An end-to-end guarantee of software correctness by formal verification must consider two sources of bugs. First, the verification tool must be correct. Second, programs are often verified at the source level, before being compiled. Hence, compilers should also be trustworthy. Verifiers and compilers' complexity is increasing. To simplify code analysis and manipulation, these tools rely on intermediate representations (IR) of programs, that provide structural and semantic properties. This thesis gives a formal, semantic account on IRs, so that they can also be leveraged in the formal proof of such tools. We first study a register-based IR of Java bytecode used in compilers and verifiers. We specify the IR generation by a semantic theorem stating what the transformation preserves, e.g. object initialization or exceptions, but also what it modifies and how, e.g. object allocation. We implement this IR in Sawja, a Java static analysis toolbench. Then, we study the Static Single Assignment (SSA) form, an IR widely used in modern compilers and verifiers. We implement and prove in Coq an SSA middle-end for the CompCert C compiler. For the proof of SSA optimizations, we identify a key semantic property of SSA, allowing for equational reasoning. Finally, we study the semantics of concurrent Java IRs. Due to instruction reorderings performed by the compiler and the hardware, the current definition of the Java Memory Model (JMM) is complex, and unfortunately formally flawed. Targetting x86 architectures, we identify a subset of the JMM that is intuitive and tractable in formal proofs. We characterize the reorderings it allows, and factor out a proof common to the IRs of a compiler.

La vérification formelle de programme n'apporte pas de garantie complète si l'outil de vérification est incorrect. Et, si un programme est vérifié au niveau source, le compilateur pourrait introduire des bugs. Les compilateurs et vérifieurs actuels sont complexes. Pour simplifier l'analyse et la transformation de code, ils utilisent des représentations intermédiaires (IR) de programme, qui ont de fortes propriétés structurelles et sémantiques. Cette thèse étudie d'un point de vue sémantique et formel les IRs, afin de faciliter la preuve de ces outils. Nous étudions d'abord une IR basée registre du bytecode Java. Nous prouvons un théorème sur sa génération, explicitant ce que la transformation préserve (l'initialisation d'objet, les exceptions) et ce qu'elle modifie et comment (l'ordre d'allocation). Nous implantons l'IR dans Sawja, un outil de développement d'analyses statiques de Java. Nous étudions aussi la forme SSA, une IR au coeur des compilateurs et vérifieurs modernes. Nous implantons et prouvons en Coq un middle-end SSA pour le compilateur C CompCert. Pour la preuve des optimisations, nous prouvons un invariant sémantique de SSA clé pour le raisonnement équationnel. Enfin, nous étudions la sémantique des IRs de Java concurrent. La définition actuelle du Java Memory Model (JMM) autorise les optimisations aggressives des compilateurs et des architectures parallèles. Complexe, elle est formellement cassée. Ciblant les architectures x86, nous proposons un sous-ensemble du JMM intuitif et adapté à la preuve formelle. Nous le caractérisons par ses réordonnancements, et factorisons cette preuve sur les IRs d'un compilateur. / An end-to-end guarantee of software correctness by formal verification must consider two sources of bugs. First, the verification tool must be correct. Second, programs are often verified at the source level, before being compiled. Hence, compilers should also be trustworthy. Verifiers and compilers' complexity is increasing. To simplify code analysis and manipulation, these tools rely on intermediate representations (IR) of programs, that provide structural and semantic properties. This thesis gives a formal, semantic account on IRs, so that they can also be leveraged in the formal proof of such tools. We first study a register-based IR of Java bytecode used in compilers and verifiers. We specify the IR generation by a semantic theorem stating what the transformation preserves, e.g. object initialization or exceptions, but also what it modifies and how, e.g. object allocation. We implement this IR in Sawja, a Java static analysis toolbench. Then, we study the Static Single Assignment (SSA) form, an IR widely used in modern compilers and verifiers. We implement and prove in Coq an SSA middle-end for the CompCert C compiler. For the proof of SSA optimizations, we identify a key semantic property of SSA, allowing for equational reasoning. Finally, we study the semantics of concurrent Java IRs. Due to instruction reorderings performed by the compiler and the hardware, the current definition of the Java Memory Model (JMM) is complex, and unfortunately formally flawed. Targetting x86 architectures, we identify a subset of the JMM that is intuitive and tractable in formal proofs. We characterize the reorderings it allows, and factor out a proof common to the IRs of a compiler.