diff options
Diffstat (limited to 'sysdeps/x86_64/dl-trampoline.S')
| -rw-r--r-- | sysdeps/x86_64/dl-trampoline.S | 105 |
1 files changed, 94 insertions, 11 deletions
diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S index 49d239f075..20da6956f1 100644 --- a/sysdeps/x86_64/dl-trampoline.S +++ b/sysdeps/x86_64/dl-trampoline.S @@ -61,6 +61,7 @@ _dl_runtime_resolve: cfi_startproc _dl_runtime_profile: + cfi_adjust_cfa_offset(16) # Incorporate PLT /* The La_x86_64_regs data structure pointed to by the fourth paramater must be 16-byte aligned. This must be explicitly enforced. We have the set up a dynamically @@ -68,7 +69,7 @@ _dl_runtime_profile: has a fixed size and preserves the original stack pointer. */ subq $32, %rsp # Allocate the local storage. - cfi_adjust_cfa_offset(48) # Incorporate PLT + cfi_adjust_cfa_offset(32) movq %rbx, (%rsp) cfi_rel_offset(%rbx, 0) @@ -203,49 +204,49 @@ L(no_avx1): vpcmpeqq (LR_SIZE)(%rsp), %xmm0, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET)(%rsp), %ymm0 1: vpcmpeqq (LR_SIZE + XMM_SIZE)(%rsp), %xmm1, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE)(%rsp), %ymm1 1: vpcmpeqq (LR_SIZE + XMM_SIZE*2)(%rsp), %xmm2, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*2)(%rsp), %ymm2 1: vpcmpeqq (LR_SIZE + XMM_SIZE*3)(%rsp), %xmm3, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*3)(%rsp), %ymm3 1: vpcmpeqq (LR_SIZE + XMM_SIZE*4)(%rsp), %xmm4, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*4)(%rsp), %ymm4 1: vpcmpeqq (LR_SIZE + XMM_SIZE*5)(%rsp), %xmm5, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*5)(%rsp), %ymm5 1: vpcmpeqq (LR_SIZE + XMM_SIZE*6)(%rsp), %xmm6, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*6)(%rsp), %ymm6 1: vpcmpeqq (LR_SIZE + XMM_SIZE*7)(%rsp), %xmm7, %xmm8 vpmovmskb %xmm8, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu (LR_VECTOR_OFFSET + VECTOR_SIZE*7)(%rsp), %ymm7 L(no_avx2): @@ -361,13 +362,13 @@ L(no_avx3): vpcmpeqq (LRV_SIZE)(%rsp), %xmm0, %xmm2 vpmovmskb %xmm2, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu LRV_VECTOR0_OFFSET(%rsp), %ymm0 1: vpcmpeqq (LRV_SIZE + XMM_SIZE)(%rsp), %xmm1, %xmm2 vpmovmskb %xmm2, %esi cmpl $0xffff, %esi - je 1f + jne 1f vmovdqu LRV_VECTOR1_OFFSET(%rsp), %ymm1 L(no_avx4): @@ -390,3 +391,85 @@ L(no_avx4): cfi_endproc .size _dl_runtime_profile, .-_dl_runtime_profile #endif + + +#ifdef SHARED + .globl _dl_x86_64_save_sse + .type _dl_x86_64_save_sse, @function + .align 16 + cfi_startproc +_dl_x86_64_save_sse: +# ifdef HAVE_AVX_SUPPORT + cmpl $0, L(have_avx)(%rip) + jne 1f + movq %rbx, %r11 # Save rbx + movl $1, %eax + cpuid + movq %r11,%rbx # Restore rbx + movl $1, %eax + testl $(1 << 28), %ecx + jne 2f + negl %eax +2: movl %eax, L(have_avx)(%rip) + cmpl $0, %eax + +1: js L(no_avx5) + +# define YMM_SIZE 32 + vmovdqa %ymm0, %fs:RTLD_SAVESPACE_SSE+0*YMM_SIZE + vmovdqa %ymm1, %fs:RTLD_SAVESPACE_SSE+1*YMM_SIZE + vmovdqa %ymm2, %fs:RTLD_SAVESPACE_SSE+2*YMM_SIZE + vmovdqa %ymm3, %fs:RTLD_SAVESPACE_SSE+3*YMM_SIZE + vmovdqa %ymm4, %fs:RTLD_SAVESPACE_SSE+4*YMM_SIZE + vmovdqa %ymm5, %fs:RTLD_SAVESPACE_SSE+5*YMM_SIZE + vmovdqa %ymm6, %fs:RTLD_SAVESPACE_SSE+6*YMM_SIZE + vmovdqa %ymm7, %fs:RTLD_SAVESPACE_SSE+7*YMM_SIZE + ret +L(no_avx5): +# endif +# define YMM_SIZE 16 + movdqa %xmm0, %fs:RTLD_SAVESPACE_SSE+0*XMM_SIZE + movdqa %xmm1, %fs:RTLD_SAVESPACE_SSE+1*XMM_SIZE + movdqa %xmm2, %fs:RTLD_SAVESPACE_SSE+2*XMM_SIZE + movdqa %xmm3, %fs:RTLD_SAVESPACE_SSE+3*XMM_SIZE + movdqa %xmm4, %fs:RTLD_SAVESPACE_SSE+4*XMM_SIZE + movdqa %xmm5, %fs:RTLD_SAVESPACE_SSE+5*XMM_SIZE + movdqa %xmm6, %fs:RTLD_SAVESPACE_SSE+6*XMM_SIZE + movdqa %xmm7, %fs:RTLD_SAVESPACE_SSE+7*XMM_SIZE + ret + cfi_endproc + .size _dl_x86_64_save_sse, .-_dl_x86_64_save_sse + + + .globl _dl_x86_64_restore_sse + .type _dl_x86_64_restore_sse, @function + .align 16 + cfi_startproc +_dl_x86_64_restore_sse: +# ifdef HAVE_AVX_SUPPORT + cmpl $0, L(have_avx)(%rip) + js L(no_avx6) + + vmovdqa %fs:RTLD_SAVESPACE_SSE+0*YMM_SIZE, %ymm0 + vmovdqa %fs:RTLD_SAVESPACE_SSE+1*YMM_SIZE, %ymm1 + vmovdqa %fs:RTLD_SAVESPACE_SSE+2*YMM_SIZE, %ymm2 + vmovdqa %fs:RTLD_SAVESPACE_SSE+3*YMM_SIZE, %ymm3 + vmovdqa %fs:RTLD_SAVESPACE_SSE+4*YMM_SIZE, %ymm4 + vmovdqa %fs:RTLD_SAVESPACE_SSE+5*YMM_SIZE, %ymm5 + vmovdqa %fs:RTLD_SAVESPACE_SSE+6*YMM_SIZE, %ymm6 + vmovdqa %fs:RTLD_SAVESPACE_SSE+7*YMM_SIZE, %ymm7 + ret +L(no_avx6): +# endif + movdqa %fs:RTLD_SAVESPACE_SSE+0*XMM_SIZE, %xmm0 + movdqa %fs:RTLD_SAVESPACE_SSE+1*XMM_SIZE, %xmm1 + movdqa %fs:RTLD_SAVESPACE_SSE+2*XMM_SIZE, %xmm2 + movdqa %fs:RTLD_SAVESPACE_SSE+3*XMM_SIZE, %xmm3 + movdqa %fs:RTLD_SAVESPACE_SSE+4*XMM_SIZE, %xmm4 + movdqa %fs:RTLD_SAVESPACE_SSE+5*XMM_SIZE, %xmm5 + movdqa %fs:RTLD_SAVESPACE_SSE+6*XMM_SIZE, %xmm6 + movdqa %fs:RTLD_SAVESPACE_SSE+7*XMM_SIZE, %xmm7 + ret + cfi_endproc + .size _dl_x86_64_restore_sse, .-_dl_x86_64_restore_sse +#endif |