From: Vladislav Shchapov <vladislav@shchapov.ru>
Date: Fri, 19 Jan 2024 18:42:19 +0000 (+0500)
Subject: Remove always true arch conditions.
X-Git-Tag: 2.2.0~119
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=0b856b7351791bf7ea175c96bb89039cdc6c9910;p=thirdparty%2Fzlib-ng.git

Remove always true arch conditions.

Signed-off-by: Vladislav Shchapov <vladislav@shchapov.ru>
---

diff --git a/arch/x86/adler32_avx2.c b/arch/x86/adler32_avx2.c
index 5b0d32d7..38e7f068 100644
--- a/arch/x86/adler32_avx2.c
+++ b/arch/x86/adler32_avx2.c
@@ -15,17 +15,9 @@
 #include "adler32_avx2_p.h"
 #include "x86_intrins.h"
 
-#ifdef X86_SSE42
 extern uint32_t adler32_fold_copy_sse42(uint32_t adler, uint8_t *dst, const uint8_t *src, size_t len);
 extern uint32_t adler32_ssse3(uint32_t adler, const uint8_t *src, size_t len);
 
-#define copy_sub32(a, b, c, d) adler32_fold_copy_sse42(a, b, c, d)
-#define sub32(a, b, c) adler32_ssse3(a, b, c)
-#else
-#define copy_sub32(a, b, c, d) adler32_copy_len_16(adler0, c, b, d, adler1)
-#define sub32(a, b, c) adler32_len_16(adler0, b, c, adler1)
-#endif
-
 static inline uint32_t adler32_fold_copy_impl(uint32_t adler, uint8_t *dst, const uint8_t *src, size_t len, const int COPY) {
     if (src == NULL) return 1L;
     if (len == 0) return adler;
@@ -43,9 +35,9 @@ rem_peel:
         }
     } else if (len < 32) {
         if (COPY) {
-            return copy_sub32(adler, dst, src, len);
+            return adler32_fold_copy_sse42(adler, dst, src, len);
         } else {
-            return sub32(adler, src, len);
+            return adler32_ssse3(adler, src, len);
         }
     }
 
diff --git a/arch/x86/adler32_avx512.c b/arch/x86/adler32_avx512.c
index 83d521ab..cc6bfa64 100644
--- a/arch/x86/adler32_avx512.c
+++ b/arch/x86/adler32_avx512.c
@@ -32,13 +32,7 @@ rem_peel:
             _mm512_mask_storeu_epi8(dst, storemask, copy_vec);
         }
 
-#ifdef X86_AVX2
         return adler32_avx2(adler, src, len);
-#elif defined(X86_SSSE3)
-        return adler32_ssse3(adler, src, len);
-#else
-        return adler32_len_16(adler0, src, len, adler1);
-#endif
     }
 
     __m512i vbuf, vs1_0, vs3;
diff --git a/arch/x86/adler32_avx512_vnni.c b/arch/x86/adler32_avx512_vnni.c
index b8ccd9ac..3dd26d7d 100644
--- a/arch/x86/adler32_avx512_vnni.c
+++ b/arch/x86/adler32_avx512_vnni.c
@@ -27,20 +27,10 @@ Z_INTERNAL uint32_t adler32_avx512_vnni(uint32_t adler, const uint8_t *src, size
 
 rem_peel:
     if (len < 32)
-#if defined(X86_SSSE3)
         return adler32_ssse3(adler, src, len);
-#else
-        return adler32_len_16(adler0, src, len, adler1);
-#endif
 
     if (len < 64)
-#ifdef X86_AVX2
         return adler32_avx2(adler, src, len);
-#elif defined(X86_SSE3)
-        return adler32_ssse3(adler, src, len);
-#else
-        return adler32_len_16(adler0, src, len, adler1);
-#endif
 
     const __m512i dot2v = _mm512_set_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
                                           20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37,
@@ -134,11 +124,7 @@ rem_peel_copy:
         __m256i copy_vec = _mm256_maskz_loadu_epi8(storemask, src);
         _mm256_mask_storeu_epi8(dst, storemask, copy_vec);
 
-#if defined(X86_SSSE3)
         return adler32_ssse3(adler, src, len);
-#else
-        return adler32_len_16(adler0, src, len, adler1);
-#endif
     }
 
     const __m256i dot2v = _mm256_set_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
diff --git a/arch/x86/chunkset_ssse3.c b/arch/x86/chunkset_ssse3.c
index c06d1b37..7ff16a4f 100644
--- a/arch/x86/chunkset_ssse3.c
+++ b/arch/x86/chunkset_ssse3.c
@@ -7,7 +7,7 @@
 /* This requires SSE2 support. While it's implicit with SSSE3, we can minimize
  * code size by sharing the chunkcopy functions, which will certainly compile
  * to identical machine code */
-#if defined(X86_SSSE3) && defined(X86_SSE2)
+#if defined(X86_SSSE3)
 #include <immintrin.h>
 #include "../generic/chunk_permute_table.h"
 
diff --git a/chunkset_tpl.h b/chunkset_tpl.h
index f909a125..250fdc36 100644
--- a/chunkset_tpl.h
+++ b/chunkset_tpl.h
@@ -5,7 +5,7 @@
 #include "zbuild.h"
 #include <stdlib.h>
 
-#if CHUNK_SIZE == 32 && defined(X86_SSSE3) && defined(X86_SSE2)
+#if CHUNK_SIZE == 32 && defined(X86_SSSE3)
 extern uint8_t* chunkmemset_ssse3(uint8_t *out, unsigned dist, unsigned len);
 #endif
 
@@ -98,7 +98,7 @@ Z_INTERNAL uint8_t* CHUNKMEMSET(uint8_t *out, unsigned dist, unsigned len) {
        Assert(len >= sizeof(uint64_t), "chunkmemset should be called on larger chunks"); */
     Assert(dist > 0, "chunkmemset cannot have a distance 0");
     /* Only AVX2 */
-#if CHUNK_SIZE == 32 && defined(X86_SSSE3) && defined(X86_SSE2)
+#if CHUNK_SIZE == 32 && defined(X86_SSSE3)
     if (len <= 16) {
         return chunkmemset_ssse3(out, dist, len);
     }
diff --git a/functable.c b/functable.c
index 37c4aeef..179126fc 100644
--- a/functable.c
+++ b/functable.c
@@ -110,10 +110,8 @@ static void init_functable(void) {
 #ifdef X86_SSSE3
     if (cf.x86.has_ssse3) {
         ft.adler32 = &adler32_ssse3;
-#  ifdef X86_SSE2
         ft.chunkmemset_safe = &chunkmemset_safe_ssse3;
         ft.inflate_fast = &inflate_fast_ssse3;
-#  endif
     }
 #endif
     // X86 - SSE4.2