If you are still interested in. Here are 3 versions of pack-float. The union version of pack-float should run faster. The code is simpler, the dependencies are easier.
But it may be less accurate or even wrong, as for signed integers (x>>2) and (x/4) are not the same. Consider x = -1. You may try pack_float_good, which gives the same asm as v3, but without warnings. - Mikhail, respectfully On Thu, Sep 08, 2016 at 01:29:36PM +0500, Andrew Borodin wrote: > >autoconf check for IEEE 754 floats > Autoconf man says folowing: > >it is safe to assume IEEE-754 in most portable code these days > https://www.gnu.org/software/autoconf/manual/autoconf.html#Floating-Point-Portability > > > A union might be more readable > Here is union version of the patch. It's slower 10% than original cube > and dereference version. Have no idea why. > Select performance is improved as in v3. >
#include <stdint.h> typedef union { float fp; int i; } U; float pack_float(const float v, const int r) { const U a = { .fp = v }; const U b = { .i = (a.i >> 2) + r * (INT32_MAX / 4) }; return b.fp; } float pack_float_av(float v, int r) { U buf; buf.fp = v; buf.i = (buf.i >> 2) + (INT32_MAX / 4) * r; return buf.fp; } float pack_float_v3(float actualValue, int realm) { /* two bits for realm, others for value */ /* we have 4 realms */ int realmAjustment = *((int*)&actualValue)/4; int realCode = realm * (INT32_MAX/4) + realmAjustment; return *((float*)&realCode); } float pack_float_good(const float v, const int r) { const U a = { .fp = v }; const U b = { .i = a.i/4 + r * (INT32_MAX / 4) }; return b.fp; }
.file "pack-float.c" .text .p2align 4,,15 .globl pack_float .type pack_float, @function pack_float: .LFB0: .cfi_startproc movd %xmm0, %eax movl %edi, %edx sall $29, %edx sarl $2, %eax subl %edi, %edx addl %edx, %eax movl %eax, -4(%rsp) movss -4(%rsp), %xmm0 ret .cfi_endproc .LFE0: .size pack_float, .-pack_float .p2align 4,,15 .globl pack_float_av .type pack_float_av, @function pack_float_av: .LFB1: .cfi_startproc movd %xmm0, %eax movl %edi, %edx sall $29, %edx sarl $2, %eax subl %edi, %edx addl %edx, %eax movl %eax, -4(%rsp) movss -4(%rsp), %xmm0 ret .cfi_endproc .LFE1: .size pack_float_av, .-pack_float_av .p2align 4,,15 .globl pack_float_v3 .type pack_float_v3, @function pack_float_v3: .LFB2: .cfi_startproc movd %xmm0, %edx leal 3(%rdx), %eax testl %edx, %edx cmovns %edx, %eax sarl $2, %eax movl %eax, %edx movl %edi, %eax sall $29, %eax subl %edi, %eax addl %edx, %eax movl %eax, -4(%rsp) movss -4(%rsp), %xmm0 ret .cfi_endproc .LFE2: .size pack_float_v3, .-pack_float_v3 .p2align 4,,15 .globl pack_float_good .type pack_float_good, @function pack_float_good: .LFB3: .cfi_startproc movd %xmm0, %edx leal 3(%rdx), %eax testl %edx, %edx cmovns %edx, %eax sarl $2, %eax movl %eax, %edx movl %edi, %eax sall $29, %eax subl %edi, %eax addl %edx, %eax movl %eax, -4(%rsp) movss -4(%rsp), %xmm0 ret .cfi_endproc .LFE3: .size pack_float_good, .-pack_float_good .ident "GCC: (GNU) 6.1.1 20160802" .section .note.GNU-stack,"",@progbits
signature.asc
Description: PGP signature