From patchwork Tue Jan 16 03:33:44 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 124585 Delivered-To: patch@linaro.org Received: by 10.46.64.148 with SMTP id r20csp876416lje; Mon, 15 Jan 2018 19:40:19 -0800 (PST) X-Google-Smtp-Source: ACJfBouOYiEBCA29wo869qxSu9tEyABwzRFbP1NzyW/D7g4iqB1afFobSYUOdnkVFJrWbp+xpbxd X-Received: by 10.37.2.79 with SMTP id 76mr10320800ybc.182.1516074019631; Mon, 15 Jan 2018 19:40:19 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1516074019; cv=none; d=google.com; s=arc-20160816; b=DZiedu56X39806llm9DntTsghPOcG8FRbrr7ymrnhf3B3uphO7k9CpoBvzWWm1PrAm G1B2UkYwP/8KZojUORuCyby/+8ry8IWKYY0fmViQzfJuGvIZBQdldFK4LDNI2CAWBehl y9ULntTUc1NxdNCuDEXEUT1u/9+MUkbzXzzu94VdrbK2nzkAJ+LxbRISf5DKZ9LvCGq5 gBHjH9bUBBlU9Vut4OHFWdfmdumn7ZchU9T6Zt7pVQurJ4gz3xvYzKy4Ya8ceYmhSbm7 4HqMevtybIwSYalmSOb3uUS88IzfgKJZTLkF/HO4sIMEIagS6Js6ySvHk5d9569tNKhe vVzQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:cc:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:to:from:dkim-signature:arc-authentication-results; bh=Zb5Iatq08S+WU5UmQxE4USUhL+pXr4eQ81ZFSAnOVyw=; b=HP9Ox5ncBLinCTXUwvJEeayF1CiKybop5lJgb60NCvMNgnjIuDdPmfjBz7rtEJ9oA/ T4NRoiNKG70UYqEXM6Ihu9MyVbtNJZWEF+r9CTy2jUidPOEdgwU4IEyR90RDY6OGmquA T1GXmmjJUqEKIuEoY96WTVaaF5JzMKQrgJ/lY5aUIhAJL1wW/88cX/6TKbXluHdhwien 92cNqQ/xdH/6pbtUJv9ifHjcHYPndtz2vnkvR18y7dzl8Mll0I28OMouuS4ESf4Fsmm0 p6ugB9RpNcffFN0p73dmm78lfZjC3QMV7AoCZJjELXw4Q8uY1fFCPp8GiIJESIflYlAR 7PDg== ARC-Authentication-Results: i=1; mx.google.com; dkim=fail header.i=@linaro.org header.s=google header.b=XT6YCsog; spf=pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 2001:4830:134:3::11 as permitted sender) smtp.mailfrom=qemu-devel-bounces+patch=linaro.org@nongnu.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=linaro.org Return-Path: Received: from lists.gnu.org (lists.gnu.org. [2001:4830:134:3::11]) by mx.google.com with ESMTPS id q82si280127ywc.232.2018.01.15.19.40.19 for (version=TLS1 cipher=AES128-SHA bits=128/128); Mon, 15 Jan 2018 19:40:19 -0800 (PST) Received-SPF: pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 2001:4830:134:3::11 as permitted sender) client-ip=2001:4830:134:3::11; Authentication-Results: mx.google.com; dkim=fail header.i=@linaro.org header.s=google header.b=XT6YCsog; spf=pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 2001:4830:134:3::11 as permitted sender) smtp.mailfrom=qemu-devel-bounces+patch=linaro.org@nongnu.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=linaro.org Received: from localhost ([::1]:59127 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ebI6k-0000Kl-Uv for patch@linaro.org; Mon, 15 Jan 2018 22:40:18 -0500 Received: from eggs.gnu.org ([2001:4830:134:3::10]:59602) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ebI1C-0004zf-Ca for qemu-devel@nongnu.org; Mon, 15 Jan 2018 22:34:37 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1ebI19-0003EB-MD for qemu-devel@nongnu.org; Mon, 15 Jan 2018 22:34:34 -0500 Received: from mail-pg0-x241.google.com ([2607:f8b0:400e:c05::241]:39486) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1ebI19-0003Dw-Cb for qemu-devel@nongnu.org; Mon, 15 Jan 2018 22:34:31 -0500 Received: by mail-pg0-x241.google.com with SMTP id w17so3009647pgv.6 for ; Mon, 15 Jan 2018 19:34:31 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=Zb5Iatq08S+WU5UmQxE4USUhL+pXr4eQ81ZFSAnOVyw=; b=XT6YCsogG7fgOixowCqGOCFzQ7ruywthL6OiBvlVH6ESCLZkh96AZtTr1r4Mufz4Dv 5eoE+MyG8qFWdA/3A1aALuBHRVP/zDBmdM+lS1F3oAFT3cJWgdJBWPE0F3j2dUM0eZZb S72hnxCjXRjOK/jGygjU4PB0FKWD2K/3p/zZA= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=Zb5Iatq08S+WU5UmQxE4USUhL+pXr4eQ81ZFSAnOVyw=; b=J0AeLdAX7vYeWdq2cYDXjdATYbLNN85kEW5gO7W8htubSDvyXdV8MYGNu7wdxivzfi Z08EPq8ItdIJfvAFDAyzHQSBhYbqap88NunBDqTYF9TOxCKqVYBKQLpWDVB4RX7StF+O LS3KlO2VuF/VKFk1Cx8toKii3ax2nM6xlzrXeM2yw9+Xw1Q7vP+hBFHSTbOCvr751iNT xj77MMa7FmoZ33TeYUdUyLlJTjqZr3LZZG+b54wwVx9ICkNLUCmdUHF3oD09BB7B4G26 krqHfosr1BKIHMh7pEkGHrj/KRVapFB+jiyPW0hBCLXj9IUTrrv/s5p48kjbKcCIZKp+ gx7Q== X-Gm-Message-State: AKGB3mLJ1+xhSaDi3cmb2FHe1faIx+E0B1ESwUAmpD/2cD3n7GKbIERZ TxYqLhRX5cQykYyIy7LoQnj267M0JsY= X-Received: by 10.99.117.18 with SMTP id q18mr24317103pgc.71.1516073669655; Mon, 15 Jan 2018 19:34:29 -0800 (PST) Received: from cloudburst.twiddle.net.com (24-181-135-57.dhcp.knwc.wa.charter.com. [24.181.135.57]) by smtp.gmail.com with ESMTPSA id c83sm1281024pfk.8.2018.01.15.19.34.24 (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Mon, 15 Jan 2018 19:34:28 -0800 (PST) From: Richard Henderson To: qemu-devel@nongnu.org Date: Mon, 15 Jan 2018 19:33:44 -0800 Message-Id: <20180116033404.31532-7-richard.henderson@linaro.org> X-Mailer: git-send-email 2.14.3 In-Reply-To: <20180116033404.31532-1-richard.henderson@linaro.org> References: <20180116033404.31532-1-richard.henderson@linaro.org> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 2607:f8b0:400e:c05::241 Subject: [Qemu-devel] [PATCH v9 06/26] tcg: Add generic vector ops for constant shifts X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: peter.maydell@linaro.org Errors-To: qemu-devel-bounces+patch=linaro.org@nongnu.org Sender: "Qemu-devel" Opcodes are added for scalar and vector shifts, but considering the varied semantics of these do not expose them to the front ends. Do go ahead and provide them in case they are needed for backend expansion. Signed-off-by: Richard Henderson --- accel/tcg/tcg-runtime.h | 15 +++ tcg/tcg-op-gvec.h | 35 ++++++ tcg/tcg-op.h | 5 + tcg/tcg-opc.h | 12 ++ tcg/tcg.h | 3 + accel/tcg/tcg-runtime-gvec.c | 149 ++++++++++++++++++++++++ tcg/tcg-op-gvec.c | 272 +++++++++++++++++++++++++++++++++++++++++++ tcg/tcg-op-vec.c | 45 +++++++ tcg/tcg.c | 12 ++ tcg/README | 29 +++++ 10 files changed, 577 insertions(+) -- 2.14.3 diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h index c6de749134..cb05a755b8 100644 --- a/accel/tcg/tcg-runtime.h +++ b/accel/tcg/tcg-runtime.h @@ -164,6 +164,21 @@ DEF_HELPER_FLAGS_4(gvec_xor, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_andc, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_orc, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shl8i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shl16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shl32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shl64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + +DEF_HELPER_FLAGS_3(gvec_shr8i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shr16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shr32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_shr64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + +DEF_HELPER_FLAGS_3(gvec_sar8i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_sar16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_sar32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) +DEF_HELPER_FLAGS_3(gvec_sar64i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(gvec_zip8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_zip16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_zip32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/tcg/tcg-op-gvec.h b/tcg/tcg-op-gvec.h index f4afcd258e..08f9c5556c 100644 --- a/tcg/tcg-op-gvec.h +++ b/tcg/tcg-op-gvec.h @@ -95,6 +95,25 @@ typedef struct { bool prefer_i64; } GVecGen2; +typedef struct { + /* Expand inline as a 64-bit or 32-bit integer. + Only one of these will be non-NULL. */ + void (*fni8)(TCGv_i64, TCGv_i64, int64_t); + void (*fni4)(TCGv_i32, TCGv_i32, int32_t); + /* Expand inline with a host vector type. */ + void (*fniv)(unsigned, TCGv_vec, TCGv_vec, int64_t); + /* Expand out-of-line helper w/descriptor. */ + gen_helper_gvec_2 *fno; + /* The opcode, if any, to which this corresponds. */ + TCGOpcode opc; + /* The vector element size, if applicable. */ + uint8_t vece; + /* Prefer i64 to v64. */ + bool prefer_i64; + /* Load dest as a 3rd source operand. */ + bool load_dest; +} GVecGen2i; + typedef struct { /* Expand inline as a 64-bit or 32-bit integer. Only one of these will be non-NULL. */ @@ -137,6 +156,8 @@ typedef struct { void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, const GVecGen2 *); +void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + uint32_t maxsz, int64_t c, const GVecGen2i *); void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz, const GVecGen3 *); void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs, @@ -179,6 +200,13 @@ void tcg_gen_gvec_dup16i(uint32_t dofs, uint32_t s, uint32_t m, uint16_t x); void tcg_gen_gvec_dup32i(uint32_t dofs, uint32_t s, uint32_t m, uint32_t x); void tcg_gen_gvec_dup64i(uint32_t dofs, uint32_t s, uint32_t m, uint64_t x); +void tcg_gen_gvec_shli(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t maxsz, unsigned shift); +void tcg_gen_gvec_shri(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t maxsz, unsigned shift); +void tcg_gen_gvec_sari(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t maxsz, unsigned shift); + void tcg_gen_gvec_zipl(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); void tcg_gen_gvec_ziph(unsigned vece, uint32_t dofs, uint32_t aofs, @@ -209,3 +237,10 @@ void tcg_gen_vec_add32_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b); void tcg_gen_vec_sub8_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b); void tcg_gen_vec_sub16_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b); void tcg_gen_vec_sub32_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b); + +void tcg_gen_vec_shl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); +void tcg_gen_vec_shl16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); +void tcg_gen_vec_shr8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); +void tcg_gen_vec_shr16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); +void tcg_gen_vec_sar8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); +void tcg_gen_vec_sar16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t); diff --git a/tcg/tcg-op.h b/tcg/tcg-op.h index 87be11a90d..c002523bea 100644 --- a/tcg/tcg-op.h +++ b/tcg/tcg-op.h @@ -927,6 +927,11 @@ void tcg_gen_andc_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_orc_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_not_vec(unsigned vece, TCGv_vec r, TCGv_vec a); void tcg_gen_neg_vec(unsigned vece, TCGv_vec r, TCGv_vec a); + +void tcg_gen_shli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); +void tcg_gen_shri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); +void tcg_gen_sari_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i); + void tcg_gen_zipl_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_ziph_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); void tcg_gen_uzpe_vec(unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b); diff --git a/tcg/tcg-opc.h b/tcg/tcg-opc.h index c911d62442..a085fc077b 100644 --- a/tcg/tcg-opc.h +++ b/tcg/tcg-opc.h @@ -229,6 +229,18 @@ DEF(andc_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_andc_vec)) DEF(orc_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_orc_vec)) DEF(not_vec, 1, 1, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_not_vec)) +DEF(shli_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec)) +DEF(shri_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec)) +DEF(sari_vec, 1, 1, 1, IMPLVEC | IMPL(TCG_TARGET_HAS_shi_vec)) + +DEF(shls_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) +DEF(shrs_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) +DEF(sars_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shs_vec)) + +DEF(shlv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) +DEF(shrv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) +DEF(sarv_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_shv_vec)) + DEF(zipl_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_zip_vec)) DEF(ziph_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_zip_vec)) DEF(uzpe_vec, 1, 2, 0, IMPLVEC | IMPL(TCG_TARGET_HAS_uzp_vec)) diff --git a/tcg/tcg.h b/tcg/tcg.h index bccdaa5eca..49ee07b02e 100644 --- a/tcg/tcg.h +++ b/tcg/tcg.h @@ -178,6 +178,9 @@ typedef uint64_t TCGRegSet; #define TCG_TARGET_HAS_not_vec 0 #define TCG_TARGET_HAS_andc_vec 0 #define TCG_TARGET_HAS_orc_vec 0 +#define TCG_TARGET_HAS_shi_vec 0 +#define TCG_TARGET_HAS_shs_vec 0 +#define TCG_TARGET_HAS_shv_vec 0 #define TCG_TARGET_HAS_zip_vec 0 #define TCG_TARGET_HAS_uzp_vec 0 #define TCG_TARGET_HAS_trn_vec 0 diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c index 628df811b2..fba62f1192 100644 --- a/accel/tcg/tcg-runtime-gvec.c +++ b/accel/tcg/tcg-runtime-gvec.c @@ -36,6 +36,11 @@ typedef uint16_t vec16 __attribute__((vector_size(16))); typedef uint32_t vec32 __attribute__((vector_size(16))); typedef uint64_t vec64 __attribute__((vector_size(16))); +typedef int8_t svec8 __attribute__((vector_size(16))); +typedef int16_t svec16 __attribute__((vector_size(16))); +typedef int32_t svec32 __attribute__((vector_size(16))); +typedef int64_t svec64 __attribute__((vector_size(16))); + static inline void clear_high(void *d, intptr_t oprsz, uint32_t desc) { intptr_t maxsz = simd_maxsz(desc); @@ -294,6 +299,150 @@ void HELPER(gvec_orc)(void *d, void *a, void *b, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_shl8i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec8)) { + *(vec8 *)(d + i) = *(vec8 *)(a + i) << shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shl16i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec16)) { + *(vec16 *)(d + i) = *(vec16 *)(a + i) << shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shl32i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec32)) { + *(vec32 *)(d + i) = *(vec32 *)(a + i) << shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shl64i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) << shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shr8i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec8)) { + *(vec8 *)(d + i) = *(vec8 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shr16i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec16)) { + *(vec16 *)(d + i) = *(vec16 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shr32i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec32)) { + *(vec32 *)(d + i) = *(vec32 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_shr64i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_sar8i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec8)) { + *(svec8 *)(d + i) = *(svec8 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_sar16i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec16)) { + *(svec16 *)(d + i) = *(svec16 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_sar32i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec32)) { + *(svec32 *)(d + i) = *(svec32 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_sar64i)(void *d, void *a, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + int shift = simd_data(desc); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(svec64 *)(d + i) = *(svec64 *)(a + i) >> shift; + } + clear_high(d, oprsz, desc); +} + /* The size of the alloca in the following is currently bounded to 2k. */ #define DO_ZIP(NAME, TYPE) \ diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index 18f5da2c8b..4210b6787a 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -534,6 +534,26 @@ static void expand_2_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz, tcg_temp_free_i32(t0); } +static void expand_2i_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + int32_t c, bool load_dest, + void (*fni)(TCGv_i32, TCGv_i32, int32_t)) +{ + TCGv_i32 t0 = tcg_temp_new_i32(); + TCGv_i32 t1 = tcg_temp_new_i32(); + uint32_t i; + + for (i = 0; i < oprsz; i += 4) { + tcg_gen_ld_i32(t0, cpu_env, aofs + i); + if (load_dest) { + tcg_gen_ld_i32(t1, cpu_env, dofs + i); + } + fni(t1, t0, c); + tcg_gen_st_i32(t1, cpu_env, dofs + i); + } + tcg_temp_free_i32(t0); + tcg_temp_free_i32(t1); +} + /* Expand OPSZ bytes worth of three-operand operations using i32 elements. */ static void expand_3_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, bool load_dest, @@ -597,6 +617,26 @@ static void expand_2_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, tcg_temp_free_i64(t0); } +static void expand_2i_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + int64_t c, bool load_dest, + void (*fni)(TCGv_i64, TCGv_i64, int64_t)) +{ + TCGv_i64 t0 = tcg_temp_new_i64(); + TCGv_i64 t1 = tcg_temp_new_i64(); + uint32_t i; + + for (i = 0; i < oprsz; i += 8) { + tcg_gen_ld_i64(t0, cpu_env, aofs + i); + if (load_dest) { + tcg_gen_ld_i64(t1, cpu_env, dofs + i); + } + fni(t1, t0, c); + tcg_gen_st_i64(t1, cpu_env, dofs + i); + } + tcg_temp_free_i64(t0); + tcg_temp_free_i64(t1); +} + /* Expand OPSZ bytes worth of three-operand operations using i64 elements. */ static void expand_3_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, bool load_dest, @@ -661,6 +701,29 @@ static void expand_2_vec(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_temp_free_vec(t0); } +/* Expand OPSZ bytes worth of two-vector operands and an immediate operand + using host vectors. */ +static void expand_2i_vec(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t tysz, TCGType type, + int64_t c, bool load_dest, + void (*fni)(unsigned, TCGv_vec, TCGv_vec, int64_t)) +{ + TCGv_vec t0 = tcg_temp_new_vec(type); + TCGv_vec t1 = tcg_temp_new_vec(type); + uint32_t i; + + for (i = 0; i < oprsz; i += tysz) { + tcg_gen_ld_vec(t0, cpu_env, aofs + i); + if (load_dest) { + tcg_gen_ld_vec(t1, cpu_env, dofs + i); + } + fni(vece, t1, t0, c); + tcg_gen_st_vec(t1, cpu_env, dofs + i); + } + tcg_temp_free_vec(t0); + tcg_temp_free_vec(t1); +} + /* Expand OPSZ bytes worth of three-operand operations using host vectors. */ static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, @@ -760,6 +823,51 @@ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, } } +void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + uint32_t maxsz, int64_t c, const GVecGen2i *g) +{ + check_size_align(oprsz, maxsz, dofs | aofs); + check_overlap_2(dofs, aofs, maxsz); + + /* Recall that ARM SVE allows vector sizes that are not a power of 2. + Expand with successively smaller host vector sizes. The intent is + that e.g. oprsz == 80 would be expanded with 2x32 + 1x16. */ + + if (TCG_TARGET_HAS_v256 && g->fniv && check_size_impl(oprsz, 32) + && (!g->opc || tcg_can_emit_vec_op(g->opc, TCG_TYPE_V256, g->vece))) { + uint32_t done = QEMU_ALIGN_DOWN(oprsz, 32); + expand_2i_vec(g->vece, dofs, aofs, done, 32, TCG_TYPE_V256, + c, g->load_dest, g->fniv); + dofs += done; + aofs += done; + oprsz -= done; + maxsz -= done; + } + + if (TCG_TARGET_HAS_v128 && g->fniv && check_size_impl(oprsz, 16) + && (!g->opc || tcg_can_emit_vec_op(g->opc, TCG_TYPE_V128, g->vece))) { + expand_2i_vec(g->vece, dofs, aofs, oprsz, 16, TCG_TYPE_V128, + c, g->load_dest, g->fniv); + } else if (TCG_TARGET_HAS_v64 && !g->prefer_i64 + && g->fniv && check_size_impl(oprsz, 8) + && (!g->opc + || tcg_can_emit_vec_op(g->opc, TCG_TYPE_V64, g->vece))) { + expand_2i_vec(g->vece, dofs, aofs, oprsz, 8, TCG_TYPE_V64, + c, g->load_dest, g->fniv); + } else if (g->fni8 && check_size_impl(oprsz, 8)) { + expand_2i_i64(dofs, aofs, oprsz, c, g->load_dest, g->fni8); + } else if (g->fni4 && check_size_impl(oprsz, 4)) { + expand_2i_i32(dofs, aofs, oprsz, c, g->load_dest, g->fni4); + } else { + tcg_gen_gvec_2_ool(dofs, aofs, oprsz, maxsz, c, g->fno); + return; + } + + if (oprsz < maxsz) { + expand_clr(dofs + oprsz, maxsz - oprsz); + } +} + /* Expand a vector three-operand operation. */ void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g) @@ -1294,6 +1402,170 @@ void tcg_gen_gvec_orc(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g); } +void tcg_gen_vec_shl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t mask = ((0xff << c) & 0xff) * (-1ull / 0xff); + tcg_gen_shli_i64(d, a, c); + tcg_gen_andi_i64(d, d, mask); +} + +void tcg_gen_vec_shl16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t mask = ((0xffff << c) & 0xffff) * (-1ull / 0xffff); + tcg_gen_shli_i64(d, a, c); + tcg_gen_andi_i64(d, d, mask); +} + +void tcg_gen_gvec_shli(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t maxsz, unsigned shift) +{ + static const GVecGen2i g[4] = { + { .fni8 = tcg_gen_vec_shl8i_i64, + .fniv = tcg_gen_shli_vec, + .fno = gen_helper_gvec_shl8i, + .opc = INDEX_op_shli_vec, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_shl16i_i64, + .fniv = tcg_gen_shli_vec, + .fno = gen_helper_gvec_shl16i, + .opc = INDEX_op_shli_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_shli_i32, + .fniv = tcg_gen_shli_vec, + .fno = gen_helper_gvec_shl32i, + .opc = INDEX_op_shli_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_shli_i64, + .fniv = tcg_gen_shli_vec, + .fno = gen_helper_gvec_shl64i, + .opc = INDEX_op_shli_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_debug_assert(shift < (8 << vece)); + if (shift == 0) { + tcg_gen_gvec_mov(vece, dofs, aofs, oprsz, maxsz); + } else { + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, shift, &g[vece]); + } +} + +void tcg_gen_vec_shr8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t mask = (0xff >> c) * (-1ull / 0xff); + tcg_gen_shri_i64(d, a, c); + tcg_gen_andi_i64(d, d, mask); +} + +void tcg_gen_vec_shr16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t mask = (0xffff >> c) * (-1ull / 0xffff); + tcg_gen_shri_i64(d, a, c); + tcg_gen_andi_i64(d, d, mask); +} + +void tcg_gen_gvec_shri(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t maxsz, unsigned shift) +{ + static const GVecGen2i g[4] = { + { .fni8 = tcg_gen_vec_shr8i_i64, + .fniv = tcg_gen_shri_vec, + .fno = gen_helper_gvec_shr8i, + .opc = INDEX_op_shri_vec, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_shr16i_i64, + .fniv = tcg_gen_shri_vec, + .fno = gen_helper_gvec_shr16i, + .opc = INDEX_op_shri_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_shri_i32, + .fniv = tcg_gen_shri_vec, + .fno = gen_helper_gvec_shr32i, + .opc = INDEX_op_shri_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_shri_i64, + .fniv = tcg_gen_shri_vec, + .fno = gen_helper_gvec_shr64i, + .opc = INDEX_op_shri_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_debug_assert(shift < (8 << vece)); + if (shift == 0) { + tcg_gen_gvec_mov(vece, dofs, aofs, oprsz, maxsz); + } else { + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, shift, &g[vece]); + } +} + +void tcg_gen_vec_sar8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t s_mask = (0x80 >> c) * (-1ull / 0xff); + uint64_t c_mask = (0xff >> c) * (-1ull / 0xff); + TCGv_i64 s = tcg_temp_new_i64(); + + tcg_gen_shri_i64(d, a, c); + tcg_gen_andi_i64(s, d, s_mask); /* isolate (shifted) sign bit */ + tcg_gen_muli_i64(s, s, (2 << c) - 2); /* replicate isolated signs */ + tcg_gen_andi_i64(d, d, c_mask); /* clear out bits above sign */ + tcg_gen_or_i64(d, d, s); /* include sign extension */ + tcg_temp_free_i64(s); +} + +void tcg_gen_vec_sar16i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) +{ + uint64_t s_mask = (0x8000 >> c) * (-1ull / 0xffff); + uint64_t c_mask = (0xffff >> c) * (-1ull / 0xffff); + TCGv_i64 s = tcg_temp_new_i64(); + + tcg_gen_shri_i64(d, a, c); + tcg_gen_andi_i64(s, d, s_mask); /* isolate (shifted) sign bit */ + tcg_gen_andi_i64(d, d, c_mask); /* clear out bits above sign */ + tcg_gen_muli_i64(s, s, (2 << c) - 2); /* replicate isolated signs */ + tcg_gen_or_i64(d, d, s); /* include sign extension */ + tcg_temp_free_i64(s); +} + +void tcg_gen_gvec_sari(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t maxsz, unsigned shift) +{ + static const GVecGen2i g[4] = { + { .fni8 = tcg_gen_vec_sar8i_i64, + .fniv = tcg_gen_sari_vec, + .fno = gen_helper_gvec_sar8i, + .opc = INDEX_op_sari_vec, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_sar16i_i64, + .fniv = tcg_gen_sari_vec, + .fno = gen_helper_gvec_sar16i, + .opc = INDEX_op_sari_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_sari_i32, + .fniv = tcg_gen_sari_vec, + .fno = gen_helper_gvec_sar32i, + .opc = INDEX_op_sari_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_sari_i64, + .fniv = tcg_gen_sari_vec, + .fno = gen_helper_gvec_sar64i, + .opc = INDEX_op_sari_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_debug_assert(shift < (8 << vece)); + if (shift == 0) { + tcg_gen_gvec_mov(vece, dofs, aofs, oprsz, maxsz); + } else { + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, shift, &g[vece]); + } +} + static void do_zip(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz, bool high) diff --git a/tcg/tcg-op-vec.c b/tcg/tcg-op-vec.c index a5d0ff89c3..b54d3de19f 100644 --- a/tcg/tcg-op-vec.c +++ b/tcg/tcg-op-vec.c @@ -381,6 +381,51 @@ void tcg_gen_neg_vec(unsigned vece, TCGv_vec r, TCGv_vec a) } } +static void do_shifti(TCGOpcode opc, unsigned vece, + TCGv_vec r, TCGv_vec a, int64_t i) +{ + TCGTemp *rt = tcgv_vec_temp(r); + TCGTemp *at = tcgv_vec_temp(a); + TCGArg ri = temp_arg(rt); + TCGArg ai = temp_arg(at); + TCGType type = rt->base_type; + int can; + + tcg_debug_assert(at->base_type == type); + tcg_debug_assert(i >= 0 && i < (8 << vece)); + + if (i == 0) { + tcg_gen_mov_vec(r, a); + return; + } + + can = tcg_can_emit_vec_op(opc, type, vece); + if (can > 0) { + vec_gen_3(opc, type, vece, ri, ai, i); + } else { + /* We leave the choice of expansion via scalar or vector shift + to the target. Often, but not always, dupi can feed a vector + shift easier than a scalar. */ + tcg_debug_assert(can < 0); + tcg_expand_vec_op(opc, type, vece, ri, ai, i); + } +} + +void tcg_gen_shli_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i) +{ + do_shifti(INDEX_op_shli_vec, vece, r, a, i); +} + +void tcg_gen_shri_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i) +{ + do_shifti(INDEX_op_shri_vec, vece, r, a, i); +} + +void tcg_gen_sari_vec(unsigned vece, TCGv_vec r, TCGv_vec a, int64_t i) +{ + do_shifti(INDEX_op_sari_vec, vece, r, a, i); +} + static void do_interleave(TCGOpcode opc, unsigned vece, TCGv_vec r, TCGv_vec a, TCGv_vec b) { diff --git a/tcg/tcg.c b/tcg/tcg.c index 33e3c03cbc..f82d6e80b0 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -1403,6 +1403,18 @@ bool tcg_op_supported(TCGOpcode op) return have_vec && TCG_TARGET_HAS_andc_vec; case INDEX_op_orc_vec: return have_vec && TCG_TARGET_HAS_orc_vec; + case INDEX_op_shli_vec: + case INDEX_op_shri_vec: + case INDEX_op_sari_vec: + return have_vec && TCG_TARGET_HAS_shi_vec; + case INDEX_op_shls_vec: + case INDEX_op_shrs_vec: + case INDEX_op_sars_vec: + return have_vec && TCG_TARGET_HAS_shs_vec; + case INDEX_op_shlv_vec: + case INDEX_op_shrv_vec: + case INDEX_op_sarv_vec: + return have_vec && TCG_TARGET_HAS_shv_vec; case INDEX_op_zipl_vec: case INDEX_op_ziph_vec: return have_vec && TCG_TARGET_HAS_zip_vec; diff --git a/tcg/README b/tcg/README index 8ab8d3ab7e..75db47922d 100644 --- a/tcg/README +++ b/tcg/README @@ -561,6 +561,35 @@ E.g. VECL=1 -> 64 << 1 -> v128, and VECE=2 -> 1 << 2 -> i32. Similarly, logical operations with and without compliment. Note that VECE is unused. +* shli_vec v0, v1, i2 +* shls_vec v0, v1, s2 + + Shift all elements from v1 by a scalar i2/s2. I.e. + + for (i = 0; i < VECL/VECE; ++i) { + v0[i] = v1[i] << s2; + } + +* shri_vec v0, v1, i2 +* sari_vec v0, v1, i2 +* shrs_vec v0, v1, s2 +* sars_vec v0, v1, s2 + + Similarly for logical and arithmetic right shift. + +* shlv_vec v0, v1, v2 + + Shift elements from v1 by elements from v2. I.e. + + for (i = 0; i < VECL/VECE; ++i) { + v0[i] = v1[i] << v2[i]; + } + +* shrv_vec v0, v1, v2 +* sarv_vec v0, v1, v2 + + Similarly for logical and arithmetic right shift. + * zipl_vec v0, v1, v2 * ziph_vec v0, v1, v2