From 8661f70c499e9a6f2d8b12241d663881452178ae Mon Sep 17 00:00:00 2001 From: Felix Willgerodt Date: Fri, 11 Sep 2020 15:43:54 +0200 Subject: [PATCH] Add half support for AVX512 register view. This adds support for the half datatype, FP16, to the AVX512 register printing. gdb/ChangeLog: 2020-07-21 Felix Willgerodt * i386-tdep.c (i386_zmm_type) : New field. (i386_ymm_type) : New field. (i386_gdbarch_init): Add set_gdbarch_half_format. * features/i386/64bit-avx512.xml: Add half type. * features/i386/64bit-avx512.c: Regenerated. * features/i386/64bit-sse.xml: Add half type. * features/i386/64bit-sse.c: Regenerated. gdb/testsuite/ChangeLog: 2021-07-21 Felix Willgerodt * gdb.arch/x86-avx512fp16.c: New file. * gdb.arch/x86-avx512fp16.exp: New file. * lib/gdb.exp (skip_avx512fp16_tests): New function. --- gdb/features/i386/32bit-sse.c | 5 + gdb/features/i386/32bit-sse.xml | 2 + gdb/features/i386/64bit-avx512.c | 5 + gdb/features/i386/64bit-avx512.xml | 2 + gdb/features/i386/64bit-sse.c | 5 + gdb/features/i386/64bit-sse.xml | 2 + gdb/i386-tdep.c | 6 + gdb/testsuite/gdb.arch/x86-avx512fp16.c | 164 ++++++++++++++++++++++ gdb/testsuite/gdb.arch/x86-avx512fp16.exp | 68 +++++++++ gdb/testsuite/lib/gdb.exp | 51 +++++++ 10 files changed, 310 insertions(+) create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16.c create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16.exp diff --git a/gdb/features/i386/32bit-sse.c b/gdb/features/i386/32bit-sse.c index b6d5fbaf0fc..adc7fe06276 100644 --- a/gdb/features/i386/32bit-sse.c +++ b/gdb/features/i386/32bit-sse.c @@ -13,6 +13,9 @@ create_feature_i386_32bit_sse (struct target_desc *result, long regnum) element_type = tdesc_named_type (feature, "bfloat16"); tdesc_create_vector (feature, "v8bf16", element_type, 8); + element_type = tdesc_named_type (feature, "ieee_half"); + tdesc_create_vector (feature, "v8h", element_type, 8); + element_type = tdesc_named_type (feature, "ieee_single"); tdesc_create_vector (feature, "v4f", element_type, 4); @@ -36,6 +39,8 @@ create_feature_i386_32bit_sse (struct target_desc *result, long regnum) tdesc_type *field_type; field_type = tdesc_named_type (feature, "v8bf16"); tdesc_add_field (type_with_fields, "v8_bfloat16", field_type); + field_type = tdesc_named_type (feature, "v8h"); + tdesc_add_field (type_with_fields, "v8_half", field_type); field_type = tdesc_named_type (feature, "v4f"); tdesc_add_field (type_with_fields, "v4_float", field_type); field_type = tdesc_named_type (feature, "v2d"); diff --git a/gdb/features/i386/32bit-sse.xml b/gdb/features/i386/32bit-sse.xml index 8710622e944..f7d1a75e754 100644 --- a/gdb/features/i386/32bit-sse.xml +++ b/gdb/features/i386/32bit-sse.xml @@ -8,6 +8,7 @@ + @@ -16,6 +17,7 @@ + diff --git a/gdb/features/i386/64bit-avx512.c b/gdb/features/i386/64bit-avx512.c index 1bd49dcd9d5..9ceadd89cda 100644 --- a/gdb/features/i386/64bit-avx512.c +++ b/gdb/features/i386/64bit-avx512.c @@ -13,6 +13,9 @@ create_feature_i386_64bit_avx512 (struct target_desc *result, long regnum) element_type = tdesc_named_type (feature, "bfloat16"); tdesc_create_vector (feature, "v8bf16", element_type, 8); + element_type = tdesc_named_type (feature, "ieee_half"); + tdesc_create_vector (feature, "v8h", element_type, 8); + element_type = tdesc_named_type (feature, "ieee_single"); tdesc_create_vector (feature, "v4f", element_type, 4); @@ -36,6 +39,8 @@ create_feature_i386_64bit_avx512 (struct target_desc *result, long regnum) tdesc_type *field_type; field_type = tdesc_named_type (feature, "v8bf16"); tdesc_add_field (type_with_fields, "v8_bfloat16", field_type); + field_type = tdesc_named_type (feature, "v8h"); + tdesc_add_field (type_with_fields, "v8_half", field_type); field_type = tdesc_named_type (feature, "v4f"); tdesc_add_field (type_with_fields, "v4_float", field_type); field_type = tdesc_named_type (feature, "v2d"); diff --git a/gdb/features/i386/64bit-avx512.xml b/gdb/features/i386/64bit-avx512.xml index a4c7f02d6d8..b7fe4d5015b 100644 --- a/gdb/features/i386/64bit-avx512.xml +++ b/gdb/features/i386/64bit-avx512.xml @@ -8,6 +8,7 @@ + @@ -16,6 +17,7 @@ + diff --git a/gdb/features/i386/64bit-sse.c b/gdb/features/i386/64bit-sse.c index 645e3141337..ffb7a45ec1a 100644 --- a/gdb/features/i386/64bit-sse.c +++ b/gdb/features/i386/64bit-sse.c @@ -13,6 +13,9 @@ create_feature_i386_64bit_sse (struct target_desc *result, long regnum) element_type = tdesc_named_type (feature, "bfloat16"); tdesc_create_vector (feature, "v8bf16", element_type, 8); + element_type = tdesc_named_type (feature, "ieee_half"); + tdesc_create_vector (feature, "v8h", element_type, 8); + element_type = tdesc_named_type (feature, "ieee_single"); tdesc_create_vector (feature, "v4f", element_type, 4); @@ -36,6 +39,8 @@ create_feature_i386_64bit_sse (struct target_desc *result, long regnum) tdesc_type *field_type; field_type = tdesc_named_type (feature, "v8bf16"); tdesc_add_field (type_with_fields, "v8_bfloat16", field_type); + field_type = tdesc_named_type (feature, "v8h"); + tdesc_add_field (type_with_fields, "v8_half", field_type); field_type = tdesc_named_type (feature, "v4f"); tdesc_add_field (type_with_fields, "v4_float", field_type); field_type = tdesc_named_type (feature, "v2d"); diff --git a/gdb/features/i386/64bit-sse.xml b/gdb/features/i386/64bit-sse.xml index f23d4d5cb4e..9ae2310af33 100644 --- a/gdb/features/i386/64bit-sse.xml +++ b/gdb/features/i386/64bit-sse.xml @@ -8,6 +8,7 @@ + @@ -16,6 +17,7 @@ + diff --git a/gdb/i386-tdep.c b/gdb/i386-tdep.c index e436437c683..c2835a2458d 100644 --- a/gdb/i386-tdep.c +++ b/gdb/i386-tdep.c @@ -3117,6 +3117,7 @@ i386_zmm_type (struct gdbarch *gdbarch) int8_t v64_int8[64]; double v8_double[8]; float v16_float[16]; + float16_t v32_half[32]; bfloat16_t v32_bfloat16[32]; }; #endif @@ -3127,6 +3128,8 @@ i386_zmm_type (struct gdbarch *gdbarch) "__gdb_builtin_type_vec512i", TYPE_CODE_UNION); append_composite_type_field (t, "v32_bfloat16", init_vector_type (bt->builtin_bfloat16, 32)); + append_composite_type_field (t, "v32_half", + init_vector_type (bt->builtin_half, 32)); append_composite_type_field (t, "v16_float", init_vector_type (bt->builtin_float, 16)); append_composite_type_field (t, "v8_double", @@ -3173,6 +3176,7 @@ i386_ymm_type (struct gdbarch *gdbarch) int8_t v32_int8[32]; double v4_double[4]; float v8_float[8]; + float16_t v16_half[16]; bfloat16_t v16_bfloat16[16]; }; #endif @@ -3183,6 +3187,8 @@ i386_ymm_type (struct gdbarch *gdbarch) "__gdb_builtin_type_vec256i", TYPE_CODE_UNION); append_composite_type_field (t, "v16_bfloat16", init_vector_type (bt->builtin_bfloat16, 16)); + append_composite_type_field (t, "v16_half", + init_vector_type (bt->builtin_half, 16)); append_composite_type_field (t, "v8_float", init_vector_type (bt->builtin_float, 8)); append_composite_type_field (t, "v4_double", diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16.c b/gdb/testsuite/gdb.arch/x86-avx512fp16.c new file mode 100644 index 00000000000..ebc72957bb3 --- /dev/null +++ b/gdb/testsuite/gdb.arch/x86-avx512fp16.c @@ -0,0 +1,164 @@ +/* Test program for bfloat16 of AVX 512 registers. + + Copyright 2021 Free Software Foundation, Inc. + + This file is part of GDB. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + This program 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 General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . */ + +typedef struct +{ + float f[4]; +} v4sd_t; + +typedef struct +{ + float f[8]; +} v8sd_t; + +typedef struct +{ + float f[16]; +} v16sd_t; + +v4sd_t xmm_data[] = +{ + { { 0.0, 0.125, 0.25, 0.375 } }, + { { 0.5, 0.625, 0.75, 0.875 } }, + { { 1.0, 1.125, 1.25, 1.375 } }, + { { 1.5, 1.625, 1.75, 1.875 } }, + { { 2.0, 2.125, 2.25, 2.375 } }, + { { 2.5, 2.625, 2.75, 2.875 } }, + { { 3.0, 3.125, 3.25, 3.375 } }, + { { 3.5, 3.625, 3.75, 3.875 } }, +}; + +v8sd_t ymm_data[] = +{ + { { 8.0, 8.25, 8.5, 8.75, 9.0, 9.25, 9.5, 9.75 } }, + { { 10.0, 10.25, 10.5, 10.75, 11.0, 11.25, 11.5, 11.75 } }, + { { 12.0, 12.25, 12.5, 12.75, 13.0, 13.25, 13.5, 13.75 } }, + { { 14.0, 14.25, 14.5, 14.75, 15.0, 15.25, 15.5, 15.75 } }, + { { 16.0, 16.25, 16.5, 16.75, 17.0, 17.25, 17.5, 17.75 } }, + { { 18.0, 18.25, 18.5, 18.75, 19.0, 19.25, 19.5, 19.75 } }, + { { 20.0, 20.25, 20.5, 20.75, 21.0, 21.25, 21.5, 21.75 } }, + { { 22.0, 22.25, 22.5, 22.75, 23.0, 23.25, 23.5, 23.75 } }, +}; + +v16sd_t zmm_data[] = +{ + { { 20.0, 20.5, 21.0, 21.5, 22.0, 22.5, 23.0, 23.5, 24.0, 24.5, + 25.0, 25.5, 26.0, 26.5, 27.0, 27.5 } }, + { { 28.0, 28.5, 29.0, 29.5, 30.0, 30.5, 31.0, 31.5, 32.0, 32.5, + 33.0, 33.5, 34.0, 34.5, 35.0, 35.5 } }, + { { 36.0, 36.5, 37.0, 37.5, 38.0, 38.5, 39.0, 39.5, 40.0, 40.5, + 41.0, 41.5, 42.0, 42.5, 43.0, 43.5 } }, + { { 44.0, 44.5, 45.0, 45.5, 46.0, 46.5, 47.0, 47.5, 48.0, 48.5, + 49.0, 49.5, 50.0, 50.5, 51.0, 51.5 } }, + { { 52.0, 52.5, 53.0, 53.5, 54.0, 54.5, 55.0, 55.5, 56.0, 56.5, + 57.0, 57.5, 58.0, 58.5, 59.0, 59.5 } }, + { { 60.0, 60.5, 61.0, 61.5, 62.0, 62.5, 63.0, 63.5, 64.0, 64.5, + 65.0, 65.5, 66.0, 66.5, 67.0, 67.5 } }, + { { 68.0, 68.5, 69.0, 69.5, 70.0, 70.5, 71.0, 71.5, 72.0, 72.5, + 73.0, 73.5, 74.0, 74.5, 75.0, 75.5 } }, + { { 76.0, 76.5, 77.0, 77.5, 78.0, 78.5, 79.0, 79.5, 80.0, 80.5, + 81.0, 81.5, 82.0, 82.5, 83.0, 83.5 } }, +}; + +void +move_data_to_xmm_reg (void) +{ + asm ("vmovups 0(%0), %%xmm0 \n\t" + "vmovups 16(%0), %%xmm1 \n\t" + "vmovups 32(%0), %%xmm2 \n\t" + "vmovups 48(%0), %%xmm3 \n\t" + "vmovups 64(%0), %%xmm4 \n\t" + "vmovups 80(%0), %%xmm5 \n\t" + "vmovups 96(%0), %%xmm6 \n\t" + "vmovups 112(%0), %%xmm7 \n\t" + : /* no output operands */ + : "r" (xmm_data)); +} + +void +move_data_to_ymm_reg (void) +{ + asm ("vmovups 0(%0), %%ymm0 \n\t" + "vmovups 32(%0), %%ymm1 \n\t" + "vmovups 64(%0), %%ymm2 \n\t" + "vmovups 96(%0), %%ymm3 \n\t" + "vmovups 128(%0), %%ymm4 \n\t" + "vmovups 160(%0), %%ymm5 \n\t" + "vmovups 192(%0), %%ymm6 \n\t" + "vmovups 224(%0), %%ymm7 \n\t" + : /* no output operands */ + : "r" (ymm_data)); +} + +void +move_data_to_zmm_reg (void) +{ + asm ("vmovups 0(%0), %%zmm0 \n\t" + "vmovups 64(%0), %%zmm1 \n\t" + "vmovups 128(%0), %%zmm2 \n\t" + "vmovups 192(%0), %%zmm3 \n\t" + "vmovups 256(%0), %%zmm4 \n\t" + "vmovups 320(%0), %%zmm5 \n\t" + "vmovups 384(%0), %%zmm6 \n\t" + "vmovups 448(%0), %%zmm7 \n\t" + : /* no output operands */ + : "r" (zmm_data)); +} + +void +convert_xmm_from_float_to_half (void) +{ + asm("vcvtps2phx %xmm1, %xmm0"); + asm("vcvtps2phx %xmm7, %xmm6"); +} + +void +convert_ymm_from_float_to_half (void) +{ + asm("vcvtps2phx %ymm1, %xmm0"); + asm("vcvtps2phx %ymm7, %xmm6"); +} + +void +convert_zmm_from_float_to_half (void) +{ + asm("vcvtps2phx %zmm1, %ymm0"); + asm("vcvtps2phx %zmm7, %ymm6"); +} + +int +main () +{ + /* Move initial values from array to registers and read from XMM regs. */ + move_data_to_xmm_reg (); + convert_xmm_from_float_to_half (); + asm ("nop"); /* first breakpoint here */ + + /* Move initial values from array to registers and read from YMM regs. */ + move_data_to_ymm_reg (); + convert_ymm_from_float_to_half (); + asm ("nop"); /* second breakpoint here */ + + /* Move initial values from array to registers and read from ZMM regs. */ + move_data_to_zmm_reg (); + convert_zmm_from_float_to_half (); + asm ("nop"); /* third breakpoint here */ + + return 0; +} diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16.exp b/gdb/testsuite/gdb.arch/x86-avx512fp16.exp new file mode 100644 index 00000000000..22575cc2fe8 --- /dev/null +++ b/gdb/testsuite/gdb.arch/x86-avx512fp16.exp @@ -0,0 +1,68 @@ +# Copyright 2021 Free Software Foundation, Inc. + +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 3 of the License, or +# (at your option) any later version. +# +# This program 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 General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see . + +# Test fp16 support in AVX512 registers. + +if { [skip_avx512fp16_tests] } { + unsupported "target does not support AVX512fp16" + return -1 +} + +standard_testfile + +if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile}] } { + return -1 +} + +if { ![runto_main] } { + unsupported "could not run to main" + return -1 +} + +# Test xmm. +set line1 [gdb_get_line_number "first breakpoint here"] +gdb_breakpoint $line1 +gdb_continue_to_breakpoint "line1" ".*$srcfile:$line1.*" + +gdb_test "print \$xmm0.v8_half" \ + "= \\{0.5, 0.625, 0.75, 0.875, 0, 0, 0, 0\\}" + +gdb_test "print \$xmm6.v8_half" \ + "= \\{3.5, 3.625, 3.75, 3.875, 0, 0, 0, 0\\}" + +# Test ymm. +set line2 [gdb_get_line_number "second breakpoint here"] +gdb_breakpoint $line2 +gdb_continue_to_breakpoint "line2" ".*$srcfile:$line2.*" + +gdb_test "print \$ymm0.v16_half\[1\]" "= 10.25" +gdb_test "print \$ymm6.v16_half\[1\]" "= 22.25" + +# Test zmm. +set line3 [gdb_get_line_number "third breakpoint here"] +gdb_breakpoint $line3 +gdb_continue_to_breakpoint "line3" ".*$srcfile:$line3.*" + +gdb_test "print \$zmm0.v32_half\[1\]" "= 28.5" +gdb_test "print \$zmm6.v32_half\[1\]" "= 76.5" + +# Test setting of half values. +gdb_test_no_output "set var \$xmm0.v8_half\[0\] = 32.25" +gdb_test_no_output "set var \$ymm3.v16_half\[1\] = 33.5" +gdb_test_no_output "set var \$zmm7.v32_half\[2\] = 22.75" + +gdb_test "p \$xmm0.v8_half\[0\]" "= 32.25" +gdb_test "p \$ymm3.v16_half\[1\]" "= 33.5" +gdb_test "p \$zmm7.v32_half\[2\]" "= 22.75" diff --git a/gdb/testsuite/lib/gdb.exp b/gdb/testsuite/lib/gdb.exp index 3aea7baaab0..ca9864b9007 100644 --- a/gdb/testsuite/lib/gdb.exp +++ b/gdb/testsuite/lib/gdb.exp @@ -3384,6 +3384,57 @@ gdb_caching_proc skip_avx512bf16_tests { return $skip_avx512bf16_tests } +# Run a test on the target to see if it supports avx512fp16. Return 0 if so, +# 1 if it does not. Based on 'check_vmx_hw_available' from the GCC testsuite. + +gdb_caching_proc skip_avx512fp16_tests { + global srcdir subdir gdb_prompt inferior_exited_re + + set me "skip_avx512fp16_tests" + if { ![istarget "i?86-*-*"] && ![istarget "x86_64-*-*"] } { + verbose "$me: target does not support avx512fp16, returning 1" 2 + return 1 + } + + # Compile a test program. + set src { + int main() { + asm volatile ("vcvtps2phx %xmm1, %xmm0"); + return 0; + } + } + if {![gdb_simple_compile $me $src executable]} { + return 1 + } + + # No error message, compilation succeeded so now run it via gdb. + + gdb_exit + gdb_start + gdb_reinitialize_dir $srcdir/$subdir + gdb_load "$obj" + gdb_run_cmd + gdb_expect { + -re ".*Illegal instruction.*${gdb_prompt} $" { + verbose -log "$me: avx512fp16 hardware not detected." + set skip_avx512fp16_tests 1 + } + -re ".*$inferior_exited_re normally.*${gdb_prompt} $" { + verbose -log "$me: avx512fp16 hardware detected." + set skip_avx512fp16_tests 0 + } + default { + warning "\n$me: default case taken." + set skip_avx512fp16_tests 1 + } + } + gdb_exit + remote_file build delete $obj + + verbose "$me: returning $skip_avx512fp16_tests" 2 + return $skip_avx512fp16_tests +} + # Run a test on the target to see if it supports btrace hardware. Return 0 if so, # 1 if it does not. Based on 'check_vmx_hw_available' from the GCC testsuite. -- 2.30.2