packages/mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
mesa/1000-add-loongarch-support-not-upstream-modified.patch
herengui 8b4ed147c6 添加龙芯架构支持 
Signed-off-by: herengui <herengui@kylinsec.com.cn>
(cherry picked from commit 7538a562243a088b99a889fad48542ecd196ae49)
2023-08-30 15:48:23 +08:00

483 lines
18 KiB
Diff
Raw Blame History

From b7ce39c71900c356b4161b5ef4d6c8a5ffa12732 Mon Sep 17 00:00:00 2001
From: herengui <herengui@kylinsec.com.cn>
Date: Tue, 29 Aug 2023 11:12:32 +0800
Subject: [PATCH 1000/1001] add loongarch support not upstream modified
Signed-off-by: herengui <herengui@kylinsec.com.cn>
---
meson.build | 14 ++++++
src/gallium/auxiliary/gallivm/lp_bld.h | 4 ++
src/gallium/auxiliary/gallivm/lp_bld_arit.c | 17 ++++++-
src/gallium/auxiliary/gallivm/lp_bld_debug.h | 8 ++++
src/gallium/auxiliary/gallivm/lp_bld_init.c | 34 ++++++++++++--
src/gallium/auxiliary/gallivm/lp_bld_misc.cpp | 44 ++++++++++++++++++-
src/gallium/auxiliary/gallivm/lp_bld_misc.h | 3 ++
src/gallium/drivers/llvmpipe/lp_screen.c | 4 +-
src/gallium/drivers/llvmpipe/lp_test_arit.c | 2 +-
src/gallium/include/pipe/p_config.h | 6 +++
src/gallium/targets/dri/meson.build | 1 +
src/gallium/targets/dri/target.c | 1 +
src/loader/pci_id_driver_map.h | 7 +++
src/util/u_cpu_detect.c | 30 +++++++++++++
src/util/u_cpu_detect.h | 2 +
15 files changed, 167 insertions(+), 10 deletions(-)
diff --git a/meson.build b/meson.build
index bd54e78..1a00889 100644
--- a/meson.build
+++ b/meson.build
@@ -176,6 +176,8 @@ if dri_drivers.contains('auto')
dri_drivers = []
elif ['mips', 'mips64', 'riscv32', 'riscv64'].contains(host_machine.cpu_family())
dri_drivers = ['r100', 'r200', 'nouveau']
+ elif ['loongarch64'].contains(host_machine.cpu_family())
+ dri_drivers = []
else
error('Unknown architecture @0@. Please pass -Ddri-drivers to set driver options. Patches gladly accepted to fix this.'.format(
host_machine.cpu_family()))
@@ -215,6 +217,11 @@ if gallium_drivers.contains('auto')
gallium_drivers = [
'r300', 'r600', 'radeonsi', 'nouveau', 'virgl', 'swrast'
]
+ elif ['loongarch64'].contains(host_machine.cpu_family())
+ gallium_drivers = [
+ 'r300', 'r600', 'radeonsi', 'nouveau', 'etnaviv', 'kmsro', 'swrast',
+ 'virgl'
+ ]
else
error('Unknown architecture @0@. Please pass -Dgallium-drivers to set driver options. Patches gladly accepted to fix this.'.format(
host_machine.cpu_family()))
@@ -268,6 +275,8 @@ if _vulkan_drivers.contains('auto')
_vulkan_drivers = ['swrast']
elif ['mips', 'mips64', 'riscv32', 'riscv64'].contains(host_machine.cpu_family())
_vulkan_drivers = ['amd', 'swrast']
+ elif ['loongarch64'].contains(host_machine.cpu_family())
+ _vulkan_drivers = ['amd']
else
error('Unknown architecture @0@. Please pass -Dvulkan-drivers to set driver options. Patches gladly accepted to fix this.'.format(
host_machine.cpu_family()))
@@ -1349,6 +1358,11 @@ elif host_machine.cpu_family() == 'mips64' and host_machine.endian() == 'little'
with_asm_arch = 'mips64el'
pre_args += ['-DUSE_MIPS64EL_ASM']
endif
+elif host_machine.cpu_family() == 'loongarch64'
+ if system_has_kms_drm
+ with_asm_arch = 'loongarch64'
+ pre_args += ['-DUSE_LOONGARCH64_ASM']
+ endif
endif
# Check for standard headers and functions
diff --git a/src/gallium/auxiliary/gallivm/lp_bld.h b/src/gallium/auxiliary/gallivm/lp_bld.h
index 9144428..2fd50dd 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld.h
@@ -82,7 +82,11 @@
#define LLVMCreateBuilder ILLEGAL_LLVM_FUNCTION
#if LLVM_VERSION_MAJOR >= 8
+#if defined(__loongarch__) || defined(__mips__)
+#define GALLIVM_HAVE_CORO 0
+#else
#define GALLIVM_HAVE_CORO 1
+#endif
#else
#define GALLIVM_HAVE_CORO 0
#endif
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index 1c71c05..99abcae 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -1887,6 +1887,15 @@ arch_rounding_available(const struct lp_type type)
return TRUE;
else if (util_get_cpu_caps()->has_neon)
return TRUE;
+ else if (util_get_cpu_caps()->has_msa &&
+ (type.width * type.length == 128))
+ return ((gallivm_perf & GALLIVM_PERF_USE_ARCH_ROUNDING) ? TRUE : FALSE);
+ else if (util_get_cpu_caps()->has_lsx &&
+ (type.width * type.length == 128))
+ return ((gallivm_perf & GALLIVM_PERF_USE_ARCH_ROUNDING) ? TRUE : FALSE);
+ else if (util_get_cpu_caps()->has_lasx &&
+ (type.width * type.length == 256))
+ return ((gallivm_perf & GALLIVM_PERF_USE_ARCH_ROUNDING) ? TRUE : FALSE);
return FALSE;
}
@@ -1994,7 +2003,8 @@ lp_build_round_arch(struct lp_build_context *bld,
LLVMValueRef a,
enum lp_build_round_mode mode)
{
- if (util_get_cpu_caps()->has_sse4_1 || util_get_cpu_caps()->has_neon) {
+ if (util_get_cpu_caps()->has_sse4_1 || util_get_cpu_caps()->has_neon || util_get_cpu_caps()->has_msa
+ || util_get_cpu_caps()->has_lsx || util_get_cpu_caps()->has_lasx) {
LLVMBuilderRef builder = bld->gallivm->builder;
const struct lp_type type = bld->type;
const char *intrinsic_root;
@@ -2403,7 +2413,10 @@ lp_build_iround(struct lp_build_context *bld,
res = LLVMBuildFAdd(builder, a, half, "");
}
- res = LLVMBuildFPToSI(builder, res, int_vec_type, "");
+ if (type.sign)
+ res = LLVMBuildFPToSI(builder, res, int_vec_type, "");
+ else
+ res = LLVMBuildFPToUI(builder, res, int_vec_type, "");
return res;
}
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_debug.h b/src/gallium/auxiliary/gallivm/lp_bld_debug.h
index a5dd7b8..6c22dc2 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_debug.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_debug.h
@@ -49,6 +49,14 @@
#define GALLIVM_PERF_NO_OPT (1 << 3)
#define GALLIVM_PERF_NO_AOS_SAMPLING (1 << 4)
+#if defined(PIPE_ARCH_MIPS64) || defined(PIPE_ARCH_LOONGARCH64)
+#define GALLIVM_PERF_OPT_O1 (1 << 5)
+#define GALLIVM_PERF_OPT_O2 (1 << 6)
+#define GALLIVM_PERF_OPT_O3 (1 << 7)
+#endif
+
+#define GALLIVM_PERF_USE_ARCH_ROUNDING (1 << 8)
+
#ifdef __cplusplus
extern "C" {
#endif
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_init.c b/src/gallium/auxiliary/gallivm/lp_bld_init.c
index 3f040ac..64740ec 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_init.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_init.c
@@ -46,7 +46,7 @@
#endif
#include <llvm-c/BitWriter.h>
#if GALLIVM_HAVE_CORO
-#if LLVM_VERSION_MAJOR <= 8 && (defined(PIPE_ARCH_AARCH64) || defined (PIPE_ARCH_ARM) || defined(PIPE_ARCH_S390) || defined(PIPE_ARCH_MIPS64))
+#if LLVM_VERSION_MAJOR <= 8 && (defined(PIPE_ARCH_AARCH64) || defined (PIPE_ARCH_ARM) || defined(PIPE_ARCH_S390) || defined(PIPE_ARCH_MIPS64) || defined(PIPE_ARCH_LOONGARCH64))
#include <llvm-c/Transforms/IPO.h>
#endif
#include <llvm-c/Transforms/Coroutines.h>
@@ -60,6 +60,12 @@ static const struct debug_named_value lp_bld_perf_flags[] = {
{ "no_quad_lod", GALLIVM_PERF_NO_QUAD_LOD, "disable quad_lod optimization" },
{ "no_aos_sampling", GALLIVM_PERF_NO_AOS_SAMPLING, "disable aos sampling optimization" },
{ "nopt", GALLIVM_PERF_NO_OPT, "disable optimization passes to speed up shader compilation" },
+#if defined(PIPE_ARCH_MIPS64) || defined(PIPE_ARCH_LOONGARCH64)
+ { "o3", GALLIVM_PERF_OPT_O3, "enable aggressive optimization passes" },
+ { "o2", GALLIVM_PERF_OPT_O2, "enable medium optimization passes" },
+ { "o1", GALLIVM_PERF_OPT_O1, "enable less optimization passes" },
+#endif
+ { "use_arch_rounding", GALLIVM_PERF_USE_ARCH_ROUNDING, "use poor arch rounding function provided by glibc" },
DEBUG_NAMED_VALUE_END
};
@@ -135,7 +141,7 @@ create_pass_manager(struct gallivm_state *gallivm)
}
#if GALLIVM_HAVE_CORO
-#if LLVM_VERSION_MAJOR <= 8 && (defined(PIPE_ARCH_AARCH64) || defined (PIPE_ARCH_ARM) || defined(PIPE_ARCH_S390) || defined(PIPE_ARCH_MIPS64))
+#if LLVM_VERSION_MAJOR <= 8 && (defined(PIPE_ARCH_AARCH64) || defined (PIPE_ARCH_ARM) || defined(PIPE_ARCH_S390) || defined(PIPE_ARCH_MIPS64) || defined(PIPE_ARCH_LOONGARCH64))
LLVMAddArgumentPromotionPass(gallivm->cgpassmgr);
LLVMAddFunctionAttrsPass(gallivm->cgpassmgr);
#endif
@@ -267,7 +273,16 @@ init_gallivm_engine(struct gallivm_state *gallivm)
optlevel = None;
}
else {
- optlevel = Default;
+#if defined(PIPE_ARCH_MIPS64) || defined(PIPE_ARCH_LOONGARCH64)
+ if (gallivm_perf & GALLIVM_PERF_OPT_O3)
+ optlevel = Aggressive;
+ else if (gallivm_perf & GALLIVM_PERF_OPT_O2)
+ optlevel = Default;
+ else if (gallivm_perf & GALLIVM_PERF_OPT_O1)
+ optlevel = Less;
+ else
+#endif
+ optlevel = Default;
}
ret = lp_build_create_jit_compiler_for_module(&gallivm->engine,
@@ -450,7 +465,14 @@ lp_build_init(void)
if (util_get_cpu_caps()->has_avx2 || util_get_cpu_caps()->has_avx) {
lp_native_vector_width = 256;
- } else {
+ } else if (util_get_cpu_caps()->has_lasx) {
+#if defined(PIPE_ARCH_LOONGARCH64)
+ if (lp_probe_lasx())
+ lp_native_vector_width = 256;
+ else
+#endif
+ lp_native_vector_width = 128;
+ } else {
/* Leave it at 128, even when no SIMD extensions are available.
* Really needs to be a multiple of 128 so can fit 4 floats.
*/
@@ -622,6 +644,10 @@ gallivm_compile_module(struct gallivm_state *gallivm)
LLVMAddTargetDependentFunctionAttr(func, "no-frame-pointer-elim-non-leaf", "true");
#endif
+#if defined(PIPE_ARCH_MIPS64)
+ LLVMAddTargetDependentFunctionAttr(func, "target-features", "+nomadd4");
+#endif
+
LLVMRunFunctionPassManager(gallivm->passmgr, func);
func = LLVMGetNextFunction(func);
}
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
index be288ab..675de06 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
+++ b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
@@ -376,7 +376,7 @@ lp_build_create_jit_compiler_for_module(LLVMExecutionEngineRef *OutJIT,
llvm::SmallVector<std::string, 16> MAttrs;
-#if LLVM_VERSION_MAJOR >= 4 && (defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64) || defined(PIPE_ARCH_ARM))
+#if LLVM_VERSION_MAJOR >= 4 && (defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64) || defined(PIPE_ARCH_ARM) || defined(PIPE_ARCH_LOONGARCH64))
/* llvm-3.3+ implements sys::getHostCPUFeatures for Arm
* and llvm-3.7+ for x86, which allows us to enable/disable
* code generation based on the results of cpuid on these
@@ -470,6 +470,17 @@ lp_build_create_jit_compiler_for_module(LLVMExecutionEngineRef *OutJIT,
MAttrs.push_back("+fp64");
#endif
+#if defined(PIPE_ARCH_MIPS64)
+ /* MSA requires a 64-bit FPU register file */
+ MAttrs.push_back(util_cpu_caps.has_msa ? "+msa" : "-msa");
+ /* Support 64-bit FP registers. */
+ MAttrs.push_back("+fp64");
+ /* General Purpose Registers are 64-bit wide */
+ MAttrs.push_back("+gp64");
+ /* Pointers are 64-bit wide */
+ MAttrs.push_back("+ptr64");
+#endif
+
builder.setMAttrs(MAttrs);
if (gallivm_debug & (GALLIVM_DEBUG_IR | GALLIVM_DEBUG_ASM | GALLIVM_DEBUG_DUMP_BC)) {
@@ -533,6 +544,13 @@ lp_build_create_jit_compiler_for_module(LLVMExecutionEngineRef *OutJIT,
*/
if (MCPU == "generic")
MCPU = util_get_cpu_caps()->has_msa ? "mips64r5" : "mips64r2";
+ else if (MCPU == "loongson3a")
+ MCPU = util_get_cpu_caps()->has_msa ? "mips64r5" : "mips64r2";
+#endif
+
+#if defined(PIPE_ARCH_LOONGARCH64)
+ if (MCPU == "generic")
+ MCPU = "gs464v";
#endif
builder.setMCPU(MCPU);
@@ -573,6 +591,30 @@ lp_build_create_jit_compiler_for_module(LLVMExecutionEngineRef *OutJIT,
return 1;
}
+#if defined(PIPE_ARCH_LOONGARCH64)
+extern "C"
+LLVMBool
+lp_probe_lasx(void)
+{
+ using namespace llvm;
+ /* our llvm-8+ implements sys::getHostCPUFeatures for loongarch,
+ * which allows us to enable/disable code generation based
+ * on the results of cpucfg.
+ */
+ llvm::StringMap<bool> features;
+ llvm::sys::getHostCPUFeatures(features);
+
+ for (StringMapIterator<bool> f = features.begin();
+ f != features.end();
+ ++f) {
+
+ if ((*f).first() == "lasx" && (*f).second)
+ return 1;
+ }
+
+ return 0;
+}
+#endif
extern "C"
void
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_misc.h b/src/gallium/auxiliary/gallivm/lp_bld_misc.h
index fa0ce90..f9eb530 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_misc.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_misc.h
@@ -64,6 +64,9 @@ gallivm_dispose_target_library_info(LLVMTargetLibraryInfoRef library_info);
extern void
lp_set_target_options(void);
+#if defined(PIPE_ARCH_LOONGARCH64)
+extern LLVMBool lp_probe_lasx(void);
+#endif
extern int
lp_build_create_jit_compiler_for_module(LLVMExecutionEngineRef *OutJIT,
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c
index 839902b..c1ce1cf 100644
--- a/src/gallium/drivers/llvmpipe/lp_screen.c
+++ b/src/gallium/drivers/llvmpipe/lp_screen.c
@@ -897,8 +897,8 @@ static void update_cache_sha1_cpu(struct mesa_sha1 *ctx)
* Don't need the cpu cache affinity stuff. The rest
* is contained in first 5 dwords.
*/
- STATIC_ASSERT(offsetof(struct util_cpu_caps_t, num_L3_caches) == 5 * sizeof(uint32_t));
- _mesa_sha1_update(ctx, cpu_caps, 5 * sizeof(uint32_t));
+ STATIC_ASSERT(offsetof(struct util_cpu_caps_t, num_L3_caches) == 6 * sizeof(uint32_t));
+ _mesa_sha1_update(ctx, cpu_caps, 6 * sizeof(uint32_t));
}
static void lp_disk_cache_create(struct llvmpipe_screen *screen)
diff --git a/src/gallium/drivers/llvmpipe/lp_test_arit.c b/src/gallium/drivers/llvmpipe/lp_test_arit.c
index cbea1e2..5d64132 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_arit.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_arit.c
@@ -479,7 +479,7 @@ test_unary(unsigned verbose, FILE *fp, const struct unary_test_t *test, unsigned
continue;
}
- if (!util_get_cpu_caps()->has_neon &&
+ if (!util_get_cpu_caps()->has_neon && !util_get_cpu_caps()->has_msa &&
test->ref == &nearbyintf && length == 2 &&
ref != roundf(testval)) {
/* FIXME: The generic (non SSE) path in lp_build_iround, which is
diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h
index 978aa45..cc55351 100644
--- a/src/gallium/include/pipe/p_config.h
+++ b/src/gallium/include/pipe/p_config.h
@@ -130,6 +130,12 @@
#define PIPE_ARCH_MIPS
#endif
+#if defined(__loongarch64__) || defined(__loongarch64)
+#define PIPE_ARCH_LOONGARCH64
+#elif defined(__loongarch__)
+#define PIPE_ARCH_LOONGARCH
+#endif
+
/*
* Endian detection.
*/
diff --git a/src/gallium/targets/dri/meson.build b/src/gallium/targets/dri/meson.build
index 86f66a6..310d839 100644
--- a/src/gallium/targets/dri/meson.build
+++ b/src/gallium/targets/dri/meson.build
@@ -77,6 +77,7 @@ foreach d : [[with_gallium_kmsro, [
'ingenic-drm_dri.so',
'kirin_dri.so',
'mali-dp_dri.so',
+ 'loongson-drm_dri.so',
'mcde_dri.so',
'mediatek_dri.so',
'meson_dri.so',
diff --git a/src/gallium/targets/dri/target.c b/src/gallium/targets/dri/target.c
index 30c9ee9..aa9f768 100644
--- a/src/gallium/targets/dri/target.c
+++ b/src/gallium/targets/dri/target.c
@@ -106,6 +106,7 @@ DEFINE_LOADER_DRM_ENTRYPOINT(imx_dcss)
DEFINE_LOADER_DRM_ENTRYPOINT(ingenic_drm)
DEFINE_LOADER_DRM_ENTRYPOINT(kirin)
DEFINE_LOADER_DRM_ENTRYPOINT(mali_dp)
+DEFINE_LOADER_DRM_ENTRYPOINT(loongson_drm)
DEFINE_LOADER_DRM_ENTRYPOINT(mcde)
DEFINE_LOADER_DRM_ENTRYPOINT(mediatek)
DEFINE_LOADER_DRM_ENTRYPOINT(meson)
diff --git a/src/loader/pci_id_driver_map.h b/src/loader/pci_id_driver_map.h
index d75b2cc..232e1b7 100644
--- a/src/loader/pci_id_driver_map.h
+++ b/src/loader/pci_id_driver_map.h
@@ -68,6 +68,12 @@ static const int vmwgfx_chip_ids[] = {
#undef CHIPSET
};
+static const int ls7a_chip_ids[] = {
+#define CHIPSET(chip, name, family) chip,
+#include "pci_ids/ls7a1000_pci_ids.h"
+#undef CHIPSET
+};
+
bool is_nouveau_vieux(int fd);
bool is_kernel_i915(int fd);
@@ -90,6 +96,7 @@ static const struct {
{ 0x1002, "radeonsi", NULL, -1 },
{ 0x10de, "nouveau_vieux", NULL, -1, is_nouveau_vieux },
{ 0x10de, "nouveau", NULL, -1, },
+ { 0x0014, "loongson-drm", ls7a_chip_ids, ARRAY_SIZE(ls7a_chip_ids) },
{ 0x1af4, "virtio_gpu", virtio_gpu_chip_ids, ARRAY_SIZE(virtio_gpu_chip_ids) },
{ 0x15ad, "vmwgfx", vmwgfx_chip_ids, ARRAY_SIZE(vmwgfx_chip_ids) },
};
diff --git a/src/util/u_cpu_detect.c b/src/util/u_cpu_detect.c
index 955d087..43675d2 100644
--- a/src/util/u_cpu_detect.c
+++ b/src/util/u_cpu_detect.c
@@ -456,6 +456,30 @@ check_os_mips64_support(void)
}
#endif /* PIPE_ARCH_MIPS64 */
+#if defined(PIPE_ARCH_LOONGARCH64)
+static void
+check_os_loongarch64_support(void)
+{
+#if defined(PIPE_OS_LINUX)
+ Elf64_auxv_t aux;
+ int fd;
+
+ fd = open("/proc/self/auxv", O_RDONLY | O_CLOEXEC);
+ if (fd >= 0) {
+ while (read(fd, &aux, sizeof(Elf64_auxv_t)) == sizeof(Elf64_auxv_t)) {
+ if (aux.a_type == AT_HWCAP) {
+ uint64_t hwcap = aux.a_un.a_val;
+
+ util_cpu_caps.has_lsx = (hwcap >> 2) & 1;
+ util_cpu_caps.has_lasx = (hwcap >> 3) & 1;
+ break;
+ }
+ }
+ close (fd);
+ }
+#endif /* PIPE_OS_LINUX */
+}
+#endif
static void
get_cpu_topology(void)
@@ -813,6 +837,10 @@ util_cpu_detect_once(void)
check_os_mips64_support();
#endif /* PIPE_ARCH_MIPS64 */
+#if defined(PIPE_ARCH_LOONGARCH64)
+ check_os_loongarch64_support();
+#endif
+
get_cpu_topology();
if (debug_get_option_dump_cpu()) {
@@ -842,6 +870,8 @@ util_cpu_detect_once(void)
printf("util_cpu_caps.has_neon = %u\n", util_cpu_caps.has_neon);
printf("util_cpu_caps.has_msa = %u\n", util_cpu_caps.has_msa);
printf("util_cpu_caps.has_daz = %u\n", util_cpu_caps.has_daz);
+ printf("util_cpu_caps.has_lsx = %u\n", util_cpu_caps.has_lsx);
+ printf("util_cpu_caps.has_lasx = %u\n", util_cpu_caps.has_lasx);
printf("util_cpu_caps.has_avx512f = %u\n", util_cpu_caps.has_avx512f);
printf("util_cpu_caps.has_avx512dq = %u\n", util_cpu_caps.has_avx512dq);
printf("util_cpu_caps.has_avx512ifma = %u\n", util_cpu_caps.has_avx512ifma);
diff --git a/src/util/u_cpu_detect.h b/src/util/u_cpu_detect.h
index 59dd230..cd4319e 100644
--- a/src/util/u_cpu_detect.h
+++ b/src/util/u_cpu_detect.h
@@ -103,6 +103,8 @@ struct util_cpu_caps_t {
unsigned has_daz:1;
unsigned has_neon:1;
unsigned has_msa:1;
+ unsigned has_lsx:1;
+ unsigned has_lasx:1;
unsigned has_avx512f:1;
unsigned has_avx512dq:1;
--
2.41.0
Reference in New Issue View Git Blame Copy Permalink