.
> > > -ifeq ($(subdir),math)
> > > -ifeq ($(build-mathvec),yes)
> > > +ifeq ($(subdir)$(build-mathvec),mathyes)
> > > libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
> > > - float-vlen4 float-vlen8 float-vlen8-avx2
> > > -tests += test-double-libmvec-sincos test-double-libmvec-sincos-avx \
> > > - test-double-libmvec-sincos-avx2 test-float-libmvec-sincosf \
> > > - test-float-libmvec-sincosf-avx test-float-libmvec-sincosf-avx2
> > > -test-extras += test-libmvec test-libmvec-avx \
> > > - test-libmvec-avx2 test-libmvec-avx512f
> > > -extra-test-objs += test-libmvec.o test-libmvec-avx.o \
> > > - test-libmvec-avx2.o test-libmvec-avx512f.o
> > > -
> > > -$(objpfx)test-double-libmvec-sincos: \
> > > - $(objpfx)test-double-libmvec-sincos.o \
> > > - $(objpfx)test-libmvec.o $(libmvec)
> > > -
> > > -$(objpfx)test-double-libmvec-sincos-avx: \
> > > - $(objpfx)test-double-libmvec-sincos-avx.o \
> > > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > > -
> > > -$(objpfx)test-double-libmvec-sincos-avx2: \
> > > - $(objpfx)test-double-libmvec-sincos-avx2.o \
> > > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > > -
> > > -$(objpfx)test-float-libmvec-sincosf: \
> > > - $(objpfx)test-float-libmvec-sincosf.o \
> > > - $(objpfx)test-libmvec.o $(libmvec)
> > > -
> > > -$(objpfx)test-float-libmvec-sincosf-avx: \
> > > - $(objpfx)test-float-libmvec-sincosf-avx.o \
> > > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > > -
> > > -$(objpfx)test-float-libmvec-sincosf-avx2: \
> > > - $(objpfx)test-float-libmvec-sincosf-avx2.o \
> > > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > > -
> > > -libmvec-tests += double-vlen8 float-vlen16
> > > -tests += test-double-libmvec-sincos-avx512f \
> > > - test-float-libmvec-sincosf-avx512f
> > > -test-extras += test-libmvec-avx512f
> > > -extra-test-objs += test-libmvec-avx512f.o
> > > -
> > > -$(objpfx)test-double-libmvec-sincos-avx512f: \
> > > - $(objpfx)test-double-libmvec-sincos-avx512f.o \
> > > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > > -
> > > -$(objpfx)test-float-libmvec-sincosf-avx512f: \
> > > - $(objpfx)test-float-libmvec-sincosf-avx512f.o \
> > > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > > -
> > > -double-vlen2-funcs = cos exp log pow sin sincos
> > > -double-vlen4-funcs = cos exp log pow sin sincos
> > > -double-vlen4-avx2-funcs = cos exp log pow sin sincos
> > > -double-vlen8-funcs = cos exp log pow sin sincos
> > > -float-vlen4-funcs = cos exp log pow sin sincos
> > > -float-vlen8-funcs = cos exp log pow sin sincos
> > > -float-vlen8-avx2-funcs = cos exp log pow sin sincos
> > > -float-vlen16-funcs = cos exp log pow sin sincos
> > > + float-vlen4 float-vlen8 float-vlen8-avx2 \
> > > + double-vlen8 float-vlen16
> > > +tests += \
> > > + $(libmvec-abi-func-tests) \
> > > + $(libmvec-abi-func-avx-tests) \
> > > + $(libmvec-abi-func-avx2-tests) \
> > > + $(libmvec-abi-func-avx512f-tests)
> > > +
> > > +double-vlen2-funcs = $(libmvec-funcs)
> > > +double-vlen4-funcs = $(libmvec-funcs)
> > > +double-vlen4-avx2-funcs = $(libmvec-funcs)
> > > +double-vlen8-funcs = $(libmvec-funcs)
> > > +float-vlen4-funcs = $(libmvec-funcs)
> > > +float-vlen8-funcs = $(libmvec-funcs)
> > > +float-vlen8-avx2-funcs = $(libmvec-funcs)
> > > +float-vlen16-funcs = $(libmvec-funcs)
> > >
> > > double-vlen4-arch-ext-cflags = -mavx
> > > double-vlen4-arch-ext2-cflags = -mavx2
> > > @@ -91,22 +52,13 @@ float-vlen8-arch-ext-cflags = -mavx
> > > float-vlen8-arch-ext2-cflags = -mavx2
> > > float-vlen16-arch-ext-cflags = -mavx512f
> > >
> > > -libmvec-sincos-cflags = $(libm-test-fast-math-cflags) -fno-inline -fopenmp -Wno-unknown-pragmas
> > > +libmvec-abi-test-cflags = \
> > > + $(libm-test-fast-math-cflags) \
> > > + -fno-inline -fopenmp -Wno-unknown-pragmas
> > >
> > > CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)
> > >
> > > CFLAGS-test-float-vlen8-avx2-wrappers.c = $(float-vlen8-arch-ext2-cflags)
> > > -
> > > -CFLAGS-test-double-libmvec-sincos.c = $(libmvec-sincos-cflags)
> > > -CFLAGS-test-double-libmvec-sincos-avx.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext-cflags)
> > > -CFLAGS-test-double-libmvec-sincos-avx2.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext2-cflags)
> > > -CFLAGS-test-double-libmvec-sincos-avx512f.c = $(libmvec-sincos-cflags) $(double-vlen8-arch-ext-cflags)
> > > -
> > > -CFLAGS-test-float-libmvec-sincosf.c = $(libmvec-sincos-cflags)
> > > -CFLAGS-test-float-libmvec-sincosf-avx.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext-cflags)
> > > -CFLAGS-test-float-libmvec-sincosf-avx2.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext2-cflags)
> > > -CFLAGS-test-float-libmvec-sincosf-avx512f.c = $(libmvec-sincos-cflags) $(float-vlen16-arch-ext-cflags)
> > > -endif
> > > endif
> > >
> > > ifeq ($(subdir)$(config-cflags-mprefer-vector-width),mathyes)
> > > diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
> > > index d425ffd6d3..ec796277a5 100644
> > > --- a/sysdeps/x86_64/fpu/multiarch/Makefile
> > > +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
> > > @@ -59,52 +59,24 @@ CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
> > > endif
> > >
> > > ifeq ($(subdir),mathvec)
> > > -libmvec-sysdep_routines += svml_d_cos2_core_sse4 svml_d_cos4_core_avx2 \
> > > - svml_d_cos8_core_avx512 svml_d_sin2_core_sse4 \
> > > - svml_d_sin4_core_avx2 svml_d_sin8_core_avx512 \
> > > - svml_d_log2_core_sse4 svml_d_log4_core_avx2 \
> > > - svml_d_log8_core_avx512 svml_d_sincos2_core_sse4 \
> > > - svml_d_sincos4_core_avx2 svml_d_sincos8_core_avx512 \
> > > - svml_s_cosf4_core_sse4 svml_s_cosf8_core_avx2 \
> > > - svml_s_cosf16_core_avx512 svml_s_sinf4_core_sse4 \
> > > - svml_s_sinf8_core_avx2 svml_s_sinf16_core_avx512 \
> > > - svml_s_logf4_core_sse4 svml_s_logf8_core_avx2 \
> > > - svml_s_logf16_core_avx512 svml_d_exp2_core_sse4 \
> > > - svml_d_exp4_core_avx2 svml_d_exp8_core_avx512 \
> > > - svml_s_expf4_core_sse4 svml_s_expf8_core_avx2 \
> > > - svml_s_expf16_core_avx512 svml_d_pow2_core_sse4 \
> > > - svml_d_pow4_core_avx2 svml_d_pow8_core_avx512 \
> > > - svml_s_powf4_core_sse4 svml_s_powf8_core_avx2 \
> > > - svml_s_powf16_core_avx512 svml_s_sincosf4_core_sse4 \
> > > - svml_s_sincosf8_core_avx2 \
> > > - svml_s_sincosf16_core_avx512 \
> > > - svml_d_cos2_core-sse2 svml_d_cos4_core-sse \
> > > - svml_d_cos8_core-avx2 svml_d_exp2_core-sse2 \
> > > - svml_d_exp4_core-sse svml_d_exp8_core-avx2 \
> > > - svml_d_log2_core-sse2 svml_d_log4_core-sse \
> > > - svml_d_log8_core-avx2 svml_d_pow2_core-sse2 \
> > > - svml_d_pow4_core-sse svml_d_pow8_core-avx2 \
> > > - svml_d_sin2_core-sse2 svml_d_sin4_core-sse \
> > > - svml_d_sin8_core-avx2 \
> > > - svml_d_sincos2_core-sse2 \
> > > - svml_d_sincos4_core-sse \
> > > - svml_d_sincos8_core-avx2 \
> > > - svml_s_cosf16_core-avx2 \
> > > - svml_s_cosf4_core-sse2 \
> > > - svml_s_cosf8_core-sse \
> > > - svml_s_expf16_core-avx2 \
> > > - svml_s_expf4_core-sse2 \
> > > - svml_s_expf8_core-sse \
> > > - svml_s_logf16_core-avx2 \
> > > - svml_s_logf4_core-sse2 \
> > > - svml_s_logf8_core-sse \
> > > - svml_s_powf16_core-avx2 \
> > > - svml_s_powf4_core-sse2 \
> > > - svml_s_powf8_core-sse \
> > > - svml_s_sincosf16_core-avx2 \
> > > - svml_s_sincosf4_core-sse2 \
> > > - svml_s_sincosf8_core-sse \
> > > - svml_s_sinf16_core-avx2 \
> > > - svml_s_sinf4_core-sse2 \
> > > - svml_s_sinf8_core-sse
> > > +libmvec-multiarch-double-func-list = \
> > > + 2_core-sse2 \
> > > + 2_core_sse4 \
> > > + 4_core-sse \
> > > + 4_core_avx2 \
> > > + 8_core-avx2 \
> > > + 8_core_avx512
> > > +libmvec-multiarch-float-func-list = \
> > > + f4_core-sse2 \
> > > + f4_core_sse4 \
> > > + f8_core-sse \
> > > + f8_core_avx2 \
> > > + f16_core-avx2 \
> > > + f16_core_avx512
> > > +
> > > +libmvec-sysdep_routines += \
> > > + $(foreach l,$(libmvec-multiarch-double-func-list), \
> > > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > > + $(foreach l,$(libmvec-multiarch-float-func-list), \
> > > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > > endif
> > > --
> > > 2.31.1
> > >
> >
> > PING.
> >
> > --
> > H.J.
>
> How can I test this patch?
This patch improves libmvec build and test. Otherwise it should be NFC.
Sunil has a followup patch to add the rest of libmvec ABI tests.
--
H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: PING [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig
2021-10-20 18:18 ` H.J. Lu
@ 2021-10-20 18:47 ` Noah Goldstein
2021-10-20 18:53 ` H.J. Lu
0 siblings, 1 reply; 6+ messages in thread
From: Noah Goldstein @ 2021-10-20 18:47 UTC (permalink / raw)
To: H.J. Lu; +Cc: GNU C Library
On Wed, Oct 20, 2021 at 1:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>
> On Wed, Oct 20, 2021 at 11:07 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> >
> > On Wed, Oct 20, 2021 at 11:01 AM H.J. Lu via Libc-alpha
> > <libc-alpha@sourceware.org> wrote:
> > >
> > > On Fri, Oct 15, 2021 at 6:12 AM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > >
> > > > 1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which
> > > > contains libmvec ABI test dependencies and CFLAGS, in the build directory.
> > > > 2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS.
> > > >
> > > > Tested on SSE4, AVX, AVX2 and AVX512 machines.
> > > > ---
> > > > sysdeps/x86_64/fpu/Makeconfig | 92 ++++++++++++++++++
> > > > sysdeps/x86_64/fpu/Makefile | 134 +++++++++-----------------
> > > > sysdeps/x86_64/fpu/multiarch/Makefile | 68 ++++---------
> > > > 3 files changed, 155 insertions(+), 139 deletions(-)
> > > > create mode 100644 sysdeps/x86_64/fpu/Makeconfig
> > > >
> > > > diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig
> > > > new file mode 100644
> > > > index 0000000000..0a9341cdd9
> > > > --- /dev/null
> > > > +++ b/sysdeps/x86_64/fpu/Makeconfig
> > > > @@ -0,0 +1,92 @@
> > > > +# Generate libmvec.mk for libmvec ABI tests.
> > > > +
> > > > +# Copyright (C) 2021 Free Software Foundation, Inc.
> > > > +# This file is part of the GNU C Library.
> > > > +
> > > > +# The GNU C Library is free software; you can redistribute it and/or
> > > > +# modify it under the terms of the GNU Lesser General Public
> > > > +# License as published by the Free Software Foundation; either
> > > > +# version 2.1 of the License, or (at your option) any later version.
> > > > +
> > > > +# The GNU C Library is distributed in the hope that it will be useful,
> > > > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > > > +# Lesser General Public License for more details.
> > > > +
> > > > +# You should have received a copy of the GNU Lesser General Public
> > > > +# License along with the GNU C Library; if not, see
> > > > +# <https://www.gnu.org/licenses/>.
> > > > +
> > > > +-include $(common-objpfx)libmvec.mk
> > > > +postclean-generated += libmvec.mk
> > > > +
> > > > +# Define for both math and mathvec directories.
> > > > +libmvec-funcs = \
> > > > + cos \
> > > > + exp \
> > > > + log \
> > > > + pow \
> > > > + sin \
> > > > + sincos \
> > > > +
> > > > +# FIXME: Add ABI tests for all libmvec functions and replace
> > > > +# libmvec-abi-funcs with libmvec-funcs.
> > > > +libmvec-abi-funcs = \
> > > > + sincos
> >
> > What if we add a function that only supports some extensions (i.e just
> > AVX2 but not AVX and AVX512)?
>
> All libmvec functions must support SSE, AVX and AVX512.
>
> > > > +
> > > > +# The base libmvec ABI tests.
> > > > +libmvec-abi-func-tests = \
> > > > + $(addprefix test-double-libmvec-,$(libmvec-abi-funcs)) \
> > > > + $(addsuffix f,$(addprefix test-float-libmvec-,$(libmvec-abi-funcs)))
> > > > +
> > > > +# The AVX libmvec ABI tests.
> > > > +libmvec-abi-func-avx-tests = \
> > > > + $(addsuffix -avx,$(libmvec-abi-func-tests))
> > > > +#
> > > > +# The AVX2 libmvec ABI tests.
> > > > +libmvec-abi-func-avx2-tests = \
> > > > + $(addsuffix -avx2,$(libmvec-abi-func-tests))
> > > > +#
> > > > +# The AVX512F libmvec ABI tests.
> > > > +libmvec-abi-func-avx512f-tests = \
> > > > + $(addsuffix -avx512f,$(libmvec-abi-func-tests))
> > > > +
> > > > +$(common-objpfx)libmvec.mk: $(common-objpfx)config.make
> > > > + (echo "ifeq (\$$(subdir)\$$(build-mathvec),mathyes)"; \
> > > > + echo; \
> > > > + for i in "" "-avx" "-avx2" "-avx512f"; do \
> > > > + for t in $(libmvec-abi-func-tests); do \
> > > > + echo "\$$(objpfx)$$t$$i: \\"; \
> > > > + echo " \$$(objpfx)$$t$$i.o \\"; \
> > > > + echo " \$$(objpfx)test-libmvec$$i.o \$$(libmvec)"; \
> > > > + echo; \
> > > > + done; \
> > > > + done; \
> > > > + for t in $(libmvec-abi-funcs); do \
> > > > + echo "CFLAGS-test-double-libmvec-$$t.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags)"; \
> > > > + echo "CFLAGS-test-double-libmvec-$$t-avx.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > + echo " \$$(double-vlen4-arch-ext-cflags)"; \
> > > > + echo "CFLAGS-test-double-libmvec-$$t-avx2.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > + echo " \$$(double-vlen4-arch-ext2-cflags)"; \
> > > > + echo "CFLAGS-test-double-libmvec-$$t-avx512f.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > + echo " \$$(double-vlen8-arch-ext-cflags)"; \
> > > > + echo; \
> > > > + echo "CFLAGS-test-float-libmvec-$${t}f.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags)"; \
> > > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > + echo " \$$(float-vlen8-arch-ext-cflags)"; \
> > > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx2.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > + echo " \$$(float-vlen8-arch-ext2-cflags)"; \
> > > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx512f.c = \\"; \
> > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > + echo " \$$(float-vlen16-arch-ext-cflags)"; \
> > > > + echo; \
> > > > + done; \
> > > > + echo "endif") > $@T
> > > > + mv -f $@T $@
> > > > diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
> > > > index e5a65750c0..d172ae815d 100644
> > > > --- a/sysdeps/x86_64/fpu/Makefile
> > > > +++ b/sysdeps/x86_64/fpu/Makefile
> > > > @@ -1,87 +1,48 @@
> > > > ifeq ($(subdir),mathvec)
> > > > -libmvec-support += svml_d_cos2_core svml_d_cos4_core_avx \
> > > > - svml_d_cos4_core svml_d_cos8_core \
> > > > - svml_d_sin2_core svml_d_sin4_core_avx \
> > > > - svml_d_sin4_core svml_d_sin8_core svml_d_trig_data \
> > > > - svml_s_cosf4_core svml_s_cosf8_core_avx \
> > > > - svml_s_cosf8_core svml_s_cosf16_core svml_s_trig_data \
> > > > - svml_s_sinf4_core svml_s_sinf8_core_avx \
> > > > - svml_s_sinf8_core svml_s_sinf16_core \
> > > > - svml_d_sincos2_core svml_d_sincos4_core_avx \
> > > > - svml_d_sincos4_core svml_d_sincos8_core \
> > > > - svml_d_log2_core svml_d_log4_core_avx svml_d_log4_core \
> > > > - svml_d_log8_core svml_d_log_data svml_s_logf4_core \
> > > > - svml_s_logf8_core_avx svml_s_logf8_core svml_s_logf16_core \
> > > > - svml_s_logf_data svml_d_exp2_core svml_d_exp4_core_avx \
> > > > - svml_d_exp4_core svml_d_exp8_core svml_d_exp_data \
> > > > - svml_s_expf4_core svml_s_expf8_core_avx svml_s_expf8_core \
> > > > - svml_s_expf16_core svml_s_expf_data svml_d_pow2_core \
> > > > - svml_d_pow4_core_avx svml_d_pow4_core svml_d_pow8_core \
> > > > - svml_d_pow_data svml_s_powf4_core svml_s_powf8_core_avx \
> > > > - svml_s_powf8_core svml_s_powf16_core svml_s_powf_data \
> > > > - svml_s_sincosf4_core svml_s_sincosf8_core_avx \
> > > > - svml_s_sincosf8_core svml_s_sincosf16_core
> > > > +libmvec-double-func-list = \
> > > > + 2_core \
> > > > + 4_core \
> > > > + 4_core_avx \
> > > > + 8_core
> > > > +libmvec-float-func-list = \
> > > > + f4_core \
> > > > + f8_core \
> > > > + f8_core_avx \
> > > > + f16_core
> > > > +libmvec-support += \
> > > > + svml_d_exp_data \
> > > > + svml_d_log_data \
> > > > + svml_d_pow_data \
> > > > + svml_d_trig_data \
> > > > + svml_s_expf_data \
> > > > + svml_s_logf_data \
> > > > + svml_s_powf_data \
> > > > + svml_s_trig_data \
> > > > + $(foreach l,$(libmvec-double-func-list), \
> > > > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > > > + $(foreach l,$(libmvec-float-func-list), \
> > > > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > > > endif
> > > >
> > > > # Variables for libmvec tests.
> > > > -ifeq ($(subdir),math)
> > > > -ifeq ($(build-mathvec),yes)
> > > > +ifeq ($(subdir)$(build-mathvec),mathyes)
> > > > libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
> > > > - float-vlen4 float-vlen8 float-vlen8-avx2
> > > > -tests += test-double-libmvec-sincos test-double-libmvec-sincos-avx \
> > > > - test-double-libmvec-sincos-avx2 test-float-libmvec-sincosf \
> > > > - test-float-libmvec-sincosf-avx test-float-libmvec-sincosf-avx2
> > > > -test-extras += test-libmvec test-libmvec-avx \
> > > > - test-libmvec-avx2 test-libmvec-avx512f
> > > > -extra-test-objs += test-libmvec.o test-libmvec-avx.o \
> > > > - test-libmvec-avx2.o test-libmvec-avx512f.o
> > > > -
> > > > -$(objpfx)test-double-libmvec-sincos: \
> > > > - $(objpfx)test-double-libmvec-sincos.o \
> > > > - $(objpfx)test-libmvec.o $(libmvec)
> > > > -
> > > > -$(objpfx)test-double-libmvec-sincos-avx: \
> > > > - $(objpfx)test-double-libmvec-sincos-avx.o \
> > > > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > > > -
> > > > -$(objpfx)test-double-libmvec-sincos-avx2: \
> > > > - $(objpfx)test-double-libmvec-sincos-avx2.o \
> > > > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > > > -
> > > > -$(objpfx)test-float-libmvec-sincosf: \
> > > > - $(objpfx)test-float-libmvec-sincosf.o \
> > > > - $(objpfx)test-libmvec.o $(libmvec)
> > > > -
> > > > -$(objpfx)test-float-libmvec-sincosf-avx: \
> > > > - $(objpfx)test-float-libmvec-sincosf-avx.o \
> > > > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > > > -
> > > > -$(objpfx)test-float-libmvec-sincosf-avx2: \
> > > > - $(objpfx)test-float-libmvec-sincosf-avx2.o \
> > > > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > > > -
> > > > -libmvec-tests += double-vlen8 float-vlen16
> > > > -tests += test-double-libmvec-sincos-avx512f \
> > > > - test-float-libmvec-sincosf-avx512f
> > > > -test-extras += test-libmvec-avx512f
> > > > -extra-test-objs += test-libmvec-avx512f.o
> > > > -
> > > > -$(objpfx)test-double-libmvec-sincos-avx512f: \
> > > > - $(objpfx)test-double-libmvec-sincos-avx512f.o \
> > > > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > > > -
> > > > -$(objpfx)test-float-libmvec-sincosf-avx512f: \
> > > > - $(objpfx)test-float-libmvec-sincosf-avx512f.o \
> > > > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > > > -
> > > > -double-vlen2-funcs = cos exp log pow sin sincos
> > > > -double-vlen4-funcs = cos exp log pow sin sincos
> > > > -double-vlen4-avx2-funcs = cos exp log pow sin sincos
> > > > -double-vlen8-funcs = cos exp log pow sin sincos
> > > > -float-vlen4-funcs = cos exp log pow sin sincos
> > > > -float-vlen8-funcs = cos exp log pow sin sincos
> > > > -float-vlen8-avx2-funcs = cos exp log pow sin sincos
> > > > -float-vlen16-funcs = cos exp log pow sin sincos
> > > > + float-vlen4 float-vlen8 float-vlen8-avx2 \
> > > > + double-vlen8 float-vlen16
> > > > +tests += \
> > > > + $(libmvec-abi-func-tests) \
> > > > + $(libmvec-abi-func-avx-tests) \
> > > > + $(libmvec-abi-func-avx2-tests) \
> > > > + $(libmvec-abi-func-avx512f-tests)
> > > > +
> > > > +double-vlen2-funcs = $(libmvec-funcs)
> > > > +double-vlen4-funcs = $(libmvec-funcs)
> > > > +double-vlen4-avx2-funcs = $(libmvec-funcs)
> > > > +double-vlen8-funcs = $(libmvec-funcs)
> > > > +float-vlen4-funcs = $(libmvec-funcs)
> > > > +float-vlen8-funcs = $(libmvec-funcs)
> > > > +float-vlen8-avx2-funcs = $(libmvec-funcs)
> > > > +float-vlen16-funcs = $(libmvec-funcs)
> > > >
> > > > double-vlen4-arch-ext-cflags = -mavx
> > > > double-vlen4-arch-ext2-cflags = -mavx2
> > > > @@ -91,22 +52,13 @@ float-vlen8-arch-ext-cflags = -mavx
> > > > float-vlen8-arch-ext2-cflags = -mavx2
> > > > float-vlen16-arch-ext-cflags = -mavx512f
> > > >
> > > > -libmvec-sincos-cflags = $(libm-test-fast-math-cflags) -fno-inline -fopenmp -Wno-unknown-pragmas
> > > > +libmvec-abi-test-cflags = \
> > > > + $(libm-test-fast-math-cflags) \
> > > > + -fno-inline -fopenmp -Wno-unknown-pragmas
> > > >
> > > > CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)
> > > >
> > > > CFLAGS-test-float-vlen8-avx2-wrappers.c = $(float-vlen8-arch-ext2-cflags)
> > > > -
> > > > -CFLAGS-test-double-libmvec-sincos.c = $(libmvec-sincos-cflags)
> > > > -CFLAGS-test-double-libmvec-sincos-avx.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext-cflags)
> > > > -CFLAGS-test-double-libmvec-sincos-avx2.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext2-cflags)
> > > > -CFLAGS-test-double-libmvec-sincos-avx512f.c = $(libmvec-sincos-cflags) $(double-vlen8-arch-ext-cflags)
> > > > -
> > > > -CFLAGS-test-float-libmvec-sincosf.c = $(libmvec-sincos-cflags)
> > > > -CFLAGS-test-float-libmvec-sincosf-avx.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext-cflags)
> > > > -CFLAGS-test-float-libmvec-sincosf-avx2.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext2-cflags)
> > > > -CFLAGS-test-float-libmvec-sincosf-avx512f.c = $(libmvec-sincos-cflags) $(float-vlen16-arch-ext-cflags)
> > > > -endif
> > > > endif
> > > >
> > > > ifeq ($(subdir)$(config-cflags-mprefer-vector-width),mathyes)
> > > > diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
> > > > index d425ffd6d3..ec796277a5 100644
> > > > --- a/sysdeps/x86_64/fpu/multiarch/Makefile
> > > > +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
> > > > @@ -59,52 +59,24 @@ CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
> > > > endif
> > > >
> > > > ifeq ($(subdir),mathvec)
> > > > -libmvec-sysdep_routines += svml_d_cos2_core_sse4 svml_d_cos4_core_avx2 \
> > > > - svml_d_cos8_core_avx512 svml_d_sin2_core_sse4 \
> > > > - svml_d_sin4_core_avx2 svml_d_sin8_core_avx512 \
> > > > - svml_d_log2_core_sse4 svml_d_log4_core_avx2 \
> > > > - svml_d_log8_core_avx512 svml_d_sincos2_core_sse4 \
> > > > - svml_d_sincos4_core_avx2 svml_d_sincos8_core_avx512 \
> > > > - svml_s_cosf4_core_sse4 svml_s_cosf8_core_avx2 \
> > > > - svml_s_cosf16_core_avx512 svml_s_sinf4_core_sse4 \
> > > > - svml_s_sinf8_core_avx2 svml_s_sinf16_core_avx512 \
> > > > - svml_s_logf4_core_sse4 svml_s_logf8_core_avx2 \
> > > > - svml_s_logf16_core_avx512 svml_d_exp2_core_sse4 \
> > > > - svml_d_exp4_core_avx2 svml_d_exp8_core_avx512 \
> > > > - svml_s_expf4_core_sse4 svml_s_expf8_core_avx2 \
> > > > - svml_s_expf16_core_avx512 svml_d_pow2_core_sse4 \
> > > > - svml_d_pow4_core_avx2 svml_d_pow8_core_avx512 \
> > > > - svml_s_powf4_core_sse4 svml_s_powf8_core_avx2 \
> > > > - svml_s_powf16_core_avx512 svml_s_sincosf4_core_sse4 \
> > > > - svml_s_sincosf8_core_avx2 \
> > > > - svml_s_sincosf16_core_avx512 \
> > > > - svml_d_cos2_core-sse2 svml_d_cos4_core-sse \
> > > > - svml_d_cos8_core-avx2 svml_d_exp2_core-sse2 \
> > > > - svml_d_exp4_core-sse svml_d_exp8_core-avx2 \
> > > > - svml_d_log2_core-sse2 svml_d_log4_core-sse \
> > > > - svml_d_log8_core-avx2 svml_d_pow2_core-sse2 \
> > > > - svml_d_pow4_core-sse svml_d_pow8_core-avx2 \
> > > > - svml_d_sin2_core-sse2 svml_d_sin4_core-sse \
> > > > - svml_d_sin8_core-avx2 \
> > > > - svml_d_sincos2_core-sse2 \
> > > > - svml_d_sincos4_core-sse \
> > > > - svml_d_sincos8_core-avx2 \
> > > > - svml_s_cosf16_core-avx2 \
> > > > - svml_s_cosf4_core-sse2 \
> > > > - svml_s_cosf8_core-sse \
> > > > - svml_s_expf16_core-avx2 \
> > > > - svml_s_expf4_core-sse2 \
> > > > - svml_s_expf8_core-sse \
> > > > - svml_s_logf16_core-avx2 \
> > > > - svml_s_logf4_core-sse2 \
> > > > - svml_s_logf8_core-sse \
> > > > - svml_s_powf16_core-avx2 \
> > > > - svml_s_powf4_core-sse2 \
> > > > - svml_s_powf8_core-sse \
> > > > - svml_s_sincosf16_core-avx2 \
> > > > - svml_s_sincosf4_core-sse2 \
> > > > - svml_s_sincosf8_core-sse \
> > > > - svml_s_sinf16_core-avx2 \
> > > > - svml_s_sinf4_core-sse2 \
> > > > - svml_s_sinf8_core-sse
> > > > +libmvec-multiarch-double-func-list = \
> > > > + 2_core-sse2 \
> > > > + 2_core_sse4 \
> > > > + 4_core-sse \
> > > > + 4_core_avx2 \
> > > > + 8_core-avx2 \
> > > > + 8_core_avx512
> > > > +libmvec-multiarch-float-func-list = \
> > > > + f4_core-sse2 \
> > > > + f4_core_sse4 \
> > > > + f8_core-sse \
> > > > + f8_core_avx2 \
> > > > + f16_core-avx2 \
> > > > + f16_core_avx512
> > > > +
> > > > +libmvec-sysdep_routines += \
> > > > + $(foreach l,$(libmvec-multiarch-double-func-list), \
> > > > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > > > + $(foreach l,$(libmvec-multiarch-float-func-list), \
> > > > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > > > endif
> > > > --
> > > > 2.31.1
> > > >
> > >
> > > PING.
> > >
> > > --
> > > H.J.
> >
> > How can I test this patch?
>
> This patch improves libmvec build and test. Otherwise it should be NFC.
> Sunil has a followup patch to add the rest of libmvec ABI tests.
Err, I meant how can I test this for review.
Build succeeds on x86_64. All tests in mathvec/math pass. As well the
exact same tests run.
If that's sufficient then LGTM.
>
> --
> H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: PING [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig
2021-10-20 18:47 ` Noah Goldstein
@ 2021-10-20 18:53 ` H.J. Lu
0 siblings, 0 replies; 6+ messages in thread
From: H.J. Lu @ 2021-10-20 18:53 UTC (permalink / raw)
To: Noah Goldstein; +Cc: GNU C Library
On Wed, Oct 20, 2021 at 11:47 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> On Wed, Oct 20, 2021 at 1:19 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> >
> > On Wed, Oct 20, 2021 at 11:07 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
> > >
> > > On Wed, Oct 20, 2021 at 11:01 AM H.J. Lu via Libc-alpha
> > > <libc-alpha@sourceware.org> wrote:
> > > >
> > > > On Fri, Oct 15, 2021 at 6:12 AM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > > >
> > > > > 1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which
> > > > > contains libmvec ABI test dependencies and CFLAGS, in the build directory.
> > > > > 2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS.
> > > > >
> > > > > Tested on SSE4, AVX, AVX2 and AVX512 machines.
> > > > > ---
> > > > > sysdeps/x86_64/fpu/Makeconfig | 92 ++++++++++++++++++
> > > > > sysdeps/x86_64/fpu/Makefile | 134 +++++++++-----------------
> > > > > sysdeps/x86_64/fpu/multiarch/Makefile | 68 ++++---------
> > > > > 3 files changed, 155 insertions(+), 139 deletions(-)
> > > > > create mode 100644 sysdeps/x86_64/fpu/Makeconfig
> > > > >
> > > > > diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig
> > > > > new file mode 100644
> > > > > index 0000000000..0a9341cdd9
> > > > > --- /dev/null
> > > > > +++ b/sysdeps/x86_64/fpu/Makeconfig
> > > > > @@ -0,0 +1,92 @@
> > > > > +# Generate libmvec.mk for libmvec ABI tests.
> > > > > +
> > > > > +# Copyright (C) 2021 Free Software Foundation, Inc.
> > > > > +# This file is part of the GNU C Library.
> > > > > +
> > > > > +# The GNU C Library is free software; you can redistribute it and/or
> > > > > +# modify it under the terms of the GNU Lesser General Public
> > > > > +# License as published by the Free Software Foundation; either
> > > > > +# version 2.1 of the License, or (at your option) any later version.
> > > > > +
> > > > > +# The GNU C Library is distributed in the hope that it will be useful,
> > > > > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > > > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > > > > +# Lesser General Public License for more details.
> > > > > +
> > > > > +# You should have received a copy of the GNU Lesser General Public
> > > > > +# License along with the GNU C Library; if not, see
> > > > > +# <https://www.gnu.org/licenses/>.
> > > > > +
> > > > > +-include $(common-objpfx)libmvec.mk
> > > > > +postclean-generated += libmvec.mk
> > > > > +
> > > > > +# Define for both math and mathvec directories.
> > > > > +libmvec-funcs = \
> > > > > + cos \
> > > > > + exp \
> > > > > + log \
> > > > > + pow \
> > > > > + sin \
> > > > > + sincos \
> > > > > +
> > > > > +# FIXME: Add ABI tests for all libmvec functions and replace
> > > > > +# libmvec-abi-funcs with libmvec-funcs.
> > > > > +libmvec-abi-funcs = \
> > > > > + sincos
> > >
> > > What if we add a function that only supports some extensions (i.e just
> > > AVX2 but not AVX and AVX512)?
> >
> > All libmvec functions must support SSE, AVX and AVX512.
> >
> > > > > +
> > > > > +# The base libmvec ABI tests.
> > > > > +libmvec-abi-func-tests = \
> > > > > + $(addprefix test-double-libmvec-,$(libmvec-abi-funcs)) \
> > > > > + $(addsuffix f,$(addprefix test-float-libmvec-,$(libmvec-abi-funcs)))
> > > > > +
> > > > > +# The AVX libmvec ABI tests.
> > > > > +libmvec-abi-func-avx-tests = \
> > > > > + $(addsuffix -avx,$(libmvec-abi-func-tests))
> > > > > +#
> > > > > +# The AVX2 libmvec ABI tests.
> > > > > +libmvec-abi-func-avx2-tests = \
> > > > > + $(addsuffix -avx2,$(libmvec-abi-func-tests))
> > > > > +#
> > > > > +# The AVX512F libmvec ABI tests.
> > > > > +libmvec-abi-func-avx512f-tests = \
> > > > > + $(addsuffix -avx512f,$(libmvec-abi-func-tests))
> > > > > +
> > > > > +$(common-objpfx)libmvec.mk: $(common-objpfx)config.make
> > > > > + (echo "ifeq (\$$(subdir)\$$(build-mathvec),mathyes)"; \
> > > > > + echo; \
> > > > > + for i in "" "-avx" "-avx2" "-avx512f"; do \
> > > > > + for t in $(libmvec-abi-func-tests); do \
> > > > > + echo "\$$(objpfx)$$t$$i: \\"; \
> > > > > + echo " \$$(objpfx)$$t$$i.o \\"; \
> > > > > + echo " \$$(objpfx)test-libmvec$$i.o \$$(libmvec)"; \
> > > > > + echo; \
> > > > > + done; \
> > > > > + done; \
> > > > > + for t in $(libmvec-abi-funcs); do \
> > > > > + echo "CFLAGS-test-double-libmvec-$$t.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags)"; \
> > > > > + echo "CFLAGS-test-double-libmvec-$$t-avx.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > > + echo " \$$(double-vlen4-arch-ext-cflags)"; \
> > > > > + echo "CFLAGS-test-double-libmvec-$$t-avx2.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > > + echo " \$$(double-vlen4-arch-ext2-cflags)"; \
> > > > > + echo "CFLAGS-test-double-libmvec-$$t-avx512f.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > > + echo " \$$(double-vlen8-arch-ext-cflags)"; \
> > > > > + echo; \
> > > > > + echo "CFLAGS-test-float-libmvec-$${t}f.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags)"; \
> > > > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > > + echo " \$$(float-vlen8-arch-ext-cflags)"; \
> > > > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx2.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > > + echo " \$$(float-vlen8-arch-ext2-cflags)"; \
> > > > > + echo "CFLAGS-test-float-libmvec-$${t}f-avx512f.c = \\"; \
> > > > > + echo " \$$(libmvec-abi-test-cflags) \\"; \
> > > > > + echo " \$$(float-vlen16-arch-ext-cflags)"; \
> > > > > + echo; \
> > > > > + done; \
> > > > > + echo "endif") > $@T
> > > > > + mv -f $@T $@
> > > > > diff --git a/sysdeps/x86_64/fpu/Makefile b/sysdeps/x86_64/fpu/Makefile
> > > > > index e5a65750c0..d172ae815d 100644
> > > > > --- a/sysdeps/x86_64/fpu/Makefile
> > > > > +++ b/sysdeps/x86_64/fpu/Makefile
> > > > > @@ -1,87 +1,48 @@
> > > > > ifeq ($(subdir),mathvec)
> > > > > -libmvec-support += svml_d_cos2_core svml_d_cos4_core_avx \
> > > > > - svml_d_cos4_core svml_d_cos8_core \
> > > > > - svml_d_sin2_core svml_d_sin4_core_avx \
> > > > > - svml_d_sin4_core svml_d_sin8_core svml_d_trig_data \
> > > > > - svml_s_cosf4_core svml_s_cosf8_core_avx \
> > > > > - svml_s_cosf8_core svml_s_cosf16_core svml_s_trig_data \
> > > > > - svml_s_sinf4_core svml_s_sinf8_core_avx \
> > > > > - svml_s_sinf8_core svml_s_sinf16_core \
> > > > > - svml_d_sincos2_core svml_d_sincos4_core_avx \
> > > > > - svml_d_sincos4_core svml_d_sincos8_core \
> > > > > - svml_d_log2_core svml_d_log4_core_avx svml_d_log4_core \
> > > > > - svml_d_log8_core svml_d_log_data svml_s_logf4_core \
> > > > > - svml_s_logf8_core_avx svml_s_logf8_core svml_s_logf16_core \
> > > > > - svml_s_logf_data svml_d_exp2_core svml_d_exp4_core_avx \
> > > > > - svml_d_exp4_core svml_d_exp8_core svml_d_exp_data \
> > > > > - svml_s_expf4_core svml_s_expf8_core_avx svml_s_expf8_core \
> > > > > - svml_s_expf16_core svml_s_expf_data svml_d_pow2_core \
> > > > > - svml_d_pow4_core_avx svml_d_pow4_core svml_d_pow8_core \
> > > > > - svml_d_pow_data svml_s_powf4_core svml_s_powf8_core_avx \
> > > > > - svml_s_powf8_core svml_s_powf16_core svml_s_powf_data \
> > > > > - svml_s_sincosf4_core svml_s_sincosf8_core_avx \
> > > > > - svml_s_sincosf8_core svml_s_sincosf16_core
> > > > > +libmvec-double-func-list = \
> > > > > + 2_core \
> > > > > + 4_core \
> > > > > + 4_core_avx \
> > > > > + 8_core
> > > > > +libmvec-float-func-list = \
> > > > > + f4_core \
> > > > > + f8_core \
> > > > > + f8_core_avx \
> > > > > + f16_core
> > > > > +libmvec-support += \
> > > > > + svml_d_exp_data \
> > > > > + svml_d_log_data \
> > > > > + svml_d_pow_data \
> > > > > + svml_d_trig_data \
> > > > > + svml_s_expf_data \
> > > > > + svml_s_logf_data \
> > > > > + svml_s_powf_data \
> > > > > + svml_s_trig_data \
> > > > > + $(foreach l,$(libmvec-double-func-list), \
> > > > > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > > > > + $(foreach l,$(libmvec-float-func-list), \
> > > > > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > > > > endif
> > > > >
> > > > > # Variables for libmvec tests.
> > > > > -ifeq ($(subdir),math)
> > > > > -ifeq ($(build-mathvec),yes)
> > > > > +ifeq ($(subdir)$(build-mathvec),mathyes)
> > > > > libmvec-tests += double-vlen2 double-vlen4 double-vlen4-avx2 \
> > > > > - float-vlen4 float-vlen8 float-vlen8-avx2
> > > > > -tests += test-double-libmvec-sincos test-double-libmvec-sincos-avx \
> > > > > - test-double-libmvec-sincos-avx2 test-float-libmvec-sincosf \
> > > > > - test-float-libmvec-sincosf-avx test-float-libmvec-sincosf-avx2
> > > > > -test-extras += test-libmvec test-libmvec-avx \
> > > > > - test-libmvec-avx2 test-libmvec-avx512f
> > > > > -extra-test-objs += test-libmvec.o test-libmvec-avx.o \
> > > > > - test-libmvec-avx2.o test-libmvec-avx512f.o
> > > > > -
> > > > > -$(objpfx)test-double-libmvec-sincos: \
> > > > > - $(objpfx)test-double-libmvec-sincos.o \
> > > > > - $(objpfx)test-libmvec.o $(libmvec)
> > > > > -
> > > > > -$(objpfx)test-double-libmvec-sincos-avx: \
> > > > > - $(objpfx)test-double-libmvec-sincos-avx.o \
> > > > > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > > > > -
> > > > > -$(objpfx)test-double-libmvec-sincos-avx2: \
> > > > > - $(objpfx)test-double-libmvec-sincos-avx2.o \
> > > > > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > > > > -
> > > > > -$(objpfx)test-float-libmvec-sincosf: \
> > > > > - $(objpfx)test-float-libmvec-sincosf.o \
> > > > > - $(objpfx)test-libmvec.o $(libmvec)
> > > > > -
> > > > > -$(objpfx)test-float-libmvec-sincosf-avx: \
> > > > > - $(objpfx)test-float-libmvec-sincosf-avx.o \
> > > > > - $(objpfx)test-libmvec-avx.o $(libmvec)
> > > > > -
> > > > > -$(objpfx)test-float-libmvec-sincosf-avx2: \
> > > > > - $(objpfx)test-float-libmvec-sincosf-avx2.o \
> > > > > - $(objpfx)test-libmvec-avx2.o $(libmvec)
> > > > > -
> > > > > -libmvec-tests += double-vlen8 float-vlen16
> > > > > -tests += test-double-libmvec-sincos-avx512f \
> > > > > - test-float-libmvec-sincosf-avx512f
> > > > > -test-extras += test-libmvec-avx512f
> > > > > -extra-test-objs += test-libmvec-avx512f.o
> > > > > -
> > > > > -$(objpfx)test-double-libmvec-sincos-avx512f: \
> > > > > - $(objpfx)test-double-libmvec-sincos-avx512f.o \
> > > > > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > > > > -
> > > > > -$(objpfx)test-float-libmvec-sincosf-avx512f: \
> > > > > - $(objpfx)test-float-libmvec-sincosf-avx512f.o \
> > > > > - $(objpfx)test-libmvec-avx512f.o $(libmvec)
> > > > > -
> > > > > -double-vlen2-funcs = cos exp log pow sin sincos
> > > > > -double-vlen4-funcs = cos exp log pow sin sincos
> > > > > -double-vlen4-avx2-funcs = cos exp log pow sin sincos
> > > > > -double-vlen8-funcs = cos exp log pow sin sincos
> > > > > -float-vlen4-funcs = cos exp log pow sin sincos
> > > > > -float-vlen8-funcs = cos exp log pow sin sincos
> > > > > -float-vlen8-avx2-funcs = cos exp log pow sin sincos
> > > > > -float-vlen16-funcs = cos exp log pow sin sincos
> > > > > + float-vlen4 float-vlen8 float-vlen8-avx2 \
> > > > > + double-vlen8 float-vlen16
> > > > > +tests += \
> > > > > + $(libmvec-abi-func-tests) \
> > > > > + $(libmvec-abi-func-avx-tests) \
> > > > > + $(libmvec-abi-func-avx2-tests) \
> > > > > + $(libmvec-abi-func-avx512f-tests)
> > > > > +
> > > > > +double-vlen2-funcs = $(libmvec-funcs)
> > > > > +double-vlen4-funcs = $(libmvec-funcs)
> > > > > +double-vlen4-avx2-funcs = $(libmvec-funcs)
> > > > > +double-vlen8-funcs = $(libmvec-funcs)
> > > > > +float-vlen4-funcs = $(libmvec-funcs)
> > > > > +float-vlen8-funcs = $(libmvec-funcs)
> > > > > +float-vlen8-avx2-funcs = $(libmvec-funcs)
> > > > > +float-vlen16-funcs = $(libmvec-funcs)
> > > > >
> > > > > double-vlen4-arch-ext-cflags = -mavx
> > > > > double-vlen4-arch-ext2-cflags = -mavx2
> > > > > @@ -91,22 +52,13 @@ float-vlen8-arch-ext-cflags = -mavx
> > > > > float-vlen8-arch-ext2-cflags = -mavx2
> > > > > float-vlen16-arch-ext-cflags = -mavx512f
> > > > >
> > > > > -libmvec-sincos-cflags = $(libm-test-fast-math-cflags) -fno-inline -fopenmp -Wno-unknown-pragmas
> > > > > +libmvec-abi-test-cflags = \
> > > > > + $(libm-test-fast-math-cflags) \
> > > > > + -fno-inline -fopenmp -Wno-unknown-pragmas
> > > > >
> > > > > CFLAGS-test-double-vlen4-avx2-wrappers.c = $(double-vlen4-arch-ext2-cflags)
> > > > >
> > > > > CFLAGS-test-float-vlen8-avx2-wrappers.c = $(float-vlen8-arch-ext2-cflags)
> > > > > -
> > > > > -CFLAGS-test-double-libmvec-sincos.c = $(libmvec-sincos-cflags)
> > > > > -CFLAGS-test-double-libmvec-sincos-avx.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext-cflags)
> > > > > -CFLAGS-test-double-libmvec-sincos-avx2.c = $(libmvec-sincos-cflags) $(double-vlen4-arch-ext2-cflags)
> > > > > -CFLAGS-test-double-libmvec-sincos-avx512f.c = $(libmvec-sincos-cflags) $(double-vlen8-arch-ext-cflags)
> > > > > -
> > > > > -CFLAGS-test-float-libmvec-sincosf.c = $(libmvec-sincos-cflags)
> > > > > -CFLAGS-test-float-libmvec-sincosf-avx.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext-cflags)
> > > > > -CFLAGS-test-float-libmvec-sincosf-avx2.c = $(libmvec-sincos-cflags) $(float-vlen8-arch-ext2-cflags)
> > > > > -CFLAGS-test-float-libmvec-sincosf-avx512f.c = $(libmvec-sincos-cflags) $(float-vlen16-arch-ext-cflags)
> > > > > -endif
> > > > > endif
> > > > >
> > > > > ifeq ($(subdir)$(config-cflags-mprefer-vector-width),mathyes)
> > > > > diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
> > > > > index d425ffd6d3..ec796277a5 100644
> > > > > --- a/sysdeps/x86_64/fpu/multiarch/Makefile
> > > > > +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
> > > > > @@ -59,52 +59,24 @@ CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
> > > > > endif
> > > > >
> > > > > ifeq ($(subdir),mathvec)
> > > > > -libmvec-sysdep_routines += svml_d_cos2_core_sse4 svml_d_cos4_core_avx2 \
> > > > > - svml_d_cos8_core_avx512 svml_d_sin2_core_sse4 \
> > > > > - svml_d_sin4_core_avx2 svml_d_sin8_core_avx512 \
> > > > > - svml_d_log2_core_sse4 svml_d_log4_core_avx2 \
> > > > > - svml_d_log8_core_avx512 svml_d_sincos2_core_sse4 \
> > > > > - svml_d_sincos4_core_avx2 svml_d_sincos8_core_avx512 \
> > > > > - svml_s_cosf4_core_sse4 svml_s_cosf8_core_avx2 \
> > > > > - svml_s_cosf16_core_avx512 svml_s_sinf4_core_sse4 \
> > > > > - svml_s_sinf8_core_avx2 svml_s_sinf16_core_avx512 \
> > > > > - svml_s_logf4_core_sse4 svml_s_logf8_core_avx2 \
> > > > > - svml_s_logf16_core_avx512 svml_d_exp2_core_sse4 \
> > > > > - svml_d_exp4_core_avx2 svml_d_exp8_core_avx512 \
> > > > > - svml_s_expf4_core_sse4 svml_s_expf8_core_avx2 \
> > > > > - svml_s_expf16_core_avx512 svml_d_pow2_core_sse4 \
> > > > > - svml_d_pow4_core_avx2 svml_d_pow8_core_avx512 \
> > > > > - svml_s_powf4_core_sse4 svml_s_powf8_core_avx2 \
> > > > > - svml_s_powf16_core_avx512 svml_s_sincosf4_core_sse4 \
> > > > > - svml_s_sincosf8_core_avx2 \
> > > > > - svml_s_sincosf16_core_avx512 \
> > > > > - svml_d_cos2_core-sse2 svml_d_cos4_core-sse \
> > > > > - svml_d_cos8_core-avx2 svml_d_exp2_core-sse2 \
> > > > > - svml_d_exp4_core-sse svml_d_exp8_core-avx2 \
> > > > > - svml_d_log2_core-sse2 svml_d_log4_core-sse \
> > > > > - svml_d_log8_core-avx2 svml_d_pow2_core-sse2 \
> > > > > - svml_d_pow4_core-sse svml_d_pow8_core-avx2 \
> > > > > - svml_d_sin2_core-sse2 svml_d_sin4_core-sse \
> > > > > - svml_d_sin8_core-avx2 \
> > > > > - svml_d_sincos2_core-sse2 \
> > > > > - svml_d_sincos4_core-sse \
> > > > > - svml_d_sincos8_core-avx2 \
> > > > > - svml_s_cosf16_core-avx2 \
> > > > > - svml_s_cosf4_core-sse2 \
> > > > > - svml_s_cosf8_core-sse \
> > > > > - svml_s_expf16_core-avx2 \
> > > > > - svml_s_expf4_core-sse2 \
> > > > > - svml_s_expf8_core-sse \
> > > > > - svml_s_logf16_core-avx2 \
> > > > > - svml_s_logf4_core-sse2 \
> > > > > - svml_s_logf8_core-sse \
> > > > > - svml_s_powf16_core-avx2 \
> > > > > - svml_s_powf4_core-sse2 \
> > > > > - svml_s_powf8_core-sse \
> > > > > - svml_s_sincosf16_core-avx2 \
> > > > > - svml_s_sincosf4_core-sse2 \
> > > > > - svml_s_sincosf8_core-sse \
> > > > > - svml_s_sinf16_core-avx2 \
> > > > > - svml_s_sinf4_core-sse2 \
> > > > > - svml_s_sinf8_core-sse
> > > > > +libmvec-multiarch-double-func-list = \
> > > > > + 2_core-sse2 \
> > > > > + 2_core_sse4 \
> > > > > + 4_core-sse \
> > > > > + 4_core_avx2 \
> > > > > + 8_core-avx2 \
> > > > > + 8_core_avx512
> > > > > +libmvec-multiarch-float-func-list = \
> > > > > + f4_core-sse2 \
> > > > > + f4_core_sse4 \
> > > > > + f8_core-sse \
> > > > > + f8_core_avx2 \
> > > > > + f16_core-avx2 \
> > > > > + f16_core_avx512
> > > > > +
> > > > > +libmvec-sysdep_routines += \
> > > > > + $(foreach l,$(libmvec-multiarch-double-func-list), \
> > > > > + $(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
> > > > > + $(foreach l,$(libmvec-multiarch-float-func-list), \
> > > > > + $(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
> > > > > endif
> > > > > --
> > > > > 2.31.1
> > > > >
> > > >
> > > > PING.
> > > >
> > > > --
> > > > H.J.
> > >
> > > How can I test this patch?
> >
> > This patch improves libmvec build and test. Otherwise it should be NFC.
> > Sunil has a followup patch to add the rest of libmvec ABI tests.
>
> Err, I meant how can I test this for review.
>
> Build succeeds on x86_64. All tests in mathvec/math pass. As well the
> exact same tests run.
> If that's sufficient then LGTM.
Great. I am checking it with
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Thanks.
--
H.J.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-10-20 18:53 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-10-15 13:12 [PATCH v2] x86-64: Add sysdeps/x86_64/fpu/Makeconfig H.J. Lu
2021-10-20 16:00 ` PING " H.J. Lu
2021-10-20 18:07 ` Noah Goldstein
2021-10-20 18:18 ` H.J. Lu
2021-10-20 18:47 ` Noah Goldstein
2021-10-20 18:53 ` H.J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).